shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.store.ll
Matt Arsenault 62d946e133 GlobalISel: Merge some AMDGPU ABI lowering code to generic code 
AMDGPU currently has a lot of pre-processing code to pre-split
argument types into 32-bit pieces before passing it to the generic
code in handleAssignments. This is a bit sloppy and also requires some
overly fancy iterator work when building the calls. It's better if all
argument marshalling code is handled directly in
handleAssignments. This handles more situations like decomposing large
element vectors into sub-element sized pieces.

This should mostly be NFC, but does change the generated code by
shifting where the initial argument packing instructions are placed. I
think this is nicer looking, since it now emits the packing code
directly after the relevant copies, rather than after the copies for
the remaining arguments.

This doubles down on gfx6/gfx7 using the gfx8+ ABI for 16-bit
types. This is ultimately the better option, but incompatible with the
DAG. Fixing this requires more work, especially for f16.
2021-02-18 17:26:55 -05:00

270 lines
20 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -stop-after=instruction-select -verify-machineinstrs -o - %s | FileCheck %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx810 -stop-after=instruction-select -verify-machineinstrs -o - %s | FileCheck %s
; Natural mapping
define amdgpu_ps void @struct_buffer_store_f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(float %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; CHECK-LABEL: name: struct_buffer_store_f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset
; CHECK: bb.1 (%ir-block.0):
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
; CHECK: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY6]], %subreg.sub1
; CHECK: BUFFER_STORE_DWORD_BOTHEN_exact [[COPY]], [[REG_SEQUENCE1]], [[REG_SEQUENCE]], [[COPY7]], 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 4 into custom "BufferResource", align 1, addrspace 4)
; CHECK: S_ENDPGM 0
call void @llvm.amdgcn.struct.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret void
}
define amdgpu_ps void @struct_buffer_store_v2f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(<2 x float> %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; CHECK-LABEL: name: struct_buffer_store_v2f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset
; CHECK: bb.1 (%ir-block.0):
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3
; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1
; CHECK: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY3]], %subreg.sub1, [[COPY4]], %subreg.sub2, [[COPY5]], %subreg.sub3
; CHECK: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
; CHECK: [[COPY8:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY7]], %subreg.sub1
; CHECK: BUFFER_STORE_DWORDX2_BOTHEN_exact [[REG_SEQUENCE]], [[REG_SEQUENCE2]], [[REG_SEQUENCE1]], [[COPY8]], 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 8 into custom "BufferResource", align 1, addrspace 4)
; CHECK: S_ENDPGM 0
call void @llvm.amdgcn.struct.buffer.store.v2f32(<2 x float> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret void
}
define amdgpu_ps void @struct_buffer_store_v3f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(<3 x float> %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; CHECK-LABEL: name: struct_buffer_store_v3f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset
; CHECK: bb.1 (%ir-block.0):
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2
; CHECK: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY6]], %subreg.sub3
; CHECK: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
; CHECK: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr4
; CHECK: [[COPY9:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY8]], %subreg.sub1
; CHECK: BUFFER_STORE_DWORDX3_BOTHEN_exact [[REG_SEQUENCE]], [[REG_SEQUENCE2]], [[REG_SEQUENCE1]], [[COPY9]], 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 12 into custom "BufferResource", align 1, addrspace 4)
; CHECK: S_ENDPGM 0
call void @llvm.amdgcn.struct.buffer.store.v3f32(<3 x float> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret void
}
define amdgpu_ps void @struct_buffer_store_v4f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(<4 x float> %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; CHECK-LABEL: name: struct_buffer_store_v4f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset
; CHECK: bb.1 (%ir-block.0):
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
; CHECK: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY6]], %subreg.sub2, [[COPY7]], %subreg.sub3
; CHECK: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr4
