[llvm-branch-commits] [llvm] release/18.x: [DAGCombiner] Fix miscompile bug in combineShiftOfShiftedLogic (#89616) (PR #89766)

[Björn Pettersson via llvm-branch-commits]

bjope wrote:

> @bjope What do you think about backporting this?

Thanks! That looks/sounds good to me.

(The fault was found when using "-fsanitize=undefined -O0" so it was not as if 
we were using fuzzy pass pipelines etc to trigger the bug. I guess the special 
thing was the use of larger BitInt types being used in the test,. Combining two 
128-bit shifts would't result in a combined shift amount larger than 255, so 
larger types than 128 bits would be needed to expose the problem.)

https://github.com/llvm/llvm-project/pull/89766
___
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] release/18.x: [TableGen] Fix ReplaceRegAction RTTI Kind (PR #89790)

[Pierre van Houtryve via llvm-branch-commits]

Pierre-vh wrote:

We don't use RTTI of that class before #89736 so unless that's also being 
backported for some reason it's not  needed.


https://github.com/llvm/llvm-project/pull/89790
___
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] [libcxx] release/18.x: [libcxx] [modules] Add _LIBCPP_USING_IF_EXISTS on aligned_alloc (#89827) (PR #89894)

[via llvm-branch-commits]

https://github.com/llvmbot created 
https://github.com/llvm/llvm-project/pull/89894

Backport 91526d64a8adb14edc55adfd5270858791822837

Requested by: @mstorsjo

>From 7fc76537bb8097ab98abe0b83479d4aeff36b28c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= 
Date: Wed, 24 Apr 2024 11:45:27 +0300
Subject: [PATCH] =?UTF-8?q?[libcxx]=20[modules]=C2=A0Add=20=5FLIBCPP=5FUSI?=
 =?UTF-8?q?NG=5FIF=5FEXISTS=20on=20aligned=5Falloc=20(#89827)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

This is missing e.g. on Windows. With this change, it's possible to make
the libcxx std module work on mingw-w64 (although that requires a few
fixes to those headers).

In the regular cstdlib header, we have _LIBCPP_USING_IF_EXISTS flagged
on every single reexported function (since
a9c9183ca42629fa83cdda297d1d30c7bc1d7c91), but the modules seem to only
have _LIBCPP_USING_IF_EXISTS set on a few individual functions, so far.

(cherry picked from commit 91526d64a8adb14edc55adfd5270858791822837)
---
 libcxx/modules/std.compat/cstdlib.inc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libcxx/modules/std.compat/cstdlib.inc 
b/libcxx/modules/std.compat/cstdlib.inc
index a45a0a1caf8ba9..4783cbf5162390 100644
--- a/libcxx/modules/std.compat/cstdlib.inc
+++ b/libcxx/modules/std.compat/cstdlib.inc
@@ -25,7 +25,7 @@ export {
   using ::system;
 
   // [c.malloc], C library memory allocation
-  using ::aligned_alloc;
+  using ::aligned_alloc _LIBCPP_USING_IF_EXISTS;
   using ::calloc;
   using ::free;
   using ::malloc;

___
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] [libcxx] release/18.x: [libcxx] [modules] Add _LIBCPP_USING_IF_EXISTS on aligned_alloc (#89827) (PR #89894)

[via llvm-branch-commits]

https://github.com/llvmbot milestoned 
https://github.com/llvm/llvm-project/pull/89894
___
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] [libcxx] release/18.x: [libcxx] [modules] Add _LIBCPP_USING_IF_EXISTS on aligned_alloc (#89827) (PR #89894)

[via llvm-branch-commits]

llvmbot wrote:

@mordante What do you think about merging this PR to the release branch?

https://github.com/llvm/llvm-project/pull/89894
___
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] [libcxx] release/18.x: [libcxx] [modules] Add _LIBCPP_USING_IF_EXISTS on aligned_alloc (#89827) (PR #89894)

[via llvm-branch-commits]

llvmbot wrote:




@llvm/pr-subscribers-libcxx

Author: None (llvmbot)


Changes

Backport 91526d64a8adb14edc55adfd5270858791822837

Requested by: @mstorsjo

---
Full diff: https://github.com/llvm/llvm-project/pull/89894.diff


1 Files Affected:

- (modified) libcxx/modules/std.compat/cstdlib.inc (+1-1) 


