AMDGPU: Account for globals in AMDGPUPromoteAlloca pass

Patch by Bas Nieuwenhuizen

llvm-svn: 267791

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp b/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
index 1b3ce61..b7d30d2 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
@@ -125,14 +125,16 @@ bool AMDGPUPromoteAlloca::runOnFunction(Function &F) {
     if (GV.getType()->getAddressSpace() != AMDGPUAS::LOCAL_ADDRESS)
       continue;
 
-    for (Use &U : GV.uses()) {
+    for (User *U : GV.users()) {
       Instruction *Use = dyn_cast<Instruction>(U);
       if (!Use)
         continue;
 
-      if (Use->getParent()->getParent() == &F)
+      if (Use->getParent()->getParent() == &F) {
         LocalMemAvailable -=
           Mod->getDataLayout().getTypeAllocSize(GV.getValueType());
+        break;
+      }
     }
   }
 

diff --git a/llvm/test/CodeGen/AMDGPU/promote-alloca-globals.ll b/llvm/test/CodeGen/AMDGPU/promote-alloca-globals.ll
new file mode 100644 (file)
index 0000000..8cfe2df
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/promote-alloca-globals.ll
@@ -0,0 +1,32 @@
+; RUN: opt -S -mtriple=amdgcn-unknown-unknown -amdgpu-promote-alloca < %s | FileCheck -check-prefix=IR %s
+; RUN: llc -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=ASM %s
+
+
+@global_array= internal unnamed_addr addrspace(3) global [1500 x [10 x i32]] undef, align 4
+
+; IR-LABEL: define void @promote_alloca_size_256(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) {
+; IR: alloca [10 x i32]
+; ASM-LABEL: {{^}}promote_alloca_size_256:
+; ASM: ; LDSByteSize: 60000 bytes/workgroup (compile time only)
+
+define void @promote_alloca_size_256(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) {
+entry:
+  %stack = alloca [10 x i32], align 4
+  %tmp = load i32, i32 addrspace(1)* %in, align 4
+  %arrayidx1 = getelementptr inbounds [10 x i32], [10 x i32]* %stack, i32 0, i32 %tmp
+  store i32 4, i32* %arrayidx1, align 4
+  %arrayidx2 = getelementptr inbounds i32, i32 addrspace(1)* %in, i32 1
+  %tmp1 = load i32, i32 addrspace(1)* %arrayidx2, align 4
+  %arrayidx3 = getelementptr inbounds [10 x i32], [10 x i32]* %stack, i32 0, i32 %tmp1
+  store i32 5, i32* %arrayidx3, align 4
+  %arrayidx10 = getelementptr inbounds [10 x i32], [10 x i32]* %stack, i32 0, i32 0
+  %tmp2 = load i32, i32* %arrayidx10, align 4
+  store i32 %tmp2, i32 addrspace(1)* %out, align 4
+  %arrayidx12 = getelementptr inbounds [10 x i32], [10 x i32]* %stack, i32 0, i32 1
+  %tmp3 = load i32, i32* %arrayidx12
+  %arrayidx13 = getelementptr inbounds i32, i32 addrspace(1)* %out, i32 1
+  store i32 %tmp3, i32 addrspace(1)* %arrayidx13
+  %v = getelementptr inbounds [1500 x [10 x i32]], [1500 x [10 x i32]] addrspace(3)* @global_array, i32 0, i32 0, i32 0
+  store i32 %tmp3, i32 addrspace(3)* %v
+  ret void
+}