diff options
| author | Justin Holewinski <jholewinski@nvidia.com> | 2013-02-11 18:56:35 +0000 |
|---|---|---|
| committer | Justin Holewinski <jholewinski@nvidia.com> | 2013-02-11 18:56:35 +0000 |
| commit | ff5adad9f39f60dc9f1caadef03b6aa7922ed168 (patch) | |
| tree | 80c4f5563a4b5e01ac6e3c04c92a31c59eb1bb7c | |
| parent | b1a82589339fed148c12b052d30861a539552f1a (diff) | |
| download | external_llvm-ff5adad9f39f60dc9f1caadef03b6aa7922ed168.zip external_llvm-ff5adad9f39f60dc9f1caadef03b6aa7922ed168.tar.gz external_llvm-ff5adad9f39f60dc9f1caadef03b6aa7922ed168.tar.bz2 | |
[NVPTX] Remove NoCapture from address space conversion intrinsics. NoCapture is not valid in this case, and was causing incorrect optimizations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@174896 91177308-0d34-0410-b5e6-96231b3b80d8
| -rw-r--r-- | include/llvm/IR/IntrinsicsNVVM.td | 18 | ||||
| -rw-r--r-- | test/CodeGen/NVPTX/intrin-nocapture.ll | 21 |
2 files changed, 30 insertions, 9 deletions
diff --git a/include/llvm/IR/IntrinsicsNVVM.td b/include/llvm/IR/IntrinsicsNVVM.td index 1853c99..6b85300 100644 --- a/include/llvm/IR/IntrinsicsNVVM.td +++ b/include/llvm/IR/IntrinsicsNVVM.td @@ -815,36 +815,36 @@ def int_nvvm_ldu_global_p : Intrinsic<[llvm_anyptr_ty], // of pointer to another type of pointer, while the address space remains // the same. def int_nvvm_ptr_local_to_gen: Intrinsic<[llvm_anyptr_ty], - [llvm_anyptr_ty], [IntrNoMem, NoCapture<0>], + [llvm_anyptr_ty], [IntrNoMem], "llvm.nvvm.ptr.local.to.gen">; def int_nvvm_ptr_shared_to_gen: Intrinsic<[llvm_anyptr_ty], - [llvm_anyptr_ty], [IntrNoMem, NoCapture<0>], + [llvm_anyptr_ty], [IntrNoMem], "llvm.nvvm.ptr.shared.to.gen">; def int_nvvm_ptr_global_to_gen: Intrinsic<[llvm_anyptr_ty], - [llvm_anyptr_ty], [IntrNoMem, NoCapture<0>], + [llvm_anyptr_ty], [IntrNoMem], "llvm.nvvm.ptr.global.to.gen">; def int_nvvm_ptr_constant_to_gen: Intrinsic<[llvm_anyptr_ty], - [llvm_anyptr_ty], [IntrNoMem, NoCapture<0>], + [llvm_anyptr_ty], [IntrNoMem], "llvm.nvvm.ptr.constant.to.gen">; def int_nvvm_ptr_gen_to_global: Intrinsic<[llvm_anyptr_ty], - [llvm_anyptr_ty], [IntrNoMem, NoCapture<0>], + [llvm_anyptr_ty], [IntrNoMem], "llvm.nvvm.ptr.gen.to.global">; def int_nvvm_ptr_gen_to_shared: Intrinsic<[llvm_anyptr_ty], - [llvm_anyptr_ty], [IntrNoMem, NoCapture<0>], + [llvm_anyptr_ty], [IntrNoMem], "llvm.nvvm.ptr.gen.to.shared">; def int_nvvm_ptr_gen_to_local: Intrinsic<[llvm_anyptr_ty], - [llvm_anyptr_ty], [IntrNoMem, NoCapture<0>], + [llvm_anyptr_ty], [IntrNoMem], "llvm.nvvm.ptr.gen.to.local">; def int_nvvm_ptr_gen_to_constant: Intrinsic<[llvm_anyptr_ty], - [llvm_anyptr_ty], [IntrNoMem, NoCapture<0>], + [llvm_anyptr_ty], [IntrNoMem], "llvm.nvvm.ptr.gen.to.constant">; // Used in nvvm internally to help address space opt and ptx code generation // This is for params that are passed to kernel functions by pointer by-val. def int_nvvm_ptr_gen_to_param: Intrinsic<[llvm_anyptr_ty], [llvm_anyptr_ty], - [IntrNoMem, NoCapture<0>], + [IntrNoMem], "llvm.nvvm.ptr.gen.to.param">; // Move intrinsics, used in nvvm internally diff --git a/test/CodeGen/NVPTX/intrin-nocapture.ll b/test/CodeGen/NVPTX/intrin-nocapture.ll new file mode 100644 index 0000000..55781bb --- /dev/null +++ b/test/CodeGen/NVPTX/intrin-nocapture.ll @@ -0,0 +1,21 @@ +; RUN: opt < %s -O3 -S | FileCheck %s + +; Address space intrinsics were erroneously marked NoCapture, leading to bad +; optimizations (such as the store below being eliminated as dead code). This +; test makes sure we don't regress. + +declare void @foo(i32 addrspace(1)*) + +declare i32 addrspace(1)* @llvm.nvvm.ptr.gen.to.global.p1i32.p0i32(i32*) + +; CHECK: @bar +define void @bar() { + %t1 = alloca i32 +; CHECK: call i32 addrspace(1)* @llvm.nvvm.ptr.gen.to.global.p1i32.p0i32(i32* %t1) +; CHECK-NEXT: store i32 10, i32* %t1 + %t2 = call i32 addrspace(1)* @llvm.nvvm.ptr.gen.to.global.p1i32.p0i32(i32* %t1) + store i32 10, i32* %t1 + call void @foo(i32 addrspace(1)* %t2) + ret void +} + |