1058 lines
44 KiB
LLVM
1058 lines
44 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=amdgcn-amd- -O0 -mcpu=gfx600 < %s | FileCheck --check-prefixes=GFX6 %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx700 < %s | FileCheck --check-prefixes=GFX7 %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1010 < %s | FileCheck --check-prefixes=GFX10-WGP %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1010 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX10-CU %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdpal -O0 -mcpu=gfx700 -amdgcn-skip-cache-invalidations < %s | FileCheck --check-prefixes=SKIP-CACHE-INV %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 < %s | FileCheck --check-prefixes=GFX11-WGP %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250 %s
|
|
|
|
define amdgpu_kernel void @global_volatile_load_0(
|
|
; GFX6-LABEL: global_volatile_load_0:
|
|
; GFX6: ; %bb.0: ; %entry
|
|
; GFX6-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; GFX6-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9
|
|
; GFX6-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xb
|
|
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-NEXT: s_mov_b32 s2, s5
|
|
; GFX6-NEXT: ; kill: def $sgpr4 killed $sgpr4 killed $sgpr4_sgpr5
|
|
; GFX6-NEXT: s_mov_b32 s8, 0xf000
|
|
; GFX6-NEXT: s_mov_b32 s9, -1
|
|
; GFX6-NEXT: ; kill: def $sgpr4 killed $sgpr4 def $sgpr4_sgpr5_sgpr6_sgpr7
|
|
; GFX6-NEXT: s_mov_b32 s5, s2
|
|
; GFX6-NEXT: s_mov_b32 s6, s9
|
|
; GFX6-NEXT: s_mov_b32 s7, s8
|
|
; GFX6-NEXT: s_mov_b32 s10, s1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; GFX6-NEXT: s_mov_b32 s1, s10
|
|
; GFX6-NEXT: s_mov_b32 s2, s9
|
|
; GFX6-NEXT: s_mov_b32 s3, s8
|
|
; GFX6-NEXT: buffer_load_dword v0, off, s[4:7], 0 glc
|
|
; GFX6-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX6-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: global_volatile_load_0:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_mov_b32 flat_scratch_lo, s13
|
|
; GFX7-NEXT: s_add_i32 s12, s12, s17
|
|
; GFX7-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
|
|
; GFX7-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX7-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x2
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s6
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s7
|
|
; GFX7-NEXT: flat_load_dword v2, v[0:1] glc
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX7-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-WGP-LABEL: global_volatile_load_0:
|
|
; GFX10-WGP: ; %bb.0: ; %entry
|
|
; GFX10-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-WGP-NEXT: global_load_dword v1, v0, s[6:7] glc dlc
|
|
; GFX10-WGP-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-WGP-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX10-CU-LABEL: global_volatile_load_0:
|
|
; GFX10-CU: ; %bb.0: ; %entry
|
|
; GFX10-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-CU-NEXT: global_load_dword v1, v0, s[6:7] glc dlc
|
|
; GFX10-CU-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-CU-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-CU-NEXT: s_endpgm
|
|
;
|
|
; SKIP-CACHE-INV-LABEL: global_volatile_load_0:
|
|
; SKIP-CACHE-INV: ; %bb.0: ; %entry
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x0
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x2
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s2, s5
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr4 killed $sgpr4 killed $sgpr4_sgpr5
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s8, 0xf000
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s9, -1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr4 killed $sgpr4 def $sgpr4_sgpr5_sgpr6_sgpr7
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s5, s2
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s6, s9
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s7, s8
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s10, s1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s1, s10
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s2, s9
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s3, s8
|
|
; SKIP-CACHE-INV-NEXT: buffer_load_dword v0, off, s[4:7], 0 glc
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt vmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; SKIP-CACHE-INV-NEXT: s_endpgm
|
|
;
|
|
; GFX11-WGP-LABEL: global_volatile_load_0:
|
|
; GFX11-WGP: ; %bb.0: ; %entry
|
|
; GFX11-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-WGP-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc
|
|
; GFX11-WGP-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-WGP-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX11-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX11-CU-LABEL: global_volatile_load_0:
|
|
; GFX11-CU: ; %bb.0: ; %entry
|
|
; GFX11-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-CU-NEXT: global_load_b32 v1, v0, s[2:3] glc dlc
|
|
; GFX11-CU-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-CU-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX11-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX12-WGP-LABEL: global_volatile_load_0:
