[OPENMP][NVPTX]Emit shared memory buffer for reduction as 128 bytes
buffer.
Seems to me, nvlink has a bug with the proper support of the weakly
linked symbols. It does not allow to define several shared memory buffer
with the different sizes even with the weak linkage. Instead we always
use 128 bytes buffer to prevent nvlink from the error message emission.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@349540 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp b/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
index 59086d2..5e2676b 100644
--- a/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
+++ b/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
@@ -4518,6 +4518,22 @@
Records.UseSharedMemory->setInitializer(
llvm::ConstantInt::get(CGM.Int16Ty, UseSharedMemory ? 1 : 0));
}
+ // Allocate SharedMemorySize buffer for the shared memory.
+ // FIXME: nvlink does not handle weak linkage correctly (object with the
+ // different size are reported as erroneous).
+ // Restore this code as sson as nvlink is fixed.
+ if (!SharedStaticRD->field_empty()) {
+ llvm::APInt ArySize(/*numBits=*/64, SharedMemorySize);
+ QualType SubTy = C.getConstantArrayType(
+ C.CharTy, ArySize, ArrayType::Normal, /*IndexTypeQuals=*/0);
+ auto *Field = FieldDecl::Create(
+ C, SharedStaticRD, SourceLocation(), SourceLocation(), nullptr, SubTy,
+ C.getTrivialTypeSourceInfo(SubTy, SourceLocation()),
+ /*BW=*/nullptr, /*Mutable=*/false,
+ /*InitStyle=*/ICIS_NoInit);
+ Field->setAccess(AS_public);
+ SharedStaticRD->addDecl(Field);
+ }
SharedStaticRD->completeDefinition();
if (!SharedStaticRD->field_empty()) {
QualType StaticTy = C.getRecordType(SharedStaticRD);
diff --git a/test/OpenMP/nvptx_data_sharing.cpp b/test/OpenMP/nvptx_data_sharing.cpp
index ed3c88b..3baee00 100644
--- a/test/OpenMP/nvptx_data_sharing.cpp
+++ b/test/OpenMP/nvptx_data_sharing.cpp
@@ -26,7 +26,7 @@
}
}
}
-// CK1: [[MEM_TY:%.+]] = type { [8 x i8] }
+// CK1: [[MEM_TY:%.+]] = type { [128 x i8] }
// CK1-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer
// CK1-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
// CK1-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i64 8
diff --git a/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp b/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp
index a84962c..4e763bd 100644
--- a/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp
+++ b/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp
@@ -21,7 +21,7 @@
return 0;
}
-// CHECK: [[MEM_TY:%.+]] = type { [84 x i8] }
+// CHECK: [[MEM_TY:%.+]] = type { [128 x i8] }
// CHECK-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer
// CHECK-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
// CHECK-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i{{64|32}} 84
diff --git a/test/OpenMP/nvptx_parallel_codegen.cpp b/test/OpenMP/nvptx_parallel_codegen.cpp
index 3dcf330..821897c 100644
--- a/test/OpenMP/nvptx_parallel_codegen.cpp
+++ b/test/OpenMP/nvptx_parallel_codegen.cpp
@@ -72,7 +72,7 @@
return a;
}
-// CHECK: [[MEM_TY:%.+]] = type { [4 x i8] }
+// CHECK: [[MEM_TY:%.+]] = type { [128 x i8] }
// CHECK-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer
// CHECK-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
// CHECK-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i{{64|32}} 4
diff --git a/test/OpenMP/nvptx_parallel_for_codegen.cpp b/test/OpenMP/nvptx_parallel_for_codegen.cpp
index 25a7a15..6a8601f 100644
--- a/test/OpenMP/nvptx_parallel_for_codegen.cpp
+++ b/test/OpenMP/nvptx_parallel_for_codegen.cpp
@@ -30,7 +30,7 @@
return a;
}
-// CHECK: [[MEM_TY:%.+]] = type { [4 x i8] }
+// CHECK: [[MEM_TY:%.+]] = type { [128 x i8] }
// CHECK-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer
// CHECK-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
