[llvm-branch-commits] [clang] release/21.x: [Clang][CMake] Use IRPGO instead of FE PGO for Cmake Caches (#155957) (PR #156271)
boomanaiden154 wrote: This makes it easier to use the new options in the CI container since we build from the release branch. This should be a relatively safe backport given how small the change is in addition to how infrequently these CMake caches change at this point. https://github.com/llvm/llvm-project/pull/156271 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] release/21.x: [Clang][CMake] Use IRPGO instead of FE PGO for Cmake Caches (#155957) (PR #156271)
https://github.com/mtrofin approved this pull request. lgtm, but you probably want the release branch owner to lgtm, too https://github.com/llvm/llvm-project/pull/156271 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] [AArch64] Split large loop dependence masks (PR #153187)
@@ -5247,50 +5247,85 @@ SDValue
AArch64TargetLowering::LowerLOOP_DEPENDENCE_MASK(SDValue Op,
SelectionDAG &DAG) const {
SDLoc DL(Op);
- uint64_t EltSize = Op.getConstantOperandVal(2);
- EVT VT = Op.getValueType();
- switch (EltSize) {
- case 1:
-if (VT != MVT::v16i8 && VT != MVT::nxv16i1)
- return SDValue();
-break;
- case 2:
-if (VT != MVT::v8i8 && VT != MVT::nxv8i1)
- return SDValue();
-break;
- case 4:
-if (VT != MVT::v4i16 && VT != MVT::nxv4i1)
- return SDValue();
-break;
- case 8:
-if (VT != MVT::v2i32 && VT != MVT::nxv2i1)
- return SDValue();
-break;
- default:
-// Other element sizes are incompatible with whilewr/rw, so expand instead
-return SDValue();
- }
+ assert((Subtarget->hasSVE2() ||
+ (Subtarget->hasSME() && Subtarget->isStreaming())) &&
+ "Lowering loop_dependence_raw_mask or loop_dependence_war_mask "
+ "requires SVE or SME");
+
+ uint64_t EltSizeInBytes = Op.getConstantOperandVal(2);
+ // Other element sizes are incompatible with whilewr/rw, so expand instead
+ if (!is_contained({1u, 2u, 4u, 8u}, EltSizeInBytes))
+return SDValue();
+
+ EVT FullVT = Op.getValueType();
+ EVT ExtractVT = FullVT;
+ EVT EltVT = MVT::getIntegerVT(EltSizeInBytes * 8);
+ unsigned NumElements = FullVT.getVectorMinNumElements();
+ unsigned PredElements =
getPackedSVEVectorVT(EltVT).getVectorMinNumElements();
+ bool Split = NumElements > PredElements;
+
+ if (EltSizeInBytes * NumElements < 16)
+// The element size and vector length combination must at least form a
+// 128-bit vector. Shorter vector lengths can be widened then extracted
+FullVT = FullVT.getDoubleNumVectorElementsVT(*DAG.getContext());
+
+ auto LowerToWhile = [&](EVT VT, unsigned AddrScale) {
+SDValue PtrA = Op.getOperand(0);
+SDValue PtrB = Op.getOperand(1);
+
+if (AddrScale > 0) {
+ unsigned Offset =
+ VT.getVectorMinNumElements() * EltSizeInBytes * AddrScale;
+ SDValue Addend;
+
+ if (VT.isScalableVT())
+Addend = DAG.getVScale(DL, MVT::i64, APInt(64, Offset));
+ else
+Addend = DAG.getConstant(Offset, DL, MVT::i64);
- SDValue PtrA = Op.getOperand(0);
- SDValue PtrB = Op.getOperand(1);
+ PtrA = DAG.getNode(ISD::ADD, DL, MVT::i64, PtrA, Addend);
+}
- if (VT.isScalableVT())
-return DAG.getNode(Op.getOpcode(), DL, VT, PtrA, PtrB, Op.getOperand(2));
+if (VT.isScalableVT())
+ return DAG.getNode(Op.getOpcode(), DL, VT, PtrA, PtrB, Op.getOperand(2));
- // We can use the SVE whilewr/whilerw instruction to lower this
- // intrinsic by creating the appropriate sequence of scalable vector
- // operations and then extracting a fixed-width subvector from the scalable
- // vector. Scalable vector variants are already legal.
- EVT ContainerVT =
- EVT::getVectorVT(*DAG.getContext(), VT.getVectorElementType(),
- VT.getVectorNumElements(), true);
- EVT WhileVT = ContainerVT.changeElementType(MVT::i1);
+// We can use the SVE whilewr/whilerw instruction to lower this
+// intrinsic by creating the appropriate sequence of scalable vector
+// operations and then extracting a fixed-width subvector from the scalable
+// vector. Scalable vector variants are already legal.