|
|
; GFX12-WGP: ; %bb.0: ; %entry
|
|
; GFX12-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
|
|
; GFX12-WGP-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-WGP-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX12-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX12-CU-LABEL: global_volatile_load_0:
|
|
; GFX12-CU: ; %bb.0: ; %entry
|
|
; GFX12-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
|
|
; GFX12-CU-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-CU-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX12-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX1250-LABEL: global_volatile_load_0:
|
|
; GFX1250: ; %bb.0: ; %entry
|
|
; GFX1250-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX1250-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX1250-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
|
|
; GFX1250-NEXT: s_wait_loadcnt 0x0
|
|
; GFX1250-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX1250-NEXT: s_endpgm
|
|
ptr addrspace(1) %in, ptr addrspace(1) %out) {
|
|
entry:
|
|
%val = load volatile i32, ptr addrspace(1) %in, align 4
|
|
store i32 %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @global_volatile_load_1(
|
|
; GFX6-LABEL: global_volatile_load_1:
|
|
; GFX6: ; %bb.0: ; %entry
|
|
; GFX6-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; GFX6-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9
|
|
; GFX6-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xb
|
|
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-NEXT: s_mov_b32 s8, s1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; GFX6-NEXT: s_mov_b32 s6, 0xf000
|
|
; GFX6-NEXT: s_mov_b32 s7, -1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; GFX6-NEXT: s_mov_b32 s1, s8
|
|
; GFX6-NEXT: s_mov_b32 s2, s7
|
|
; GFX6-NEXT: s_mov_b32 s3, s6
|
|
; GFX6-NEXT: s_mov_b32 s8, 0
|
|
; GFX6-NEXT: ; kill: def $sgpr8 killed $sgpr8 def $sgpr8_sgpr9
|
|
; GFX6-NEXT: s_mov_b32 s9, s6
|
|
; GFX6-NEXT: ; kill: def $sgpr4_sgpr5 killed $sgpr4_sgpr5 def $sgpr4_sgpr5_sgpr6_sgpr7
|
|
; GFX6-NEXT: s_mov_b64 s[6:7], s[8:9]
|
|
; GFX6-NEXT: s_mov_b32 s8, 2
|
|
; GFX6-NEXT: v_lshlrev_b32_e64 v0, s8, v0
|
|
; GFX6-NEXT: s_mov_b32 s8, 0
|
|
; GFX6-NEXT: ; implicit-def: $sgpr8
|
|
; GFX6-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX6-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
|
|
; GFX6-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64 glc
|
|
; GFX6-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX6-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: global_volatile_load_1:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_mov_b32 flat_scratch_lo, s13
|
|
; GFX7-NEXT: s_add_i32 s12, s12, s17
|
|
; GFX7-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], s[8:9]
|
|
; GFX7-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0x0
|
|
; GFX7-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x2
|
|
; GFX7-NEXT: s_mov_b32 s6, 2
|
|
; GFX7-NEXT: v_lshlrev_b32_e64 v1, s6, v0
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: ; implicit-def: $sgpr6
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX7-NEXT: ; kill: def $vgpr1 killed $vgpr1 def $vgpr1_vgpr2 killed $exec
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_mov_b32 s6, s8
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX7-NEXT: s_mov_b32 s8, s9
|
|
; GFX7-NEXT: ; kill: def $vgpr2 killed $vgpr2 killed $vgpr1_vgpr2 killed $exec
|
|
; GFX7-NEXT: v_add_i32_e64 v0, s[6:7], s6, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s8
|
|
; GFX7-NEXT: v_addc_u32_e64 v2, s[6:7], v1, v2, s[6:7]
|
|
; GFX7-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX7-NEXT: flat_load_dword v2, v[0:1] glc
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX7-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-WGP-LABEL: global_volatile_load_1:
|
|
; GFX10-WGP: ; %bb.0: ; %entry
|
|
; GFX10-WGP-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-WGP-NEXT: s_mov_b32 s8, 2
|
|
; GFX10-WGP-NEXT: v_lshlrev_b32_e64 v1, s8, v1
|
|
; GFX10-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-WGP-NEXT: global_load_dword v1, v1, s[6:7] glc dlc
|
|
; GFX10-WGP-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-WGP-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX10-CU-LABEL: global_volatile_load_1:
|
|
; GFX10-CU: ; %bb.0: ; %entry
|
|
; GFX10-CU-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-CU-NEXT: s_mov_b32 s8, 2
|
|
; GFX10-CU-NEXT: v_lshlrev_b32_e64 v1, s8, v1
|
|
; GFX10-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-CU-NEXT: global_load_dword v1, v1, s[6:7] glc dlc
|
|
; GFX10-CU-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-CU-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-CU-NEXT: s_endpgm
|
|
;
|
|
; SKIP-CACHE-INV-LABEL: global_volatile_load_1:
|
|
; SKIP-CACHE-INV: ; %bb.0: ; %entry
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x0
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x2
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s8, s1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s6, 0xf000
