[llvm] 862f504 - [AMDGPU] Enable AMDGPUAttributorPass in full LTO (#102673)

[via llvm-commits]

Author: Shilei Tian
Date: 2024-08-12T13:39:23-04:00
New Revision: 862f5040fb854b5ca0aba2d3088396201bf7db9a

URL: https://github.com/llvm/llvm-project/commit/862f5040fb854b5ca0aba2d3088396201bf7db9a
DIFF: https://github.com/llvm/llvm-project/commit/862f5040fb854b5ca0aba2d3088396201bf7db9a.diff

LOG: [AMDGPU] Enable AMDGPUAttributorPass in full LTO (#102673)

This is basically same as
https://github.com/llvm/llvm-project/pull/102086 but reverts some test
case changes that are no longer needed.

Added: 
    llvm/test/CodeGen/AMDGPU/print-pipeline-passes.ll

Modified: 
    llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
index 0523fee5bcf9f..d82be9a7e9041 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -383,6 +383,11 @@ static cl::opt<bool> EnableHipStdPar(
   cl::desc("Enable HIP Standard Parallelism Offload support"), cl::init(false),
   cl::Hidden);
 
+static cl::opt<bool>
+    EnableAMDGPUAttributor("amdgpu-attributor-enable",
+                           cl::desc("Enable AMDGPUAttributorPass"),
+                           cl::init(true), cl::Hidden);
+
 extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAMDGPUTarget() {
   // Register the target
   RegisterTargetMachine<R600TargetMachine> X(getTheR600Target());
@@ -762,6 +767,8 @@ void AMDGPUTargetMachine::registerPassBuilderCallbacks(PassBuilder &PB) {
         // module is partitioned for codegen.
         if (EnableLowerModuleLDS)
           PM.addPass(AMDGPULowerModuleLDSPass(*this));
+        if (EnableAMDGPUAttributor && Level != OptimizationLevel::O0)
+          PM.addPass(AMDGPUAttributorPass(*this));
       });
 
   PB.registerRegClassFilterParsingCallback(

diff  --git a/llvm/test/CodeGen/AMDGPU/print-pipeline-passes.ll b/llvm/test/CodeGen/AMDGPU/print-pipeline-passes.ll
new file mode 100644
index 0000000000000..b9eda0c1cd3bb
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/print-pipeline-passes.ll
@@ -0,0 +1,12 @@
+; RUN: opt -mtriple=amdgcn--amdhsa -S -passes="lto<O0>" -print-pipeline-passes %s -o - | FileCheck --check-prefix=O0 %s
+; RUN: opt -mtriple=amdgcn--amdhsa -S -passes="lto<O1>" -print-pipeline-passes %s -o - | FileCheck %s
+; RUN: opt -mtriple=amdgcn--amdhsa -S -passes="lto<O2>" -print-pipeline-passes %s -o - | FileCheck %s
+; RUN: opt -mtriple=amdgcn--amdhsa -S -passes="lto<O3>" -print-pipeline-passes %s -o - | FileCheck %s
+
+; CHECK: amdgpu-attributor
+; O0-NOT: amdgpu-attributor
+
+define amdgpu_kernel void @kernel() {
+entry:
+  ret void
+}