recreate this patch from https://github.com/llvm/llvm-project/pull/153894 It seems ISel sliently ignore the `i64 = zext i16` with a chained `reg_sequence` pattern and thus this is causing a selection failure in hip test. Recreate a new patch with an alternative pattern, and added a ll test global-extload-gfx11plus.ll
This commit is contained in:
Brox Chen
GitHub
parent
2f237670b1
commit
c50ed05cad
@ -3078,6 +3078,8 @@ def : GCNPat<
|
||||
}
|
||||
} // AddedComplexity = 1
|
||||
|
||||
foreach p = [NotHasTrue16BitInsts, UseFakeTrue16Insts] in
|
||||
let True16Predicate = p in {
|
||||
def : GCNPat<
|
||||
(i32 (DivergentUnaryFrag<zext> i16:$src)),
|
||||
(V_AND_B32_e64 (S_MOV_B32 (i32 0xffff)), $src)
|
||||
@ -3092,7 +3094,28 @@ def : GCNPat<
|
||||
|
||||
def : GCNPat<
|
||||
(i32 (zext (i16 (bitconvert fp16_zeros_high_16bits:$src)))),
|
||||
(COPY VSrc_b16:$src)>;
|
||||
(COPY VSrc_b16:$src)
|
||||
>;
|
||||
}
|
||||
|
||||
let True16Predicate = UseRealTrue16Insts in {
|
||||
def : GCNPat<
|
||||
(i32 (DivergentUnaryFrag<zext> i16:$src)),
|
||||
(REG_SEQUENCE VGPR_32, VGPR_16:$src, lo16, (V_MOV_B16_t16_e64 0, (i16 0), 0), hi16)
|
||||
>;
|
||||
|
||||
def : GCNPat<
|
||||
(i64 (DivergentUnaryFrag<zext> i16:$src)),
|
||||
(REG_SEQUENCE VReg_64,
|
||||
(INSERT_SUBREG (i32 (V_MOV_B32_e32 (i32 0))), VGPR_16:$src, lo16), sub0,
|
||||
(S_MOV_B32 (i32 0)), sub1)
|
||||
>;
|
||||
|
||||
def : GCNPat<
|
||||
(i32 (zext (i16 (bitconvert fp16_zeros_high_16bits:$src)))),
|
||||
(REG_SEQUENCE VGPR_32, VGPR_16:$src, lo16, (V_MOV_B16_t16_e64 0, (i16 0), 0), hi16)
|
||||
>;
|
||||
}
|
||||
|
||||
def : GCNPat <
|
||||
(i32 (trunc i64:$a)),
|
||||
|
||||
@ -164,7 +164,7 @@ define zeroext i16 @v_mul_i16_zeroext(i16 zeroext %num, i16 zeroext %den) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: v_mul_i16_zeroext:
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
@ -2257,8 +2257,8 @@ define i32 @bitcast_v4i8_to_i32(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v4
|
||||
@ -2273,19 +2273,17 @@ define i32 @bitcast_v4i8_to_i32(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB22_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v2, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB22_2
|
||||
; GFX11-TRUE16-NEXT: .LBB22_4: ; %cmp.true
|
||||
@ -2295,16 +2293,14 @@ define i32 @bitcast_v4i8_to_i32(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
@ -4506,8 +4502,8 @@ define float @bitcast_v4i8_to_f32(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v4
|
||||
@ -4522,19 +4518,17 @@ define float @bitcast_v4i8_to_f32(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB42_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v2, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB42_2
|
||||
; GFX11-TRUE16-NEXT: .LBB42_4: ; %cmp.true
|
||||
@ -4544,16 +4538,14 @@ define float @bitcast_v4i8_to_f32(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
@ -6467,8 +6459,8 @@ define <2 x i16> @bitcast_v4i8_to_v2i16(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v4
|
||||
@ -6483,19 +6475,17 @@ define <2 x i16> @bitcast_v4i8_to_v2i16(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB58_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v2, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB58_2
|
||||
; GFX11-TRUE16-NEXT: .LBB58_4: ; %cmp.true
|
||||
@ -6505,16 +6495,14 @@ define <2 x i16> @bitcast_v4i8_to_v2i16(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
@ -8116,8 +8104,8 @@ define <2 x half> @bitcast_v4i8_to_v2f16(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v4
|
||||
@ -8132,19 +8120,17 @@ define <2 x half> @bitcast_v4i8_to_v2f16(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB70_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v2, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB70_2
|
||||
; GFX11-TRUE16-NEXT: .LBB70_4: ; %cmp.true
|
||||
@ -8154,16 +8140,14 @@ define <2 x half> @bitcast_v4i8_to_v2f16(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
@ -9479,8 +9463,8 @@ define <2 x bfloat> @bitcast_v4i8_to_v2bf16(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v4
|
||||
@ -9495,19 +9479,17 @@ define <2 x bfloat> @bitcast_v4i8_to_v2bf16(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB78_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v2, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB78_2
|
||||
; GFX11-TRUE16-NEXT: .LBB78_4: ; %cmp.true
|
||||
@ -9517,16 +9499,14 @@ define <2 x bfloat> @bitcast_v4i8_to_v2bf16(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
@ -10193,8 +10173,8 @@ define <1 x i32> @bitcast_v4i8_to_v1i32(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v4
|
||||
@ -10209,19 +10189,17 @@ define <1 x i32> @bitcast_v4i8_to_v1i32(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB82_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v2, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB82_2
|
||||
; GFX11-TRUE16-NEXT: .LBB82_4: ; %cmp.true
|
||||
@ -10231,16 +10209,14 @@ define <1 x i32> @bitcast_v4i8_to_v1i32(<4 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
@ -3065,13 +3065,12 @@ define i64 @bitcast_v8i8_to_i64(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v8
|
||||
@ -3085,61 +3084,53 @@ define i64 @bitcast_v8i8_to_i64(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
; GFX11-TRUE16-NEXT: .LBB26_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v4
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v1.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v5, v2
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB26_2
|
||||
; GFX11-TRUE16-NEXT: .LBB26_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v5.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v3.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v3.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v4.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v2, v5
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v3.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v3.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v2.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -6214,13 +6205,12 @@ define double @bitcast_v8i8_to_f64(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v8
|
||||
@ -6234,61 +6224,53 @@ define double @bitcast_v8i8_to_f64(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
; GFX11-TRUE16-NEXT: .LBB50_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v4
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v1.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v5, v2
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB50_2
|
||||
; GFX11-TRUE16-NEXT: .LBB50_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v5.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v3.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v3.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v4.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v2, v5
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v3.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v3.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v2.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -9063,13 +9045,12 @@ define <2 x i32> @bitcast_v8i8_to_v2i32(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v8
|
||||
@ -9083,61 +9064,53 @@ define <2 x i32> @bitcast_v8i8_to_v2i32(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
; GFX11-TRUE16-NEXT: .LBB70_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v4
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v1.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v5, v2
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB70_2
|
||||
; GFX11-TRUE16-NEXT: .LBB70_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v5.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v3.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v3.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v4.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v2, v5
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v3.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v3.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v2.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -11603,13 +11576,12 @@ define <2 x float> @bitcast_v8i8_to_v2f32(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v8
|
||||
@ -11623,61 +11595,53 @@ define <2 x float> @bitcast_v8i8_to_v2f32(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
; GFX11-TRUE16-NEXT: .LBB86_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v4
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v1.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v5, v2
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB86_2
|
||||
; GFX11-TRUE16-NEXT: .LBB86_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v5.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v3.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v3.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v4.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v2, v5
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v3.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v3.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v2.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -13829,13 +13793,12 @@ define <4 x i16> @bitcast_v8i8_to_v4i16(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v8
|
||||
@ -13849,61 +13812,53 @@ define <4 x i16> @bitcast_v8i8_to_v4i16(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
; GFX11-TRUE16-NEXT: .LBB98_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v4
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v1.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v5, v2
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB98_2
|
||||
; GFX11-TRUE16-NEXT: .LBB98_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v5.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v3.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v3.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v4.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v2, v5
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v3.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v3.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v2.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -15655,13 +15610,12 @@ define <4 x half> @bitcast_v8i8_to_v4f16(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v8
|
||||
@ -15675,61 +15629,53 @@ define <4 x half> @bitcast_v8i8_to_v4f16(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
; GFX11-TRUE16-NEXT: .LBB106_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v4
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v1.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v5, v2
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB106_2
|
||||
; GFX11-TRUE16-NEXT: .LBB106_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v5.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v3.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v3.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v4.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v2, v5
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v3.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v3.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v2.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -16966,13 +16912,12 @@ define <4 x bfloat> @bitcast_v8i8_to_v4bf16(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v8
|
||||
@ -16986,61 +16931,53 @@ define <4 x bfloat> @bitcast_v8i8_to_v4bf16(<8 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
; GFX11-TRUE16-NEXT: .LBB110_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v4
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v0.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v1.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v5, v2
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB110_2
|
||||
; GFX11-TRUE16-NEXT: .LBB110_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v5.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v3.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v4.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v3.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v4.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v2, v5
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v2.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v3.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v3.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v2.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
|
||||
@ -1102,15 +1102,16 @@ define <3 x i32> @bitcast_v12i8_to_v3i32(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v8.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v9.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v8.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v12
|
||||
@ -1125,80 +1126,74 @@ define <3 x i32> @bitcast_v12i8_to_v3i32(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB6_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v6, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v2.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v0.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v4, v7
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v7, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v1.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v2.h, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v7, v4
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v2.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v7, v2
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB6_2
|
||||
; GFX11-TRUE16-NEXT: .LBB6_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v7.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v7.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v6.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, v10.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v5.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v5.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v5.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v4.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v4.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v3.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v7, v5
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.l, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v7
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v7
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.h, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v3.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v7, v1
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.l, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v7, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -4241,15 +4236,16 @@ define <3 x float> @bitcast_v12i8_to_v3f32(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v8.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v9.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v8.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v12
|
||||
@ -4264,80 +4260,74 @@ define <3 x float> @bitcast_v12i8_to_v3f32(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB22_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v4, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v6, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v2.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v0.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v4, v7
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v7, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v1.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v2.h, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v7, v4
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v2.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr3_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v7, v2
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_lo16
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB22_2
|
||||
; GFX11-TRUE16-NEXT: .LBB22_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v7.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v7.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v6.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, v10.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v5.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v5.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v5.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v4.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v4.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v4.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v3.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v7, v5
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.l, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v7
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v7
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.h, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v3.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v7, v1
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.l, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v7, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -6885,16 +6875,16 @@ define <6 x bfloat> @bitcast_v12i8_to_v6bf16(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.h, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v10.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v9.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v10.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v9.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v12
|
||||
@ -6909,80 +6899,74 @@ define <6 x bfloat> @bitcast_v12i8_to_v6bf16(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB36_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v8.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.h, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.l, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.h, v1.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr9_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v6, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v2.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v7
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v1.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v2.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v2.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v2
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB36_2
|
||||
; GFX11-TRUE16-NEXT: .LBB36_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v9.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v8.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v7.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v7.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, v10.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v7.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v6.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v6.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v5.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v6.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v5.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v9.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v4.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v9
