shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity

AMDGPU/GlobalISel: RegBankLegalize rules for ds_ordered_add/swap (#189137)

Browse Source
This commit is contained in:
vangthao95 2026-03-30 09:57:04 -07:00 committed by GitHub
parent 27e3c43d74
commit c32d670757
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
8 changed files with 27 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
View File

@ -1712,6 +1712,9 @@ RegBankLegalizeRules::RegBankLegalizeRules(const GCNSubtarget &_ST,
addRulesForIOpcs({amdgcn_ds_bvh_stack_push8_pop2_rtn}, Standard)
.Div(S64, {{Vgpr64, Vgpr32}, {IntrId, Vgpr32, Vgpr32, VgprV8S32}});
addRulesForIOpcs({amdgcn_ds_ordered_add, amdgcn_ds_ordered_swap}, Standard)
.Div(S32, {{Vgpr32}, {IntrId, SgprB32_M0, Vgpr32}});
addRulesForIOpcs({amdgcn_ds_swizzle}, Standard)
.Uni(S32, {{UniInVgprS32}, {IntrId, Vgpr32}})
.Div(S32, {{Vgpr32}, {IntrId, Vgpr32}});

15
llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.ds.ordered.add.mir
View File

@ -1,6 +1,5 @@
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=regbankselect -regbankselect-fast -verify-machineinstrs -o - %s | FileCheck %s
# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=regbankselect -regbankselect-greedy -verify-machineinstrs -o - %s | FileCheck %s
# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass='amdgpu-regbankselect,amdgpu-regbanklegalize' -o - %s | FileCheck %s
---
name: ds_ordered_add_ss
@ -31,11 +30,11 @@ body: |
; CHECK-LABEL: name: ds_ordered_add_vs
; CHECK: liveins: $vgpr0, $sgpr0
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0
; CHECK-NEXT: [[INTRINSIC_CONVERGENT:%[0-9]+]]:sgpr(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.readfirstlane), [[COPY]](s32)
; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
; CHECK-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY]](s32), implicit $exec
; CHECK-NEXT: [[INT:%[0-9]+]]:vgpr(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.add), [[V_READFIRSTLANE_B32_]](s32), [[COPY2]](s32), 0, 0, 0, 0, 0, 0
; CHECK-NEXT: [[INT:%[0-9]+]]:vgpr(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.add), [[INTRINSIC_CONVERGENT]](s32), [[COPY2]](s32), 0, 0, 0, 0, 0, 0
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $sgpr0
%2:_(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.add), %0, %1, 0, 0, 0, 0, 0, 0
@ -51,10 +50,10 @@ body: |
; CHECK-LABEL: name: ds_ordered_add_vv
; CHECK: liveins: $vgpr0, $vgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
; CHECK-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY]](s32), implicit $exec
; CHECK-NEXT: [[INT:%[0-9]+]]:vgpr(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.add), [[V_READFIRSTLANE_B32_]](s32), [[COPY1]](s32), 0, 0, 0, 0, 0, 0
; CHECK-NEXT: [[INTRINSIC_CONVERGENT:%[0-9]+]]:sgpr(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.readfirstlane), [[COPY]](s32)
; CHECK-NEXT: [[INT:%[0-9]+]]:vgpr(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.add), [[INTRINSIC_CONVERGENT]](s32), [[COPY1]](s32), 0, 0, 0, 0, 0, 0
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.add), %0, %1, 0, 0, 0, 0, 0, 0

15
llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.ds.ordered.swap.mir
View File

