You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
108 lines
7.3 KiB
108 lines
7.3 KiB
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
|
|
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 -verify-machineinstrs -run-pass=regbankselect -o - %s | FileCheck %s
|
|
|
|
# Make sure that an arbitrary AGPR is treated as a divergent value
|
|
# that needs to be copied to VGPR, and then waterfalled
|
|
|
|
# 32-bit case
|
|
---
|
|
name: raw_buffer_store__sgpr_rsrc__vgpr_val__vgpr_voffset__agpr_soffset
|
|
legalized: true
|
|
tracksRegLiveness: true
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $vgpr0, $vgpr1, $agpr0
|
|
|
|
; CHECK-LABEL: name: raw_buffer_store__sgpr_rsrc__vgpr_val__vgpr_voffset__agpr_soffset
|
|
; CHECK: successors: %bb.1(0x80000000)
|
|
; CHECK: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $vgpr0, $vgpr1, $agpr0
|
|
; CHECK: %val:vgpr(s32) = COPY $vgpr0
|
|
; CHECK: %rsrc:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
|
|
; CHECK: %agpr:agpr(s32) = COPY $agpr0
|
|
; CHECK: %voffset:vgpr(s32) = COPY $vgpr1
|
|
; CHECK: %zero:sgpr(s32) = G_CONSTANT i32 0
|
|
; CHECK: [[COPY:%[0-9]+]]:vgpr(s32) = COPY %zero(s32)
|
|
; CHECK: [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
|
|
; CHECK: [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY %agpr(s32)
|
|
; CHECK: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
|
|
; CHECK: .1:
|
|
; CHECK: successors: %bb.2(0x40000000), %bb.1(0x40000000)
|
|
; CHECK: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.0, %9, %bb.1
|
|
; CHECK: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY1]](s32), implicit $exec
|
|
; CHECK: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_]](s32), [[COPY1]](s32), implicit $exec
|
|
; CHECK: G_AMDGPU_BUFFER_STORE %val(s32), %rsrc(<4 x s32>), [[COPY]](s32), %voffset, [[V_READFIRSTLANE_B32_]], 0, 0, 0 :: (dereferenceable store 4, addrspace 4)
|
|
; CHECK: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[V_CMP_EQ_U32_e64_]], 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.1, implicit $exec
|
|
; CHECK: .2:
|
|
; CHECK: successors: %bb.3(0x80000000)
|
|
; CHECK: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
|
|
; CHECK: .3:
|
|
; CHECK: S_ENDPGM 0
|
|
%val:_(s32) = COPY $vgpr0
|
|
%rsrc:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
|
|
%agpr:_(s32) = COPY $agpr0
|
|
%voffset:_(s32) = COPY $vgpr1
|
|
%zero:_(s32) = G_CONSTANT i32 0
|
|
G_AMDGPU_BUFFER_STORE %val, %rsrc, %zero, %voffset, %agpr, 0, 0, 0 :: (dereferenceable store 4, addrspace 4)
|
|
S_ENDPGM 0
|
|
|
|
...
|
|
|
|
# Register tuple case
|
|
---
|
|
name: load_1d_vgpr_vaddr__agpr_srsrc
|
|
legalized: true
|
|
tracksRegLiveness: true
|
|
body: |
|
|
bb.0:
|
|
liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7, $vgpr0
|
|
; CHECK-LABEL: name: load_1d_vgpr_vaddr__agpr_srsrc
|
|
; CHECK: successors: %bb.1(0x80000000)
|
|
; CHECK: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7, $vgpr0
|
|
; CHECK: [[COPY:%[0-9]+]]:agpr(<8 x s32>) = COPY $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7
|
|
; CHECK: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
|
|
; CHECK: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
|
|
; CHECK: [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
|
|
; CHECK: [[COPY2:%[0-9]+]]:vgpr(<8 x s32>) = COPY [[COPY]](<8 x s32>)
|
|
; CHECK: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY2]](<8 x s32>)
|
|
; CHECK: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
|
|
; CHECK: .1:
|
|
; CHECK: successors: %bb.2(0x40000000), %bb.1(0x40000000)
|
|
; CHECK: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.0, %8, %bb.1
|
|
; CHECK: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %2(<4 x s32>), %bb.1
|
|
; CHECK: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
|
|
; CHECK: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
|
|
; CHECK: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
|
|
; CHECK: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
|
|
; CHECK: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
|
|
; CHECK: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
|
|
; CHECK: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
|
|
; CHECK: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), 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: [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub0(s64), implicit $exec
|
|
; CHECK: [[V_READFIRSTLANE_B32_5:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub1(s64), implicit $exec
|
|
; CHECK: [[MV2:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32)
|
|
; CHECK: [[V_CMP_EQ_U64_e64_2:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV2]](s64), [[UV2]](s64), implicit $exec
|
|
; CHECK: [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_2]], [[S_AND_B64_]], implicit-def $scc
|
|
; CHECK: [[V_READFIRSTLANE_B32_6:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub0(s64), implicit $exec
|
|
; CHECK: [[V_READFIRSTLANE_B32_7:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub1(s64), implicit $exec
|
|
; CHECK: [[MV3:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
|
|
; CHECK: [[V_CMP_EQ_U64_e64_3:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV3]](s64), [[UV3]](s64), implicit $exec
|
|
; CHECK: [[S_AND_B64_2:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_3]], [[S_AND_B64_1]], implicit-def $scc
|
|
; CHECK: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32), [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32), [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
|
|
; CHECK: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY1]](s32), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable load 16)
|
|
; CHECK: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_2]], 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.1, implicit $exec
|
|
; CHECK: .2:
|
|
; CHECK: successors: %bb.3(0x80000000)
|
|
; CHECK: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
|
|
; CHECK: .3:
|
|
; CHECK: S_ENDPGM 0, implicit [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
|
|
%0:_(<8 x s32>) = COPY $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7
|
|
%1:_(s32) = COPY $vgpr0
|
|
%2:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, %1(s32), %0(<8 x s32>), 0, 0, 0 :: (dereferenceable load 16)
|
|
S_ENDPGM 0, implicit %2
|
|
...
|