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s7, -1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s1, s8
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s2, s7
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s3, s6
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s8, 0
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr8 killed $sgpr8 def $sgpr8_sgpr9
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s9, s6
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr4_sgpr5 killed $sgpr4_sgpr5 def $sgpr4_sgpr5_sgpr6_sgpr7
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b64 s[6:7], s[8:9]
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s8, 2
|
|
; SKIP-CACHE-INV-NEXT: v_lshlrev_b32_e64 v0, s8, v0
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s8, 0
|
|
; SKIP-CACHE-INV-NEXT: ; implicit-def: $sgpr8
|
|
; SKIP-CACHE-INV-NEXT: v_mov_b32_e32 v2, 0
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
|
|
; SKIP-CACHE-INV-NEXT: v_mov_b32_e32 v1, v2
|
|
; SKIP-CACHE-INV-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64 glc
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt vmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; SKIP-CACHE-INV-NEXT: s_endpgm
|
|
;
|
|
; GFX11-WGP-LABEL: global_volatile_load_1:
|
|
; GFX11-WGP: ; %bb.0: ; %entry
|
|
; GFX11-WGP-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX11-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-WGP-NEXT: s_mov_b32 s4, 0x3ff
|
|
; GFX11-WGP-NEXT: v_and_b32_e64 v1, v1, s4
|
|
; GFX11-WGP-NEXT: s_mov_b32 s4, 2
|
|
; GFX11-WGP-NEXT: v_lshlrev_b32_e64 v1, s4, v1
|
|
; GFX11-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-WGP-NEXT: global_load_b32 v1, v1, s[2:3] glc dlc
|
|
; GFX11-WGP-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-WGP-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX11-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX11-CU-LABEL: global_volatile_load_1:
|
|
; GFX11-CU: ; %bb.0: ; %entry
|
|
; GFX11-CU-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX11-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-CU-NEXT: s_mov_b32 s4, 0x3ff
|
|
; GFX11-CU-NEXT: v_and_b32_e64 v1, v1, s4
|
|
; GFX11-CU-NEXT: s_mov_b32 s4, 2
|
|
; GFX11-CU-NEXT: v_lshlrev_b32_e64 v1, s4, v1
|
|
; GFX11-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-CU-NEXT: global_load_b32 v1, v1, s[2:3] glc dlc
|
|
; GFX11-CU-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-CU-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX11-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX12-WGP-LABEL: global_volatile_load_1:
|
|
; GFX12-WGP: ; %bb.0: ; %entry
|
|
; GFX12-WGP-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX12-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-WGP-NEXT: s_mov_b32 s4, 0x3ff
|
|
; GFX12-WGP-NEXT: v_and_b32_e64 v1, v1, s4
|
|
; GFX12-WGP-NEXT: s_mov_b32 s4, 2
|
|
; GFX12-WGP-NEXT: s_wait_alu 0xfffe
|
|
; GFX12-WGP-NEXT: v_lshlrev_b32_e64 v1, s4, v1
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: global_load_b32 v1, v1, s[2:3] scope:SCOPE_SYS
|
|
; GFX12-WGP-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-WGP-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX12-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX12-CU-LABEL: global_volatile_load_1:
|
|
; GFX12-CU: ; %bb.0: ; %entry
|
|
; GFX12-CU-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX12-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-CU-NEXT: s_mov_b32 s4, 0x3ff
|
|
; GFX12-CU-NEXT: v_and_b32_e64 v1, v1, s4
|
|
; GFX12-CU-NEXT: s_mov_b32 s4, 2
|
|
; GFX12-CU-NEXT: s_wait_alu 0xfffe
|
|
; GFX12-CU-NEXT: v_lshlrev_b32_e64 v1, s4, v1
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: global_load_b32 v1, v1, s[2:3] scope:SCOPE_SYS
|
|
; GFX12-CU-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-CU-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX12-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX1250-LABEL: global_volatile_load_1:
|
|
; GFX1250: ; %bb.0: ; %entry
|
|
; GFX1250-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX1250-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX1250-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX1250-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX1250-NEXT: s_wait_xcnt 0x0
|
|
; GFX1250-NEXT: s_mov_b32 s4, 0x3ff
|
|
; GFX1250-NEXT: v_and_b32_e64 v1, v1, s4
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: global_load_b32 v1, v1, s[2:3] scale_offset scope:SCOPE_SYS
|
|
; GFX1250-NEXT: s_wait_loadcnt 0x0
|
|
; GFX1250-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX1250-NEXT: s_endpgm
|
|
ptr addrspace(1) %in, ptr addrspace(1) %out) {
|
|
entry:
|
|
%tid = call i32 @llvm.amdgcn.workitem.id.x()
|
|
%val.gep = getelementptr inbounds i32, ptr addrspace(1) %in, i32 %tid
|
|
%val = load volatile i32, ptr addrspace(1) %val.gep, align 4
|
|
store i32 %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @global_volatile_store_0(
|
|
; GFX6-LABEL: global_volatile_store_0:
|
|
; GFX6: ; %bb.0: ; %entry
|
|
; GFX6-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; GFX6-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9
|
|
; GFX6-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xb