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v9.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v9
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v9.h, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v4, v9
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v4.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v1
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -8651,16 +8635,16 @@ define <6 x half> @bitcast_v12i8_to_v6f16(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.h, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v10.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v9.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v10.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v9.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v12
|
||||
@ -8675,80 +8659,74 @@ define <6 x half> @bitcast_v12i8_to_v6f16(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB40_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v8.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.h, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.l, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.h, v1.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr9_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v6, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v2.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v7
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v1.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v2.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v2.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v2
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB40_2
|
||||
; GFX11-TRUE16-NEXT: .LBB40_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v9.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v8.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v7.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v7.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, v10.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v7.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v6.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v6.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v5.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v6.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v5.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v9.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v4.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v9
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v9.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v9
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v9.h, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v4, v9
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v4.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v1
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -10065,16 +10043,16 @@ define <6 x i16> @bitcast_v12i8_to_v6i16(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.h, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v10.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v9.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v10.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v9.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3
|
||||
; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v12
|
||||
@ -10089,80 +10067,74 @@ define <6 x i16> @bitcast_v12i8_to_v6i16(<12 x i8> %a, i32 %b) {
|
||||
; GFX11-TRUE16-NEXT: .LBB44_3: ; %cmp.false
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v8.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.h, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v7.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.l, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.h, v1.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr9_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v1.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v6, v7
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v7.h, v2.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v7
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v1.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v2.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v2.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_hi16
|
||||
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_lo16
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v2
|
||||
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
|
||||
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB44_2
|
||||
; GFX11-TRUE16-NEXT: .LBB44_4: ; %cmp.true
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v9.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v8.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, v7.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v6.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, v7.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.l, v8.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, v10.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v7.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v6.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v6.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v5.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v6.h, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v5.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x300, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v9.h, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.l, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v4.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v1.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v5, v9
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v9.h, 0x300, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v9
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v9.h, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v4, v9
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v4.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v3, v1
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v3.l, 0x300, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v2.h, 0x300, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v3.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v2
|
||||
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
|
||||
@ -1,3 +1,4 @@
|
||||
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
||||
; RUN: llc %s -o %t.o -mcpu=gfx1030 -filetype=obj -O0
|
||||
; RUN: llvm-debuginfo-analyzer %t.o --print=all --attribute=all | FileCheck %s
|
||||
|
||||
|
||||
@ -9022,13 +9022,12 @@ define amdgpu_kernel void @uniform_or_i16(ptr addrspace(1) %result, ptr addrspac
|
||||
; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s7, v0
|
||||
; GFX1164-TRUE16-NEXT: .LBB15_2:
|
||||
; GFX1164-TRUE16-NEXT: s_or_b64 exec, exec, s[4:5]
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX1164-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1164-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1164-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX1164-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX1164-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
|
||||
; GFX1164-TRUE16-NEXT: v_cndmask_b16 v0.l, s6, 0, vcc
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1164-TRUE16-NEXT: v_or_b16 v0.l, s2, v0.l
|
||||
; GFX1164-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX1164-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
|
||||
@ -9101,13 +9100,12 @@ define amdgpu_kernel void @uniform_or_i16(ptr addrspace(1) %result, ptr addrspac
|
||||
; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s6, v0
|
||||
; GFX1132-TRUE16-NEXT: .LBB15_2:
|
||||
; GFX1132-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1132-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1132-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX1132-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
|
||||
; GFX1132-TRUE16-NEXT: v_cndmask_b16 v0.l, s4, 0, vcc_lo
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: v_or_b16 v0.l, s2, v0.l
|
||||
; GFX1132-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX1132-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
|
||||
@ -9180,13 +9178,12 @@ define amdgpu_kernel void @uniform_or_i16(ptr addrspace(1) %result, ptr addrspac
|
||||
; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s7, v0
|
||||
; GFX1264-TRUE16-NEXT: .LBB15_2:
|
||||
; GFX1264-TRUE16-NEXT: s_or_b64 exec, exec, s[4:5]
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX1264-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1264-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1264-TRUE16-NEXT: s_wait_kmcnt 0x0
|
||||
; GFX1264-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX1264-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
|
||||
; GFX1264-TRUE16-NEXT: v_cndmask_b16 v0.l, s6, 0, vcc
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1264-TRUE16-NEXT: v_or_b16 v0.l, s2, v0.l
|
||||
; GFX1264-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX1264-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
|
||||
@ -9259,13 +9256,12 @@ define amdgpu_kernel void @uniform_or_i16(ptr addrspace(1) %result, ptr addrspac
|
||||
; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s6, v0
|
||||
; GFX1232-TRUE16-NEXT: .LBB15_2:
|
||||
; GFX1232-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1232-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1232-TRUE16-NEXT: s_wait_kmcnt 0x0
|
||||
; GFX1232-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
|
||||
; GFX1232-TRUE16-NEXT: v_cndmask_b16 v0.l, s4, 0, vcc_lo
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: v_or_b16 v0.l, s2, v0.l
|
||||
; GFX1232-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX1232-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
|
||||
@ -9662,12 +9658,11 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
|
||||
; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
|
||||
; GFX1164-TRUE16-NEXT: .LBB16_4: ; %Flow
|
||||
; GFX1164-TRUE16-NEXT: s_or_b64 exec, exec, s[8:9]
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX1164-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1164-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1164-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX1164-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1164-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1164-TRUE16-NEXT: v_mad_u16 v0.l, s10, v4.l, s2
|
||||
; GFX1164-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX1164-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
|
||||
@ -9789,12 +9784,11 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
|
||||
; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
|
||||
; GFX1132-TRUE16-NEXT: .LBB16_4: ; %Flow
|
||||
; GFX1132-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1132-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1132-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX1132-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: v_mad_u16 v0.l, s8, v4.l, s2
|
||||
; GFX1132-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX1132-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
|
||||
@ -9916,13 +9910,12 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
|
||||
; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
|
||||
; GFX1264-TRUE16-NEXT: .LBB16_4: ; %Flow
|
||||
; GFX1264-TRUE16-NEXT: s_or_b64 exec, exec, s[8:9]
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX1264-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1264-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1264-TRUE16-NEXT: s_wait_kmcnt 0x0
|
||||
; GFX1264-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX1264-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
|
||||
; GFX1264-TRUE16-NEXT: s_wait_alu 0xf1ff
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1264-TRUE16-NEXT: v_mad_u16 v0.l, s10, v4.l, s2
|
||||
; GFX1264-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX1264-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
|
||||
@ -10048,13 +10041,12 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
|
||||
; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
|
||||
; GFX1232-TRUE16-NEXT: .LBB16_4: ; %Flow
|
||||
; GFX1232-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1232-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1232-TRUE16-NEXT: s_wait_kmcnt 0x0
|
||||
; GFX1232-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
|
||||
; GFX1232-TRUE16-NEXT: s_wait_alu 0xf1ff
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: v_mad_u16 v0.l, s8, v4.l, s2
|
||||
; GFX1232-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX1232-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
|
||||
@ -10734,15 +10726,15 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
|
||||
; GFX1164-TRUE16-NEXT: s_mov_b64 s[2:3], 0
|
||||
; GFX1164-TRUE16-NEXT: .LBB18_1: ; %atomicrmw.start
|
||||
; GFX1164-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v1
|
||||
; GFX1164-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1164-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1164-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1164-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s9, v0
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX1164-TRUE16-NEXT: v_and_or_b32 v0, v1, s10, v0
|
||||
; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v3, v1
|
||||
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
||||
; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v2, v0
|
||||
; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
|
||||
; GFX1164-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
@ -10828,14 +10820,14 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
|
||||
; GFX1132-TRUE16-NEXT: s_mov_b32 s6, -1
|
||||
; GFX1132-TRUE16-NEXT: .LBB18_1: ; %atomicrmw.start
|
||||
; GFX1132-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
|
||||
; GFX1132-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1132-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1132-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: v_and_or_b32 v0, v1, s3, v0
|
||||
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1132-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
|
||||
; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
|
||||
; GFX1132-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
@ -10920,15 +10912,15 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
|
||||
; GFX1264-TRUE16-NEXT: s_mov_b64 s[2:3], 0
|
||||
; GFX1264-TRUE16-NEXT: .LBB18_1: ; %atomicrmw.start
|
||||
; GFX1264-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v1
|
||||
; GFX1264-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1264-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1264-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1264-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s9, v0
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX1264-TRUE16-NEXT: v_and_or_b32 v0, v1, s10, v0
|
||||
; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v3, v1
|
||||
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
||||
; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v2, v0
|
||||
; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
|
||||
; GFX1264-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
@ -11014,14 +11006,14 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
|
||||
; GFX1232-TRUE16-NEXT: s_mov_b32 s6, -1
|
||||
; GFX1232-TRUE16-NEXT: .LBB18_1: ; %atomicrmw.start
|
||||
; GFX1232-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
|
||||
; GFX1232-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1232-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1232-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: v_and_or_b32 v0, v1, s3, v0
|
||||
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1232-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
|
||||
; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
|
||||
; GFX1232-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
|
||||
@ -37774,9 +37774,10 @@ define bfloat @v_uitofp_i16_to_bf16(i16 %x) {
|
||||
; GFX11TRUE16-LABEL: v_uitofp_i16_to_bf16:
|
||||
; GFX11TRUE16: ; %bb.0:
|
||||
; GFX11TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX11TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11TRUE16-NEXT: v_cvt_f32_u32_e32 v0, v0
|
||||
; GFX11TRUE16-NEXT: v_cvt_f32_u32_e32 v0, v1
|
||||
; GFX11TRUE16-NEXT: v_bfe_u32 v1, v0, 16, 1
|
||||
; GFX11TRUE16-NEXT: v_or_b32_e32 v2, 0x400000, v0
|
||||
; GFX11TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
|
||||
@ -40750,12 +40751,11 @@ define amdgpu_ps i32 @s_select_bf16(bfloat inreg %a, bfloat inreg %b, i32 %c) {
|
||||
;
|
||||
; GFX11TRUE16-LABEL: s_select_bf16:
|
||||
; GFX11TRUE16: ; %bb.0:
|
||||
; GFX11TRUE16-NEXT: v_mov_b16_e32 v1.l, s0
|
||||
; GFX11TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
|
||||
; GFX11TRUE16-NEXT: v_mov_b16_e32 v0.l, s0
|
||||
; GFX11TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11TRUE16-NEXT: v_cndmask_b16 v0.l, s1, v0.l, vcc_lo
|
||||
; GFX11TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11TRUE16-NEXT: v_cndmask_b16 v0.l, s1, v1.l, vcc_lo
|
||||
; GFX11TRUE16-NEXT: v_readfirstlane_b32 s0, v0
|
||||
; GFX11TRUE16-NEXT: ; return to shader part epilog
|
||||
;
|
||||
|
||||
@ -3443,15 +3443,14 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v1.l, v1.l, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
|
||||
; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], null offen th:TH_ATOMIC_RETURN
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
@ -3569,14 +3568,13 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v1.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
|
||||
; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], 0 offen glc
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
@ -3884,15 +3882,14 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v1.l, v1.l, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
|
||||
; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
@ -4007,14 +4004,13 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v1.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
|
||||
; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
@ -4328,15 +4324,14 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
|
||||
; GFX12-TRUE16-NEXT: ; Child Loop BB15_4 Depth 2
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v6, v4, v7
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v6.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_mov_b32 s2, exec_lo
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v6.l, v6.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v6, v4, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v6, v7, v11, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v9, v7 :: v_dual_mov_b32 v8, v6
|
||||
; GFX12-TRUE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