; CHECK: [[COPY9:%[0-9]+]]:vgpr_32 = COPY $vgpr5
; CHECK: [[COPY10:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY8]], %subreg.sub0, [[COPY9]], %subreg.sub1
; CHECK: BUFFER_STORE_DWORDX4_BOTHEN_exact [[REG_SEQUENCE]], [[REG_SEQUENCE2]], [[REG_SEQUENCE1]], [[COPY10]], 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 16 into custom "BufferResource", align 1, addrspace 4)
; CHECK: S_ENDPGM 0
call void @llvm.amdgcn.struct.buffer.store.v4f32(<4 x float> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret void
}
define amdgpu_ps void @struct_buffer_store_v4f32_vgpr_rsrc__sgpr_val__sgpr_vindex__sgpr_voffset__vgpr_soffset(<4 x float> inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
; CHECK-LABEL: name: struct_buffer_store_v4f32_vgpr_rsrc__sgpr_val__sgpr_vindex__sgpr_voffset__vgpr_soffset
; CHECK: bb.1 (%ir-block.0):
; CHECK: successors: %bb.2(0x80000000)
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
; CHECK: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
; CHECK: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY6]], %subreg.sub2, [[COPY7]], %subreg.sub3
; CHECK: [[COPY8:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[COPY9:%[0-9]+]]:sreg_32 = COPY $sgpr7
; CHECK: [[COPY10:%[0-9]+]]:vgpr_32 = COPY $vgpr4
; CHECK: [[COPY11:%[0-9]+]]:vreg_128 = COPY [[REG_SEQUENCE]]
; CHECK: [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[COPY8]]
; CHECK: [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[COPY9]]
; CHECK: [[COPY14:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
; CHECK: [[COPY15:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
; CHECK: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
; CHECK: bb.2:
; CHECK: successors: %bb.3(0x40000000), %bb.2(0x40000000)
; CHECK: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY14]].sub0, implicit $exec
; CHECK: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY14]].sub1, implicit $exec
; CHECK: [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
; CHECK: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY14]], implicit $exec
; CHECK: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY15]].sub0, implicit $exec
; CHECK: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY15]].sub1, implicit $exec
; CHECK: [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
; CHECK: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY15]], implicit $exec
; CHECK: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
; CHECK: [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
; CHECK: [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]], implicit $exec
; CHECK: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY10]], implicit $exec
; CHECK: [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
; CHECK: [[REG_SEQUENCE5:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY12]], %subreg.sub0, [[COPY13]], %subreg.sub1
; CHECK: BUFFER_STORE_DWORDX4_BOTHEN_exact [[COPY11]], [[REG_SEQUENCE5]], [[REG_SEQUENCE4]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 16 into custom "BufferResource", align 1, addrspace 4)
; CHECK: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
; CHECK: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
; CHECK: S_CBRANCH_EXECNZ %bb.2, implicit $exec
; CHECK: bb.3:
; CHECK: successors: %bb.4(0x80000000)
; CHECK: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
; CHECK: bb.4:
; CHECK: S_ENDPGM 0
call void @llvm.amdgcn.struct.buffer.store.v4f32(<4 x float> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret void
}
define amdgpu_ps void @struct_buffer_store_i8_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(i32 %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; CHECK-LABEL: name: struct_buffer_store_i8_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset
; CHECK: bb.1 (%ir-block.0):
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
; CHECK: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY6]], %subreg.sub1
; CHECK: BUFFER_STORE_BYTE_BOTHEN_exact [[COPY]], [[REG_SEQUENCE1]], [[REG_SEQUENCE]], [[COPY7]], 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 1 into custom "BufferResource", addrspace 4)
; CHECK: S_ENDPGM 0
%val.trunc = trunc i32 %val to i8
call void @llvm.amdgcn.struct.buffer.store.i8(i8 %val.trunc, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret void
}
define amdgpu_ps void @struct_buffer_store_i16_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(i32 %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; CHECK-LABEL: name: struct_buffer_store_i16_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset
; CHECK: bb.1 (%ir-block.0):
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