+EVT ContainerVT =
+EVT::getVectorVT(*DAG.getContext(), VT.getVectorElementType(),
+ VT.getVectorNumElements(), true);
+EVT WhileVT = ContainerVT.changeElementType(MVT::i1);
- SDValue Mask =
- DAG.getNode(Op.getOpcode(), DL, WhileVT, PtrA, PtrB, Op.getOperand(2));
- SDValue MaskAsInt = DAG.getNode(ISD::SIGN_EXTEND, DL, ContainerVT, Mask);
- return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, VT, MaskAsInt,
- DAG.getVectorIdxConstant(0, DL));
+SDValue Mask =
+DAG.getNode(Op.getOpcode(), DL, WhileVT, PtrA, PtrB, Op.getOperand(2));
+SDValue MaskAsInt = DAG.getNode(ISD::SIGN_EXTEND, DL, ContainerVT, Mask);
+return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, VT, MaskAsInt,
+ DAG.getVectorIdxConstant(0, DL));
+ };
+
+ SDValue Result;
+ if (!Split) {
+Result = LowerToWhile(FullVT, 0);
SamTebbs33 wrote:
It was being called to re-use the containerisation, not needed now with the new
approach.
https://github.com/llvm/llvm-project/pull/153187
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] [AArch64] Split large loop dependence masks (PR #153187)
@@ -5247,50 +5247,85 @@ SDValue
AArch64TargetLowering::LowerLOOP_DEPENDENCE_MASK(SDValue Op,
SelectionDAG &DAG) const {
SDLoc DL(Op);
- uint64_t EltSize = Op.getConstantOperandVal(2);
- EVT VT = Op.getValueType();
- switch (EltSize) {
- case 1:
-if (VT != MVT::v16i8 && VT != MVT::nxv16i1)
- return SDValue();
-break;
- case 2:
-if (VT != MVT::v8i8 && VT != MVT::nxv8i1)
- return SDValue();
-break;
- case 4:
-if (VT != MVT::v4i16 && VT != MVT::nxv4i1)
- return SDValue();
-break;
- case 8:
-if (VT != MVT::v2i32 && VT != MVT::nxv2i1)
- return SDValue();
-break;
- default:
-// Other element sizes are incompatible with whilewr/rw, so expand instead
-return SDValue();
- }
+ assert((Subtarget->hasSVE2() ||
+ (Subtarget->hasSME() && Subtarget->isStreaming())) &&
+ "Lowering loop_dependence_raw_mask or loop_dependence_war_mask "
+ "requires SVE or SME");
+
+ uint64_t EltSizeInBytes = Op.getConstantOperandVal(2);
+ // Other element sizes are incompatible with whilewr/rw, so expand instead
+ if (!is_contained({1u, 2u, 4u, 8u}, EltSizeInBytes))
+return SDValue();
+
+ EVT FullVT = Op.getValueType();
+ EVT ExtractVT = FullVT;
+ EVT EltVT = MVT::getIntegerVT(EltSizeInBytes * 8);
+ unsigned NumElements = FullVT.getVectorMinNumElements();
+ unsigned PredElements =
getPackedSVEVectorVT(EltVT).getVectorMinNumElements();
+ bool Split = NumElements > PredElements;
+
+ if (EltSizeInBytes * NumElements < 16)
+// The element size and vector length combination must at least form a
+// 128-bit vector. Shorter vector lengths can be widened then extracted
+FullVT = FullVT.getDoubleNumVectorElementsVT(*DAG.getContext());
+
+ auto LowerToWhile = [&](EVT VT, unsigned AddrScale) {
SamTebbs33 wrote:
Done.
https://github.com/llvm/llvm-project/pull/153187
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] [AArch64] Split large loop dependence masks (PR #153187)
@@ -5286,41 +5285,44 @@
AArch64TargetLowering::LowerLOOP_DEPENDENCE_MASK(SDValue Op,
PtrA = DAG.getNode(ISD::ADD, DL, MVT::i64, PtrA, Addend);
}
-if (VT.isScalableVT())
- return DAG.getNode(Op.getOpcode(), DL, VT, PtrA, PtrB, Op.getOperand(2));
-
-// We can use the SVE whilewr/whilerw instruction to lower this
-// intrinsic by creating the appropriate sequence of scalable vector
-// operations and then extracting a fixed-width subvector from the scalable
-// vector. Scalable vector variants are already legal.
-EVT ContainerVT =
-EVT::getVectorVT(*DAG.getContext(), VT.getVectorElementType(),
- VT.getVectorNumElements(), true);
-EVT WhileVT = ContainerVT.changeElementType(MVT::i1);
-
-SDValue Mask =
-DAG.getNode(Op.getOpcode(), DL, WhileVT, PtrA, PtrB, Op.getOperand(2));
-SDValue MaskAsInt = DAG.getNode(ISD::SIGN_EXTEND, DL, ContainerVT, Mask);
-return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, VT, MaskAsInt,
- DAG.getVectorIdxConstant(0, DL));
+return DAG.getNode(Op.getOpcode(), DL, VT, PtrA, PtrB, Op.getOperand(2));
};
SDValue Result;
- if (!Split) {
-Result = LowerToWhile(FullVT, 0);
- } else {
-
+ if (Split) {
SamTebbs33 wrote:
Yeah I separated the splitting and containerisation logic as per the request
[here](https://github.com/llvm/llvm-project/pull/153187/commits/85b52942d61712aba884c12a60b98cbdaee2b233#r2304356700).