|
||||
; GFX12-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
|
||||
@ -4556,15 +4551,14 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
|
||||
; GFX11-TRUE16-NEXT: ; Child Loop BB15_4 Depth 2
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v6, v4, v7
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v6.l, v6.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v6, v4, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v6, v7, v11, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v9, v7 :: v_dual_mov_b32 v8, v6
|
||||
; GFX11-TRUE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
|
||||
; GFX11-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
|
||||
|
||||
@ -2512,16 +2512,16 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
|
||||
; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], null offen th:TH_ATOMIC_RETURN
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
@ -2640,20 +2640,19 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
|
||||
; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v5, s[0:3], 0 offen
|
||||
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
|
||||
; GFX11-TRUE16-NEXT: .p2align 6
|
||||
; GFX11-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v1.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
|
||||
; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], 0 offen glc
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
@ -2973,16 +2972,16 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
|
||||
; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
@ -3098,20 +3097,19 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
|
||||
; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v3, s[0:3], 0 offen
|
||||
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
|
||||
; GFX11-TRUE16-NEXT: .p2align 6
|
||||
; GFX11-TRUE16-NEXT: .LBB11_1: ; %atomicrmw.start
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v1.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
|
||||
; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
@ -3437,16 +3435,16 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
|
||||
; GFX12-TRUE16-NEXT: ; Child Loop BB12_4 Depth 2
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_mov_b32 s2, exec_lo
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v4.h, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v11, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v8, v6 :: v_dual_mov_b32 v7, v5
|
||||
; GFX12-TRUE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
|
||||
; GFX12-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
|
||||
@ -3672,16 +3670,16 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
|
||||
; GFX11-TRUE16-NEXT: ; Child Loop BB12_4 Depth 2
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v4.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v11, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v8, v6 :: v_dual_mov_b32 v7, v5
|
||||
; GFX11-TRUE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
|
||||
; GFX11-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
|
||||
|
||||
@ -2512,16 +2512,16 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
|
||||
; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], null offen th:TH_ATOMIC_RETURN
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
@ -2640,20 +2640,19 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
|
||||
; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v5, s[0:3], 0 offen
|
||||
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
|
||||
; GFX11-TRUE16-NEXT: .p2align 6
|
||||
; GFX11-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v1.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
|
||||
; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], 0 offen glc
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
@ -2973,16 +2972,16 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
|
||||
; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
@ -3098,20 +3097,19 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
|
||||
; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v3, s[0:3], 0 offen
|
||||
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
|
||||
; GFX11-TRUE16-NEXT: .p2align 6
|
||||
; GFX11-TRUE16-NEXT: .LBB11_1: ; %atomicrmw.start
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v1.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
|
||||
; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
@ -3437,16 +3435,16 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
|
||||
; GFX12-TRUE16-NEXT: ; Child Loop BB12_4 Depth 2
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_mov_b32 s2, exec_lo
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v4.h, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v11, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v8, v6 :: v_dual_mov_b32 v7, v5
|
||||
; GFX12-TRUE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
|
||||
; GFX12-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
|
||||
@ -3672,16 +3670,16 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
|
||||
; GFX11-TRUE16-NEXT: ; Child Loop BB12_4 Depth 2
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v4.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v11, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v8, v6 :: v_dual_mov_b32 v7, v5
|
||||
; GFX11-TRUE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
|
||||
; GFX11-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
|
||||
|
||||
@ -2745,6 +2745,15 @@ define amdgpu_cs void @amdgpu_cs_v32i1(<32 x i1> %arg0) {
|
||||
;
|
||||
; GFX11-TRUE16-LABEL: amdgpu_cs_v32i1:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v26.l, v26.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v25.l, 1, v25.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v24.l, v24.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v20.h, v22.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v21.l, 1, v21.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v20.l, v20.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v17.h, v18.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.l, 1, v17.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v16.l, v16.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v10.l, v10.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.l, 1, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v8.l, v8.l, 1
|
||||
@ -2754,6 +2763,18 @@ define amdgpu_cs void @amdgpu_cs_v32i1(<32 x i1> %arg0) {
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, v2.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 1, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, v0.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v30.l, v30.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v29.l, 1, v29.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v28.l, v28.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v25.h, 3, v27.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v26.l, 2, v26.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v22.h, v24.l, v25.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v22.l, 3, v23.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v20.h, 2, v20.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v16.h, v20.l, v21.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v18.h, 3, v19.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.h, 2, v17.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v16.l, v16.l, v17.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v14.l, v14.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v13.l, 1, v13.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v12.l, v12.l, 1
|
||||
@ -2766,15 +2787,15 @@ define amdgpu_cs void @amdgpu_cs_v32i1(<32 x i1> %arg0) {
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 3, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 2, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v25.h, v26.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v25.l, 1, v25.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v24.l, v24.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v22.l, v22.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v21.l, 1, v21.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v20.l, v20.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v18.l, v18.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.l, 1, v17.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v16.l, v16.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v28.h, 3, v31.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v29.h, 2, v30.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v28.l, v28.l, v29.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v25.h, v25.h, v26.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v21.h, v22.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v18.l, v22.l, v20.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v16.h, v16.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v17.h, v18.h, v17.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v16.l, v16.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 3, v15.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 2, v14.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v8.h, v12.l, v13.l
|
||||
@ -2784,65 +2805,42 @@ define amdgpu_cs void @amdgpu_cs_v32i1(<32 x i1> %arg0) {
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, v0.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v2.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, v0.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v30.l, v30.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v29.l, 1, v29.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v28.l, v28.l, 1
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v22.h, 3, v27.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v25.h, 2, v25.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v24.l, v24.l, v25.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v23.l, 3, v23.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v22.l, 2, v22.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v16.h, v20.l, v21.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.h, 3, v19.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v18.l, 2, v18.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v15.h, v16.l, v17.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v26.h, v28.h, v29.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v24.h, v28.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v19.l, v21.h, v25.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v16.h, v16.h, v18.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v15.h, v16.l, v17.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v12.h, v10.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v3.l, v8.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v1.l, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v0.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v30.h, 3, v31.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v28.h, 2, v30.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v24.h, v28.l, v29.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v22.h, v22.h, v25.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v24.l, v24.l, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v18.h, v23.l, v22.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v14.h, v16.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v16.h, v17.h, v18.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, v15.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v24.h, v24.h, v26.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v14.h, v19.l, 15
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v16.h, 4, v16.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, v15.h, 15
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v3.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, v1.l, 15
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 4, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, v0.l, 15
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v28.h, v30.h, v28.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v24.h, v24.h, 3
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v20.h, v24.l, v22.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v14.h, v18.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.l, 12, v24.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v14.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v1.h, v16.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 12, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v23.h, v24.h, v28.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, v20.h, 15
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 4, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, v1.h, 15
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v17.l, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v2.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 12, v23.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.h, v1.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v2.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v2
|
||||
; GFX11-TRUE16-NEXT: global_store_b32 v[0:1], v0, off
|
||||
; GFX11-TRUE16-NEXT: s_endpgm
|
||||
;
|
||||
|
||||
@ -1561,10 +1561,10 @@ define amdgpu_kernel void @v_no_clamp_add_src_v2f16_f16_src(ptr addrspace(1) %ou
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 2, v1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: global_load_d16_b16 v0, v0, s[2:3]
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v0.l, 1.0, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_pk_max_f16 v0, v0, v0 clamp
|
||||
; GFX11-TRUE16-NEXT: global_store_b32 v1, v0, s[0:1]
|
||||
; GFX11-TRUE16-NEXT: s_endpgm
|
||||
|
||||
@ -946,9 +946,9 @@ define double @v_uitofp_i8_to_f64(i8 %arg0) nounwind {
|
||||
; GFX11-TRUE16-LABEL: v_uitofp_i8_to_f64:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f64_u32_e32 v[0:1], v0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -1770,40 +1770,38 @@ define amdgpu_kernel void @load_v4i8_to_v4f32_2_uses(ptr addrspace(1) noalias %o
|
||||
; GFX11-TRUE16-LABEL: load_v4i8_to_v4f32_2_uses:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_load_b64 s[0:1], s[4:5], 0x34
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, 0 :: v_dual_and_b32 v0, 0x3ff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0x3ff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 2, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v5.h
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: global_load_b32 v4, v0, s[0:1]
|
||||
; GFX11-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v4.l, 9
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff00, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff00, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f32_ubyte3_e32 v3, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.h, v4.h, 9
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x900, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff00, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff00, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f32_ubyte3_e32 v3, v4
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f32_ubyte2_e32 v2, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f32_ubyte1_e32 v1, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v7, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v6.h, 0x900, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f32_ubyte2_e32 v2, v4
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v5.l, 0x900, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v7.h, 0x900, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f32_ubyte0_e32 v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, v7, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, v5, v7
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: s_clause 0x1
|
||||
; GFX11-TRUE16-NEXT: global_store_b128 v5, v[0:3], s[0:1]
|
||||
; GFX11-TRUE16-NEXT: global_store_b32 v5, v4, s[2:3]
|
||||
; GFX11-TRUE16-NEXT: global_store_b128 v6, v[0:3], s[0:1]
|
||||
; GFX11-TRUE16-NEXT: global_store_b32 v6, v4, s[2:3]
|
||||
; GFX11-TRUE16-NEXT: s_endpgm
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: load_v4i8_to_v4f32_2_uses:
|
||||
|
||||
@ -2536,12 +2536,13 @@ define void @test_dynamic_stackalloc_device_divergent_non_standard_size_i16(i16
|
||||
; GFX11-SDAG-LABEL: test_dynamic_stackalloc_device_divergent_non_standard_size_i16:
|
||||
; GFX11-SDAG: ; %bb.0:
|
||||
; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-SDAG-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-SDAG-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-SDAG-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX11-SDAG-NEXT: s_mov_b32 s4, s33
|
||||
; GFX11-SDAG-NEXT: s_mov_b32 s1, exec_lo
|
||||
; GFX11-SDAG-NEXT: s_mov_b32 s0, 0
|
||||
; GFX11-SDAG-NEXT: s_mov_b32 s33, s32
|
||||
; GFX11-SDAG-NEXT: v_lshl_add_u32 v0, v0, 2, 15
|
||||
; GFX11-SDAG-NEXT: v_lshl_add_u32 v0, v1, 2, 15
|
||||
; GFX11-SDAG-NEXT: s_add_i32 s32, s32, 16
|
||||
; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-SDAG-NEXT: v_and_b32_e32 v0, 0x7fff0, v0
|
||||
|
||||
@ -8410,13 +8410,12 @@ define half @flat_agent_atomic_fadd_ret_f16__amdgpu_no_fine_grained_memory(ptr %
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8529,13 +8528,12 @@ define half @flat_agent_atomic_fadd_ret_f16__amdgpu_no_fine_grained_memory(ptr %
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -8785,13 +8783,12 @@ define half @flat_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_grain
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8908,13 +8905,12 @@ define half @flat_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_grain
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -9171,13 +9167,12 @@ define half @flat_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_grain
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -9295,13 +9290,12 @@ define half @flat_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_grain
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -9557,11 +9551,11 @@ define void @flat_agent_atomic_fadd_noret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -9671,11 +9665,11 @@ define void @flat_agent_atomic_fadd_noret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
@ -9917,11 +9911,11 @@ define void @flat_agent_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -10035,11 +10029,11 @@ define void @flat_agent_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
@ -10288,11 +10282,11 @@ define void @flat_agent_atomic_fadd_noret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -10407,11 +10401,11 @@ define void @flat_agent_atomic_fadd_noret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
@ -10651,8 +10645,8 @@ define void @flat_agent_atomic_fadd_noret_f16__offset12b__align4_pos__amdgpu_no_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -10735,8 +10729,8 @@ define void @flat_agent_atomic_fadd_noret_f16__offset12b__align4_pos__amdgpu_no_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 glc
|
||||