``diff
diff --git a/libcxx/modules/std.compat/cstdlib.inc 
b/libcxx/modules/std.compat/cstdlib.inc
index a45a0a1caf8ba9..4783cbf5162390 100644
--- a/libcxx/modules/std.compat/cstdlib.inc
+++ b/libcxx/modules/std.compat/cstdlib.inc
@@ -25,7 +25,7 @@ export {
   using ::system;
 
   // [c.malloc], C library memory allocation
-  using ::aligned_alloc;
+  using ::aligned_alloc _LIBCPP_USING_IF_EXISTS;
   using ::calloc;
   using ::free;
   using ::malloc;

``




https://github.com/llvm/llvm-project/pull/89894
___
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] [OpenMP][MLIR] Extend record member map support for omp dialect to LLVM-IR (PR #82852)

[Sergio Afonso via llvm-branch-commits]

@@ -2081,6 +2083,79 @@ void collectMapDataFromMapOperands(MapInfoData &mapData,
   }
 }
 
+static int getMapDataMemberIdx(MapInfoData &mapData,
+   mlir::omp::MapInfoOp memberOp) {
+  auto *res = llvm::find(mapData.MapClause, memberOp);
+  assert(res != mapData.MapClause.end());
+  return std::distance(mapData.MapClause.begin(), res);
+}
+
+static mlir::omp::MapInfoOp
+getFirstOrLastMappedMemberPtr(mlir::omp::MapInfoOp mapInfo, bool first) {
+  // Only 1 member has been mapped, we can return it.
+  if (mapInfo.getMembersIndex()->size() == 1)
+if (auto mapOp = mlir::dyn_cast(
+mapInfo.getMembers()[0].getDefiningOp()))
+  return mapOp;
+
+  std::vector indices(
+  mapInfo.getMembersIndexAttr().getShapedType().getShape()[0]);
+  std::iota(indices.begin(), indices.end(), 0);

skatrak wrote:

Disregard this suggestion, it's not the same. Sorry about that!

https://github.com/llvm/llvm-project/pull/82852
___
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] [OpenMP][MLIR] Extend record member map support for omp dialect to LLVM-IR (PR #82852)

[Sergio Afonso via llvm-branch-commits]

https://github.com/skatrak edited 
https://github.com/llvm/llvm-project/pull/82852
___
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] [mlir] [MLIR][OpenMP] Update op verifiers dependent on omp.wsloop (2/5) (PR #89211)

[Sergio Afonso via llvm-branch-commits]

https://github.com/skatrak updated 
https://github.com/llvm/llvm-project/pull/89211

>From f9b14e37a6f437768c405291c064f541f0655b1c Mon Sep 17 00:00:00 2001
From: Sergio Afonso 
Date: Wed, 17 Apr 2024 16:40:03 +0100
Subject: [PATCH 1/2] [MLIR][OpenMP] Update op verifiers dependent on
 omp.wsloop (2/5)

This patch updates verifiers for `omp.ordered.region`, `omp.cancel` and
`omp.cancellation_point`, which check for a parent `omp.wsloop`.

After transitioning to a loop wrapper-based approach, the expected direct
parent will become `omp.loop_nest` instead, so verifiers need to take this into
account.

This PR on its own will not pass premerge tests. All patches in the stack are
needed before it can be compiled and passes tests.
---
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp | 23 
 1 file changed, 14 insertions(+), 9 deletions(-)

diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp 
b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index d66186effc31d6..d014c27e1ee157 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1977,9 +1977,10 @@ LogicalResult OrderedRegionOp::verify() {
   if (getSimd())
 return failure();
 