|
|
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-NEXT: s_mov_b32 s8, s1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; GFX6-NEXT: s_mov_b32 s6, 0xf000
|
|
; GFX6-NEXT: s_mov_b32 s7, -1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; GFX6-NEXT: s_mov_b32 s1, s8
|
|
; GFX6-NEXT: s_mov_b32 s2, s7
|
|
; GFX6-NEXT: s_mov_b32 s3, s6
|
|
; GFX6-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX6-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX6-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: global_volatile_store_0:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_mov_b32 flat_scratch_lo, s13
|
|
; GFX7-NEXT: s_add_i32 s12, s12, s17
|
|
; GFX7-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
|
|
; GFX7-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x0
|
|
; GFX7-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x2
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s6
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s7
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX7-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-WGP-LABEL: global_volatile_store_0:
|
|
; GFX10-WGP: ; %bb.0: ; %entry
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-WGP-NEXT: s_load_dword s6, s[6:7], 0x0
|
|
; GFX10-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-WGP-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX10-WGP-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-WGP-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX10-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX10-CU-LABEL: global_volatile_store_0:
|
|
; GFX10-CU: ; %bb.0: ; %entry
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-CU-NEXT: s_load_dword s6, s[6:7], 0x0
|
|
; GFX10-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-CU-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX10-CU-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-CU-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX10-CU-NEXT: s_endpgm
|
|
;
|
|
; SKIP-CACHE-INV-LABEL: global_volatile_store_0:
|
|
; SKIP-CACHE-INV: ; %bb.0: ; %entry
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x0
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x2
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s8, s1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s6, 0xf000
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s7, -1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s1, s8
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s2, s7
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s3, s6
|
|
; SKIP-CACHE-INV-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: v_mov_b32_e32 v0, s4
|
|
; SKIP-CACHE-INV-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt vmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: s_endpgm
|
|
;
|
|
; GFX11-WGP-LABEL: global_volatile_store_0:
|
|
; GFX11-WGP: ; %bb.0: ; %entry
|
|
; GFX11-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-WGP-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX11-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-WGP-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX11-WGP-NEXT: global_store_b32 v0, v1, s[0:1] dlc
|
|
; GFX11-WGP-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX11-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX11-CU-LABEL: global_volatile_store_0:
|
|
; GFX11-CU: ; %bb.0: ; %entry
|
|
; GFX11-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-CU-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX11-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-CU-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX11-CU-NEXT: global_store_b32 v0, v1, s[0:1] dlc
|
|
; GFX11-CU-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX11-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX12-WGP-LABEL: global_volatile_store_0:
|
|
; GFX12-WGP: ; %bb.0: ; %entry
|
|
; GFX12-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX12-WGP-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_storecnt 0x0
|
|
; GFX12-WGP-NEXT: global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
|
|
; GFX12-WGP-NEXT: s_wait_storecnt 0x0
|
|
; GFX12-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX12-CU-LABEL: global_volatile_store_0:
|
|
; GFX12-CU: ; %bb.0: ; %entry
|
|
; GFX12-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX12-CU-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_storecnt 0x0
|
|
; GFX12-CU-NEXT: global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
|
|
; GFX12-CU-NEXT: s_wait_storecnt 0x0
|
|
; GFX12-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX1250-LABEL: global_volatile_store_0:
|
|
; GFX1250: ; %bb.0: ; %entry
|
|
; GFX1250-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX1250-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX1250-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX1250-NEXT: global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
|
|
; GFX1250-NEXT: s_wait_storecnt 0x0
|
|
; GFX1250-NEXT: s_endpgm
|
|
ptr addrspace(1) %in, ptr addrspace(1) %out) {
|
|
entry:
|
|
%val = load i32, ptr addrspace(1) %in, align 4
|
|