I've experimented with adding another lambda that containerises without
re-entering and that seems to have fixed the codegen.
https://github.com/llvm/llvm-project/pull/153187
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang-tools-extra] [clangd] Use HeuristicResolver to try to resolve dependent 'auto' (PR #156283)
llvmbot wrote:
@llvm/pr-subscribers-clangd
Author: Nathan Ridge (HighCommander4)
Changes
Fixes https://github.com/clangd/clangd/issues/2431
---
Full diff: https://github.com/llvm/llvm-project/pull/156283.diff
9 Files Affected:
- (modified) clang-tools-extra/clangd/AST.cpp (+17-4)
- (modified) clang-tools-extra/clangd/AST.h (+3-1)
- (modified) clang-tools-extra/clangd/Hover.cpp (+3-1)
- (modified) clang-tools-extra/clangd/XRefs.cpp (+7-4)
- (modified) clang-tools-extra/clangd/refactor/tweaks/ExpandDeducedType.cpp
(+2-1)
- (modified) clang-tools-extra/clangd/unittests/ASTTests.cpp (+2-1)
- (modified) clang-tools-extra/clangd/unittests/HoverTests.cpp (+2-2)
- (modified) clang-tools-extra/clangd/unittests/XRefsTests.cpp (+9-1)
- (modified)
clang-tools-extra/clangd/unittests/tweaks/ExpandDeducedTypeTests.cpp (+2-2)
``diff
diff --git a/clang-tools-extra/clangd/AST.cpp b/clang-tools-extra/clangd/AST.cpp
index 2f46ecc92576c..b96a84519e78c 100644
--- a/clang-tools-extra/clangd/AST.cpp
+++ b/clang-tools-extra/clangd/AST.cpp
@@ -29,6 +29,7 @@
#include "clang/Basic/SourceManager.h"
#include "clang/Basic/Specifiers.h"
#include "clang/Index/USRGeneration.h"
+#include "clang/Sema/HeuristicResolver.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallSet.h"
@@ -479,10 +480,12 @@ namespace {
/// a deduced type set. The AST should be improved to simplify this scenario.
class DeducedTypeVisitor : public RecursiveASTVisitor {
SourceLocation SearchedLocation;
+ const HeuristicResolver *Resolver;
public:
- DeducedTypeVisitor(SourceLocation SearchedLocation)
- : SearchedLocation(SearchedLocation) {}
+ DeducedTypeVisitor(SourceLocation SearchedLocation,
+ const HeuristicResolver *Resolver)
+ : SearchedLocation(SearchedLocation), Resolver(Resolver) {}
// Handle auto initializers:
//- auto i = 1;
@@ -499,6 +502,14 @@ class DeducedTypeVisitor : public
RecursiveASTVisitor {
return true;
if (auto *AT = D->getType()->getContainedAutoType()) {
+ if (AT->isUndeducedAutoType()) {
+if (const auto *VD = dyn_cast(D)) {
+ if (Resolver && VD->hasInit()) {
+DeducedType = Resolver->resolveExprToType(VD->getInit());
+return true;
+ }
+}
+ }
DeducedType = AT->desugar();
}
return true;
@@ -608,10 +619,12 @@ class DeducedTypeVisitor : public
RecursiveASTVisitor {
};
} // namespace
-std::optional getDeducedType(ASTContext &ASTCtx, SourceLocation Loc)
{
+std::optional getDeducedType(ASTContext &ASTCtx,
+ const HeuristicResolver *Resolver,
+ SourceLocation Loc) {
if (!Loc.isValid())
return {};
- DeducedTypeVisitor V(Loc);
+ DeducedTypeVisitor V(Loc, Resolver);
V.TraverseAST(ASTCtx);
if (V.DeducedType.isNull())
return std::nullopt;
diff --git a/clang-tools-extra/clangd/AST.h b/clang-tools-extra/clangd/AST.h
index 1538d12172593..2b83595e5b8e9 100644
--- a/clang-tools-extra/clangd/AST.h
+++ b/clang-tools-extra/clangd/AST.h
@@ -31,6 +31,7 @@ namespace clang {
class SourceManager;
class Decl;
class DynTypedNode;
+class HeuristicResolver;
namespace clangd {
@@ -167,7 +168,8 @@ QualType declaredType(const TypeDecl *D);
/// Retrieves the deduced type at a given location (auto, decltype).
/// It will return the underlying type.
/// If the type is an undeduced auto, returns the type itself.
-std::optional getDeducedType(ASTContext &, SourceLocation Loc);
+std::optional getDeducedType(ASTContext &, const HeuristicResolver *,
+ SourceLocation Loc);
// Find the abbreviated-function-template `auto` within a type, or returns
null.