@ -10925,10 +10919,9 @@ define half @flat_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_fi
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -11014,10 +11007,9 @@ define half @flat_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_fi
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 glc
|
||||
@ -11220,13 +11212,12 @@ define half @flat_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_grai
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -11345,13 +11336,12 @@ define half @flat_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_grai
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -11610,11 +11600,11 @@ define void @flat_system_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -11730,11 +11720,11 @@ define void @flat_system_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
|
||||
@ -6043,14 +6043,14 @@ define half @flat_agent_atomic_fmax_ret_f16__amdgpu_no_fine_grained_memory(ptr %
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6168,14 +6168,14 @@ define half @flat_agent_atomic_fmax_ret_f16__amdgpu_no_fine_grained_memory(ptr %
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -6438,14 +6438,14 @@ define half @flat_agent_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_grain
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6570,14 +6570,14 @@ define half @flat_agent_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_grain
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -6847,14 +6847,14 @@ define half @flat_agent_atomic_fmax_ret_f16__offset12b_neg__amdgpu_no_fine_grain
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6980,14 +6980,14 @@ define half @flat_agent_atomic_fmax_ret_f16__offset12b_neg__amdgpu_no_fine_grain
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -7254,13 +7254,12 @@ define void @flat_agent_atomic_fmax_noret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v3.l, v3.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7376,13 +7375,12 @@ define void @flat_agent_atomic_fmax_noret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v3.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
@ -7638,13 +7636,12 @@ define void @flat_agent_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7767,13 +7764,12 @@ define void @flat_agent_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -8036,13 +8032,12 @@ define void @flat_agent_atomic_fmax_noret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8166,13 +8161,12 @@ define void @flat_agent_atomic_fmax_noret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -8424,11 +8418,11 @@ define half @flat_agent_atomic_fmax_ret_f16__offset12b_pos__align4__amdgpu_no_fi
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8519,11 +8513,11 @@ define half @flat_agent_atomic_fmax_ret_f16__offset12b_pos__align4__amdgpu_no_fi
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 glc
|
||||
@ -8728,10 +8722,9 @@ define void @flat_agent_atomic_fmax_noret_f16__offset12b__align4_pos__amdgpu_no_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8820,10 +8813,9 @@ define void @flat_agent_atomic_fmax_noret_f16__offset12b__align4_pos__amdgpu_no_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 glc
|
||||
@ -9035,14 +9027,14 @@ define half @flat_system_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_grai
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -9169,14 +9161,14 @@ define half @flat_system_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_grai
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -9448,13 +9440,12 @@ define void @flat_system_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -9579,13 +9570,12 @@ define void @flat_system_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
|
||||
@ -6043,14 +6043,14 @@ define half @flat_agent_atomic_fmin_ret_f16__amdgpu_no_fine_grained_memory(ptr %
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6168,14 +6168,14 @@ define half @flat_agent_atomic_fmin_ret_f16__amdgpu_no_fine_grained_memory(ptr %
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -6438,14 +6438,14 @@ define half @flat_agent_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_grain
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6570,14 +6570,14 @@ define half @flat_agent_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_grain
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -6847,14 +6847,14 @@ define half @flat_agent_atomic_fmin_ret_f16__offset12b_neg__amdgpu_no_fine_grain
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6980,14 +6980,14 @@ define half @flat_agent_atomic_fmin_ret_f16__offset12b_neg__amdgpu_no_fine_grain
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -7254,13 +7254,12 @@ define void @flat_agent_atomic_fmin_noret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v3.l, v3.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7376,13 +7375,12 @@ define void @flat_agent_atomic_fmin_noret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v3.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
@ -7638,13 +7636,12 @@ define void @flat_agent_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7767,13 +7764,12 @@ define void @flat_agent_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -8036,13 +8032,12 @@ define void @flat_agent_atomic_fmin_noret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8166,13 +8161,12 @@ define void @flat_agent_atomic_fmin_noret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -8424,11 +8418,11 @@ define half @flat_agent_atomic_fmin_ret_f16__offset12b_pos__align4__amdgpu_no_fi
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8519,11 +8513,11 @@ define half @flat_agent_atomic_fmin_ret_f16__offset12b_pos__align4__amdgpu_no_fi
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 glc
|
||||
@ -8728,10 +8722,9 @@ define void @flat_agent_atomic_fmin_noret_f16__offset12b__align4_pos__amdgpu_no_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8820,10 +8813,9 @@ define void @flat_agent_atomic_fmin_noret_f16__offset12b__align4_pos__amdgpu_no_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 glc
|
||||
@ -9035,14 +9027,14 @@ define half @flat_system_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_grai
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -9169,14 +9161,14 @@ define half @flat_system_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_grai
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
@ -9448,13 +9440,12 @@ define void @flat_system_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -9579,13 +9570,12 @@ define void @flat_system_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[3:4], v[5:6] glc
|
||||
|
||||
@ -5855,13 +5855,12 @@ define half @flat_agent_atomic_fsub_ret_f16(ptr %ptr, half %val) #0 {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -5974,13 +5973,12 @@ define half @flat_agent_atomic_fsub_ret_f16(ptr %ptr, half %val) #0 {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -6230,13 +6228,12 @@ define half @flat_agent_atomic_fsub_ret_f16__offset12b_pos(ptr %ptr, half %val)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6353,13 +6350,12 @@ define half @flat_agent_atomic_fsub_ret_f16__offset12b_pos(ptr %ptr, half %val)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -6616,13 +6612,12 @@ define half @flat_agent_atomic_fsub_ret_f16__offset12b_neg(ptr %ptr, half %val)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6740,13 +6735,12 @@ define half @flat_agent_atomic_fsub_ret_f16__offset12b_neg(ptr %ptr, half %val)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -7002,11 +6996,11 @@ define void @flat_agent_atomic_fsub_noret_f16(ptr %ptr, half %val) #0 {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7116,11 +7110,11 @@ define void @flat_agent_atomic_fsub_noret_f16(ptr %ptr, half %val) #0 {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
@ -7362,11 +7356,11 @@ define void @flat_agent_atomic_fsub_noret_f16__offset12b_pos(ptr %ptr, half %val
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7480,11 +7474,11 @@ define void @flat_agent_atomic_fsub_noret_f16__offset12b_pos(ptr %ptr, half %val
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
@ -7733,11 +7727,11 @@ define void @flat_agent_atomic_fsub_noret_f16__offset12b_neg(ptr %ptr, half %val
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7852,11 +7846,11 @@ define void @flat_agent_atomic_fsub_noret_f16__offset12b_neg(ptr %ptr, half %val
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
@ -8096,10 +8090,9 @@ define half @flat_agent_atomic_fsub_ret_f16__offset12b_pos__align4(ptr %ptr, hal
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8185,10 +8178,9 @@ define half @flat_agent_atomic_fsub_ret_f16__offset12b_pos__align4(ptr %ptr, hal
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 glc
|
||||
@ -8382,8 +8374,8 @@ define void @flat_agent_atomic_fsub_noret_f16__offset12b__align4_pos(ptr %ptr, h
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8466,8 +8458,8 @@ define void @flat_agent_atomic_fsub_noret_f16__offset12b__align4_pos(ptr %ptr, h
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] offset:2046 glc
|
||||
@ -8665,13 +8657,12 @@ define half @flat_system_atomic_fsub_ret_f16__offset12b_pos(ptr %ptr, half %val)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -8790,13 +8781,12 @@ define half @flat_system_atomic_fsub_ret_f16__offset12b_pos(ptr %ptr, half %val)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v5, v[0:1], v[5:6] glc
|
||||
@ -9055,11 +9045,11 @@ define void @flat_system_atomic_fsub_noret_f16__offset12b_pos(ptr %ptr, half %va
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -9175,11 +9165,11 @@ define void @flat_system_atomic_fsub_noret_f16__offset12b_pos(ptr %ptr, half %va
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
|
||||
|
||||
@ -4238,7 +4238,7 @@ define amdgpu_ps i32 @s_mul_32_f16(half inreg %x, half inreg %y) {
|
||||
; GFX11-GISEL-TRUE16-LABEL: s_mul_32_f16:
|
||||
; GFX11-GISEL-TRUE16: ; %bb.0:
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mul_f16_e64 v0.l, 0x5000, s0
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
|
||||
; GFX11-GISEL-TRUE16-NEXT: ; return to shader part epilog
|
||||
;
|
||||
|
||||
@ -644,11 +644,10 @@ define double @fmul_pow_mul_max_pow2(i16 %cnt) nounwind {
|
||||
; GFX11-TRUE16-LABEL: fmul_pow_mul_max_pow2:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, v0.l, 2
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f64_u32_e32 v[0:1], v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mul_f64 v[0:1], 0x40080000, v[0:1]
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -1194,13 +1193,12 @@ define double @fmul_pow_shl_cnt_safe(i16 %cnt) nounwind {
|
||||
; GFX11-TRUE16-LABEL: fmul_pow_shl_cnt_safe:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, v0.l, 1
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s0, 0xff5f3992
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s1, 0x7befffff
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f64_u32_e32 v[0:1], v0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mul_f64 v[0:1], v[0:1], s[0:1]
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
|
||||
@ -4372,14 +4372,13 @@ define amdgpu_kernel void @fptrunc_f32_to_f16_zext_i32(
|
||||
; GFX11-GISEL-TRUE16-LABEL: fptrunc_f32_to_f16_zext_i32:
|
||||
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_load_b32 s2, s[2:3], 0x0
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, s2
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-GISEL-TRUE16-NEXT: buffer_store_b32 v0, off, s[0:3], 0
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_endpgm
|
||||
;
|
||||
@ -4607,14 +4606,13 @@ define amdgpu_kernel void @fptrunc_fabs_f32_to_f16_zext_i32(
|
||||
; GFX11-GISEL-TRUE16-LABEL: fptrunc_fabs_f32_to_f16_zext_i32:
|
||||
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_load_b32 s2, s[2:3], 0x0
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_cvt_f16_f32_e64 v0.l, |s2|
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-GISEL-TRUE16-NEXT: buffer_store_b32 v0, off, s[0:3], 0
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_endpgm
|
||||
;
|
||||
|
||||
@ -1107,21 +1107,19 @@ define void @void_func_v4i8(<4 x i8> %arg0) #0 {
|
||||
; GFX11-TRUE16-LABEL: void_func_v4i8:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b64 s[0:1], 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b64 s[0:1], 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v2
|
||||
; GFX11-TRUE16-NEXT: buffer_store_b32 v0, off, s[0:3], 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -1190,22 +1188,20 @@ define void @void_func_v5i8(<5 x i8> %arg0) #0 {
|
||||
; GFX11-TRUE16-LABEL: void_func_v5i8:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b64 s[0:1], 4
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b64 s[0:1], 4
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX11-TRUE16-NEXT: buffer_store_b8 v4, off, s[0:3], 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b64 s[0:1], 0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v3, v2
|
||||
; GFX11-TRUE16-NEXT: buffer_store_b32 v0, off, s[0:3], 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -1285,29 +1281,27 @@ define void @void_func_v8i8(<8 x i8> %arg0) #0 {
|
||||
; GFX11-TRUE16-LABEL: void_func_v8i8:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v5.h, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.l, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v6.h, v5.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v4.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v4.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: s_mov_b64 s[0:1], 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v0.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v4, v6
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v6.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v6.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v5, v4
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v0, v6
|
||||
; GFX11-TRUE16-NEXT: buffer_store_b64 v[1:2], off, s[0:3], 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -1422,47 +1416,44 @@ define void @void_func_v16i8(<16 x i8> %arg0) #0 {
|
||||
; GFX11-TRUE16-LABEL: void_func_v16i8:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v13.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v15.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v13.h, 0xff, v14.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v13.l, 8, v13.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v13.l, 8, v15.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v8.h, 0xff, v14.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.l, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v8.l, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.l, v12.l, v12.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v14.h, v13.h, v12.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v8.h, v13.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v8.l, v8.l, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.l, v12.l, v13.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.h, v14.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.l, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v10.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v5.h, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, v12, v14
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v14.h, v10.l, v9.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v9.l, v10.h, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v14.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.l, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.l, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v13
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, v9, v14
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v14.h, v5.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v4.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v14.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v4.l, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, v9, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v10.l, v8.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v8, 0xffff, v11
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, v8, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b64 s[0:1], 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, v4, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, v4, v14
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v14.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v14.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, v0, v12
|
||||
; GFX11-TRUE16-NEXT: buffer_store_b128 v[6:9], off, s[0:3], 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, v0, v14
|
||||
; GFX11-TRUE16-NEXT: buffer_store_b128 v[5:8], off, s[0:3], 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: void_func_v16i8:
|
||||
@ -1658,83 +1649,77 @@ define void @void_func_v32i8(<32 x i8> %arg0) #0 {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: scratch_load_d16_u8 v31, off, s32
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v13.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v7.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v15.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v14.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v3.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.l, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v13.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.h, v32.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v5.h, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.l, v7.h, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v7.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v4.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v13, 0xffff, v13
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.l, v3.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v32.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v6.l, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v8.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v29.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v9.h, 0xff, v28.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, v13, v32
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.l, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.l, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v9.l, v7.h, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, v12, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v5.h, v4.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v11.l, 8, v25.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v11.h, 0xff, v24.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v9.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v13, 0xffff, v14
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v12, v32