store volatile i32 %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @global_volatile_store_1(
|
|
; GFX6-LABEL: global_volatile_store_1:
|
|
; GFX6: ; %bb.0: ; %entry
|
|
; GFX6-NEXT: s_load_dwordx2 s[2:3], s[4:5], 0x9
|
|
; GFX6-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xb
|
|
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-NEXT: s_load_dword s4, s[2:3], 0x0
|
|
; GFX6-NEXT: s_mov_b32 s2, 0xf000
|
|
; GFX6-NEXT: s_mov_b32 s6, 0
|
|
; GFX6-NEXT: ; kill: def $sgpr6 killed $sgpr6 def $sgpr6_sgpr7
|
|
; GFX6-NEXT: s_mov_b32 s7, s2
|
|
; GFX6-NEXT: ; kill: def $sgpr0_sgpr1 killed $sgpr0_sgpr1 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; GFX6-NEXT: s_mov_b64 s[2:3], s[6:7]
|
|
; GFX6-NEXT: s_mov_b32 s5, 2
|
|
; GFX6-NEXT: v_lshlrev_b32_e64 v1, s5, v0
|
|
; GFX6-NEXT: s_mov_b32 s5, 0
|
|
; GFX6-NEXT: ; implicit-def: $sgpr5
|
|
; GFX6-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX6-NEXT: ; kill: def $vgpr1 killed $vgpr1 def $vgpr1_vgpr2 killed $exec
|
|
; GFX6-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX6-NEXT: buffer_store_dword v0, v[1:2], s[0:3], 0 addr64
|
|
; GFX6-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX6-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: global_volatile_store_1:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_mov_b32 flat_scratch_lo, s13
|
|
; GFX7-NEXT: s_add_i32 s12, s12, s17
|
|
; GFX7-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
|
|
; GFX7-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x0
|
|
; GFX7-NEXT: s_load_dwordx2 s[8:9], s[8:9], 0x2
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_load_dword s4, s[4:5], 0x0
|
|
; GFX7-NEXT: s_mov_b32 s5, 2
|
|
; GFX7-NEXT: v_lshlrev_b32_e64 v1, s5, v0
|
|
; GFX7-NEXT: s_mov_b32 s5, 0
|
|
; GFX7-NEXT: ; implicit-def: $sgpr5
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX7-NEXT: ; kill: def $vgpr1 killed $vgpr1 def $vgpr1_vgpr2 killed $exec
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX7-NEXT: s_mov_b32 s6, s8
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX7-NEXT: s_mov_b32 s5, s9
|
|
; GFX7-NEXT: ; kill: def $vgpr2 killed $vgpr2 killed $vgpr1_vgpr2 killed $exec
|
|
; GFX7-NEXT: v_add_i32_e64 v0, s[6:7], s6, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX7-NEXT: v_addc_u32_e64 v2, s[6:7], v1, v2, s[6:7]
|
|
; GFX7-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX7-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-WGP-LABEL: global_volatile_store_1:
|
|
; GFX10-WGP: ; %bb.0: ; %entry
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-WGP-NEXT: s_load_dword s6, s[6:7], 0x0
|
|
; GFX10-WGP-NEXT: s_mov_b32 s7, 2
|
|
; GFX10-WGP-NEXT: v_lshlrev_b32_e64 v0, s7, v0
|
|
; GFX10-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-WGP-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX10-WGP-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-WGP-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX10-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX10-CU-LABEL: global_volatile_store_1:
|
|
; GFX10-CU: ; %bb.0: ; %entry
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-CU-NEXT: s_load_dword s6, s[6:7], 0x0
|
|
; GFX10-CU-NEXT: s_mov_b32 s7, 2
|
|
; GFX10-CU-NEXT: v_lshlrev_b32_e64 v0, s7, v0
|
|
; GFX10-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-CU-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX10-CU-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-CU-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX10-CU-NEXT: s_endpgm
|
|
;
|
|
; SKIP-CACHE-INV-LABEL: global_volatile_store_1:
|
|
; SKIP-CACHE-INV: ; %bb.0: ; %entry
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[2:3], s[4:5], 0x0
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x2
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: s_load_dword s4, s[2:3], 0x0
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s2, 0xf000
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s6, 0
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr6 killed $sgpr6 def $sgpr6_sgpr7
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s7, s2
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0_sgpr1 killed $sgpr0_sgpr1 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b64 s[2:3], s[6:7]
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s5, 2
|
|
; SKIP-CACHE-INV-NEXT: v_lshlrev_b32_e64 v1, s5, v0
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s5, 0
|
|
; SKIP-CACHE-INV-NEXT: ; implicit-def: $sgpr5
|
|
; SKIP-CACHE-INV-NEXT: v_mov_b32_e32 v0, 0
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $vgpr1 killed $vgpr1 def $vgpr1_vgpr2 killed $exec
|
|
; SKIP-CACHE-INV-NEXT: v_mov_b32_e32 v2, v0
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: v_mov_b32_e32 v0, s4
|
|
; SKIP-CACHE-INV-NEXT: buffer_store_dword v0, v[1:2], s[0:3], 0 addr64
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt vmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: s_endpgm
|
|
;
|
|
; GFX11-WGP-LABEL: global_volatile_store_1:
|
|
; GFX11-WGP: ; %bb.0: ; %entry
|
|
; GFX11-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-WGP-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX11-WGP-NEXT: s_mov_b32 s3, 0x3ff
