This revision was landed with ongoing or failed builds.
This revision was automatically updated to reflect the committed changes.
Closed by commit rG3274cdc83ecd: [Clang][OpenMP] Remove the mandatory flush for
capture for OpenMP 5.1 (authored by tianshilei1992).
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D100768/new/
https://reviews.llvm.org/D100768
Files:
clang/lib/CodeGen/CGStmtOpenMP.cpp
clang/test/OpenMP/atomic_capture_codegen.cpp
Index: clang/test/OpenMP/atomic_capture_codegen.cpp
===================================================================
--- clang/test/OpenMP/atomic_capture_codegen.cpp
+++ clang/test/OpenMP/atomic_capture_codegen.cpp
@@ -1,11 +1,19 @@
// RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -target-cpu core2 -fopenmp -x c -emit-llvm %s -o - | FileCheck %s
// RUN: %clang_cc1 -fopenmp -x c -triple x86_64-apple-darwin10 -target-cpu core2 -emit-pch -o %t %s
-// RUN: %clang_cc1 -fopenmp -x c -triple x86_64-apple-darwin10 -target-cpu core2 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 -fopenmp -x c -triple x86_64-apple-darwin10 -target-cpu core2 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix=CHECK --check-prefix=CHECK-50 %s
// RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -target-cpu core2 -fopenmp-simd -x c -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// RUN: %clang_cc1 -fopenmp-simd -x c -triple x86_64-apple-darwin10 -target-cpu core2 -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -x c -triple x86_64-apple-darwin10 -target-cpu core2 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
+
+// RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -target-cpu core2 -fopenmp -fopenmp-version=51 -x c -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=51 -x c -triple x86_64-apple-darwin10 -target-cpu core2 -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=51 -x c -triple x86_64-apple-darwin10 -target-cpu core2 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
+
+// RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -target-cpu core2 -fopenmp-simd -fopenmp-version=51 -x c -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=51 -x c -triple x86_64-apple-darwin10 -target-cpu core2 -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=51 -x c -triple x86_64-apple-darwin10 -target-cpu core2 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
// expected-no-diagnostics
#ifndef HEADER
@@ -343,7 +351,7 @@
// CHECK: [[IM_CAST:%.+]] = fptrunc double [[NEW_IM]] to float
// CHECK: store float [[RE_CAST]], float* getelementptr inbounds ({ float, float }, { float, float }* @{{.+}}, i32 0, i32 0),
// CHECK: store float [[IM_CAST]], float* getelementptr inbounds ({ float, float }, { float, float }* @{{.+}}, i32 0, i32 1),
-// CHECK: call{{.*}} @__kmpc_flush(
+// CHECK-50: call{{.*}} @__kmpc_flush(
#pragma omp atomic capture seq_cst
{cdx = cdx - cdv; cfv = cdx;}
// CHECK: [[BV:%.+]] = load i8, i8* @{{.+}}
@@ -393,7 +401,7 @@
// CHECK: br i1 [[SUCCESS_FAIL]], label %[[EXIT:.+]], label %[[CONT]]
// CHECK: [[EXIT]]
// CHECK: store i8 [[NEW]], i8* @{{.+}},
-// CHECK: call{{.*}} @__kmpc_flush(
+// CHECK-50: call{{.*}} @__kmpc_flush(
#pragma omp atomic capture, seq_cst
{cx = cx >> ucv; cv = cx;}
// CHECK: [[SV:%.+]] = load i16, i16* @{{.+}},
@@ -436,7 +444,7 @@
// CHECK: [[OLD:%.+]] = atomicrmw or i32* @{{.+}}, i32 [[EXPR]] seq_cst, align 4
// CHECK: [[DESIRED:%.+]] = or i32 [[EXPR]], [[OLD]]
// CHECK: store i32 [[DESIRED]], i32* @{{.+}},
-// CHECK: call{{.*}} @__kmpc_flush(
+// CHECK-50: call{{.*}} @__kmpc_flush(
#pragma omp atomic seq_cst, capture
{uix = iv | uix; uiv = uix;}
// CHECK: [[EXPR:%.+]] = load i32, i32* @{{.+}}
@@ -904,7 +912,7 @@
// CHECK: [[EXIT]]
// CHECK: [[NEW_VAL:%.+]] = trunc i64 [[CONV]] to i32
// CHECK: store i32 [[NEW_VAL]], i32* @{{.+}},
-// CHECK: call{{.*}} @__kmpc_flush(
+// CHECK-50: call{{.*}} @__kmpc_flush(
#pragma omp atomic capture release
{bfx4.b /= ldv; iv = bfx4.b;}