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.h, v32.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v10.l, v4.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v9, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v6.l, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v30.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v21.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v20.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v11.h, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v11, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v13, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v8.h, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.l, 8, v27.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v10.h, 0xff, v26.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v17.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.h, 0xff, v16.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v6.l, v0.h, v6.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v7, 0xffff, v7
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.h, v32.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.h, 0xff, v30.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v29.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v6.h, 0xff, v28.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v11.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v10, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v0.h, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.h, v32.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v27.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v5.l, 0xff, v26.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.l, 8, v25.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v7.h, 0xff, v24.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.l, v6.h, v5.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v11, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v4.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.h, 0xff, v22.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v11, 0xffff, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.h, v32.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.l, 8, v23.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v21.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v6.h, 0xff, v20.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v14.l, v7.h, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.h, v32.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v8.h, 0xff, v16.l
|
||||
; GFX11-TRUE16-NEXT: s_mov_b64 s[0:1], 16
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v15.l, v6.h, v5.h
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s2, -1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v31.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v9.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v23.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v8
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.l, 8, v19.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, v7, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v10.h, v10.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.l, 0xff, v18.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, v9, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v4.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v10
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, v11, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v4.l, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.l, 8, v31.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v4.h, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.h, 0xff, v22.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, v13, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v5.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v19.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v17.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, v14, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v4.h, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.h, 0xff, v18.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v9.l, v8.h, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, v15, v32
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v32.h, v4.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, v9, v32
|
||||
; GFX11-TRUE16-NEXT: buffer_store_b128 v[4:7], off, s[0:3], 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b64 s[0:1], 0
|
||||
|
||||
@ -4896,23 +4896,22 @@ define amdgpu_gfx void @test_call_external_void_func_v4i8_ret() #0 {
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 24, v0
|
||||
; GFX11-TRUE16-NEXT: s_swappc_b64 s[30:31], s[0:1]
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s31, v42, 1
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s30, v42, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v2
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s32, s33
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s0, v42, 2
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v1, v0
|
||||
; GFX11-TRUE16-NEXT: global_store_b32 v[40:41], v0, off
|
||||
; GFX11-TRUE16-NEXT: s_clause 0x1
|
||||
; GFX11-TRUE16-NEXT: scratch_load_b32 v41, off, s33
|
||||
@ -5156,30 +5155,29 @@ define amdgpu_gfx void @test_call_external_void_func_v5i8_ret() #0 {
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v0, v5
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v6
|
||||
; GFX11-TRUE16-NEXT: s_swappc_b64 s[30:31], s[0:1]
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v0, 4
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s31, v42, 1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v2, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s30, v42, 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s32, s33
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s0, v42, 2
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v2, v3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v2
|
||||
; GFX11-TRUE16-NEXT: s_clause 0x1
|
||||
; GFX11-TRUE16-NEXT: global_store_b8 v[0:1], v4, off
|
||||
; GFX11-TRUE16-NEXT: global_store_b32 v[40:41], v2, off
|
||||
; GFX11-TRUE16-NEXT: s_clause 0x1
|
||||
; GFX11-TRUE16-NEXT: scratch_load_b32 v41, off, s33
|
||||
; GFX11-TRUE16-NEXT: scratch_load_b32 v40, off, s33 offset:4
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s31, v42, 1
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s30, v42, 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s32, s33
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s0, v42, 2
|
||||
; GFX11-TRUE16-NEXT: s_or_saveexec_b32 s1, -1
|
||||
; GFX11-TRUE16-NEXT: scratch_load_b32 v42, off, s33 offset:8 ; 4-byte Folded Reload
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 exec_lo, s1
|
||||
@ -5441,36 +5439,34 @@ define amdgpu_gfx void @test_call_external_void_func_v8i8_ret() #0 {
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 24, v1
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, v1 :: v_dual_mov_b32 v1, v8
|
||||
; GFX11-TRUE16-NEXT: s_swappc_b64 s[30:31], s[0:1]
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.h, v1.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v3.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v3, v5
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v5.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v5, v4
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v0, v4
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s31, v42, 1
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v0, v5
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s30, v42, 0
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s32, s33
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s0, v42, 2
|
||||
; GFX11-TRUE16-NEXT: global_store_b64 v[40:41], v[1:2], off
|
||||
; GFX11-TRUE16-NEXT: s_clause 0x1
|
||||
; GFX11-TRUE16-NEXT: scratch_load_b32 v41, off, s33
|
||||
; GFX11-TRUE16-NEXT: scratch_load_b32 v40, off, s33 offset:4
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 s32, s33
|
||||
; GFX11-TRUE16-NEXT: v_readlane_b32 s0, v42, 2
|
||||
; GFX11-TRUE16-NEXT: s_or_saveexec_b32 s1, -1
|
||||
; GFX11-TRUE16-NEXT: scratch_load_b32 v42, off, s33 offset:8 ; 4-byte Folded Reload
|
||||
; GFX11-TRUE16-NEXT: s_mov_b32 exec_lo, s1
|
||||
@ -5910,85 +5906,77 @@ define amdgpu_gfx void @test_call_external_void_func_v32i8_ret() #0 {
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v17, v32 :: v_dual_mov_b32 v18, v33
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v19, v34
|
||||
; GFX11-TRUE16-NEXT: s_swappc_b64 s[30:31], s[0:1]
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v13.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v15.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v15.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v14.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v13.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, 0
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.l, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v14.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.h, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.h, v0.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v8.l, v3.h, v1.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.l, v3.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.h, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v9.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v8.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, v13, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v9.l, v3.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v11.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v10.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, v9, v13
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v8, 0xffff, v8
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v3.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v7.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v6.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, v8, v13
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, v9, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v29.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v3.h, v2.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v12.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, v4, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v1.h, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v31.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v30.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v29.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v28.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, v6, v13
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, v2, v12
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v17.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v27.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v1.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v26.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v25.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v24.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v31.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v6.l, 0xff, v30.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v2, v13
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v26.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.h, v6.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v27.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v7
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v21.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v4.h, 0xff, v16.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, v2, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v23.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v1.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v22.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v21.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v20.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, v6, v13
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v17.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v16.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v7, 0xffff, v7
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v23.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v6.l, 0xff, v22.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, v7, v13
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v18.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.h, v6.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v10
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v19.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, v6, v13
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v13.h, v1.l, v0.h
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v10
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, v0, v13
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v2, v12
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v19.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v1.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v18.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v1, v12
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v12.h, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v4.h, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v12.l
|
||||
; GFX11-TRUE16-NEXT: v_or_b32_e32 v0, v0, v12
|
||||
; GFX11-TRUE16-NEXT: s_clause 0x1
|
||||
; GFX11-TRUE16-NEXT: global_store_b128 v[42:43], v[6:9], off
|
||||
; GFX11-TRUE16-NEXT: global_store_b128 v[40:41], v[2:5], off
|
||||
; GFX11-TRUE16-NEXT: global_store_b128 v[42:43], v[0:3], off
|
||||
; GFX11-TRUE16-NEXT: global_store_b128 v[40:41], v[5:8], off
|
||||
; GFX11-TRUE16-NEXT: s_clause 0x3
|
||||
; GFX11-TRUE16-NEXT: scratch_load_b32 v43, off, s33
|
||||
; GFX11-TRUE16-NEXT: scratch_load_b32 v42, off, s33 offset:4
|
||||
|
||||
@ -8275,13 +8275,12 @@ define half @global_agent_atomic_fadd_ret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8394,13 +8393,12 @@ define half @global_agent_atomic_fadd_ret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -8700,13 +8698,12 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8823,13 +8820,12 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -9138,13 +9134,12 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -9262,13 +9257,12 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -9576,11 +9570,11 @@ define void @global_agent_atomic_fadd_noret_f16__amdgpu_no_fine_grained_memory(p
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -9690,11 +9684,11 @@ define void @global_agent_atomic_fadd_noret_f16__amdgpu_no_fine_grained_memory(p
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
@ -9985,11 +9979,11 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_g
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -10103,11 +10097,11 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_g
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
@ -10406,11 +10400,11 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_neg__amdgpu_no_fine_g
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -10525,11 +10519,11 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_neg__amdgpu_no_fine_g
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
@ -10819,10 +10813,9 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -10908,10 +10901,9 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 glc
|
||||
@ -11144,8 +11136,8 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b__align4_pos__amdgpu_n
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -11228,8 +11220,8 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b__align4_pos__amdgpu_n
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 glc
|
||||
@ -11464,13 +11456,12 @@ define half @global_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -11589,13 +11580,12 @@ define half @global_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -11906,11 +11896,11 @@ define void @global_system_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -12026,11 +12016,11 @@ define void @global_system_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
|
||||
@ -4467,14 +4467,14 @@ define half @global_agent_atomic_fmax_ret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -4592,14 +4592,14 @@ define half @global_agent_atomic_fmax_ret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -4912,14 +4912,14 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -5044,14 +5044,14 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -5373,14 +5373,14 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -5506,14 +5506,14 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -5832,13 +5832,12 @@ define void @global_agent_atomic_fmax_noret_f16__amdgpu_no_fine_grained_memory(p
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v3.l, v3.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -5954,13 +5953,12 @@ define void @global_agent_atomic_fmax_noret_f16__amdgpu_no_fine_grained_memory(p
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v3.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
@ -6265,13 +6263,12 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_g
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6394,13 +6391,12 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_g
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -6713,13 +6709,12 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b_neg__amdgpu_no_fine_g
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6843,13 +6838,12 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b_neg__amdgpu_no_fine_g
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -7151,11 +7145,11 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_pos__align4__amdgpu_no_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7246,11 +7240,11 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_pos__align4__amdgpu_no_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 glc
|
||||
@ -7494,10 +7488,9 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b__align4_pos__amdgpu_n
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7586,10 +7579,9 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b__align4_pos__amdgpu_n
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 glc
|
||||
@ -7838,14 +7830,14 @@ define half @global_system_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -7972,14 +7964,14 @@ define half @global_system_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -8303,13 +8295,12 @@ define void @global_system_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -8434,13 +8425,12 @@ define void @global_system_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
|
||||
@ -4467,14 +4467,14 @@ define half @global_agent_atomic_fmin_ret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -4592,14 +4592,14 @@ define half @global_agent_atomic_fmin_ret_f16__amdgpu_no_fine_grained_memory(ptr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -4912,14 +4912,14 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -5044,14 +5044,14 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -5373,14 +5373,14 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -5506,14 +5506,14 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_neg__amdgpu_no_fine_gra
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -5832,13 +5832,12 @@ define void @global_agent_atomic_fmin_noret_f16__amdgpu_no_fine_grained_memory(p
