commit ae51b96f99bb0fb4956465633e8b036c57c21007
author Alexey Bataev <a.bataev@hotmail.com> Fri Dec 14 21:00:58 2018 +0000
committer Alexey Bataev <a.bataev@hotmail.com> Fri Dec 14 21:00:58 2018 +0000
tree 139af5b2763752afaa0baec0a6997e78b17d8085
parent 11cadc3e6b42467f6c5388aa6bf5ee619d8146c5

[OPENMP][NVPTX]Improved interwarp copy function.

Inlined runtime with the current implementation of the interwarp copy
function leads to the undefined behavior because of the not quite
correct implementation of the barriers. Start using generic
__kmpc_barier function instead of the custom made barriers.

llvm-svn: 349192

clang/test/OpenMP/nvptx_teams_reduction_codegen.cpp

diff --git a/clang/test/OpenMP/nvptx_teams_reduction_codegen.cpp b/clang/test/OpenMP/nvptx_teams_reduction_codegen.cpp
index 97a1288..65c147b 100644
--- a/clang/test/OpenMP/nvptx_teams_reduction_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_teams_reduction_codegen.cpp
@@ -412,9 +412,8 @@
   //
   // Barrier after copy to shared memory storage medium.
   // CHECK: [[COPY_CONT]]
-  // CHECK: [[WS:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
-  // CHECK: [[ACTIVE_THREADS:%.+]] = mul nsw i32 [[ACTIVE_WARPS:%.+]], [[WS]]
-  // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]])
+  // CHECK: call void @__kmpc_barrier(%struct.ident_t* @
+  // CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32*
   //
   // Read into warp 0.
   // CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]]
@@ -433,7 +432,7 @@
   // CHECK: br label {{%?}}[[READ_CONT]]
   //
   // CHECK: [[READ_CONT]]
-  // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]])
+  // CHECK: call void @__kmpc_barrier(%struct.ident_t* @
   // CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0
   // CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]]
   //
@@ -453,9 +452,8 @@
   //
   // Barrier after copy to shared memory storage medium.
   // CHECK: [[COPY_CONT]]
-  // CHECK: [[WS:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
-  // CHECK: [[ACTIVE_THREADS:%.+]] = mul nsw i32 [[ACTIVE_WARPS:%.+]], [[WS]]
-  // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]])
+  // CHECK: call void @__kmpc_barrier(%struct.ident_t* @
+  // CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32*
   //
   // Read into warp 0.
   // CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]]
@@ -475,7 +473,7 @@
   // CHECK: br label {{%?}}[[READ_CONT]]
   //
   // CHECK: [[READ_CONT]]
-  // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]])
+  // CHECK: call void @__kmpc_barrier(%struct.ident_t* @
   // CHECK: ret
 
 #endif