derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
278 lines
16 KiB
YAML
278 lines
16 KiB
YAML
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 2
|
|
# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs -run-pass=prologepilog -o - %s | FileCheck -check-prefix=GCN %s
|
|
|
|
--- |
|
|
declare amdgpu_cs_chain void @callee()
|
|
declare amdgpu_gfx void @gfx_callee()
|
|
|
|
define amdgpu_cs_chain_preserve void @preserve_active_lanes_above_args() #0 {ret void}
|
|
define amdgpu_cs_chain_preserve void @preserve_all_lanes_wwm_above_args() #0 {ret void}
|
|
define amdgpu_cs_chain_preserve void @dont_preserve_args() #0 {ret void}
|
|
define amdgpu_cs_chain_preserve void @preserve_inactive_lanes_wwm_args() #0 {ret void}
|
|
define amdgpu_cs_chain_preserve void @dont_preserve_if_no_chain_calls() #0 {ret void}
|
|
define amdgpu_cs_chain_preserve void @dont_preserve_v0_v7() #0 {ret void}
|
|
define amdgpu_cs_chain_preserve void @dont_preserve_sgpr() #0 {ret void}
|
|
|
|
attributes #0 = { nounwind }
|
|
...
|
|
---
|
|
|
|
# NOTE: Since we don't know what the args are, we rely on the fact that we can't
|
|
# call llvm.amdgcn.cs.chain with more parameters than we received - so anything
|
|
# that is used by the SI_CS_CHAIN_TC_W32 is assumed to have been an arg and therefore
|
|
# not preserved.
|
|
|
|
---
|
|
name: preserve_active_lanes_above_args
|
|
tracksRegLiveness: true
|
|
frameInfo:
|
|
hasTailCall: true
|
|
machineFunctionInfo:
|
|
stackPtrOffsetReg: '$sgpr32'
|
|
isChainFunction: true
|
|
returnsVoid: true
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0, $vgpr8, $vgpr9
|
|
|
|
; GCN-LABEL: name: preserve_active_lanes_above_args
|
|
; GCN: liveins: $sgpr0, $vgpr8, $vgpr9, $vgpr10
|
|
; GCN-NEXT: {{ $}}
|
|
; GCN-NEXT: SCRATCH_STORE_DWORD_SADDR killed $vgpr10, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: ("amdgpu-thread-private" store (s32) into %stack.0, addrspace 5)
|
|
; GCN-NEXT: frame-setup CFI_INSTRUCTION llvm_vector_offset $vgpr10, 32, $exec_lo, 32, 0
|
|
; GCN-NEXT: renamable $vgpr10 = V_MOV_B32_e32 10, implicit $exec
|
|
; GCN-NEXT: $vgpr8 = COPY killed renamable $vgpr10
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
; GCN-NEXT: $vgpr10 = SCRATCH_LOAD_DWORD_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: ("amdgpu-thread-private" load (s32) from %stack.0, addrspace 5)
|
|
; GCN-NEXT: SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
renamable $vgpr10 = V_MOV_B32_e32 10, implicit $exec
|
|
$vgpr8 = COPY renamable killed $vgpr10
|
|
renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
|
|
...