; CHECK: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY6]], %subreg.sub1
; CHECK: BUFFER_STORE_SHORT_BOTHEN_exact [[COPY]], [[REG_SEQUENCE1]], [[REG_SEQUENCE]], [[COPY7]], 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 2 into custom "BufferResource", align 1, addrspace 4)
; CHECK: S_ENDPGM 0
%val.trunc = trunc i32 %val to i16
call void @llvm.amdgcn.struct.buffer.store.i16(i16 %val.trunc, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret void
}
define amdgpu_ps void @struct_buffer_store_f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset_glc(float %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; CHECK-LABEL: name: struct_buffer_store_f32_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset_glc
; CHECK: bb.1 (%ir-block.0):
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
; CHECK: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY6]], %subreg.sub1
; CHECK: BUFFER_STORE_DWORD_BOTHEN_exact [[COPY]], [[REG_SEQUENCE1]], [[REG_SEQUENCE]], [[COPY7]], 0, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 4 into custom "BufferResource", align 1, addrspace 4)
; CHECK: S_ENDPGM 0
call void @llvm.amdgcn.struct.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 1)
ret void
}
define amdgpu_ps void @struct_buffer_store_v2f16_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(<2 x half> %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; CHECK-LABEL: name: struct_buffer_store_v2f16_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset
; CHECK: bb.1 (%ir-block.0):
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
; CHECK: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY6]], %subreg.sub1
; CHECK: BUFFER_STORE_DWORD_BOTHEN_exact [[COPY]], [[REG_SEQUENCE1]], [[REG_SEQUENCE]], [[COPY7]], 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 4 into custom "BufferResource", align 1, addrspace 4)
; CHECK: S_ENDPGM 0
call void @llvm.amdgcn.struct.buffer.store.v2f16(<2 x half> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret void
}
; FIXME:
; define amdgpu_ps void @struct_buffer_store_v3f16_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(<3 x half> %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; call void @llvm.amdgcn.struct.buffer.store.v3f16(<3 x half> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
; ret void
; }
define amdgpu_ps void @struct_buffer_store_v4f16_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(<4 x half> %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
; CHECK-LABEL: name: struct_buffer_store_v4f16_sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset
; CHECK: bb.1 (%ir-block.0):
; CHECK: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3
; CHECK: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1
; CHECK: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr2
; CHECK: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr3
; CHECK: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr4
; CHECK: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr5
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY3]], %subreg.sub1, [[COPY4]], %subreg.sub2, [[COPY5]], %subreg.sub3
; CHECK: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
; CHECK: [[COPY8:%[0-9]+]]:sreg_32 = COPY $sgpr6
; CHECK: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY7]], %subreg.sub1
; CHECK: BUFFER_STORE_DWORDX2_BOTHEN_exact [[REG_SEQUENCE]], [[REG_SEQUENCE2]], [[REG_SEQUENCE1]], [[COPY8]], 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store 8 into custom "BufferResource", align 1, addrspace 4)
; CHECK: S_ENDPGM 0
call void @llvm.amdgcn.struct.buffer.store.v4f16(<4 x half> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret void
}
declare void @llvm.amdgcn.struct.buffer.store.i8(i8, <4 x i32>, i32, i32, i32, i32 immarg)
declare void @llvm.amdgcn.struct.buffer.store.i16(i16, <4 x i32>, i32, i32, i32, i32 immarg)
declare void @llvm.amdgcn.struct.buffer.store.f16(half, <4 x i32>, i32, i32, i32, i32 immarg)
declare void @llvm.amdgcn.struct.buffer.store.v2f16(<2 x half>, <4 x i32>, i32, i32, i32, i32 immarg)
declare void @llvm.amdgcn.struct.buffer.store.v3f16(<3 x half>, <4 x i32>, i32, i32, i32, i32 immarg)
declare void @llvm.amdgcn.struct.buffer.store.v4f16(<4 x half>, <4 x i32>, i32, i32, i32, i32 immarg)
declare void @llvm.amdgcn.struct.buffer.store.f32(float, <4 x i32>, i32, i32, i32, i32 immarg)
declare void @llvm.amdgcn.struct.buffer.store.v2f32(<2 x float>, <4 x i32>, i32, i32, i32, i32 immarg)
declare void @llvm.amdgcn.struct.buffer.store.v3f32(<3 x float>, <4 x i32>, i32, i32, i32, i32 immarg)
declare void @llvm.amdgcn.struct.buffer.store.v4f32(<4 x float>, <4 x i32>, i32, i32, i32, i32 immarg)
Reference in New Issue View Git Blame Copy Permalink