|
|
; GFX11-WGP-NEXT: v_and_b32_e64 v0, v0, s3
|
|
; GFX11-WGP-NEXT: s_mov_b32 s3, 2
|
|
; GFX11-WGP-NEXT: v_lshlrev_b32_e64 v0, s3, v0
|
|
; GFX11-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-WGP-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX11-WGP-NEXT: global_store_b32 v0, v1, s[0:1] dlc
|
|
; GFX11-WGP-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX11-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX11-CU-LABEL: global_volatile_store_1:
|
|
; GFX11-CU: ; %bb.0: ; %entry
|
|
; GFX11-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-CU-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX11-CU-NEXT: s_mov_b32 s3, 0x3ff
|
|
; GFX11-CU-NEXT: v_and_b32_e64 v0, v0, s3
|
|
; GFX11-CU-NEXT: s_mov_b32 s3, 2
|
|
; GFX11-CU-NEXT: v_lshlrev_b32_e64 v0, s3, v0
|
|
; GFX11-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-CU-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX11-CU-NEXT: global_store_b32 v0, v1, s[0:1] dlc
|
|
; GFX11-CU-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX11-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX12-WGP-LABEL: global_volatile_store_1:
|
|
; GFX12-WGP: ; %bb.0: ; %entry
|
|
; GFX12-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX12-WGP-NEXT: s_mov_b32 s3, 0x3ff
|
|
; GFX12-WGP-NEXT: v_and_b32_e64 v0, v0, s3
|
|
; GFX12-WGP-NEXT: s_mov_b32 s3, 2
|
|
; GFX12-WGP-NEXT: s_wait_alu 0xfffe
|
|
; GFX12-WGP-NEXT: v_lshlrev_b32_e64 v0, s3, v0
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX12-WGP-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_storecnt 0x0
|
|
; GFX12-WGP-NEXT: global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
|
|
; GFX12-WGP-NEXT: s_wait_storecnt 0x0
|
|
; GFX12-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX12-CU-LABEL: global_volatile_store_1:
|
|
; GFX12-CU: ; %bb.0: ; %entry
|
|
; GFX12-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX12-CU-NEXT: s_mov_b32 s3, 0x3ff
|
|
; GFX12-CU-NEXT: v_and_b32_e64 v0, v0, s3
|
|
; GFX12-CU-NEXT: s_mov_b32 s3, 2
|
|
; GFX12-CU-NEXT: s_wait_alu 0xfffe
|
|
; GFX12-CU-NEXT: v_lshlrev_b32_e64 v0, s3, v0
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX12-CU-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: s_wait_storecnt 0x0
|
|
; GFX12-CU-NEXT: global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
|
|
; GFX12-CU-NEXT: s_wait_storecnt 0x0
|
|
; GFX12-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX1250-LABEL: global_volatile_store_1:
|
|
; GFX1250: ; %bb.0: ; %entry
|
|
; GFX1250-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX1250-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: s_load_b32 s2, s[2:3], 0x0
|
|
; GFX1250-NEXT: s_wait_xcnt 0x0
|
|
; GFX1250-NEXT: s_mov_b32 s3, 0x3ff
|
|
; GFX1250-NEXT: v_and_b32_e64 v0, v0, s3
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX1250-NEXT: global_store_b32 v0, v1, s[0:1] scale_offset scope:SCOPE_SYS
|
|
; GFX1250-NEXT: s_wait_storecnt 0x0
|
|
; GFX1250-NEXT: s_endpgm
|
|
ptr addrspace(1) %in, ptr addrspace(1) %out) {
|
|
entry:
|
|
%tid = call i32 @llvm.amdgcn.workitem.id.x()
|
|
%val = load i32, ptr addrspace(1) %in, align 4
|
|
%out.gep = getelementptr inbounds i32, ptr addrspace(1) %out, i32 %tid
|
|
store volatile i32 %val, ptr addrspace(1) %out.gep
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @global_volatile_workgroup_acquire_load(
|
|
; GFX6-LABEL: global_volatile_workgroup_acquire_load:
|
|
; GFX6: ; %bb.0: ; %entry
|
|
; GFX6-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; GFX6-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9
|
|
; GFX6-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xb
|
|
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-NEXT: s_mov_b32 s2, s5
|
|
; GFX6-NEXT: ; kill: def $sgpr4 killed $sgpr4 killed $sgpr4_sgpr5
|
|
; GFX6-NEXT: s_mov_b32 s8, 0xf000
|
|
; GFX6-NEXT: s_mov_b32 s9, -1
|
|
; GFX6-NEXT: ; kill: def $sgpr4 killed $sgpr4 def $sgpr4_sgpr5_sgpr6_sgpr7
|
|
; GFX6-NEXT: s_mov_b32 s5, s2
|
|
; GFX6-NEXT: s_mov_b32 s6, s9
|
|
; GFX6-NEXT: s_mov_b32 s7, s8
|
|
; GFX6-NEXT: s_mov_b32 s10, s1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; GFX6-NEXT: s_mov_b32 s1, s10
|
|
; GFX6-NEXT: s_mov_b32 s2, s9
|
|
; GFX6-NEXT: s_mov_b32 s3, s8
|
|
; GFX6-NEXT: buffer_load_dword v0, off, s[4:7], 0
|
|
; GFX6-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX6-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: global_volatile_workgroup_acquire_load:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_mov_b32 flat_scratch_lo, s13
|
|
; GFX7-NEXT: s_add_i32 s12, s12, s17
|
|
; GFX7-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
|
|
; GFX7-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX7-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x2
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s6
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s7
|
|
; GFX7-NEXT: flat_load_dword v2, v[0:1]
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-WGP-LABEL: global_volatile_workgroup_acquire_load:
|
|
; GFX10-WGP: ; %bb.0: ; %entry
|
|
; GFX10-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-WGP-NEXT: global_load_dword v1, v0, s[6:7] glc