// CHECK-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i{{64|32}} 4
diff --git a/test/OpenMP/nvptx_target_codegen.cpp b/test/OpenMP/nvptx_target_codegen.cpp
index db608ef..033ed1e 100644
--- a/test/OpenMP/nvptx_target_codegen.cpp
+++ b/test/OpenMP/nvptx_target_codegen.cpp
@@ -18,7 +18,7 @@
// CHECK-DAG: {{@__omp_offloading_.+l362}}_exec_mode = weak constant i8 1
// CHECK-DAG: {{@__omp_offloading_.+l380}}_exec_mode = weak constant i8 1
// CHECK-DAG: {{@__omp_offloading_.+l345}}_exec_mode = weak constant i8 1
-// CHECK-DAG: [[MAP_TY:%.+]] = type { [{{8|4}} x i8] }
+// CHECK-DAG: [[MAP_TY:%.+]] = type { [128 x i8] }
// CHECK-DAG: [[GLOB_TY:%.+]] = type { i32* }
__thread int id;
diff --git a/test/OpenMP/nvptx_target_teams_distribute_codegen.cpp b/test/OpenMP/nvptx_target_teams_distribute_codegen.cpp
index ba99d17..a137625 100644
--- a/test/OpenMP/nvptx_target_teams_distribute_codegen.cpp
+++ b/test/OpenMP/nvptx_target_teams_distribute_codegen.cpp
@@ -8,7 +8,7 @@
#ifndef HEADER
#define HEADER
-// CHECK: [[MEM_TY:%.+]] = type { [4 x i8] }
+// CHECK: [[MEM_TY:%.+]] = type { [128 x i8] }
// CHECK-DAG: {{@__omp_offloading_.+}}_l19_exec_mode = weak constant i8 1
// CHECK-DAG: internal unnamed_addr constant i{{64|32}} 4
diff --git a/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp b/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp
index 0c06a60..5f003dd 100644
--- a/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp
+++ b/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp
@@ -67,7 +67,7 @@
return a;
}
-// CHECK-DAG: [[MEM_TY:%.+]] = type { [4 x i8] }
+// CHECK-DAG: [[MEM_TY:%.+]] = type { [128 x i8] }
// CHECK-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer
// CHECK-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
// CHECK-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i{{64|32}} 4
diff --git a/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp b/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp
index 5a828af..c4db1bc 100644
--- a/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp
+++ b/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp
@@ -62,7 +62,7 @@
return a;
}
-// CHECK-DAG: [[MEM_TY:%.+]] = type { [4 x i8] }
+// CHECK-DAG: [[MEM_TY:%.+]] = type { [128 x i8] }
// CHECK-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer
// CHECK-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
// CHECK-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i{{64|32}} 4
diff --git a/test/OpenMP/nvptx_teams_codegen.cpp b/test/OpenMP/nvptx_teams_codegen.cpp
index 4965a50..b259a4b 100644
--- a/test/OpenMP/nvptx_teams_codegen.cpp
+++ b/test/OpenMP/nvptx_teams_codegen.cpp
@@ -27,7 +27,7 @@
return tmain(argv);
}
-// CK1: [[MEM_TY:%.+]] = type { [{{4|8}} x i8] }
+// CK1: [[MEM_TY:%.+]] = type { [128 x i8] }
// CK1-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer
// CK1-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
// CK1-DAG: [[KERNEL_SIZE1:@.+]] = internal unnamed_addr constant i{{64|32}} 4
@@ -114,7 +114,7 @@
return tmain(argv);
}
-// CK2: [[MEM_TY:%.+]] = type { [{{4|8}} x i8] }
+// CK2: [[MEM_TY:%.+]] = type { [128 x i8] }
// CK2-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer
// CK2-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
// CK2-DAG: [[KERNEL_SIZE1:@.+]] = internal unnamed_addr constant i{{64|32}} 4
diff --git a/test/OpenMP/nvptx_teams_reduction_codegen.cpp b/test/OpenMP/nvptx_teams_reduction_codegen.cpp
index 65c147b..818f073 100644
--- a/test/OpenMP/nvptx_teams_reduction_codegen.cpp
+++ b/test/OpenMP/nvptx_teams_reduction_codegen.cpp
@@ -8,7 +8,7 @@
#ifndef HEADER
#define HEADER
-// CHECK: [[MAP_TY:%.+]] = type { [16 x i8] }
+// CHECK: [[MAP_TY:%.+]] = type { [128 x i8] }
// CHECK-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
// CHECK-DAG: [[KERNEL_SHARED1:@.+]] = internal unnamed_addr constant i16 1