-  if (auto container = (*this)->getParentOfType()) {
-if (!container.getOrderedValAttr() ||
-container.getOrderedValAttr().getInt() != 0)
+  if (auto loopOp = dyn_cast((*this)->getParentOp())) {
+auto wsloopOp = llvm::dyn_cast_if_present(loopOp->getParentOp());
+if (!wsloopOp || !wsloopOp.getOrderedValAttr() ||
+wsloopOp.getOrderedValAttr().getInt() != 0)
   return emitOpError() << "ordered region must be closely nested inside "
<< "a worksharing-loop region with an ordered "
<< "clause without parameter present";
@@ -2130,15 +2131,19 @@ LogicalResult CancelOp::verify() {
  << "inside a parallel region";
   }
   if (cct == ClauseCancellationConstructType::Loop) {
-if (!isa(parentOp)) {
-  return emitOpError() << "cancel loop must appear "
-   << "inside a worksharing-loop region";
+auto loopOp = dyn_cast(parentOp);
+auto wsloopOp = llvm::dyn_cast_if_present(
+loopOp ? loopOp->getParentOp() : nullptr);
+
+if (!wsloopOp) {
+  return emitOpError()
+ << "cancel loop must appear inside a worksharing-loop region";
 }
-if (cast(parentOp).getNowaitAttr()) {
+if (wsloopOp.getNowaitAttr()) {
   return emitError() << "A worksharing construct that is canceled "
  << "must not have a nowait clause";
 }
-if (cast(parentOp).getOrderedValAttr()) {
+if (wsloopOp.getOrderedValAttr()) {
   return emitError() << "A worksharing construct that is canceled "
  << "must not have an ordered clause";
 }
@@ -2176,7 +2181,7 @@ LogicalResult CancellationPointOp::verify() {
  << "inside a parallel region";
   }
   if ((cct == ClauseCancellationConstructType::Loop) &&
-  !isa(parentOp)) {
+  (!isa(parentOp) || !isa(parentOp->getParentOp( 
{
 return emitOpError() << "cancellation point loop must appear "
  << "inside a worksharing-loop region";
   }

>From 18c8bda112d60824c74904d8c27a16f7f016c020 Mon Sep 17 00:00:00 2001
From: Sergio Afonso 
Date: Fri, 19 Apr 2024 11:37:21 +0100
Subject: [PATCH 2/2] Address review comment, improve tests

---
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp | 57 +++--
 mlir/test/Dialect/OpenMP/invalid.mlir| 67 
 2 files changed, 94 insertions(+), 30 deletions(-)

diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp 
b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index d014c27e1ee157..c546d3a9044de1 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1945,6 +1945,39 @@ LogicalResult 
CriticalOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
 // Ordered construct
 
//===--===//
 
+static LogicalResult verifyOrderedParent(Operation &op) {
+  bool hasRegion = op.getNumRegions() > 0;
+  auto loopOp = op.getParentOfType();
+  if (!loopOp) {
+if (hasRegion)
+  return success();
+
+// TODO: Consider if this needs to be the case only for the standalone
+// variant of the ordered construct.
+return op.emitOpError() << "must be nested inside of a loop";
+  }
+
+  Operation *wrapper = loopOp->getParentOp();
+  if (auto wsloopOp = dyn_cast(wrapper)) {
+IntegerAttr orderedAttr = wsloopOp.getOrderedValAttr();
+if (!orderedAttr)
+  return op.emitOpError() << "the enclosing worksharing-loop region must "
+ "have an ordered clause";
+
+if (hasRegion && orderedAttr.getInt() != 0)
+  return op.emitOpError() << "the enclosing loop's ordered clause must not 

[llvm-branch-commits] [mlir] [MLIR][SCF] Update scf.parallel lowering to OpenMP (3/5) (PR #89212)

[Sergio Afonso via llvm-branch-commits]

https://github.com/skatrak updated 
https://github.com/llvm/llvm-project/pull/89212

>From fdee8cf17cd7d2dbdb6cf872574776f02e70be7c Mon Sep 17 00:00:00 2001
From: Sergio Afonso 
Date: Thu, 18 Apr 2024 10:55:16 +0100
Subject: [PATCH 1/2] [MLIR][SCF] Update scf.parallel lowering to OpenMP (3/5)

This patch makes changes to the `scf.parallel` to `omp.parallel` + `omp.wsloop`
lowering pass in order to introduce a nested `omp.loop_nest` as well, and to
follow the new loop wrapper role for `omp.wsloop`.