|
|
; GFX10-WGP-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-WGP-NEXT: buffer_gl0_inv
|
|
; GFX10-WGP-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX10-CU-LABEL: global_volatile_workgroup_acquire_load:
|
|
; GFX10-CU: ; %bb.0: ; %entry
|
|
; GFX10-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-CU-NEXT: global_load_dword v1, v0, s[6:7]
|
|
; GFX10-CU-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-CU-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-CU-NEXT: s_endpgm
|
|
;
|
|
; SKIP-CACHE-INV-LABEL: global_volatile_workgroup_acquire_load:
|
|
; SKIP-CACHE-INV: ; %bb.0: ; %entry
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x0
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x2
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s2, s5
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr4 killed $sgpr4 killed $sgpr4_sgpr5
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s8, 0xf000
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s9, -1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr4 killed $sgpr4 def $sgpr4_sgpr5_sgpr6_sgpr7
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s5, s2
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s6, s9
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s7, s8
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s10, s1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s1, s10
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s2, s9
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s3, s8
|
|
; SKIP-CACHE-INV-NEXT: buffer_load_dword v0, off, s[4:7], 0
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt vmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; SKIP-CACHE-INV-NEXT: s_endpgm
|
|
;
|
|
; GFX11-WGP-LABEL: global_volatile_workgroup_acquire_load:
|
|
; GFX11-WGP: ; %bb.0: ; %entry
|
|
; GFX11-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-WGP-NEXT: global_load_b32 v1, v0, s[2:3] glc
|
|
; GFX11-WGP-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-WGP-NEXT: buffer_gl0_inv
|
|
; GFX11-WGP-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX11-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX11-CU-LABEL: global_volatile_workgroup_acquire_load:
|
|
; GFX11-CU: ; %bb.0: ; %entry
|
|
; GFX11-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX11-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-CU-NEXT: global_load_b32 v1, v0, s[2:3]
|
|
; GFX11-CU-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-CU-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX11-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX12-WGP-LABEL: global_volatile_workgroup_acquire_load:
|
|
; GFX12-WGP: ; %bb.0: ; %entry
|
|
; GFX12-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-WGP-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: global_load_b32 v1, v0, s[2:3] scope:SCOPE_SE
|
|
; GFX12-WGP-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-WGP-NEXT: global_inv scope:SCOPE_SE
|
|
; GFX12-WGP-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX12-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX12-CU-LABEL: global_volatile_workgroup_acquire_load:
|
|
; GFX12-CU: ; %bb.0: ; %entry
|
|
; GFX12-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-CU-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX12-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: global_load_b32 v1, v0, s[2:3]
|
|
; GFX12-CU-NEXT: s_wait_loadcnt 0x0
|
|
; GFX12-CU-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX12-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX1250-LABEL: global_volatile_workgroup_acquire_load:
|
|
; GFX1250: ; %bb.0: ; %entry
|
|
; GFX1250-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX1250-NEXT: s_load_b64 s[2:3], s[4:5], 0x0
|
|
; GFX1250-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: global_load_b32 v1, v0, s[2:3]
|
|
; GFX1250-NEXT: s_wait_loadcnt 0x0
|
|
; GFX1250-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX1250-NEXT: s_endpgm
|
|
ptr addrspace(1) %in, ptr addrspace(1) %out) {
|
|
entry:
|
|
%val = load atomic volatile i32, ptr addrspace(1) %in syncscope("workgroup") acquire, align 4
|
|
store i32 %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @global_volatile_workgroup_release_store(
|
|
; GFX6-LABEL: global_volatile_workgroup_release_store:
|
|
; GFX6: ; %bb.0: ; %entry
|
|
; GFX6-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; GFX6-NEXT: s_load_dword s4, s[0:1], 0x9
|
|
; GFX6-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0xb
|
|
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-NEXT: s_mov_b32 s7, s1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; GFX6-NEXT: s_mov_b32 s5, 0xf000
|
|
; GFX6-NEXT: s_mov_b32 s6, -1
|
|
; GFX6-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; GFX6-NEXT: s_mov_b32 s1, s7
|
|
; GFX6-NEXT: s_mov_b32 s2, s6
|
|
; GFX6-NEXT: s_mov_b32 s3, s5
|
|
; GFX6-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: global_volatile_workgroup_release_store:
|
|
; GFX7: ; %bb.0: ; %entry
|
|
; GFX7-NEXT: s_mov_b32 flat_scratch_lo, s13
|
|
; GFX7-NEXT: s_add_i32 s12, s12, s17
|
|