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v3.l, v3.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -5954,13 +5953,12 @@ define void @global_agent_atomic_fmin_noret_f16__amdgpu_no_fine_grained_memory(p
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v3.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
@ -6265,13 +6263,12 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_g
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6394,13 +6391,12 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_g
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -6713,13 +6709,12 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b_neg__amdgpu_no_fine_g
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6843,13 +6838,12 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b_neg__amdgpu_no_fine_g
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -7151,11 +7145,11 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_pos__align4__amdgpu_no_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7246,11 +7240,11 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_pos__align4__amdgpu_no_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 glc
|
||||
@ -7494,10 +7488,9 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b__align4_pos__amdgpu_n
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7586,10 +7579,9 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b__align4_pos__amdgpu_n
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v3.l, v2.h, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 glc
|
||||
@ -7838,14 +7830,14 @@ define half @global_system_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -7972,14 +7964,14 @@ define half @global_system_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_gr
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
@ -8303,13 +8295,12 @@ define void @global_system_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -8434,13 +8425,12 @@ define void @global_system_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v1, v6
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v5.l, v5.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v1, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v2, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[3:4], v[5:6], off glc
|
||||
|
||||
@ -5221,13 +5221,12 @@ define half @global_agent_atomic_fsub_ret_f16(ptr addrspace(1) %ptr, half %val)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -5340,13 +5339,12 @@ define half @global_agent_atomic_fsub_ret_f16(ptr addrspace(1) %ptr, half %val)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -5646,13 +5644,12 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_pos(ptr addrspace(1) %p
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -5769,13 +5766,12 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_pos(ptr addrspace(1) %p
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -6084,13 +6080,12 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_neg(ptr addrspace(1) %p
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6208,13 +6203,12 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_neg(ptr addrspace(1) %p
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -6522,11 +6516,11 @@ define void @global_agent_atomic_fsub_noret_f16(ptr addrspace(1) %ptr, half %val
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -6636,11 +6630,11 @@ define void @global_agent_atomic_fsub_noret_f16(ptr addrspace(1) %ptr, half %val
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
@ -6931,11 +6925,11 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b_pos(ptr addrspace(1)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7049,11 +7043,11 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b_pos(ptr addrspace(1)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
@ -7352,11 +7346,11 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b_neg(ptr addrspace(1)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7471,11 +7465,11 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b_neg(ptr addrspace(1)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
@ -7765,10 +7759,9 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_pos__align4(ptr addrspa
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -7854,10 +7847,9 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_pos__align4(ptr addrspa
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 glc
|
||||
@ -8090,8 +8082,8 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b__align4_pos(ptr addrs
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
|
||||
@ -8174,8 +8166,8 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b__align4_pos(ptr addrs
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v4.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, 0xffff0000, v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off offset:2046 glc
|
||||
@ -8410,13 +8402,12 @@ define half @global_system_atomic_fsub_ret_f16__offset12b_pos(ptr addrspace(1) %
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -8535,13 +8526,12 @@ define half @global_system_atomic_fsub_ret_f16__offset12b_pos(ptr addrspace(1) %
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v3, v6
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v5.l, v5.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v3, v5
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v4, v5
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v5, v[0:1], v[5:6], off glc
|
||||
@ -8852,11 +8842,11 @@ define void @global_system_atomic_fsub_noret_f16__offset12b_pos(ptr addrspace(1)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX12-TRUE16-NEXT: global_wb scope:SCOPE_SYS
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
@ -8972,11 +8962,11 @@ define void @global_system_atomic_fsub_noret_f16__offset12b_pos(ptr addrspace(1)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v5, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_sub_f16_e32 v3.l, v3.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v5, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v6, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: global_atomic_cmpswap_b32 v3, v[0:1], v[3:4], off glc
|
||||
|
||||
207
llvm/test/CodeGen/AMDGPU/global-extload-gfx11plus.ll
Normal file
207
llvm/test/CodeGen/AMDGPU/global-extload-gfx11plus.ll
Normal file
@ -0,0 +1,207 @@
|
||||
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
||||
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 < %s | FileCheck -check-prefixes=GFX11,GFX11-REAL16 %s
|
||||
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 < %s | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
|
||||
|
||||
define amdgpu_kernel void @zextload_global_i8_to_i16(ptr addrspace(1) %out, ptr addrspace(1) %in) nounwind {
|
||||
; GFX11-REAL16-LABEL: zextload_global_i8_to_i16:
|
||||
; GFX11-REAL16: ; %bb.0:
|
||||
; GFX11-REAL16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-REAL16-NEXT: v_mov_b32_e32 v1, 0
|
||||
; GFX11-REAL16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-REAL16-NEXT: global_load_d16_u8 v0, v1, s[2:3]
|
||||
; GFX11-REAL16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-REAL16-NEXT: global_store_b16 v1, v0, s[0:1]
|
||||
; GFX11-REAL16-NEXT: s_endpgm
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: zextload_global_i8_to_i16:
|
||||
; GFX11-FAKE16: ; %bb.0:
|
||||
; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: global_load_u8 v1, v0, s[2:3]
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: global_store_b16 v0, v1, s[0:1]
|
||||
; GFX11-FAKE16-NEXT: s_endpgm
|
||||
%a = load i8, ptr addrspace(1) %in
|
||||
%ext = zext i8 %a to i16
|
||||
store i16 %ext, ptr addrspace(1) %out
|
||||
ret void
|
||||
}
|
||||
|
||||
define amdgpu_kernel void @sextload_global_i8_to_i16(ptr addrspace(1) %out, ptr addrspace(1) %in) nounwind {
|
||||
; GFX11-REAL16-LABEL: sextload_global_i8_to_i16:
|
||||
; GFX11-REAL16: ; %bb.0:
|
||||
; GFX11-REAL16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-REAL16-NEXT: v_mov_b32_e32 v1, 0
|
||||
; GFX11-REAL16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-REAL16-NEXT: global_load_d16_i8 v0, v1, s[2:3]
|
||||
; GFX11-REAL16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-REAL16-NEXT: global_store_b16 v1, v0, s[0:1]
|
||||
; GFX11-REAL16-NEXT: s_endpgm
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: sextload_global_i8_to_i16:
|
||||
; GFX11-FAKE16: ; %bb.0:
|
||||
; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: global_load_i8 v1, v0, s[2:3]
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: global_store_b16 v0, v1, s[0:1]
|
||||
; GFX11-FAKE16-NEXT: s_endpgm
|
||||
%a = load i8, ptr addrspace(1) %in
|
||||
%ext = sext i8 %a to i16
|
||||
store i16 %ext, ptr addrspace(1) %out
|
||||
ret void
|
||||
}
|
||||
|
||||
define amdgpu_kernel void @zextload_global_i8_to_i64(ptr addrspace(1) %out, ptr addrspace(1) %in) nounwind {
|
||||
; GFX11-REAL16-LABEL: zextload_global_i8_to_i64:
|
||||
; GFX11-REAL16: ; %bb.0:
|
||||
; GFX11-REAL16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-REAL16-NEXT: v_mov_b32_e32 v1, 0
|
||||
; GFX11-REAL16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-REAL16-NEXT: global_load_d16_u8 v0, v1, s[2:3]
|
||||
; GFX11-REAL16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-REAL16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-REAL16-NEXT: global_store_b64 v1, v[0:1], s[0:1]
|
||||
; GFX11-REAL16-NEXT: s_endpgm
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: zextload_global_i8_to_i64:
|
||||
; GFX11-FAKE16: ; %bb.0:
|
||||
; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: global_load_u8 v0, v1, s[2:3]
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-FAKE16-NEXT: global_store_b64 v1, v[0:1], s[0:1]
|
||||
; GFX11-FAKE16-NEXT: s_endpgm
|
||||
%a = load i8, ptr addrspace(1) %in
|
||||
%ext = zext i8 %a to i64
|
||||
store i64 %ext, ptr addrspace(1) %out
|
||||
ret void
|
||||
}
|
||||
|
||||
define amdgpu_kernel void @sextload_global_i8_to_i64(ptr addrspace(1) %out, ptr addrspace(1) %in) nounwind {
|
||||
; GFX11-REAL16-LABEL: sextload_global_i8_to_i64:
|
||||
; GFX11-REAL16: ; %bb.0:
|
||||
; GFX11-REAL16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-REAL16-NEXT: v_mov_b32_e32 v2, 0
|
||||
; GFX11-REAL16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-REAL16-NEXT: global_load_d16_i8 v0, v2, s[2:3]
|
||||
; GFX11-REAL16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-REAL16-NEXT: v_bfe_i32 v0, v0, 0, 16
|
||||
; GFX11-REAL16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-REAL16-NEXT: v_ashrrev_i32_e32 v1, 31, v0
|
||||
; GFX11-REAL16-NEXT: global_store_b64 v2, v[0:1], s[0:1]
|
||||
; GFX11-REAL16-NEXT: s_endpgm
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: sextload_global_i8_to_i64:
|
||||
; GFX11-FAKE16: ; %bb.0:
|
||||
; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, 0
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: global_load_i8 v0, v2, s[2:3]
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: v_bfe_i32 v0, v0, 0, 16
|
||||
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-FAKE16-NEXT: v_ashrrev_i32_e32 v1, 31, v0
|
||||
; GFX11-FAKE16-NEXT: global_store_b64 v2, v[0:1], s[0:1]
|
||||
; GFX11-FAKE16-NEXT: s_endpgm
|
||||
%a = load i8, ptr addrspace(1) %in
|
||||
%ext = sext i8 %a to i64
|
||||
store i64 %ext, ptr addrspace(1) %out
|
||||
ret void
|
||||
}
|
||||
|
||||
define amdgpu_kernel void @zextload_global_i16_to_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) nounwind {
|
||||
; GFX11-LABEL: zextload_global_i16_to_i32:
|
||||
; GFX11: ; %bb.0:
|
||||
; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-NEXT: v_mov_b32_e32 v0, 0
|
||||
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-NEXT: global_load_u16 v1, v0, s[2:3]
|
||||
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
|
||||
; GFX11-NEXT: s_endpgm
|
||||
%a = load i16, ptr addrspace(1) %in
|
||||
%ext = zext i16 %a to i32
|
||||
store i32 %ext, ptr addrspace(1) %out
|
||||
ret void
|
||||
}
|
||||
|
||||
define amdgpu_kernel void @sextload_global_i16_to_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) nounwind {
|
||||
; GFX11-LABEL: sextload_global_i16_to_i32:
|
||||
; GFX11: ; %bb.0:
|
||||
; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-NEXT: v_mov_b32_e32 v0, 0
|
||||
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-NEXT: global_load_i16 v1, v0, s[2:3]
|
||||
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
|
||||
; GFX11-NEXT: s_endpgm
|
||||
%a = load i16, ptr addrspace(1) %in
|
||||
%ext = sext i16 %a to i32
|
||||
store i32 %ext, ptr addrspace(1) %out
|
||||
ret void
|
||||
}
|
||||
|
||||
define amdgpu_kernel void @zextload_global_i16_to_i64(ptr addrspace(1) %out, ptr addrspace(1) %in) nounwind {
|
||||
; GFX11-REAL16-LABEL: zextload_global_i16_to_i64:
|
||||
; GFX11-REAL16: ; %bb.0:
|
||||
; GFX11-REAL16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-REAL16-NEXT: v_mov_b32_e32 v1, 0
|
||||
; GFX11-REAL16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-REAL16-NEXT: global_load_d16_b16 v0, v1, s[2:3]
|
||||
; GFX11-REAL16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-REAL16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-REAL16-NEXT: global_store_b64 v1, v[0:1], s[0:1]
|
||||
; GFX11-REAL16-NEXT: s_endpgm
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: zextload_global_i16_to_i64:
|
||||
; GFX11-FAKE16: ; %bb.0:
|
||||
; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: global_load_u16 v0, v1, s[2:3]
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-FAKE16-NEXT: global_store_b64 v1, v[0:1], s[0:1]
|
||||
; GFX11-FAKE16-NEXT: s_endpgm
|
||||
%a = load i16, ptr addrspace(1) %in
|
||||
%ext = zext i16 %a to i64
|
||||
store i64 %ext, ptr addrspace(1) %out
|
||||
ret void
|
||||
}
|
||||
|
||||
define amdgpu_kernel void @sextload_global_i16_to_i64(ptr addrspace(1) %out, ptr addrspace(1) %in) nounwind {
|
||||
; GFX11-REAL16-LABEL: sextload_global_i16_to_i64:
|
||||
; GFX11-REAL16: ; %bb.0:
|
||||
; GFX11-REAL16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-REAL16-NEXT: v_mov_b32_e32 v2, 0
|
||||
; GFX11-REAL16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-REAL16-NEXT: global_load_d16_b16 v0, v2, s[2:3]
|
||||
; GFX11-REAL16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-REAL16-NEXT: v_bfe_i32 v0, v0, 0, 16
|
||||
; GFX11-REAL16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-REAL16-NEXT: v_ashrrev_i32_e32 v1, 31, v0
|
||||
; GFX11-REAL16-NEXT: global_store_b64 v2, v[0:1], s[0:1]
|
||||
; GFX11-REAL16-NEXT: s_endpgm
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: sextload_global_i16_to_i64:
|
||||
; GFX11-FAKE16: ; %bb.0:
|
||||
; GFX11-FAKE16-NEXT: s_load_b128 s[0:3], s[4:5], 0x24
|
||||
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, 0
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: global_load_u16 v0, v2, s[2:3]
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: v_bfe_i32 v0, v0, 0, 16
|
||||
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-FAKE16-NEXT: v_ashrrev_i32_e32 v1, 31, v0
|
||||
; GFX11-FAKE16-NEXT: global_store_b64 v2, v[0:1], s[0:1]
|
||||
; GFX11-FAKE16-NEXT: s_endpgm
|
||||
%a = load i16, ptr addrspace(1) %in
|
||||
%ext = sext i16 %a to i64
|
||||
store i64 %ext, ptr addrspace(1) %out
|
||||
ret void
|
||||
}
|
||||
@ -1693,12 +1693,11 @@ define amdgpu_kernel void @notdot4_mixedtypes(ptr addrspace(1) %src1,
|
||||
; GFX11-DL-TRUE16-NEXT: v_mov_b16_e32 v3.l, v7.l
|
||||
; GFX11-DL-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-DL-TRUE16-NEXT: v_mad_u16 v0.l, v0.h, v1.l, v0.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-DL-TRUE16-NEXT: v_perm_b32 v1, v5, v5, 0xc0c0302
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-DL-TRUE16-NEXT: v_mad_u16 v0.l, v2.l, v3.l, v0.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_perm_b32 v2, v4, v4, 0xc0c0302
|
||||
; GFX11-DL-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-DL-TRUE16-NEXT: v_dot4_u32_u8 v0, v2, v1, v0
|
||||
; GFX11-DL-TRUE16-NEXT: global_store_b16 v6, v0, s[4:5]
|
||||
; GFX11-DL-TRUE16-NEXT: s_endpgm
|
||||
@ -2724,32 +2723,32 @@ define amdgpu_kernel void @udot4_acc8_vecMul(ptr addrspace(1) %src1,
|
||||
; GFX11-DL-TRUE16-NEXT: global_load_b32 v4, v0, s[2:3]
|
||||
; GFX11-DL-TRUE16-NEXT: global_load_d16_u8 v0, v5, s[4:5]
|
||||
; GFX11-DL-TRUE16-NEXT: s_waitcnt vmcnt(2)
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshrrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-DL-TRUE16-NEXT: s_waitcnt vmcnt(1)
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshrrev_b16 v1.l, 8, v4.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 24, v3
|
||||
; GFX11-DL-TRUE16-NEXT: s_waitcnt vmcnt(1)
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 24, v4
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshrrev_b16 v0.h, 8, v3.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_mul_lo_u16 v1.l, v3.h, v4.h
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshrrev_b16 v1.h, 8, v4.l
|
||||
; GFX11-DL-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-DL-TRUE16-NEXT: v_mad_u16 v0.l, v3.l, v4.l, v0.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_mul_lo_u16 v0.h, v0.h, v1.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_mul_lo_u16 v1.l, v3.h, v4.h
|
||||
; GFX11-DL-TRUE16-NEXT: v_mul_lo_u16 v1.h, v2.l, v6.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_mul_lo_u16 v2.l, v2.l, v6.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_mov_b16_e32 v6.l, 0
|
||||
; GFX11-DL-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_mul_lo_u16 v0.h, v0.h, v1.h
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v0.h
|
||||
; GFX11-DL-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.h
|
||||
; GFX11-DL-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-DL-TRUE16-NEXT: v_or_b16 v6.h, v0.h, v1.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v2.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_mov_b16_e32 v7.h, v6.l
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshlrev_b16 v7.l, 8, v0.h
|
||||
; GFX11-DL-TRUE16-NEXT: v_or_b16 v6.h, v1.l, v2.l
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-DL-TRUE16-NEXT: v_or_b32_e32 v1, v7, v6
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX11-DL-TRUE16-NEXT: v_or_b32_e32 v2, v2, v6
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 8, v2
|
||||
; GFX11-DL-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v2.l
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-DL-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-DL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v2
|
||||
; GFX11-DL-TRUE16-NEXT: v_mad_u16 v0.l, v3.h, v4.h, v0.l
|
||||
; GFX11-DL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-DL-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-DL-TRUE16-NEXT: global_store_b8 v5, v0, s[4:5]
|
||||
; GFX11-DL-TRUE16-NEXT: s_endpgm
|
||||
|
||||
@ -1715,9 +1715,9 @@ define zeroext i16 @clpeak_umad_pat_i16(i16 zeroext %x, i16 zeroext %y) {
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mad_u16 v0.h, v0.l, v1.l, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mad_u16 v0.l, v0.h, v0.l, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mad_u16 v0.l, v0.l, v0.h, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-SDAG-FAKE16-LABEL: clpeak_umad_pat_i16:
