[PATCH] D16416: AMDGPU: Fix crash with invariant markers

[Matt Arsenault via llvm-commits]

arsenm created this revision.
arsenm added a reviewer: tstellarAMD.
arsenm added a subscriber: llvm-commits.
Herald added a subscriber: arsenm.

The promote alloca pass didn't handle these intrinsics and crashed.
These intrinsics should accept any address space, but for now just
erase them to avoid breaking.

http://reviews.llvm.org/D16416

Files:
  lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
  test/CodeGen/AMDGPU/promote-alloca-invariant-markers.ll

Index: test/CodeGen/AMDGPU/promote-alloca-invariant-markers.ll
===================================================================
--- /dev/null
+++ test/CodeGen/AMDGPU/promote-alloca-invariant-markers.ll
@@ -0,0 +1,25 @@
+; RUN: llc -march=amdgcn -mattr=+promote-alloca -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+
+declare {}* @llvm.invariant.start(i64, i8* nocapture) #0
+declare void @llvm.invariant.end({}*, i64, i8* nocapture) #0
+declare i8* @llvm.invariant.group.barrier(i8*) #1
+
+; GCN-LABEL: {{^}}use_invariant_promotable_lds:
+; GCN: buffer_load_dword
+; GCN: ds_write_b32
+define void @use_invariant_promotable_lds(i32 addrspace(1)* %arg) #2 {
+bb:
+  %tmp = alloca i32, align 4
+  %tmp1 = bitcast i32* %tmp to i8*
+  %tmp2 = getelementptr inbounds i32, i32 addrspace(1)* %arg, i64 1
+  %tmp3 = load i32, i32 addrspace(1)* %tmp2
+  store i32 %tmp3, i32* %tmp
+  %tmp4 = call {}* @llvm.invariant.start(i64 4, i8* %tmp1) #0
+  call void @llvm.invariant.end({}* %tmp4, i64 4, i8* %tmp1) #0
+  %tmp5 = call i8* @llvm.invariant.group.barrier(i8* %tmp1) #1
+  ret void
+}
+
+attributes #0 = { argmemonly nounwind }
+attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind }
Index: lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
+++ lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
@@ -412,6 +412,14 @@
       Intr->eraseFromParent();
       continue;
     }
+    case Intrinsic::invariant_start:
+    case Intrinsic::invariant_end:
+    case Intrinsic::invariant_group_barrier:
+      Intr->eraseFromParent();
+      // FIXME: I think the invariant marker should still theoretically apply,
+      // but the intrinsics need to be changed to accept pointers with any
+      // address space.
+      continue;
     default:
       Intr->dump();
       llvm_unreachable("Don't know how to promote alloca intrinsic use.");


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D16416.45562.patch
Type: text/x-patch
Size: 1955 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160121/15cfc591/attachment.bin>