// Similar to getContainedAutoTypeLoc, but these `auto`s are
diff --git a/clang-tools-extra/clangd/Hover.cpp
b/clang-tools-extra/clangd/Hover.cpp
index 9eec322fe5963..138544dea99a1 100644
--- a/clang-tools-extra/clangd/Hover.cpp
+++ b/clang-tools-extra/clangd/Hover.cpp
@@ -1309,7 +1309,9 @@ std::optional getHover(ParsedAST &AST,
Position Pos,
}
} else if (Tok.kind() == tok::kw_auto || Tok.kind() == tok::kw_decltype) {
HoverCountMetric.record(1, "keyword");
- if (auto Deduced = getDeducedType(AST.getASTContext(), Tok.location())) {
+ if (auto Deduced =
+ getDeducedType(AST.getASTContext(), AST.getHeuristicResolver(),
+ Tok.location())) {
HI = getDeducedTypeHoverContents(*Deduced, Tok, AST.getASTContext(),
PP,
Index);
HighlightRange = Tok.range(SM).toCharRange(SM);
diff --git a/clang-tools-extra/clangd/XRefs.cpp
b/clang-tools-extra/clangd/XRefs.cpp
index a253a630a48cc..18bc29426df29 100644
--- a/clang-tools-extra/clangd/XRefs.cpp
+++ b/clang-tools-extra/clangd/XRefs.cpp
@@ -806,7 +806,9 @@ std::vector locateSymbolAt(ParsedAST &AST,
Po
[llvm-branch-commits] [clang-tools-extra] [clangd] Use HeuristicResolver to try to resolve dependent 'auto' (PR #156283)
HighCommander4 wrote: Depends on https://github.com/llvm/llvm-project/pull/156282 https://github.com/llvm/llvm-project/pull/156283 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [flang] [flang][OpenMP] Extend `do concurrent` mapping to device (PR #155987)
https://github.com/ergawy updated
https://github.com/llvm/llvm-project/pull/155987
>From 0373863b919e59130dcf57593f4283ece0dff12a Mon Sep 17 00:00:00 2001
From: ergawy
Date: Fri, 29 Aug 2025 02:04:49 -0500
Subject: [PATCH] [flang][OpenMP] Extend `do concurrent` mapping to device
Upstreams further parts of `do concurrent` to OpenMP conversion pass
from AMD's fork. This PR extends the pass by adding support for mapping
to the device.
---
flang/lib/Optimizer/OpenMP/CMakeLists.txt | 1 +
.../OpenMP/DoConcurrentConversion.cpp | 400 +-
.../Transforms/DoConcurrent/basic_device.f90 | 83
.../Transforms/DoConcurrent/basic_device.mlir | 10 +-
4 files changed, 476 insertions(+), 18 deletions(-)
create mode 100644 flang/test/Transforms/DoConcurrent/basic_device.f90
diff --git a/flang/lib/Optimizer/OpenMP/CMakeLists.txt
b/flang/lib/Optimizer/OpenMP/CMakeLists.txt
index e0aebd0714c8f..b85ee7e861a4f 100644
--- a/flang/lib/Optimizer/OpenMP/CMakeLists.txt
+++ b/flang/lib/Optimizer/OpenMP/CMakeLists.txt
@@ -26,6 +26,7 @@ add_flang_library(FlangOpenMPTransforms
FIRSupport
FortranSupport
HLFIRDialect
+ FortranUtils
MLIR_DEPS
${dialect_libs}
diff --git a/flang/lib/Optimizer/OpenMP/DoConcurrentConversion.cpp
b/flang/lib/Optimizer/OpenMP/DoConcurrentConversion.cpp
index c928b76065ade..e975b86a6ba0d 100644
--- a/flang/lib/Optimizer/OpenMP/DoConcurrentConversion.cpp
+++ b/flang/lib/Optimizer/OpenMP/DoConcurrentConversion.cpp
@@ -6,17 +6,22 @@
//
//===--===//
+#include "flang/Optimizer/Builder/DirectivesCommon.h"
#include "flang/Optimizer/Builder/FIRBuilder.h"
+#include "flang/Optimizer/Builder/HLFIRTools.h"
#include "flang/Optimizer/Builder/Todo.h"
#include "flang/Optimizer/Dialect/FIROps.h"
+#include "flang/Optimizer/HLFIR/HLFIROps.h"
#include "flang/Optimizer/OpenMP/Passes.h"
#include "flang/Optimizer/OpenMP/Utils.h"
#include "flang/Support/OpenMP-utils.h"
+#include "flang/Utils/OpenMP.h"
#include "mlir/Analysis/SliceAnalysis.h"
#include "mlir/Dialect/OpenMP/OpenMPDialect.h"
#include "mlir/IR/IRMapping.h"
#include "mlir/Transforms/DialectConversion.h"
#include "mlir/Transforms/RegionUtils.h"
+#include "llvm/Frontend/OpenMP/OMPConstants.h"
namespace flangomp {
#define GEN_PASS_DEF_DOCONCURRENTCONVERSIONPASS
@@ -107,6 +112,33 @@ struct InductionVariableInfo {
using InductionVariableInfos = llvm::SmallVector;
+/// Collect the list of values used inside the loop but defined outside of it.
+void collectLoopLiveIns(fir::DoConcurrentLoopOp loop,
+llvm::SmallVectorImpl &liveIns) {
+ llvm::SmallDenseSet seenValues;
+ llvm::SmallDenseSet seenOps;
+
+ for (auto [lb, ub, st] : llvm::zip_equal(
+ loop.getLowerBound(), loop.getUpperBound(), loop.getStep())) {
+liveIns.push_back(lb);
+liveIns.push_back(ub);
+liveIns.push_back(st);
+ }
+
+ mlir::visitUsedValuesDefinedAbove(
+ loop.getRegion(), [&](mlir::OpOperand *operand) {
+if (!seenValues.insert(operand->get()).second)
+ return;
+
+mlir::Operation *definingOp = operand->get().getDefiningOp();
+// We want to collect ops corresponding to live-ins only once.