@ -1,6 +1,5 @@
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=regbankselect -regbankselect-fast -verify-machineinstrs -o - %s | FileCheck %s
# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=regbankselect -regbankselect-greedy -verify-machineinstrs -o - %s | FileCheck %s
# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass='amdgpu-regbankselect,amdgpu-regbanklegalize' -o - %s | FileCheck %s
---
name: ds_ordered_swap_ss
@ -31,11 +30,11 @@ body: |
; CHECK-LABEL: name: ds_ordered_swap_vs
; CHECK: liveins: $vgpr0, $sgpr0
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0
; CHECK-NEXT: [[INTRINSIC_CONVERGENT:%[0-9]+]]:sgpr(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.readfirstlane), [[COPY]](s32)
; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
; CHECK-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY]](s32), implicit $exec
; CHECK-NEXT: [[INT:%[0-9]+]]:vgpr(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.swap), [[V_READFIRSTLANE_B32_]](s32), [[COPY2]](s32), 0, 0, 0, 0, 0, 0
; CHECK-NEXT: [[INT:%[0-9]+]]:vgpr(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.swap), [[INTRINSIC_CONVERGENT]](s32), [[COPY2]](s32), 0, 0, 0, 0, 0, 0
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $sgpr0
%2:_(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.swap), %0, %1, 0, 0, 0, 0, 0, 0
@ -51,10 +50,10 @@ body: |
; CHECK-LABEL: name: ds_ordered_swap_vv
; CHECK: liveins: $vgpr0, $vgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
; CHECK-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY]](s32), implicit $exec
; CHECK-NEXT: [[INT:%[0-9]+]]:vgpr(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.swap), [[V_READFIRSTLANE_B32_]](s32), [[COPY1]](s32), 0, 0, 0, 0, 0, 0
; CHECK-NEXT: [[INTRINSIC_CONVERGENT:%[0-9]+]]:sgpr(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.readfirstlane), [[COPY]](s32)
; CHECK-NEXT: [[INT:%[0-9]+]]:vgpr(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.swap), [[INTRINSIC_CONVERGENT]](s32), [[COPY1]](s32), 0, 0, 0, 0, 0, 0
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(s32) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.ds.ordered.swap), %0, %1, 0, 0, 0, 0, 0, 0

2
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.ordered.add-errors.ll
View File

@ -1,5 +1,5 @@
; RUN: not llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1010 -filetype=null %s 2>&1 | FileCheck %s
; RUN: not llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1010 -filetype=null %s 2>&1 | FileCheck %s
; RUN: not llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1010 -filetype=null %s 2>&1 | FileCheck %s
; CHECK: error: <unknown>:0:0: in function ds_ordered_add_dword_count_too_low void (ptr addrspace(2), ptr addrspace(1)): ds_ordered_count: dword count must be between 1 and 4
define amdgpu_kernel void @ds_ordered_add_dword_count_too_low(ptr addrspace(2) inreg %gds, ptr addrspace(1) %out) {

4
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.ordered.add.gfx10.ll
View File

@ -1,7 +1,7 @@
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GCN %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GCN %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GCN %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1100 < %s | FileCheck -check-prefixes=GCN %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1100 < %s | FileCheck -check-prefixes=GCN %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1100 < %s | FileCheck -check-prefixes=GCN %s
; GCN-LABEL: {{^}}ds_ordered_add:
; GCN-DAG: v_{{(dual_)?}}mov_b32{{(_e32)?}} v[[INCR:[0-9]+]], 31

2
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.ordered.add.gfx11.ll
View File

@ -1,5 +1,5 @@
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GCN,FUNC %s
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GCN,FUNC %s
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GCN,FUNC %s
; RUN: not --crash llc -mtriple=amdgcn -mcpu=gfx1200 -amdgpu-enable-vopd=0 < %s 2>&1 | FileCheck -check-prefix=GFX12-ERR %s
; GFX12-ERR: LLVM ERROR: Cannot select: {{.*}} = AMDGPUISD::DS_ORDERED_COUNT

6
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.ordered.add.ll
View File

@ -1,9 +1,9 @@
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefixes=GCN,FUNC %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefixes=GCN,FUNC %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefixes=GCN,FUNC %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; FUNC-LABEL: {{^}}ds_ordered_add:
; GCN-DAG: v_mov_b32_e32 v[[INCR:[0-9]+]], 31

6
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.ordered.swap.ll
View File

@ -1,9 +1,9 @@
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefixes=GCN,FUNC %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefixes=GCN,FUNC %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefixes=GCN,FUNC %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,VIGFX9,FUNC %s
; FUNC-LABEL: {{^}}ds_ordered_swap:
; GCN: s_mov_b32 m0, s0