|
|
|
|
---
|
|
name: preserve_all_lanes_wwm_above_args
|
|
tracksRegLiveness: true
|
|
frameInfo:
|
|
hasTailCall: true
|
|
machineFunctionInfo:
|
|
stackPtrOffsetReg: '$sgpr32'
|
|
isChainFunction: true
|
|
returnsVoid: true
|
|
wwmReservedRegs:
|
|
- '$vgpr10'
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0, $sgpr35, $vgpr8, $vgpr9, $vgpr10
|
|
|
|
; GCN-LABEL: name: preserve_all_lanes_wwm_above_args
|
|
; GCN: liveins: $sgpr0, $sgpr35, $vgpr8, $vgpr9, $vgpr10
|
|
; GCN-NEXT: {{ $}}
|
|
; GCN-NEXT: $sgpr1 = S_OR_SAVEEXEC_B32 -1, implicit-def $exec, implicit-def dead $scc, implicit $exec
|
|
; GCN-NEXT: SCRATCH_STORE_DWORD_SADDR $vgpr10, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: ("amdgpu-thread-private" store (s32) into %stack.0, addrspace 5)
|
|
; GCN-NEXT: $exec_lo = S_MOV_B32 killed $sgpr1
|
|
; GCN-NEXT: S_NOP 0, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
|
|
; GCN-NEXT: $vgpr10 = SI_SPILL_S32_TO_VGPR $sgpr35, 0, killed $vgpr10
|
|
; GCN-NEXT: $sgpr35 = S_MOV_B32 5
|
|
; GCN-NEXT: $sgpr35 = SI_RESTORE_S32_FROM_VGPR $vgpr10, 0
|
|
; GCN-NEXT: $vgpr10 = V_MOV_B32_e32 10, implicit $exec
|
|
; GCN-NEXT: $vgpr8 = COPY killed $vgpr10
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
; GCN-NEXT: $sgpr1 = S_OR_SAVEEXEC_B32 -1, implicit-def $exec, implicit-def dead $scc, implicit $exec
|
|
; GCN-NEXT: $vgpr10 = SCRATCH_LOAD_DWORD_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: ("amdgpu-thread-private" load (s32) from %stack.0, addrspace 5)
|
|
; GCN-NEXT: $exec_lo = S_MOV_B32 killed $sgpr1
|
|
; GCN-NEXT: SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
S_NOP 0, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
|
|
$vgpr10 = SI_SPILL_S32_TO_VGPR $sgpr35, 0, killed $vgpr10
|
|
$sgpr35 = S_MOV_B32 5
|
|
$sgpr35 = SI_RESTORE_S32_FROM_VGPR $vgpr10, 0
|
|
$vgpr10 = V_MOV_B32_e32 10, implicit $exec
|
|
$vgpr8 = COPY killed $vgpr10
|
|
renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
|
|
...
|
|
|
|
---
|
|
name: dont_preserve_args
|
|
tracksRegLiveness: true
|
|
frameInfo:
|
|
hasTailCall: true
|
|
machineFunctionInfo:
|
|
stackPtrOffsetReg: '$sgpr32'
|
|
isChainFunction: true
|
|
returnsVoid: true
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0, $vgpr8, $vgpr9
|
|
|
|
; GCN-LABEL: name: dont_preserve_args
|
|
; GCN: liveins: $sgpr0, $vgpr8, $vgpr9
|
|
; GCN-NEXT: {{ $}}
|
|
; GCN-NEXT: S_NOP 0, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
|
|
; GCN-NEXT: renamable $vgpr8 = V_MOV_B32_e32 10, implicit $exec
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
; GCN-NEXT: SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
S_NOP 0, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
|
|
renamable $vgpr8 = V_MOV_B32_e32 10, implicit $exec
|
|
renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
|
|
...