// CHECK: [[EXPR:%.+]] = load x86_fp80, x86_fp80* @{{.+}}
@@ -937,7 +945,7 @@
// CHECK: [[EXIT]]
// CHECK: [[NEW_VAL_I32:%.+]] = trunc i64 [[NEW_VAL]] to i32
// CHECK: store i32 [[NEW_VAL_I32]], i32* @{{.+}},
-// CHECK: call{{.*}} @__kmpc_flush(
+// CHECK-50: call{{.*}} @__kmpc_flush(
#pragma omp atomic capture acquire
iv = bfx4_packed.b += ldv;
// CHECK: load i64, i64*
@@ -963,7 +971,7 @@
// CHECK: br i1 [[FAIL_SUCCESS]], label %[[EXIT:.+]], label %[[CONT]]
// CHECK: [[EXIT]]
// CHECK: store float [[X]], float* @{{.+}},
-// CHECK: call{{.*}} @__kmpc_flush(
+// CHECK-50: call{{.*}} @__kmpc_flush(
#pragma omp atomic capture acq_rel
{fv = float2x.x; float2x.x = ulv - float2x.x;}
// CHECK: [[EXPR:%.+]] = load double, double* @{{.+}},
@@ -973,7 +981,7 @@
// CHECK: [[NEW_VAL:%.+]] = fptosi double [[DIV]] to i32
// CHECK: call void @llvm.write_register.i32([[REG]], i32 [[NEW_VAL]])
// CHECK: store i32 [[NEW_VAL]], i32* @{{.+}},
-// CHECK: call{{.*}} @__kmpc_flush(
+// CHECK-50: call{{.*}} @__kmpc_flush(
#pragma omp atomic capture seq_cst
{rix = dv / rix; iv = rix;}
// CHECK: [[OLD_VAL:%.+]] = atomicrmw xchg i32* @{{.+}}, i32 5 monotonic, align 4
Index: clang/lib/CodeGen/CGStmtOpenMP.cpp
===================================================================
--- clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -5725,32 +5725,35 @@
// Emit post-update store to 'v' of old/new 'x' value.
CGF.emitOMPSimpleStore(VLValue, NewVVal, NewVValType, Loc);
CGF.CGM.getOpenMPRuntime().checkAndEmitLastprivateConditional(CGF, V);
- // OpenMP, 2.17.7, atomic Construct
- // If the write, update, or capture clause is specified and the release,
- // acq_rel, or seq_cst clause is specified then the strong flush on entry to
- // the atomic operation is also a release flush.
- // If the read or capture clause is specified and the acquire, acq_rel, or
- // seq_cst clause is specified then the strong flush on exit from the atomic
- // operation is also an acquire flush.
- switch (AO) {
- case llvm::AtomicOrdering::Release:
- CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc,
- llvm::AtomicOrdering::Release);
- break;
- case llvm::AtomicOrdering::Acquire:
- CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc,
- llvm::AtomicOrdering::Acquire);
- break;
- case llvm::AtomicOrdering::AcquireRelease:
- case llvm::AtomicOrdering::SequentiallyConsistent:
- CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc,
- llvm::AtomicOrdering::AcquireRelease);
- break;
- case llvm::AtomicOrdering::Monotonic:
- break;
- case llvm::AtomicOrdering::NotAtomic:
- case llvm::AtomicOrdering::Unordered:
- llvm_unreachable("Unexpected ordering.");
+ // OpenMP 5.1 removes the required flush for capture clause.
+ if (CGF.CGM.getLangOpts().OpenMP < 51) {
+ // OpenMP, 2.17.7, atomic Construct
+ // If the write, update, or capture clause is specified and the release,
+ // acq_rel, or seq_cst clause is specified then the strong flush on entry to
+ // the atomic operation is also a release flush.
+ // If the read or capture clause is specified and the acquire, acq_rel, or
+ // seq_cst clause is specified then the strong flush on exit from the atomic
+ // operation is also an acquire flush.
+ switch (AO) {
+ case llvm::AtomicOrdering::Release:
+ CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc,
+ llvm::AtomicOrdering::Release);
+ break;
+ case llvm::AtomicOrdering::Acquire:
+ CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc,
+ llvm::AtomicOrdering::Acquire);
+ break;
+ case llvm::AtomicOrdering::AcquireRelease:
+ case llvm::AtomicOrdering::SequentiallyConsistent:
+ CGF.CGM.getOpenMPRuntime().emitFlush(
+ CGF, llvm::None, Loc, llvm::AtomicOrdering::AcquireRelease);
+ break;
+ case llvm::AtomicOrdering::Monotonic:
+ break;
+ case llvm::AtomicOrdering::NotAtomic:
+ case llvm::AtomicOrdering::Unordered:
+ llvm_unreachable("Unexpected ordering.");
+ }
}
}
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
