[PATCH] D156682: AMDGPU: Partially respect nobuiltin in libcall simplifier

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
arsenm added reviewers: AMDGPU, jhuber6, yaxunl, jmmartinez, rampitec.
Herald added subscribers: foad, kerbowa, hiraditya, tpr, dstuttard, jvesely, kzhuravl.
Herald added a project: All.
arsenm requested review of this revision.
Herald added a subscriber: wdng.
Herald added a project: LLVM.

There are more contexts where it's not handled correctly but this is
the simplest one.


https://reviews.llvm.org/D156682

Files:
  llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
  llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll


Index: llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll
===================================================================
--- llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll
+++ llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll
@@ -634,6 +634,18 @@
   ret void
 }
 
+; GCN-LABEL: {{^}}define amdgpu_kernel void @test_use_native_powr_nobuiltin
+; GCN: %call = tail call fast float @_Z4powrff(float %tmp, float %tmp1)
+define amdgpu_kernel void @test_use_native_powr_nobuiltin(ptr addrspace(1) nocapture %a) {
+entry:
+  %tmp = load float, ptr addrspace(1) %a, align 4
+  %arrayidx1 = getelementptr inbounds float, ptr addrspace(1) %a, i64 1
+  %tmp1 = load float, ptr addrspace(1) %arrayidx1, align 4
+  %call = call fast float @_Z4powrff(float %tmp, float %tmp1) nobuiltin
+  store float %call, ptr addrspace(1) %a, align 4
+  ret void
+}
+
 ; GCN-LABEL: {{^}}define amdgpu_kernel void @test_use_native_sqrt
 ; GCN-NATIVE: call fast float @_Z11native_sqrtf(float %tmp)
 define amdgpu_kernel void @test_use_native_sqrt(ptr addrspace(1) nocapture %a) {
Index: llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
@@ -492,7 +492,7 @@
 bool AMDGPULibCalls::useNative(CallInst *aCI) {
   CI = aCI;
   Function *Callee = aCI->getCalledFunction();
-  if (!Callee)
+  if (!Callee || aCI->isNoBuiltin())
     return false;
 
   FuncInfo FInfo;
@@ -583,7 +583,7 @@
   this->CI = CI;
   Function *Callee = CI->getCalledFunction();
   // Ignore indirect calls.
-  if (!Callee)
+  if (!Callee || CI->isNoBuiltin())
     return false;
 
   IRBuilder<> B(CI);


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D156682.545616.patch
Type: text/x-patch
Size: 1687 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230731/4e8b03a3/attachment.bin>