|
|
|
|
---
|
|
name: preserve_inactive_lanes_wwm_args
|
|
tracksRegLiveness: true
|
|
frameInfo:
|
|
hasTailCall: true
|
|
machineFunctionInfo:
|
|
stackPtrOffsetReg: '$sgpr32'
|
|
isChainFunction: true
|
|
returnsVoid: true
|
|
wwmReservedRegs:
|
|
- '$vgpr9'
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0, $sgpr35, $vgpr8, $vgpr9, $vgpr10
|
|
|
|
; GCN-LABEL: name: preserve_inactive_lanes_wwm_args
|
|
; GCN: liveins: $sgpr0, $sgpr35, $vgpr8, $vgpr9, $vgpr10
|
|
; GCN-NEXT: {{ $}}
|
|
; GCN-NEXT: $sgpr1 = S_OR_SAVEEXEC_B32 -1, implicit-def $exec, implicit-def dead $scc, implicit $exec
|
|
; GCN-NEXT: SCRATCH_STORE_DWORD_SADDR $vgpr9, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: ("amdgpu-thread-private" store (s32) into %stack.0, addrspace 5)
|
|
; GCN-NEXT: $exec_lo = S_MOV_B32 killed $sgpr1
|
|
; GCN-NEXT: S_NOP 0, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
|
|
; GCN-NEXT: $vgpr8 = SI_SPILL_S32_TO_VGPR $sgpr35, 0, killed $vgpr8
|
|
; GCN-NEXT: $sgpr35 = S_MOV_B32 5
|
|
; GCN-NEXT: $sgpr35 = SI_RESTORE_S32_FROM_VGPR $vgpr8, 0
|
|
; GCN-NEXT: renamable $vgpr8 = V_MOV_B32_e32 10, implicit $exec
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
; GCN-NEXT: $sgpr1 = S_OR_SAVEEXEC_B32 -1, implicit-def $exec, implicit-def dead $scc, implicit $exec
|
|
; GCN-NEXT: $vgpr9 = SCRATCH_LOAD_DWORD_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr, implicit $vgpr9(tied-def 0) :: ("amdgpu-thread-private" load (s32) from %stack.0, addrspace 5)
|
|
; GCN-NEXT: $exec_lo = S_MOV_B32 killed $sgpr1
|
|
; GCN-NEXT: SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
S_NOP 0, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
|
|
renamable $vgpr8 = SI_SPILL_S32_TO_VGPR $sgpr35, 0, killed $vgpr8
|
|
$sgpr35 = S_MOV_B32 5
|
|
$sgpr35 = SI_RESTORE_S32_FROM_VGPR $vgpr8, 0
|
|
renamable $vgpr8 = V_MOV_B32_e32 10, implicit $exec
|
|
renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
|
|
...
|
|
|
|
---
|
|
name: dont_preserve_if_no_chain_calls
|
|
tracksRegLiveness: true
|
|
frameInfo:
|
|
hasTailCall: false
|
|
machineFunctionInfo:
|
|
stackPtrOffsetReg: '$sgpr32'
|
|
isChainFunction: true
|
|
returnsVoid: true
|
|
wwmReservedRegs:
|
|
- '$vgpr9'
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0, $sgpr35, $vgpr0, $vgpr8, $vgpr9
|
|
|
|
; GCN-LABEL: name: dont_preserve_if_no_chain_calls
|
|
; GCN: liveins: $sgpr0, $sgpr35, $vgpr0, $vgpr8, $vgpr9
|
|
; GCN-NEXT: {{ $}}
|
|
; GCN-NEXT: S_NOP 0, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
|
|
; GCN-NEXT: $vgpr8 = SI_SPILL_S32_TO_VGPR $sgpr35, 0, killed $vgpr8
|
|
; GCN-NEXT: $sgpr35 = S_MOV_B32 5
|
|
; GCN-NEXT: $sgpr35 = SI_RESTORE_S32_FROM_VGPR $vgpr8, 0
|
|
; GCN-NEXT: renamable $vgpr8 = V_MOV_B32_e32 10, implicit $exec
|
|
; GCN-NEXT: $vgpr9 = V_MOV_B32_e32 20, implicit $exec
|
|
; GCN-NEXT: $vgpr10 = V_MOV_B32_e32 30, implicit $exec
|
|
; GCN-NEXT: S_ENDPGM 0
|
|
S_NOP 0, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
|
|
renamable $vgpr8 = SI_SPILL_S32_TO_VGPR $sgpr35, 0, killed $vgpr8