+if (definingOp && !seenOps.insert(definingOp).second)
+ return;
+
+liveIns.push_back(operand->get());
+ });
+}
+
/// Collects values that are local to a loop: "loop-local values". A loop-local
/// value is one that is used exclusively inside the loop but allocated outside
/// of it. This usually corresponds to temporary values that are used inside
the
@@ -182,10 +214,6 @@ class DoConcurrentConversion
mlir::LogicalResult
matchAndRewrite(fir::DoConcurrentOp doLoop, OpAdaptor adaptor,
mlir::ConversionPatternRewriter &rewriter) const override {
-if (mapToDevice)
- return doLoop.emitError(
- "not yet implemented: Mapping `do concurrent` loops to device");
-
looputils::InductionVariableInfos ivInfos;
auto loop = mlir::cast(
doLoop.getRegion().back().getTerminator());
@@ -196,20 +224,72 @@ class DoConcurrentConversion
for (mlir::Value indVar : *indVars)
ivInfos.emplace_back(loop, indVar);
+llvm::SmallVector loopNestLiveIns;
+looputils::collectLoopLiveIns(loop, loopNestLiveIns);
+assert(!loopNestLiveIns.empty());
+
llvm::SetVector locals;
looputils::collectLoopLocalValues(loop, locals);
+// We do not want to map "loop-local" values to the device through
+// `omp.map.info` ops. Therefore, we remove them from the list of live-ins.
+loopNestLiveIns.erase(llvm::remove_if(loopNestLiveIns,
+ [&](mlir::Value liveIn) {
+return locals.contains(liveIn);
+ }),
+ loopNestLiveIns.end());
+
+mlir::omp::TargetOp targetOp;
[llvm-branch-commits] [flang] [flang][OpenMP] `do concurrent` to device mapping lit tests (PR #155992)
https://github.com/ergawy updated
https://github.com/llvm/llvm-project/pull/155992
>From 3dd383be4ab7c3a03da321f1127e056f52eecff0 Mon Sep 17 00:00:00 2001
From: ergawy
Date: Fri, 29 Aug 2025 03:53:51 -0500
Subject: [PATCH] [flang][OpenMP] `do concurrent` to device mapping lit tests
Adds more lit tests for `do concurrent` device mapping.
---
.../Transforms/DoConcurrent/allocatable.f90 | 29 +
.../Transforms/DoConcurrent/host_eval.f90 | 63 +++
.../DoConcurrent/locally_destroyed_temp.f90 | 43 ---
.../DoConcurrent/map_shape_info.f90 | 104 +
.../multiple_iteration_ranges.f90 | 106 +++---
.../DoConcurrent/non_reference_to_device.f90 | 34 ++
.../DoConcurrent/not_perfectly_nested.f90 | 66 +++
.../DoConcurrent/runtime_sized_array.f90 | 42 +++
.../DoConcurrent/skip_all_nested_loops.f90| 68 +++
9 files changed, 478 insertions(+), 77 deletions(-)
create mode 100644 flang/test/Transforms/DoConcurrent/allocatable.f90
create mode 100644 flang/test/Transforms/DoConcurrent/host_eval.f90
create mode 100644 flang/test/Transforms/DoConcurrent/map_shape_info.f90
create mode 100644
flang/test/Transforms/DoConcurrent/non_reference_to_device.f90
create mode 100644 flang/test/Transforms/DoConcurrent/runtime_sized_array.f90
create mode 100644 flang/test/Transforms/DoConcurrent/skip_all_nested_loops.f90
diff --git a/flang/test/Transforms/DoConcurrent/allocatable.f90
b/flang/test/Transforms/DoConcurrent/allocatable.f90
new file mode 100644
index 0..03962f150eb95
--- /dev/null
+++ b/flang/test/Transforms/DoConcurrent/allocatable.f90
@@ -0,0 +1,29 @@
+! Verifies that proper `omp.map.bounds` ops are emitted when an allocatable is
+! implicitly mapped by a `do concurrent` loop.
+
+! RUN: %flang_fc1 -emit-hlfir -fopenmp -fdo-concurrent-to-openmp=device %s -o
- \
+! RUN: | FileCheck %s
+program main
+ implicit none
+
+ integer,parameter :: n = 100
+ real, allocatable, dimension(:) :: y
+ integer :: i
+
+ allocate(y(1:n))
+
+ do concurrent(i=1:n)
+ y(i) = 42
+ end do
+
+ deallocate(y)
+end program main
+
+! CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %{{.*}} {fortran_attrs =
#fir.var_attrs, uniq_name = "_QFEy"}
+! CHECK: %[[Y_VAL:.*]] = fir.load %[[Y_DECL]]#0
+! CHECK: %[[Y_DIM0:.*]]:3 = fir.box_dims %[[Y_VAL]], %{{c0_.*}}
+! CHECK: %[[Y_LB:.*]] = arith.constant 0 : index
+! CHECK: %[[Y_UB:.*]] = arith.subi %[[Y_DIM0]]#1, %{{c1_.*}} : index
+! CHECK: %[[Y_BOUNDS:.*]] = omp.map.bounds lower_bound(%[[Y_LB]] : index)
upper_bound(%[[Y_UB]] : index) extent(%[[Y_DIM0]]#1 : index)
+! CHECK: %[[MEM_MAP:.*]] = omp.map.info {{.*}} bounds(%[[Y_BOUNDS]])
+! CHECK: omp.map.info var_ptr(%[[Y_DECL]]#1 : {{.*}}) {{.*}}
members(%[[MEM_MAP]] : {{.*}})
diff --git a/flang/test/Transforms/DoConcurrent/host_eval.f90
b/flang/test/Transforms/DoConcurrent/host_eval.f90
new file mode 100644
index 0..7d16a91ae6941
--- /dev/null
+++ b/flang/test/Transforms/DoConcurrent/host_eval.f90
@@ -0,0 +1,63 @@
+! Tests `host_eval` clause code-gen and loop nest bounds on host vs. device.