|
||||
@ -1745,8 +1745,7 @@ define zeroext i16 @clpeak_umad_pat_i16(i16 zeroext %x, i16 zeroext %y) {
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v0.h
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-GISEL-FAKE16-LABEL: clpeak_umad_pat_i16:
|
||||
@ -1777,9 +1776,9 @@ define zeroext i16 @clpeak_umad_pat_i16(i16 zeroext %x, i16 zeroext %y) {
|
||||
; GFX1200-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_mad_u16 v0.h, v0.l, v1.l, v0.l
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_mad_u16 v0.l, v0.h, v0.l, v0.h
|
||||
; GFX1200-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1200-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_mad_u16 v0.l, v0.l, v0.h, v0.l
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1200-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX1200-SDAG-FAKE16-LABEL: clpeak_umad_pat_i16:
|
||||
@ -1815,8 +1814,7 @@ define zeroext i16 @clpeak_umad_pat_i16(i16 zeroext %x, i16 zeroext %y) {
|
||||
; GFX1200-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1200-GISEL-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v1.l
|
||||
; GFX1200-GISEL-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v0.h
|
||||
; GFX1200-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1200-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1200-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1200-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX1200-GISEL-FAKE16-LABEL: clpeak_umad_pat_i16:
|
||||
@ -9363,9 +9361,9 @@ define zeroext i16 @clpeak_umad_pat_i16_x2(i16 zeroext %x, i16 zeroext %y) {
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mad_u16 v0.h, v0.l, v0.h, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mad_u16 v0.l, v0.h, v0.l, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mad_u16 v0.l, v0.l, v0.h, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-SDAG-FAKE16-LABEL: clpeak_umad_pat_i16_x2:
|
||||
@ -9409,8 +9407,7 @@ define zeroext i16 @clpeak_umad_pat_i16_x2(i16 zeroext %x, i16 zeroext %y) {
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v0.h
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-GISEL-FAKE16-LABEL: clpeak_umad_pat_i16_x2:
|
||||
@ -9457,9 +9454,9 @@ define zeroext i16 @clpeak_umad_pat_i16_x2(i16 zeroext %x, i16 zeroext %y) {
|
||||
; GFX1200-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_mad_u16 v0.h, v0.l, v0.h, v0.l
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_mad_u16 v0.l, v0.h, v0.l, v0.h
|
||||
; GFX1200-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1200-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_mad_u16 v0.l, v0.l, v0.h, v0.l
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1200-SDAG-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1200-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX1200-SDAG-FAKE16-LABEL: clpeak_umad_pat_i16_x2:
|
||||
@ -9511,8 +9508,7 @@ define zeroext i16 @clpeak_umad_pat_i16_x2(i16 zeroext %x, i16 zeroext %y) {
|
||||
; GFX1200-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX1200-GISEL-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v1.l
|
||||
; GFX1200-GISEL-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v0.h
|
||||
; GFX1200-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1200-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1200-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX1200-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX1200-GISEL-FAKE16-LABEL: clpeak_umad_pat_i16_x2:
|
||||
|
||||
@ -1259,13 +1259,12 @@ define half @local_atomic_fadd_ret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v0, v4
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v0, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v1, v3, v4
|
||||
@ -1371,13 +1370,12 @@ define half @local_atomic_fadd_ret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v0, v4
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v0, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v1, v3, v4
|
||||
@ -1646,13 +1644,12 @@ define half @local_atomic_fadd_ret_f16__offset(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v1, v4
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v1, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v0, v3, v4
|
||||
@ -1763,13 +1760,12 @@ define half @local_atomic_fadd_ret_f16__offset(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v1, v4
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v1, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v0, v3, v4
|
||||
@ -2044,13 +2040,12 @@ define void @local_atomic_fadd_noret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v0, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v0, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v2, v3, v4
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v1, v4, v2
|
||||
@ -2153,13 +2148,12 @@ define void @local_atomic_fadd_noret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v0, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v2, v3, v4
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v1, v4, v2
|
||||
@ -2419,11 +2413,11 @@ define void @local_atomic_fadd_noret_f16__offset(ptr addrspace(3) %ptr) nounwind
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v1, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v1, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v3, v2, v4
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v0, v4, v3
|
||||
@ -2531,11 +2525,11 @@ define void @local_atomic_fadd_noret_f16__offset(ptr addrspace(3) %ptr) nounwind
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v1, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v1, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v3, v2, v4
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v0, v4, v3
|
||||
@ -2795,10 +2789,9 @@ define half @local_atomic_fadd_ret_f16__offset__align4(ptr addrspace(3) %ptr) no
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v1.l, 4.0, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, 0xffff0000, v2, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:65534
|
||||
@ -2882,10 +2875,9 @@ define half @local_atomic_fadd_ret_f16__offset__align4(ptr addrspace(3) %ptr) no
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v1.l, 4.0, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, 0xffff0000, v2, v1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:65534
|
||||
@ -3095,8 +3087,8 @@ define void @local_atomic_fadd_noret_f16__offset__align4(ptr addrspace(3) %ptr)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v2.l, 4.0, v1.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v2, 0xffff0000, v1, v2
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v2, v0, v2, v1 offset:65534
|
||||
@ -3177,8 +3169,8 @@ define void @local_atomic_fadd_noret_f16__offset__align4(ptr addrspace(3) %ptr)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v2.l, 4.0, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v2, 0xffff0000, v1, v2
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v2, v0, v2, v1 offset:65534
|
||||
|
||||
@ -803,14 +803,14 @@ define half @local_atomic_fmax_ret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v0, v4
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v3.l, v3.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v0, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v1, v3, v4
|
||||
@ -918,14 +918,14 @@ define half @local_atomic_fmax_ret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v0, v4
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v3.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v0, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v1, v3, v4
|
||||
@ -1199,14 +1199,14 @@ define half @local_atomic_fmax_ret_f16__offset(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v1, v4
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v3.l, v3.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v1, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v0, v3, v4
|
||||
@ -1319,14 +1319,14 @@ define half @local_atomic_fmax_ret_f16__offset(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v1, v4
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v3.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v1, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v0, v3, v4
|
||||
@ -1606,14 +1606,14 @@ define void @local_atomic_fmax_noret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v0, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v0, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v2, v3, v4
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v1, v4, v2
|
||||
@ -1718,14 +1718,14 @@ define void @local_atomic_fmax_noret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v0, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.l, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v2, v3, v4
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v1, v4, v2
|
||||
@ -1990,13 +1990,12 @@ define void @local_atomic_fmax_noret_f16__offset(ptr addrspace(3) %ptr) nounwind
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v1, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v1, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v3, v2, v4
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v0, v4, v3
|
||||
@ -2107,13 +2106,12 @@ define void @local_atomic_fmax_noret_f16__offset(ptr addrspace(3) %ptr) nounwind
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v1, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.l, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v1, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v3, v2, v4
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v0, v4, v3
|
||||
@ -2379,11 +2377,11 @@ define half @local_atomic_fmax_ret_f16__offset__align4(ptr addrspace(3) %ptr) no
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v1.l, 4.0, v1.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, 0xffff0000, v2, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:65534
|
||||
@ -2469,11 +2467,11 @@ define half @local_atomic_fmax_ret_f16__offset__align4(ptr addrspace(3) %ptr) no
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v1.l, 4.0, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, 0xffff0000, v2, v1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:65534
|
||||
@ -2688,10 +2686,9 @@ define void @local_atomic_fmax_noret_f16__offset__align4(ptr addrspace(3) %ptr)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v1.l, v1.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.l, 4.0, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v2, 0xffff0000, v1, v2
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v2, v0, v2, v1 offset:65534
|
||||
@ -2775,10 +2772,9 @@ define void @local_atomic_fmax_noret_f16__offset__align4(ptr addrspace(3) %ptr)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.l, v1.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.l, 4.0, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v2, 0xffff0000, v1, v2
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v2, v0, v2, v1 offset:65534
|
||||
|
||||
@ -803,14 +803,14 @@ define half @local_atomic_fmin_ret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v0, v4
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v3.l, v3.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v0, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v1, v3, v4
|
||||
@ -918,14 +918,14 @@ define half @local_atomic_fmin_ret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v0, v4
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v3.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v0, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v1, v3, v4
|
||||
@ -1199,14 +1199,14 @@ define half @local_atomic_fmin_ret_f16__offset(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v1, v4
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v3.l, v3.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v1, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v0, v3, v4
|
||||
@ -1319,14 +1319,14 @@ define half @local_atomic_fmin_ret_f16__offset(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v1, v4
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v3.l, v3.l, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v3.l, 4.0, v3.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v1, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v0, v3, v4
|
||||
@ -1606,14 +1606,14 @@ define void @local_atomic_fmin_noret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v0, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v0, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v2, v3, v4
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v1, v4, v2
|
||||
@ -1718,14 +1718,14 @@ define void @local_atomic_fmin_noret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v0, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.l, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v2, v3, v4
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v1, v4, v2
|
||||
@ -1990,13 +1990,12 @@ define void @local_atomic_fmin_noret_f16__offset(ptr addrspace(3) %ptr) nounwind
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v1, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v4.l, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v1, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v3, v2, v4
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v0, v4, v3
|
||||
@ -2107,13 +2106,12 @@ define void @local_atomic_fmin_noret_f16__offset(ptr addrspace(3) %ptr) nounwind
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v1, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.l, v4.l, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v4.l, 4.0, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v1, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v3, v2, v4
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v0, v4, v3
|
||||
@ -2379,11 +2377,11 @@ define half @local_atomic_fmin_ret_f16__offset__align4(ptr addrspace(3) %ptr) no
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v1.l, 4.0, v1.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, 0xffff0000, v2, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:65534
|
||||
@ -2469,11 +2467,11 @@ define half @local_atomic_fmin_ret_f16__offset__align4(ptr addrspace(3) %ptr) no
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v1.l, 4.0, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, 0xffff0000, v2, v1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:65534
|
||||
@ -2688,10 +2686,9 @@ define void @local_atomic_fmin_noret_f16__offset__align4(ptr addrspace(3) %ptr)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v1.l, v1.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v2.l, 4.0, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v2, 0xffff0000, v1, v2
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v2, v0, v2, v1 offset:65534
|
||||
@ -2775,10 +2772,9 @@ define void @local_atomic_fmin_noret_f16__offset__align4(ptr addrspace(3) %ptr)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_max_f16_e32 v2.l, v1.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_min_f16_e32 v2.l, 4.0, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v2, 0xffff0000, v1, v2
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v2, v0, v2, v1 offset:65534
|
||||
|
||||
@ -1721,13 +1721,12 @@ define half @local_atomic_fsub_ret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v0, v4
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, -4.0, v3.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v0, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v1, v3, v4
|
||||
@ -1833,13 +1832,12 @@ define half @local_atomic_fsub_ret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v0, v4
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, -4.0, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v0, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v1, v3, v4
|
||||
@ -2108,13 +2106,12 @@ define half @local_atomic_fsub_ret_f16__offset(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v1, v4
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v3.l, -4.0, v3.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v1, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v0, v3, v4
|
||||
@ -2225,13 +2222,12 @@ define half @local_atomic_fsub_ret_f16__offset(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, v1, v4
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v3.l, -4.0, v3.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, v1, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v3, v4, v2, v3
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v3, v0, v3, v4
|
||||
@ -2506,13 +2502,12 @@ define void @local_atomic_fsub_noret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX12-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v0, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v4.l, -4.0, v4.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v0, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v2, v3, v4
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v1, v4, v2
|
||||
@ -2615,13 +2610,12 @@ define void @local_atomic_fsub_noret_f16(ptr addrspace(3) %ptr) nounwind {
|
||||
; GFX11-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v0, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v4.l, -4.0, v4.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v0, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v2, v3, v4
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v1, v4, v2
|
||||
@ -2881,11 +2875,11 @@ define void @local_atomic_fsub_noret_f16__offset(ptr addrspace(3) %ptr) nounwind
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v1, v3
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v4.l, -4.0, v4.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v1, v4
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v3, v2, v4
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v0, v4, v3
|
||||
@ -2993,11 +2987,11 @@ define void @local_atomic_fsub_noret_f16__offset(ptr addrspace(3) %ptr) nounwind
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v1, v3
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v4.l, -4.0, v4.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v1, v4
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v3, v2, v4
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v4, v0, v4, v3
|
||||
@ -3257,10 +3251,9 @@ define half @local_atomic_fsub_ret_f16__offset__align4(ptr addrspace(3) %ptr) no
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v1.l, -4.0, v2.l
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v1, 0xffff0000, v2, v1
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:65534
|
||||
@ -3344,10 +3337,9 @@ define half @local_atomic_fsub_ret_f16__offset__align4(ptr addrspace(3) %ptr) no
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v1.l, -4.0, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v1, 0xffff0000, v2, v1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:65534
|
||||
@ -3557,8 +3549,8 @@ define void @local_atomic_fsub_noret_f16__offset__align4(ptr addrspace(3) %ptr)