; GFX7-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
|
|
; GFX7-NEXT: s_load_dword s4, s[8:9], 0x0
|
|
; GFX7-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x2
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s6
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s7
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-WGP-LABEL: global_volatile_workgroup_release_store:
|
|
; GFX10-WGP: ; %bb.0: ; %entry
|
|
; GFX10-WGP-NEXT: s_load_dword s6, s[8:9], 0x0
|
|
; GFX10-WGP-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-WGP-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX10-WGP-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
|
; GFX10-WGP-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX10-WGP-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX10-CU-LABEL: global_volatile_workgroup_release_store:
|
|
; GFX10-CU: ; %bb.0: ; %entry
|
|
; GFX10-CU-NEXT: s_load_dword s6, s[8:9], 0x0
|
|
; GFX10-CU-NEXT: s_load_dwordx2 s[4:5], s[8:9], 0x8
|
|
; GFX10-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-CU-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX10-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-CU-NEXT: global_store_dword v0, v1, s[4:5]
|
|
; GFX10-CU-NEXT: s_endpgm
|
|
;
|
|
; SKIP-CACHE-INV-LABEL: global_volatile_workgroup_release_store:
|
|
; SKIP-CACHE-INV: ; %bb.0: ; %entry
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b64 s[0:1], s[4:5]
|
|
; SKIP-CACHE-INV-NEXT: s_load_dword s4, s[0:1], 0x0
|
|
; SKIP-CACHE-INV-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x2
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s7, s1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 killed $sgpr0_sgpr1
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s5, 0xf000
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s6, -1
|
|
; SKIP-CACHE-INV-NEXT: ; kill: def $sgpr0 killed $sgpr0 def $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s1, s7
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s2, s6
|
|
; SKIP-CACHE-INV-NEXT: s_mov_b32 s3, s5
|
|
; SKIP-CACHE-INV-NEXT: v_mov_b32_e32 v0, s4
|
|
; SKIP-CACHE-INV-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SKIP-CACHE-INV-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; SKIP-CACHE-INV-NEXT: s_endpgm
|
|
;
|
|
; GFX11-WGP-LABEL: global_volatile_workgroup_release_store:
|
|
; GFX11-WGP: ; %bb.0: ; %entry
|
|
; GFX11-WGP-NEXT: s_load_b32 s2, s[4:5], 0x0
|
|
; GFX11-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-WGP-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-WGP-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX11-WGP-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
|
|
; GFX11-WGP-NEXT: s_waitcnt_vscnt null, 0x0
|
|
; GFX11-WGP-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX11-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX11-CU-LABEL: global_volatile_workgroup_release_store:
|
|
; GFX11-CU: ; %bb.0: ; %entry
|
|
; GFX11-CU-NEXT: s_load_b32 s2, s[4:5], 0x0
|
|
; GFX11-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX11-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-CU-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX11-CU-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-CU-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX11-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX12-WGP-LABEL: global_volatile_workgroup_release_store:
|
|
; GFX12-WGP: ; %bb.0: ; %entry
|
|
; GFX12-WGP-NEXT: s_load_b32 s2, s[4:5], 0x0
|
|
; GFX12-WGP-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-WGP-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-WGP-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-WGP-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX12-WGP-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_storecnt 0x0
|
|
; GFX12-WGP-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-WGP-NEXT: global_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
|
|
; GFX12-WGP-NEXT: s_endpgm
|
|
;
|
|
; GFX12-CU-LABEL: global_volatile_workgroup_release_store:
|
|
; GFX12-CU: ; %bb.0: ; %entry
|
|
; GFX12-CU-NEXT: s_load_b32 s2, s[4:5], 0x0
|
|
; GFX12-CU-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX12-CU-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX12-CU-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-CU-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX12-CU-NEXT: s_wait_dscnt 0x0
|
|
; GFX12-CU-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX12-CU-NEXT: s_endpgm
|
|
;
|
|
; GFX1250-LABEL: global_volatile_workgroup_release_store:
|
|
; GFX1250: ; %bb.0: ; %entry
|
|
; GFX1250-NEXT: s_load_b32 s2, s[4:5], 0x0
|
|
; GFX1250-NEXT: s_load_b64 s[0:1], s[4:5], 0x8
|
|
; GFX1250-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX1250-NEXT: s_wait_storecnt 0x0
|
|
; GFX1250-NEXT: s_wait_xcnt 0x0
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: global_store_b32 v0, v1, s[0:1]
|
|
; GFX1250-NEXT: s_endpgm
|
|
i32 %in, ptr addrspace(1) %out) {
|
|
entry:
|
|
store atomic volatile i32 %in, ptr addrspace(1) %out syncscope("workgroup") release, align 4
|
|
ret void
|
|
}
|
|
|
|
declare i32 @llvm.amdgcn.workitem.id.x()
|