+
+! RUN: %flang_fc1 -emit-hlfir -fopenmp -fopenmp-targets=amdgcn-amd-amdhsa \
+! RUN: -fdo-concurrent-to-openmp=device %s -o - \
+! RUN: | FileCheck %s --check-prefix=HOST -vv
+
+! RUN: %flang_fc1 -triple amdgcn-amd-amdhsa -emit-hlfir -fopenmp\
+! RUN: -fopenmp-is-target-device -fdo-concurrent-to-openmp=device %s -o - \
+! RUN: | FileCheck %s --check-prefix=DEVICE
+
+program do_concurrent_host_eval
+implicit none
+integer :: i, j
+
+do concurrent (i=1:10, j=1:20)
+end do
+end program do_concurrent_host_eval
+
+! HOST: omp.target host_eval(
+! HOST-SAME:%{{[^[:space:]]+}} -> %[[I_LB:[^,]+]],
+! HOST-SAME:%{{[^[:space:]]+}} -> %[[I_UB:[^,]+]],
+! HOST-SAME:%{{[^[:space:]]+}} -> %[[I_ST:[^,]+]],
+! HOST-SAME:%{{[^[:space:]]+}} -> %[[J_LB:[^,]+]],
+! HOST-SAME:%{{[^[:space:]]+}} -> %[[J_UB:[^,]+]],
+! HOST-SAME:%{{[^[:space:]]+}} -> %[[J_ST:[^,]+]] : {{.*}}) map_entries
+
+! HOST: omp.loop_nest ({{.*}}, {{.*}}) : index = (%[[I_LB]], %[[J_LB]]) to
+! HOST-SAME:(%[[I_UB]], %[[J_UB]]) inclusive step
+! HOST-SAME:(%[[I_ST]], %[[J_ST]])
+
+! DEVICE: omp.target map_entries(
+! DEVICE-SAME: %{{[^[:space:]]+}} -> %[[I_LB_MAP:[^,]+]],
+! DEVICE-SAME: %{{[^[:space:]]+}} -> %[[I_UB_MAP:[^,]+]],
+! DEVICE-SAME: %{{[^[:space:]]+}} -> %[[I_ST_MAP:[^,]+]],
+
+! DEVICE-SAME: %{{[^[:space:]]+}} -> %[[J_LB_MAP:[^,]+]],
+! DEVICE-SAME: %{{[^[:space:]]+}} -> %[[J_UB_MAP:[^,]+]],
+! DEVICE-SAME: %{{[^[:space:]]+}} -> %[[J_ST_MAP:[^,]+]],
+
+! DEVICE-SAME: %{{[^[:space:]]+}} -> %{{[^,]+}},
+! DEVICE-SAME: %{{[^[:space:]]+}} -> %{{[^,]+}} : {{.*}})
+
+! DEVICE: %[[I_LB_DECL:.*]]:2 = hlfir.declare %[[I_LB_MAP]]
+! DEVICE: %[[I_LB:.*]] = fir.load %[[I_LB_DECL]]#1 : !fir.ref
+
+! DEVICE: %[[I_UB_DECL:.*]]:2 = hlfir.declare %[[I_UB_MAP]
[llvm-branch-commits] [llvm] [flang][do concurent] Add saxpy offload tests for OpenMP mapping (PR #155993)
https://github.com/ergawy updated
https://github.com/llvm/llvm-project/pull/155993
>From f1bbd24a5069458c008736404c000c5334424b46 Mon Sep 17 00:00:00 2001
From: ergawy
Date: Fri, 29 Aug 2025 04:04:07 -0500
Subject: [PATCH] [flang][do concurent] Add saxpy offload tests for OpenMP
mapping
Adds end-to-end tests for `do concurrent` offloading to the device.