|
||||
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX12-TRUE16-NEXT: s_wait_dscnt 0x0
|
||||
; GFX12-TRUE16-NEXT: v_add_f16_e32 v2.l, -4.0, v1.l
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-TRUE16-NEXT: v_and_or_b32 v2, 0xffff0000, v1, v2
|
||||
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
|
||||
; GFX12-TRUE16-NEXT: ds_cmpstore_rtn_b32 v2, v0, v2, v1 offset:65534
|
||||
@ -3639,8 +3631,8 @@ define void @local_atomic_fsub_noret_f16__offset__align4(ptr addrspace(3) %ptr)
|
||||
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v2.l, -4.0, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_or_b32 v2, 0xffff0000, v1, v2
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
|
||||
; GFX11-TRUE16-NEXT: ds_cmpstore_rtn_b32 v2, v0, v2, v1 offset:65534
|
||||
|
||||
@ -2382,13 +2382,22 @@ define <4 x half> @v_mad_mix_v4f32_clamp_precvt(<4 x half> %src0, <4 x half> %sr
|
||||
}
|
||||
|
||||
define i32 @mixlo_zext(float %src0, float %src1, float %src2) #0 {
|
||||
; GFX1100-LABEL: mixlo_zext:
|
||||
; GFX1100: ; %bb.0:
|
||||
; GFX1100-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX1100-NEXT: v_fma_mixlo_f16 v0, v0, v1, v2
|
||||
; GFX1100-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX1100-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX1100-NEXT: s_setpc_b64 s[30:31]
|
||||
; SDAG-GFX1100-TRUE16-LABEL: mixlo_zext:
|
||||
; SDAG-GFX1100-TRUE16: ; %bb.0:
|
||||
; SDAG-GFX1100-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; SDAG-GFX1100-TRUE16-NEXT: v_fma_mixlo_f16 v1, v0, v1, v2
|
||||
; SDAG-GFX1100-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; SDAG-GFX1100-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
||||
; SDAG-GFX1100-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.l
|
||||
; SDAG-GFX1100-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; SDAG-GFX1100-FAKE16-LABEL: mixlo_zext:
|
||||
; SDAG-GFX1100-FAKE16: ; %bb.0:
|
||||
; SDAG-GFX1100-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; SDAG-GFX1100-FAKE16-NEXT: v_fma_mixlo_f16 v0, v0, v1, v2
|
||||
; SDAG-GFX1100-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; SDAG-GFX1100-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; SDAG-GFX1100-FAKE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX900-LABEL: mixlo_zext:
|
||||
; GFX900: ; %bb.0:
|
||||
@ -2418,6 +2427,14 @@ define i32 @mixlo_zext(float %src0, float %src1, float %src2) #0 {
|
||||
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v0, v2
|
||||
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GISEL-GFX1100-LABEL: mixlo_zext:
|
||||
; GISEL-GFX1100: ; %bb.0:
|
||||
; GISEL-GFX1100-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GISEL-GFX1100-NEXT: v_fma_mixlo_f16 v0, v0, v1, v2
|
||||
; GISEL-GFX1100-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GISEL-GFX1100-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GISEL-GFX1100-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GISEL-CI-LABEL: mixlo_zext:
|
||||
; GISEL-CI: ; %bb.0:
|
||||
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
|
||||
@ -179,8 +179,7 @@ define i32 @v_mad_u16_zext(i16 %arg0, i16 %arg1, i16 %arg2) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mad_u16 v0.l, v0.l, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: v_mad_u16_zext:
|
||||
@ -222,9 +221,9 @@ define i64 @v_mad_u16_zext64(i16 %arg0, i16 %arg1, i16 %arg2) {
|
||||
; GFX11-TRUE16-LABEL: v_mad_u16_zext64:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_mad_u16 v0.l, v0.l, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: v_mad_u16_zext64:
|
||||
|
||||
@ -374,7 +374,7 @@ define i32 @shl_i16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: shl_i16_zext_i32:
|
||||
@ -412,7 +412,7 @@ define i32 @lshr_i16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_lshrrev_b16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: lshr_i16_zext_i32:
|
||||
@ -450,7 +450,7 @@ define i32 @ashr_i16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_ashrrev_i16 v0.l, v1.l, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: ashr_i16_zext_i32:
|
||||
@ -488,7 +488,7 @@ define i32 @add_u16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: add_u16_zext_i32:
|
||||
@ -526,7 +526,7 @@ define i32 @sub_u16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_sub_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: sub_u16_zext_i32:
|
||||
@ -564,7 +564,7 @@ define i32 @mul_lo_u16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_mul_lo_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: mul_lo_u16_zext_i32:
|
||||
@ -602,7 +602,7 @@ define i32 @min_u16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_min_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: min_u16_zext_i32:
|
||||
@ -641,7 +641,7 @@ define i32 @min_i16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_min_i16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: min_i16_zext_i32:
|
||||
@ -680,7 +680,7 @@ define i32 @max_u16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_max_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: max_u16_zext_i32:
|
||||
@ -719,7 +719,7 @@ define i32 @max_i16_zext_i32(i16 %x, i16 %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_max_i16 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: max_i16_zext_i32:
|
||||
@ -758,7 +758,7 @@ define i32 @zext_fadd_f16(half %x, half %y) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: zext_fadd_f16:
|
||||
@ -797,8 +797,10 @@ define i32 @zext_fma_f16(half %x, half %y, half %z) {
|
||||
; GFX11-TRUE16-LABEL: zext_fma_f16:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_fmac_f16_e32 v2.l, v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v0.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_fmac_f16_e32 v0.l, v0.h, v1.l
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: zext_fma_f16:
|
||||
@ -838,7 +840,7 @@ define i32 @zext_div_fixup_f16(half %x, half %y, half %z) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_div_fixup_f16 v0.l, v0.l, v1.l, v2.l
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: zext_div_fixup_f16:
|
||||
@ -880,7 +882,7 @@ define i32 @zext_fptrunc_f16(float %x) {
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v0
|
||||
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: zext_fptrunc_f16:
|
||||
@ -924,12 +926,20 @@ define i32 @zext_fptrunc_fma_f16(float %x, float %y, float %z) {
|
||||
; GFX10-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX10-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-LABEL: zext_fptrunc_fma_f16:
|
||||
; GFX11: ; %bb.0:
|
||||
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-NEXT: v_fma_mixlo_f16 v0, v0, v1, v2
|
||||
; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
||||
; GFX11-TRUE16-LABEL: zext_fptrunc_fma_f16:
|
||||
; GFX11-TRUE16: ; %bb.0:
|
||||
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-TRUE16-NEXT: v_fma_mixlo_f16 v1, v0, v1, v2
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.l
|
||||
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
; GFX11-FAKE16-LABEL: zext_fptrunc_fma_f16:
|
||||
; GFX11-FAKE16: ; %bb.0:
|
||||
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-FAKE16-NEXT: v_fma_mixlo_f16 v0, v0, v1, v2
|
||||
; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
|
||||
%fma = call float @llvm.fma.f32(float %x, float %y, float %z)
|
||||
%fptrunc = fptrunc float %fma to half
|
||||
%cast = bitcast half %fptrunc to i16
|
||||
@ -940,3 +950,5 @@ define i32 @zext_fptrunc_fma_f16(float %x, float %y, float %z) {
|
||||
declare half @llvm.amdgcn.div.fixup.f16(half, half, half)
|
||||
declare half @llvm.fma.f16(half, half, half)
|
||||
declare float @llvm.fma.f32(float, float, float)
|
||||
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
||||
; GFX11: {{.*}}
|
||||
|
||||
@ -1528,10 +1528,9 @@ define amdgpu_kernel void @v_test_i16_x_sub_64_zext_to_i32(ptr addrspace(1) %out
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 2, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-SDAG-TRUE16-NEXT: global_load_d16_b16 v0, v0, s[2:3]
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_sub_nc_u16 v0.l, v0.l, 64
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-SDAG-TRUE16-NEXT: global_store_b32 v1, v0, s[0:1]
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_endpgm
|
||||
;
|
||||
@ -1560,10 +1559,9 @@ define amdgpu_kernel void @v_test_i16_x_sub_64_zext_to_i32(ptr addrspace(1) %out
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 2, v1
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
|
||||
; GFX11-GISEL-TRUE16-NEXT: global_load_d16_b16 v0, v0, s[2:3]
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u16 v0.l, 0xffc0, v0.l
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
|
||||
; GFX11-GISEL-TRUE16-NEXT: global_store_b32 v1, v0, s[0:1]
|
||||
; GFX11-GISEL-TRUE16-NEXT: s_endpgm
|
||||
;
|
||||
|
||||
@ -300,17 +300,15 @@ define i8 @test_vector_reduce_add_v4i8(<4 x i8> %v) {
|
||||
; GFX11-SDAG-TRUE16-LABEL: test_vector_reduce_add_v4i8:
|
||||
; GFX11-SDAG-TRUE16: ; %bb.0: ; %entry
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v3.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v1.l, v3.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v2.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v2
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
@ -348,17 +346,15 @@ define i8 @test_vector_reduce_add_v4i8(<4 x i8> %v) {
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_wait_samplecnt 0x0
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_wait_bvhcnt 0x0
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_wait_kmcnt 0x0
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v3.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v1.l, v3.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v2.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v2
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
@ -518,21 +514,19 @@ define i8 @test_vector_reduce_add_v8i8(<8 x i8> %v) {
|
||||
; GFX11-SDAG-TRUE16-LABEL: test_vector_reduce_add_v8i8:
|
||||
; GFX11-SDAG-TRUE16: ; %bb.0: ; %entry
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v2.l, v6.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.h, v3.l, v7.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v5.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v4.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v1.l, v1.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v2.l, v6.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.l, v2.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v1.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v2.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
@ -581,21 +575,19 @@ define i8 @test_vector_reduce_add_v8i8(<8 x i8> %v) {
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_wait_samplecnt 0x0
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_wait_bvhcnt 0x0
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_wait_kmcnt 0x0
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v2.l, v6.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.h, v3.l, v7.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v5.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v4.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v1.l, v1.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v2.l, v6.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.l, v2.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v1.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v2.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_4)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
@ -832,28 +824,25 @@ define i8 @test_vector_reduce_add_v16i8(<16 x i8> %v) {
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.h, v5.l, v13.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v9.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v3.h, v7.l, v15.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v3.l, v3.l, v11.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v2.h, v6.l, v14.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v5.l, v7.l, v15.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v6.l, v14.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v2.l, v2.l, v10.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v1.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.h, v4.l, v12.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v3.l, v3.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v2.h, v3.l, v11.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v3.l, v4.l, v12.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v8.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v1.l, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v2.l, v2.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v1.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v2.l, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v2.l, v2.h, v5.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v3.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v2.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.l, v2.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v2.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
@ -922,28 +911,25 @@ define i8 @test_vector_reduce_add_v16i8(<16 x i8> %v) {
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_wait_kmcnt 0x0
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.h, v5.l, v13.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v9.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v3.h, v7.l, v15.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v3.l, v3.l, v11.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v2.h, v6.l, v14.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v5.l, v7.l, v15.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v6.l, v14.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v2.l, v2.l, v10.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v1.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.h, v4.l, v12.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v3.l, v3.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v2.h, v3.l, v11.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v3.l, v4.l, v12.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v8.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v1.l, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v2.l, v2.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v1.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.h, v2.l, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v2.l, v2.h, v5.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v3.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v1.l, v1.l, v2.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.l, v2.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v2.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
|
||||
@ -374,13 +374,12 @@ define i8 @test_vector_reduce_umin_v4i8(<4 x i8> %v) {
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min_u16 v1.l, v1.l, v1.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min3_u16 v0.l, v0.l, v0.h, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -427,13 +426,12 @@ define i8 @test_vector_reduce_umin_v4i8(<4 x i8> %v) {
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min_u16 v1.l, v1.l, v1.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min3_u16 v0.l, v0.l, v0.h, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -624,22 +622,20 @@ define i8 @test_vector_reduce_umin_v8i8(<8 x i8> %v) {
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v7.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v3.l, 0xff, v3.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min_u16 v1.l, v1.l, v1.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min_u16 v0.h, v1.l, v1.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v4.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v6.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min_u16 v0.l, v0.l, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min3_u16 v1.l, v1.l, v3.l, v3.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v3.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v3
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min3_u16 v0.l, v0.l, v1.h, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min3_u16 v0.h, v0.h, v3.l, v3.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min3_u16 v0.l, v0.l, v2.l, v1.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v2
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v3
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -703,22 +699,20 @@ define i8 @test_vector_reduce_umin_v8i8(<8 x i8> %v) {
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v7.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v3.l, 0xff, v3.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v4.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min_u16 v1.l, v1.l, v1.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v2.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v2.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min_u16 v0.h, v1.l, v1.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v4.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v6.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min_u16 v0.l, v0.l, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min3_u16 v1.l, v1.l, v3.l, v3.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v1.l, 8, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v3.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v6.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v3
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min3_u16 v0.l, v0.l, v1.h, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min3_u16 v0.h, v0.h, v3.l, v3.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min_u16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v3.h, 0
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min3_u16 v0.l, v0.l, v2.l, v1.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v2
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v3
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min_u16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -1047,14 +1041,12 @@ define i8 @test_vector_reduce_umin_v16i8(<16 x i8> %v) {
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min3_u16 v0.l, v0.l, v1.h, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_or_b16 v1.l, v0.l, v0.h
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX11-SDAG-TRUE16-NEXT: v_min_u16 v0.l, v0.l, v1.l
|
||||
; GFX11-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
@ -1176,14 +1168,12 @@ define i8 @test_vector_reduce_umin_v16i8(<16 x i8> %v) {
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min3_u16 v0.l, v0.l, v1.h, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshlrev_b16 v0.h, 8, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_or_b16 v1.l, v0.l, v0.h
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
||||
; GFX12-SDAG-TRUE16-NEXT: v_min_u16 v0.l, v0.l, v1.l
|
||||
; GFX12-SDAG-TRUE16-NEXT: s_setpc_b64 s[30:31]
|
||||
;
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user