commit 9ea3c385973c395c4798145f71ff3ae34ccaabeb
author Alexey Bataev <a.bataev@hotmail.com> Tue Oct 09 14:49:00 2018 +0000
committer Alexey Bataev <a.bataev@hotmail.com> Tue Oct 09 14:49:00 2018 +0000
tree 0a2f511ca7fb094143065d6ee32b2976e7079a80
parent f5fac1826a867d5e65310456670cc1d5a495ed02

[OPENMP][NVPTX] Support memory coalescing for globalized variables.

Added support for memory coalescing for better performance for
globalized variables. From now on all the globalized variables are
represented as arrays of 32 elements and each thread accesses these
elements using `tid & 31` as index.

llvm-svn: 344049

clang/test/OpenMP/nvptx_target_codegen.cpp

diff --git a/clang/test/OpenMP/nvptx_target_codegen.cpp b/clang/test/OpenMP/nvptx_target_codegen.cpp
index 01f3461..69c3389 100644
--- a/clang/test/OpenMP/nvptx_target_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_target_codegen.cpp
@@ -554,7 +554,8 @@
   // CHECK: ret void
 
   // CHECK: define i32 [[BAZ]](i32 [[F:%.*]], double* dereferenceable{{.*}})
-  // CHECK: [[STACK:%.+]] = alloca [[GLOBAL_ST:%.+]],
+  // CHECK: alloca i32,
+  // CHECK: [[LOCAL_F_PTR:%.+]] = alloca i32,
   // CHECK: [[ZERO_ADDR:%.+]] = alloca i32,
   // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t*
   // CHECK: store i32 0, i32* [[ZERO_ADDR]]
@@ -562,11 +563,15 @@
   // CHECK: [[IS_SPMD:%.+]] = icmp ne i8 [[RES]], 0
   // CHECK: br i1 [[IS_SPMD]], label
   // CHECK: br label
-  // CHECK: [[PTR:%.+]] = call i8* @__kmpc_data_sharing_push_stack(i{{64|32}} 4, i16 0)
-  // CHECK: [[REC_ADDR:%.+]] = bitcast i8* [[PTR]] to [[GLOBAL_ST]]*
+  // CHECK: [[PTR:%.+]] = call i8* @__kmpc_data_sharing_push_stack(i{{64|32}} 128, i16 0)
+  // CHECK: [[REC_ADDR:%.+]] = bitcast i8* [[PTR]] to [[GLOBAL_ST:%.+]]*
   // CHECK: br label
-  // CHECK: [[ITEMS:%.+]] = phi [[GLOBAL_ST]]* [ [[STACK]], {{.+}} ], [ [[REC_ADDR]], {{.+}} ]
-  // CHECK: [[F_PTR:%.+]] = getelementptr inbounds [[GLOBAL_ST]], [[GLOBAL_ST]]* [[ITEMS]], i32 0, i32 0
+  // CHECK: [[ITEMS:%.+]] = phi [[GLOBAL_ST]]* [ null, {{.+}} ], [ [[REC_ADDR]], {{.+}} ]
+  // CHECK: [[F_PTR_ARR:%.+]] = getelementptr inbounds [[GLOBAL_ST]], [[GLOBAL_ST]]* [[ITEMS]], i32 0, i32 0
+  // CHECK: [[TID:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
+  // CHECK: [[LID:%.+]] = and i32 [[TID]], 31
+  // CHECK: [[GLOBAL_F_PTR:%.+]] = getelementptr inbounds [32 x i32], [32 x i32]* [[F_PTR_ARR]], i32 0, i32 [[LID]]
+  // CHECK: [[F_PTR:%.+]] = select i1 [[IS_SPMD]], i32* [[LOCAL_F_PTR]], i32* [[GLOBAL_F_PTR]]
   // CHECK: store i32 %{{.+}}, i32* [[F_PTR]],
 
   // CHECK: [[RES:%.+]] = call i8 @__kmpc_is_spmd_exec_mode()