|
|
$sgpr35 = S_MOV_B32 5
|
|
$sgpr35 = SI_RESTORE_S32_FROM_VGPR $vgpr8, 0
|
|
renamable $vgpr8 = V_MOV_B32_e32 10, implicit $exec
|
|
$vgpr9 = V_MOV_B32_e32 20, implicit $exec
|
|
$vgpr10 = V_MOV_B32_e32 30, implicit $exec
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
---
|
|
name: dont_preserve_v0_v7
|
|
tracksRegLiveness: true
|
|
frameInfo:
|
|
hasTailCall: true
|
|
machineFunctionInfo:
|
|
stackPtrOffsetReg: '$sgpr32'
|
|
isChainFunction: true
|
|
returnsVoid: true
|
|
wwmReservedRegs:
|
|
- '$vgpr1'
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0, $sgpr35, $vgpr0, $vgpr8
|
|
|
|
; GCN-LABEL: name: dont_preserve_v0_v7
|
|
; GCN: liveins: $sgpr0, $sgpr35, $vgpr0, $vgpr8
|
|
; GCN-NEXT: {{ $}}
|
|
; GCN-NEXT: $vgpr0 = SI_SPILL_S32_TO_VGPR $sgpr35, 0, killed $vgpr0
|
|
; GCN-NEXT: $sgpr35 = S_MOV_B32 5
|
|
; GCN-NEXT: $sgpr35 = SI_RESTORE_S32_FROM_VGPR $vgpr0, 0
|
|
; GCN-NEXT: renamable $vgpr0 = V_MOV_B32_e32 10, implicit $exec
|
|
; GCN-NEXT: renamable $vgpr7 = V_MOV_B32_e32 16, implicit $exec
|
|
; GCN-NEXT: renamable $vgpr8 = COPY killed renamable $vgpr0
|
|
; GCN-NEXT: renamable $vgpr9 = COPY killed renamable $vgpr7
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
; GCN-NEXT: SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
renamable $vgpr0 = SI_SPILL_S32_TO_VGPR $sgpr35, 0, killed $vgpr0
|
|
$sgpr35 = S_MOV_B32 5
|
|
$sgpr35 = SI_RESTORE_S32_FROM_VGPR $vgpr0, 0
|
|
renamable $vgpr0 = V_MOV_B32_e32 10, implicit $exec
|
|
renamable $vgpr7 = V_MOV_B32_e32 16, implicit $exec
|
|
renamable $vgpr8 = COPY killed renamable $vgpr0
|
|
renamable $vgpr9 = COPY killed renamable $vgpr7
|
|
renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0, implicit $vgpr8, implicit $vgpr9
|
|
...
|
|
|
|
---
|
|
name: dont_preserve_sgpr
|
|
tracksRegLiveness: true
|
|
frameInfo:
|
|
hasTailCall: true
|
|
machineFunctionInfo:
|
|
stackPtrOffsetReg: '$sgpr32'
|
|
isChainFunction: true
|
|
returnsVoid: true
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0
|
|
|
|
; GCN-LABEL: name: dont_preserve_sgpr
|
|
; GCN: liveins: $sgpr0
|
|
; GCN-NEXT: {{ $}}
|
|
; GCN-NEXT: renamable $sgpr1 = S_ADD_I32 killed renamable $sgpr0, renamable $sgpr0, implicit-def dead $scc
|
|
; GCN-NEXT: $sgpr0 = COPY killed renamable $sgpr1
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
; GCN-NEXT: renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
; GCN-NEXT: SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0
|
|
renamable $sgpr1 = S_ADD_I32 killed renamable $sgpr0, renamable $sgpr0, implicit-def dead $scc
|
|
$sgpr0 = COPY killed renamable $sgpr1
|
|
renamable $sgpr4_sgpr5 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-gotprel32-lo) @callee + 4, target-flags(amdgpu-gotprel32-hi) @callee + 12, implicit-def dead $scc
|
|
renamable $sgpr4_sgpr5 = S_LOAD_DWORDX2_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (p0) from got, addrspace 4)
|
|
SI_CS_CHAIN_TC_W32 killed renamable $sgpr4_sgpr5, @callee, 0, -1, amdgpu_allvgprs, implicit $sgpr0
|
|
|
|
...
|