---
.../fortran/do-concurrent-to-omp-saxpy-2d.f90 | 53 +++
.../fortran/do-concurrent-to-omp-saxpy.f90| 53 +++
2 files changed, 106 insertions(+)
create mode 100644
offload/test/offloading/fortran/do-concurrent-to-omp-saxpy-2d.f90
create mode 100644
offload/test/offloading/fortran/do-concurrent-to-omp-saxpy.f90
diff --git a/offload/test/offloading/fortran/do-concurrent-to-omp-saxpy-2d.f90
b/offload/test/offloading/fortran/do-concurrent-to-omp-saxpy-2d.f90
new file mode 100644
index 0..c6f576acb90b6
--- /dev/null
+++ b/offload/test/offloading/fortran/do-concurrent-to-omp-saxpy-2d.f90
@@ -0,0 +1,53 @@
+! REQUIRES: flang, amdgpu
+
+! RUN: %libomptarget-compile-fortran-generic -fdo-concurrent-to-openmp=device
+! RUN: env LIBOMPTARGET_INFO=16 %libomptarget-run-generic 2>&1 |
%fcheck-generic
+module saxpymod
+ use iso_fortran_env
+ public :: saxpy
+contains
+
+subroutine saxpy(a, x, y, n, m)
+ use iso_fortran_env
+ implicit none
+ integer,intent(in) :: n, m
+ real(kind=real32),intent(in) :: a
+ real(kind=real32), dimension(:,:),intent(in) :: x
+ real(kind=real32), dimension(:,:),intent(inout) :: y
+ integer :: i, j
+
+ do concurrent(i=1:n, j=1:m)
+ y(i,j) = a * x(i,j) + y(i,j)
+ end do
+
+ write(*,*) "plausibility check:"
+ write(*,'("y(1,1) ",f8.6)') y(1,1)
+ write(*,'("y(n,m) ",f8.6)') y(n,m)
+end subroutine saxpy
+
+end module saxpymod
+
+program main
+ use iso_fortran_env
+ use saxpymod, ONLY:saxpy
+ implicit none
+
+ integer,parameter :: n = 1000, m=1
+ real(kind=real32), allocatable, dimension(:,:) :: x, y
+ real(kind=real32) :: a
+ integer :: i
+
+ allocate(x(1:n,1:m), y(1:n,1:m))
+ a = 2.0_real32
+ x(:,:) = 1.0_real32
+ y(:,:) = 2.0_real32
+
+ call saxpy(a, x, y, n, m)
+
+ deallocate(x,y)
+end program main
+
+! CHECK: "PluginInterface" device {{[0-9]+}} info: Launching kernel {{.*}}
+! CHECK: plausibility check:
+! CHECK: y(1,1) 4.0
+! CHECK: y(n,m) 4.0
diff --git a/offload/test/offloading/fortran/do-concurrent-to-omp-saxpy.f90
b/offload/test/offloading/fortran/do-concurrent-to-omp-saxpy.f90
new file mode 100644
index 0..e094a1d7459ef
--- /dev/null
+++ b/offload/test/offloading/fortran/do-concurrent-to-omp-saxpy.f90
@@ -0,0 +1,53 @@
+! REQUIRES: flang, amdgpu
+
+! RUN: %libomptarget-compile-fortran-generic -fdo-concurrent-to-openmp=device
+! RUN: env LIBOMPTARGET_INFO=16 %libomptarget-run-generic 2>&1 |
%fcheck-generic
+module saxpymod
+ use iso_fortran_env
+ public :: saxpy
+contains
+
+subroutine saxpy(a, x, y, n)
+ use iso_fortran_env
+ implicit none
+ integer,intent(in) :: n
+ real(kind=real32),intent(in) :: a
+ real(kind=real32), dimension(:),intent(in) :: x
+ real(kind=real32), dimension(:),intent(inout) :: y
+ integer :: i
+
+ do concurrent(i=1:n)
+ y(i) = a * x(i) + y(i)
+ end do
+
+ write(*,*) "plausibility check:"
+ write(*,'("y(1) ",f8.6)') y(1)
+ write(*,'("y(n) ",f8.6)') y(n)
+end subroutine saxpy
+
+end module saxpymod
+
+program main
+ use iso_fortran_env
+ use saxpymod, ONLY:saxpy
+ implicit none
+
+ integer,parameter :: n = 1000
+ real(kind=real32), allocatable, dimension(:) :: x, y
+ real(kind=real32) :: a
+ integer :: i
+
+ allocate(x(1:n), y(1:n))
+ a = 2.0_real32
+ x(:) = 1.0_real32
+ y(:) = 2.0_real32
+
+ call saxpy(a, x, y, n)
+
+ deallocate(x,y)
+end program main
+
+! CHECK: "PluginInterface" device {{[0-9]+}} info: Launching kernel {{.*}}
+! CHECK: plausibility check:
+! CHECK: y(1) 4.0
+! CHECK: y(n) 4.0
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] release/21.x: [Clang][CMake] Use IRPGO instead of FE PGO for Cmake Caches (#155957) (PR #156271)
https://github.com/llvmbot created https://github.com/llvm/llvm-project/pull/156271 Backport 7fca1f841b4c226d50ab7bad64de5db225d4193b Requested by: @boomanaiden154 >From 08383fb44eba580749dc2254effc5b1b1f96f4c4 Mon Sep 17 00:00:00 2001 From: Aiden Grossman Date: Fri, 29 Aug 2025 10:06:26 -0700 Subject: [PATCH] [Clang][CMake] Use IRPGO instead of FE PGO for Cmake Caches (#155957) Currently the clang CMake caches use FE PGO for instrumentation (LLVM_BUILD_INSTRUMENTED=ON). However, IRPGO is generally regarded as better for performance. I am measuring about a 1.5% performance gain when building libLLVMSupport.a using this configuration versus what existed before this commit. I would suspect the gains are larger on other platforms like Windows where we cannot subsume any gains using PLO. (cherry picked from commit 7fca1f841b4c226d50ab7bad64de5db225d4193b) --- clang/cmake/caches/PGO.cmake | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/clang/cmake/caches/PGO.cmake b/clang/cmake/caches/PGO.cmake index 15bc755d110d1..d6471160037c1 100644 --- a/clang/cmake/caches/PGO.cmake +++ b/clang/cmake/caches/PGO.cmake @@ -5,7 +5,7 @@ set(LLVM_ENABLE_PROJECTS "clang;lld" CACHE STRING "") set(LLVM_ENABLE_RUNTIMES "compiler-rt;libcxx;libcxxabi;libunwind" CACHE STRING "") set(LLVM_TARGETS_TO_BUILD Native CACHE STRING "") -set(BOOTSTRAP_LLVM_BUILD_INSTRUMENTED ON CACHE BOOL "") +set(BOOTSTRAP_LLVM_BUILD_INSTRUMENTED IR CACHE BOOL "") set(CLANG_BOOTSTRAP_TARGETS generate-profdata stage2 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] release/21.x: [Clang][CMake] Use IRPGO instead of FE PGO for Cmake Caches (#155957) (PR #156271)
