[llvm] AMDGPU/GlobalISel: RegBankLegalize rules for USUB_COND/USUB_SAT (PR #197800)
via llvm-commits
llvm-commits at lists.llvm.org
Thu May 14 14:02:53 PDT 2026
llvmorg-github-actions[bot] wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-llvm-globalisel
Author: vangthao95
<details>
<summary>Changes</summary>
---
Full diff: https://github.com/llvm/llvm-project/pull/197800.diff
3 Files Affected:
- (modified) llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp (+5)
- (modified) llvm/test/CodeGen/AMDGPU/atomicrmw_usub_cond.ll (+8-4)
- (modified) llvm/test/CodeGen/AMDGPU/atomicrmw_usub_sat.ll (+4-4)
``````````diff
diff --git a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
index 2ce2545f57893..d924a363665f7 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
@@ -897,6 +897,11 @@ RegBankLegalizeRules::RegBankLegalizeRules(const GCNSubtarget &_ST,
.Any({{DivS32, P3, S32}, {{Vgpr32}, {VgprP3, Vgpr32}}})
.Any({{DivS64, P3, S64}, {{Vgpr64}, {VgprP3, Vgpr64}}});
+ addRulesForGOpcs({G_ATOMICRMW_USUB_SAT, G_ATOMICRMW_USUB_COND})
+ .Any({{DivS32, P0}, {{Vgpr32}, {VgprP0, Vgpr32}}})
+ .Any({{DivS32, P1}, {{Vgpr32}, {VgprP1, Vgpr32}}})
+ .Any({{DivS32, P3}, {{Vgpr32}, {VgprP3, Vgpr32}}});
+
bool HasAtomicFlatPkAdd16Insts = ST->hasAtomicFlatPkAdd16Insts();
bool HasAtomicBufferGlobalPkAddF16Insts =
ST->hasAtomicBufferGlobalPkAddF16NoRtnInsts() ||
diff --git a/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_cond.ll b/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_cond.ll
index e58ccfd331845..0df8b3f8bdde3 100644
--- a/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_cond.ll
+++ b/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_cond.ll
@@ -1,8 +1,8 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9-SDAG %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GFX12-SDAG %s
-; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9-GISEL %s
-; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GFX12-GISEL %s
+; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9-GISEL %s
+; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GFX12-GISEL %s
define amdgpu_kernel void @flat_atomic_usub_cond_no_rtn_u32(ptr %addr, i32 %in) {
; GFX9-SDAG-LABEL: flat_atomic_usub_cond_no_rtn_u32:
@@ -447,9 +447,11 @@ define amdgpu_kernel void @ds_usub_cond_no_rtn_u32(ptr addrspace(3) %addr, i32 %
; GFX9-GISEL-NEXT: s_add_u32 s0, s0, -16
; GFX9-GISEL-NEXT: v_mov_b32_e32 v0, s0
; GFX9-GISEL-NEXT: ds_read_b32 v1, v0
+; GFX9-GISEL-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-GISEL-NEXT: v_readfirstlane_b32 s0, v1
+; GFX9-GISEL-NEXT: v_mov_b32_e32 v1, s0
; GFX9-GISEL-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX9-GISEL-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-GISEL-NEXT: v_subrev_u32_e32 v2, s1, v1
; GFX9-GISEL-NEXT: v_cmp_le_u32_e32 vcc, s1, v1
; GFX9-GISEL-NEXT: v_cndmask_b32_e32 v2, v1, v2, vcc
@@ -527,9 +529,11 @@ define amdgpu_kernel void @ds_usub_cond_rtn_u32(ptr addrspace(3) %addr, i32 %in,
; GFX9-GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-GISEL-NEXT: v_mov_b32_e32 v0, s0
; GFX9-GISEL-NEXT: ds_read_b32 v1, v0 offset:16
+; GFX9-GISEL-NEXT: s_waitcnt lgkmcnt(0)
+; GFX9-GISEL-NEXT: v_readfirstlane_b32 s0, v1
+; GFX9-GISEL-NEXT: v_mov_b32_e32 v1, s0
; GFX9-GISEL-NEXT: .LBB5_1: ; %atomicrmw.start
; GFX9-GISEL-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-GISEL-NEXT: v_mov_b32_e32 v2, v1
; GFX9-GISEL-NEXT: v_subrev_u32_e32 v1, s1, v2
; GFX9-GISEL-NEXT: v_cmp_le_u32_e32 vcc, s1, v2
diff --git a/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_sat.ll b/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_sat.ll
index b0fae21a88484..a49537c51adae 100644
--- a/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_sat.ll
+++ b/llvm/test/CodeGen/AMDGPU/atomicrmw_usub_sat.ll
@@ -1,8 +1,8 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck %s -check-prefix=GFX9-GISEL
-; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 < %s | FileCheck %s -check-prefix=GFX10-GISEL
-; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 < %s | FileCheck %s -check-prefix=GFX11-GISEL
-; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 < %s | FileCheck %s -check-prefix=GFX12-GISEL
+; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck %s -check-prefix=GFX9-GISEL
+; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 < %s | FileCheck %s -check-prefix=GFX10-GISEL
+; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 < %s | FileCheck %s -check-prefix=GFX11-GISEL
+; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 < %s | FileCheck %s -check-prefix=GFX12-GISEL
; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck %s -check-prefix=GFX9-SDAG
; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 < %s | FileCheck %s -check-prefix=GFX10-SDAG
; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 < %s | FileCheck %s -check-prefix=GFX11-SDAG
``````````
</details>
https://github.com/llvm/llvm-project/pull/197800
More information about the llvm-commits
mailing list