108 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py2;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 --stop-after=greedy,2 < %s | FileCheck -check-prefix=REGALLOC-GFX908 %s3;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 --stop-after=prologepilog < %s | FileCheck -check-prefix=PEI-GFX908 %s4;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a --stop-after=greedy,2 < %s | FileCheck -check-prefix=REGALLOC-GFX90A %s5;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a --stop-after=prologepilog < %s | FileCheck -check-prefix=PEI-GFX90A %s6 7; Partial reg copy and spill missed during regalloc handled later at frame lowering.8define amdgpu_kernel void @partial_copy(<4 x i32> %arg) #0 {9 10 ; REGALLOC-GFX908-LABEL: name: partial_copy11 ; REGALLOC-GFX908: bb.0 (%ir-block.0):12 ; REGALLOC-GFX908-NEXT: liveins: $sgpr4_sgpr513 ; REGALLOC-GFX908-NEXT: {{ $}}14 ; REGALLOC-GFX908-NEXT: INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 1638409 /* reguse:AGPR_32 */, undef %6:agpr_3215 ; REGALLOC-GFX908-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 7012362 /* regdef:VReg_128 */, def %2516 ; REGALLOC-GFX908-NEXT: [[COPY:%[0-9]+]]:av_128 = COPY %2517 ; REGALLOC-GFX908-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 2818058 /* regdef:VReg_64 */, def %2718 ; REGALLOC-GFX908-NEXT: SI_SPILL_AV64_SAVE %27, %stack.0, $sgpr32, 0, implicit $exec :: (store (s64) into %stack.0, align 4, addrspace 5)19 ; REGALLOC-GFX908-NEXT: [[COPY1:%[0-9]+]]:vreg_128 = COPY [[COPY]]20 ; REGALLOC-GFX908-NEXT: GLOBAL_STORE_DWORDX4 undef %15:vreg_64, [[COPY1]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)21 ; REGALLOC-GFX908-NEXT: renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (s128) from %ir.arg.kernarg.offset1, addrspace 4)22 ; REGALLOC-GFX908-NEXT: [[COPY2:%[0-9]+]]:areg_128 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr323 ; REGALLOC-GFX908-NEXT: [[AV_MOV_:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 1, implicit $exec24 ; REGALLOC-GFX908-NEXT: [[AV_MOV_1:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 2, implicit $exec25 ; REGALLOC-GFX908-NEXT: [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY2]], 0, 0, 0, implicit $mode, implicit $exec26 ; REGALLOC-GFX908-NEXT: [[SI_SPILL_AV64_RESTORE:%[0-9]+]]:vreg_64 = SI_SPILL_AV64_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load (s64) from %stack.0, align 4, addrspace 5)27 ; REGALLOC-GFX908-NEXT: GLOBAL_STORE_DWORDX2 undef %17:vreg_64, [[SI_SPILL_AV64_RESTORE]], 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)28 ; REGALLOC-GFX908-NEXT: [[COPY3:%[0-9]+]]:vreg_128 = COPY [[V_MFMA_I32_4X4X4I8_e64_]]29 ; REGALLOC-GFX908-NEXT: GLOBAL_STORE_DWORDX4 undef %19:vreg_64, [[COPY3]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)30 ; REGALLOC-GFX908-NEXT: S_ENDPGM 031 ;32 ; PEI-GFX908-LABEL: name: partial_copy33 ; PEI-GFX908: bb.0 (%ir-block.0):34 ; PEI-GFX908-NEXT: liveins: $agpr4, $sgpr4_sgpr5, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr935 ; PEI-GFX908-NEXT: {{ $}}36 ; PEI-GFX908-NEXT: $sgpr12_sgpr13_sgpr14_sgpr15 = COPY killed $sgpr0_sgpr1_sgpr2_sgpr337 ; PEI-GFX908-NEXT: $sgpr12 = S_ADD_U32 $sgpr12, $sgpr9, implicit-def $scc, implicit-def $sgpr12_sgpr13_sgpr14_sgpr1538 ; PEI-GFX908-NEXT: $sgpr13 = S_ADDC_U32 $sgpr13, 0, implicit-def dead $scc, implicit $scc, implicit-def $sgpr12_sgpr13_sgpr14_sgpr1539 ; PEI-GFX908-NEXT: INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 1638409 /* reguse:AGPR_32 */, undef renamable $agpr040 ; PEI-GFX908-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 7012362 /* regdef:VReg_128 */, def renamable $vgpr0_vgpr1_vgpr2_vgpr341 ; PEI-GFX908-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = COPY killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, implicit $exec42 ; PEI-GFX908-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 2818058 /* regdef:VReg_64 */, def renamable $vgpr0_vgpr143 ; PEI-GFX908-NEXT: BUFFER_STORE_DWORD_OFFSET killed $vgpr0, $sgpr12_sgpr13_sgpr14_sgpr15, 0, 0, 0, 0, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $vgpr0_vgpr1 :: (store (s32) into %stack.0, addrspace 5)44 ; PEI-GFX908-NEXT: $agpr4 = V_ACCVGPR_WRITE_B32_e64 killed $vgpr1, implicit $exec, implicit killed $vgpr0_vgpr145 ; PEI-GFX908-NEXT: renamable $vgpr0_vgpr1_vgpr2_vgpr3 = COPY killed renamable $agpr0_agpr1_agpr2_agpr3, implicit $exec46 ; PEI-GFX908-NEXT: GLOBAL_STORE_DWORDX4 undef renamable $vgpr0_vgpr1, killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)47 ; PEI-GFX908-NEXT: renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (s128) from %ir.arg.kernarg.offset1, addrspace 4)48 ; PEI-GFX908-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, implicit $exec49 ; PEI-GFX908-NEXT: renamable $vgpr0 = AV_MOV_B32_IMM_PSEUDO 1, implicit $exec50 ; PEI-GFX908-NEXT: renamable $vgpr1 = AV_MOV_B32_IMM_PSEUDO 2, implicit $exec51 ; PEI-GFX908-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_I32_4X4X4I8_e64 killed $vgpr0, killed $vgpr1, killed $agpr0_agpr1_agpr2_agpr3, 0, 0, 0, implicit $mode, implicit $exec52 ; PEI-GFX908-NEXT: $vgpr0 = BUFFER_LOAD_DWORD_OFFSET $sgpr12_sgpr13_sgpr14_sgpr15, 0, 0, 0, 0, implicit $exec, implicit-def $vgpr0_vgpr1 :: (load (s32) from %stack.0, addrspace 5)53 ; PEI-GFX908-NEXT: $vgpr1 = V_ACCVGPR_READ_B32_e64 $agpr4, implicit $exec, implicit $vgpr0_vgpr154 ; PEI-GFX908-NEXT: GLOBAL_STORE_DWORDX2 undef renamable $vgpr0_vgpr1, killed renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)55 ; PEI-GFX908-NEXT: renamable $vgpr0_vgpr1_vgpr2_vgpr3 = COPY killed renamable $agpr0_agpr1_agpr2_agpr3, implicit $exec56 ; PEI-GFX908-NEXT: GLOBAL_STORE_DWORDX4 undef renamable $vgpr0_vgpr1, killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)57 ; PEI-GFX908-NEXT: S_ENDPGM 058 ;59 ; REGALLOC-GFX90A-LABEL: name: partial_copy60 ; REGALLOC-GFX90A: bb.0 (%ir-block.0):61 ; REGALLOC-GFX90A-NEXT: liveins: $sgpr4_sgpr562 ; REGALLOC-GFX90A-NEXT: {{ $}}63 ; REGALLOC-GFX90A-NEXT: INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 1638409 /* reguse:AGPR_32 */, undef %6:agpr_3264 ; REGALLOC-GFX90A-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 7340042 /* regdef:VReg_128_Align2 */, def %2365 ; REGALLOC-GFX90A-NEXT: [[COPY:%[0-9]+]]:av_128_align2 = COPY %2366 ; REGALLOC-GFX90A-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3080202 /* regdef:VReg_64_Align2 */, def %2167 ; REGALLOC-GFX90A-NEXT: [[COPY1:%[0-9]+]]:av_64_align2 = COPY %2168 ; REGALLOC-GFX90A-NEXT: GLOBAL_STORE_DWORDX4 undef %15:vreg_64_align2, [[COPY]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)69 ; REGALLOC-GFX90A-NEXT: renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (s128) from %ir.arg.kernarg.offset1, addrspace 4)70 ; REGALLOC-GFX90A-NEXT: [[COPY2:%[0-9]+]]:areg_128_align2 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr371 ; REGALLOC-GFX90A-NEXT: [[AV_MOV_:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 1, implicit $exec72 ; REGALLOC-GFX90A-NEXT: [[AV_MOV_1:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 2, implicit $exec73 ; REGALLOC-GFX90A-NEXT: [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128_align2 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY2]], 0, 0, 0, implicit $mode, implicit $exec74 ; REGALLOC-GFX90A-NEXT: GLOBAL_STORE_DWORDX2 undef %17:vreg_64_align2, [[COPY1]], 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)75 ; REGALLOC-GFX90A-NEXT: GLOBAL_STORE_DWORDX4 undef %19:vreg_64_align2, [[V_MFMA_I32_4X4X4I8_e64_]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)76 ; REGALLOC-GFX90A-NEXT: S_ENDPGM 077 ;78 ; PEI-GFX90A-LABEL: name: partial_copy79 ; PEI-GFX90A: bb.0 (%ir-block.0):80 ; PEI-GFX90A-NEXT: liveins: $sgpr4_sgpr581 ; PEI-GFX90A-NEXT: {{ $}}82 ; PEI-GFX90A-NEXT: INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 1638409 /* reguse:AGPR_32 */, undef renamable $agpr083 ; PEI-GFX90A-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 7340042 /* regdef:VReg_128_Align2 */, def renamable $vgpr0_vgpr1_vgpr2_vgpr384 ; PEI-GFX90A-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = COPY killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, implicit $exec85 ; PEI-GFX90A-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3080202 /* regdef:VReg_64_Align2 */, def renamable $vgpr2_vgpr386 ; PEI-GFX90A-NEXT: GLOBAL_STORE_DWORDX4 undef renamable $vgpr0_vgpr1, killed renamable $agpr0_agpr1_agpr2_agpr3, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)87 ; PEI-GFX90A-NEXT: renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (s128) from %ir.arg.kernarg.offset1, addrspace 4)88 ; PEI-GFX90A-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, implicit $exec89 ; PEI-GFX90A-NEXT: renamable $vgpr0 = AV_MOV_B32_IMM_PSEUDO 1, implicit $exec90 ; PEI-GFX90A-NEXT: renamable $vgpr1 = AV_MOV_B32_IMM_PSEUDO 2, implicit $exec91 ; PEI-GFX90A-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_I32_4X4X4I8_e64 killed $vgpr0, killed $vgpr1, killed $agpr0_agpr1_agpr2_agpr3, 0, 0, 0, implicit $mode, implicit $exec92 ; PEI-GFX90A-NEXT: GLOBAL_STORE_DWORDX2 undef renamable $vgpr0_vgpr1, killed renamable $vgpr2_vgpr3, 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)93 ; PEI-GFX90A-NEXT: GLOBAL_STORE_DWORDX4 undef renamable $vgpr0_vgpr1, killed renamable $agpr0_agpr1_agpr2_agpr3, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)94 ; PEI-GFX90A-NEXT: S_ENDPGM 095 call void asm sideeffect "; use $0", "a" (i32 poison)96 %v0 = call <4 x i32> asm sideeffect "; def $0", "=v" ()97 %v1 = call <2 x i32> asm sideeffect "; def $0", "=v" ()98 %mai = tail call <4 x i32> @llvm.amdgcn.mfma.i32.4x4x4i8(i32 1, i32 2, <4 x i32> %arg, i32 0, i32 0, i32 0)99 store volatile <4 x i32> %v0, ptr addrspace(1) poison100 store volatile <2 x i32> %v1, ptr addrspace(1) poison101 store volatile <4 x i32> %mai, ptr addrspace(1) poison102 ret void103}104 105declare <4 x i32> @llvm.amdgcn.mfma.i32.4x4x4i8(i32, i32, <4 x i32>, i32, i32, i32)106 107attributes #0 = { nounwind "amdgpu-num-vgpr"="5" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-cluster-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-cluster-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-cluster-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" }108