https://github.com/llvmbot milestoned https://github.com/llvm/llvm-project/pull/156271 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] release/21.x: [Clang][CMake] Use IRPGO instead of FE PGO for Cmake Caches (#155957) (PR #156271)
llvmbot wrote: @llvm/pr-subscribers-clang Author: None (llvmbot) Changes Backport 7fca1f841b4c226d50ab7bad64de5db225d4193b Requested by: @boomanaiden154 --- Full diff: https://github.com/llvm/llvm-project/pull/156271.diff 1 Files Affected: - (modified) clang/cmake/caches/PGO.cmake (+1-1) ``diff diff --git a/clang/cmake/caches/PGO.cmake b/clang/cmake/caches/PGO.cmake index 15bc755d110d1..d6471160037c1 100644 --- a/clang/cmake/caches/PGO.cmake +++ b/clang/cmake/caches/PGO.cmake @@ -5,7 +5,7 @@ set(LLVM_ENABLE_PROJECTS "clang;lld" CACHE STRING "") set(LLVM_ENABLE_RUNTIMES "compiler-rt;libcxx;libcxxabi;libunwind" CACHE STRING "") set(LLVM_TARGETS_TO_BUILD Native CACHE STRING "") -set(BOOTSTRAP_LLVM_BUILD_INSTRUMENTED ON CACHE BOOL "") +set(BOOTSTRAP_LLVM_BUILD_INSTRUMENTED IR CACHE BOOL "") set(CLANG_BOOTSTRAP_TARGETS generate-profdata stage2 `` https://github.com/llvm/llvm-project/pull/156271 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] release/21.x: [Clang][CMake] Use IRPGO instead of FE PGO for Cmake Caches (#155957) (PR #156271)
llvmbot wrote: @mtrofin What do you think about merging this PR to the release branch? https://github.com/llvm/llvm-project/pull/156271 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang-tools-extra] [clangd] Show type hint for simple cases of dependent 'auto' (PR #156284)
https://github.com/HighCommander4 created
https://github.com/llvm/llvm-project/pull/156284
Fixes https://github.com/clangd/clangd/issues/2275
>From 67b81f57250655c41112b3c3601ca537119df5cc Mon Sep 17 00:00:00 2001
From: Nathan Ridge
Date: Mon, 1 Sep 2025 02:48:44 -0400
Subject: [PATCH] [clangd] Show type hint for simple cases of dependent 'auto'
Fixes https://github.com/clangd/clangd/issues/2275
---
clang-tools-extra/clangd/InlayHints.cpp | 27 ++-
.../clangd/unittests/InlayHintTests.cpp | 3 ++-
2 files changed, 22 insertions(+), 8 deletions(-)
diff --git a/clang-tools-extra/clangd/InlayHints.cpp
b/clang-tools-extra/clangd/InlayHints.cpp
index cd479e1b7c9bc..819e597873d83 100644
--- a/clang-tools-extra/clangd/InlayHints.cpp
+++ b/clang-tools-extra/clangd/InlayHints.cpp
@@ -633,13 +633,26 @@ class InlayHintVisitor : public
RecursiveASTVisitor {
}
if (auto *AT = D->getType()->getContainedAutoType()) {
- if (AT->isDeduced() && !D->getType()->isDependentType()) {
-// Our current approach is to place the hint on the variable
-// and accordingly print the full type
-// (e.g. for `const auto& x = 42`, print `const int&`).
-// Alternatively, we could place the hint on the `auto`
-// (and then just print the type deduced for the `auto`).
-addTypeHint(D->getLocation(), D->getType(), /*Prefix=*/": ");
+ if (AT->isDeduced()) {
+QualType T;
+if (D->getType()->isDependentType()) {
+ if (D->hasInit()) {
+QualType Resolved = Resolver->resolveExprToType(D->getInit());
+if (Resolved != AST.DependentTy) {
+ T = Resolved;
+}
+ }
+} else {
+ T = D->getType();
+}
+if (!T.isNull()) {
+ // Our current approach is to place the hint on the variable
+ // and accordingly print the full type
+ // (e.g. for `const auto& x = 42`, print `const int&`).
+ // Alternatively, we could place the hint on the `auto`
+ // (and then just print the type deduced for the `auto`).
+ addTypeHint(D->getLocation(), T, /*Prefix=*/": ");
+}
}
}
diff --git a/clang-tools-extra/clangd/unittests/InlayHintTests.cpp
b/clang-tools-extra/clangd/unittests/InlayHintTests.cpp
index 99e728c40063d..feb4404b3d2bf 100644
--- a/clang-tools-extra/clangd/unittests/InlayHintTests.cpp
+++ b/clang-tools-extra/clangd/unittests/InlayHintTests.cpp
@@ -1441,7 +1441,8 @@ TEST(TypeHints, DependentType) {
void bar(T arg) {
auto [a, b] = arg;
}
- )cpp");
+ )cpp",
+ ExpectedHint{": T", "var2"});
}
TEST(TypeHints, LongTypeName) {
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
