maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[OPENMP] Enable codegen of the ‘private’ clause for ‘omp simd’ directive 

llvm-svn: 232353
This commit is contained in:
Alexander Musman 2015-03-16 07:14:41 +00:00
parent c01ab519e6
commit 7931b98735
2 changed files with 27 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
clang/lib/CodeGen/CGStmtOpenMP.cpp
View File

@ -189,7 +189,7 @@ void CodeGenFunction::EmitOMPFirstprivateClause(
EmitDecl(*VD);
return GetAddrOfLocalVar(VD);
});
assert(IsRegistered && "counter already registered as private");
assert(IsRegistered && "firstprivate var already registered as private");
// Silence the warning about unused variable.
(void)IsRegistered;
++IRef, ++InitsRef;
@ -216,7 +216,7 @@ void CodeGenFunction::EmitOMPPrivateClause(
EmitDecl(*VD);
return GetAddrOfLocalVar(VD);
});
assert(IsRegistered && "counter already registered as private");
assert(IsRegistered && "private var already registered as private");
// Silence the warning about unused variable.
(void)IsRegistered;
++IRef;
@ -379,7 +379,6 @@ static void EmitPrivateLoopCounters(CodeGenFunction &CGF,
// Silence the warning about unused variable.
(void)IsRegistered;
}
(void)LoopScope.Privatize();
}
void CodeGenFunction::EmitOMPSimdDirective(const OMPSimdDirective &S) {
@ -456,6 +455,8 @@ void CodeGenFunction::EmitOMPSimdDirective(const OMPSimdDirective &S) {
{
OMPPrivateScope LoopScope(*this);
EmitPrivateLoopCounters(*this, LoopScope, S.counters());
EmitOMPPrivateClause(S, LoopScope);
(void)LoopScope.Privatize();
EmitOMPInnerLoop(S, LoopScope.requiresCleanups(),
S.getCond(/*SeparateIter=*/true), S.getInc(),
[&S, this]() {
@ -472,6 +473,8 @@ void CodeGenFunction::EmitOMPSimdDirective(const OMPSimdDirective &S) {
{
OMPPrivateScope LoopScope(*this);
EmitPrivateLoopCounters(*this, LoopScope, S.counters());
EmitOMPPrivateClause(S, LoopScope);
(void)LoopScope.Privatize();
EmitOMPInnerLoop(S, LoopScope.requiresCleanups(),
S.getCond(/*SeparateIter=*/false), S.getInc(),
[&S, this]() {
@ -670,6 +673,7 @@ void CodeGenFunction::EmitOMPWorksharingLoop(const OMPLoopDirective &S) {
OMPPrivateScope LoopScope(*this);
EmitPrivateLoopCounters(*this, LoopScope, S.counters());
(void)LoopScope.Privatize();
// Detect the loop schedule kind and chunk.
auto ScheduleKind = OMPC_SCHEDULE_unknown;

24
clang/test/OpenMP/simd_codegen.cpp
View File

@ -362,15 +362,19 @@ void collapsed(float *a, float *b, float *c, float *d) {
}
extern char foo();
extern double globalfloat;
// CHECK-LABEL: define {{.*void}} @{{.*}}widened{{.*}}
void widened(float *a, float *b, float *c, float *d) {
int i; // outer loop counter
short j; // inner loop counter
globalfloat = 1.0;
int localint = 1;
// CHECK: store double {{.+}}, double* [[GLOBALFLOAT:@.+]]
// Counter is widened to 64 bits.
// CHECK: store i64 0, i64* [[OMP_IV:[^,]+]]
//
#pragma omp simd collapse(2)
#pragma omp simd collapse(2) private(globalfloat, localint)
// CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]{{.+}}!llvm.mem.parallel_loop_access ![[WIDE1_LOOP_ID:[0-9]+]]
// CHECK-NEXT: [[LI:%.+]] = load i64, i64* [[OMP_LI:%[^,]+]]{{.+}}!llvm.mem.parallel_loop_access ![[WIDE1_LOOP_ID]]
@ -388,19 +392,31 @@ void widened(float *a, float *b, float *c, float *d) {
// CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.+}}!llvm.mem.parallel_loop_access ![[WIDE1_LOOP_ID]]
// CHECK: store i16 {{[^,]+}}, i16* [[LC_J:.+]]
// ... loop body ...
// End of body: store into a[i]:
// CHECK: store float [[RESULT:%.+]], float* [[RESULT_ADDR:%.+]]{{.+}}!llvm.mem.parallel_loop_access ![[WIDE1_LOOP_ID]]
//
// Here we expect store into private double var, not global
// CHECK-NOT: store double {{.+}}, double* [[GLOBALFLOAT]]
globalfloat = (float)j/i;
float res = b[j] * c[j];
// Store into a[i]:
// CHECK: store float [[RESULT:%.+]], float* [[RESULT_ADDR:%.+]]{{.+}}!llvm.mem.parallel_loop_access ![[WIDE1_LOOP_ID]]
a[i] = res * d[i];
// Then there's a store into private var localint:
// CHECK: store i32 {{.+}}, i32* [[LOCALINT:%[^,]+]]{{.+}}!llvm.mem.parallel_loop_access ![[WIDE1_LOOP_ID]]
localint = (int)j;
// CHECK: [[IV2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}!llvm.mem.parallel_loop_access ![[WIDE1_LOOP_ID]]
// CHECK-NEXT: [[ADD2:%.+]] = add nsw i64 [[IV2]], 1
// CHECK-NEXT: store i64 [[ADD2]], i64* [[OMP_IV]]{{.*}}!llvm.mem.parallel_loop_access ![[WIDE1_LOOP_ID]]
//
// br label %{{[^,]+}}, !llvm.loop ![[WIDE1_LOOP_ID]]
// CHECK: [[WIDE1_END]]
}
// i,j are updated.
// CHECK: store i32 3, i32* [[I:%[^,]+]]
// CHECK: store i16
//
// Here we expect store into original localint, not its privatized version.
// CHECK-NOT: store i32 {{.+}}, i32* [[LOCALINT]]
localint = (int)j;
// CHECK: ret void
}
@ -421,6 +437,6 @@ void parallel_simd(float *a) {
for (unsigned i = 131071; i <= 2147483647; i += 127)
a[i] += bar();
}
// TERM_DEBUG: !{{[0-9]+}} = !MDLocation(line: 413,
// TERM_DEBUG: !{{[0-9]+}} = !MDLocation(line: [[@LINE-11]],
#endif // HEADER