This PR on its own will not pass premerge tests. All patches in the stack are
needed before it can be compiled and passes tests.
---
 .../Conversion/SCFToOpenMP/SCFToOpenMP.cpp| 52 ++-
 .../Conversion/SCFToOpenMP/reductions.mlir|  5 ++
 .../Conversion/SCFToOpenMP/scf-to-openmp.mlir | 31 ---
 3 files changed, 68 insertions(+), 20 deletions(-)

diff --git a/mlir/lib/Conversion/SCFToOpenMP/SCFToOpenMP.cpp 
b/mlir/lib/Conversion/SCFToOpenMP/SCFToOpenMP.cpp
index 7f91367ad427a2..f0b8d6c5309357 100644
--- a/mlir/lib/Conversion/SCFToOpenMP/SCFToOpenMP.cpp
+++ b/mlir/lib/Conversion/SCFToOpenMP/SCFToOpenMP.cpp
@@ -461,18 +461,51 @@ struct ParallelOpLowering : public 
OpRewritePattern {
   // Replace the loop.
   {
 OpBuilder::InsertionGuard allocaGuard(rewriter);
-auto loop = rewriter.create(
+// Create worksharing loop wrapper.
+auto wsloopOp = rewriter.create(parallelOp.getLoc());
+if (!reductionVariables.empty()) {
+  wsloopOp.setReductionsAttr(
+  ArrayAttr::get(rewriter.getContext(), reductionDeclSymbols));
+  wsloopOp.getReductionVarsMutable().append(reductionVariables);
+}
+rewriter.create(loc); // omp.parallel terminator.
+
+// The wrapper's entry block arguments will define the reduction
+// variables.
+llvm::SmallVector reductionTypes;
+reductionTypes.reserve(reductionVariables.size());
+llvm::transform(reductionVariables, std::back_inserter(reductionTypes),
+[](mlir::Value v) { return v.getType(); });
+rewriter.createBlock(
+&wsloopOp.getRegion(), {}, reductionTypes,
+llvm::SmallVector(reductionVariables.size(),
+  parallelOp.getLoc()));
+
+rewriter.setInsertionPoint(
+rewriter.create(parallelOp.getLoc()));
+
+// Create loop nest and populate region with contents of scf.parallel.
+auto loopOp = rewriter.create(
 parallelOp.getLoc(), parallelOp.getLowerBound(),
 parallelOp.getUpperBound(), parallelOp.getStep());
-rewriter.create(loc);
 
-rewriter.inlineRegionBefore(parallelOp.getRegion(), loop.getRegion(),
-loop.getRegion().begin());
+rewriter.inlineRegionBefore(parallelOp.getRegion(), loopOp.getRegion(),
+loopOp.getRegion().begin());
+
+// Remove reduction-related block arguments from omp.loop_nest and
+// redirect uses to the corresponding omp.wsloop block argument.
+mlir::Block &loopOpEntryBlock = loopOp.getRegion().front();
+unsigned numLoops = parallelOp.getNumLoops();
+rewriter.replaceAllUsesWith(
+loopOpEntryBlock.getArguments().drop_front(numLoops),
+wsloopOp.getRegion().getArguments());
+loopOpEntryBlock.eraseArguments(
+numLoops, loopOpEntryBlock.getNumArguments() - numLoops);
 
-Block *ops = rewriter.splitBlock(&*loop.getRegion().begin(),
- loop.getRegion().begin()->begin());
+Block *ops = rewriter.splitBlock(&*loopOp.getRegion().begin(),
+ loopOp.getRegion().begin()->begin());
 
-rewriter.setInsertionPointToStart(&*loop.getRegion().begin());
+rewriter.setInsertionPointToStart(&*loopOp.getRegion().begin());
 
 auto scope = 
rewriter.create(parallelOp.getLoc(),
 TypeRange());
@@ -481,11 +514,6 @@ struct ParallelOpLowering : public 
OpRewritePattern {
 rewriter.mergeBlocks(ops, scopeBlock);
 rewriter.setInsertionPointToEnd(&*scope.getBodyRegion().begin());
 rewriter.create(loc, ValueRange());
-if (!reductionVariables.empty()) {
-  loop.setReductionsAttr(
-  ArrayAttr::get(rewriter.getContext(), reductionDeclSymbols));
-  loop.getReductionVarsMutable().append(reductionVariables);
-}
   }
 }
 
diff --git a/mlir/test/Conversion/SCFToOpenMP/reductions.mlir 
b/mlir/test/Conversion/SCFToOpenMP/reductions.mlir
index 3b6c145d62f1a8..fc6d56559c2618 100644
--- a/mlir/test/Conversion/SCFToOpenMP/reductions.mlir
+++ b/mlir/test/Conversion/SCFToOpenMP/reductions.mlir
@@ -28,6 +28,7 @@ func.func @reduction1(%arg0 : index, %arg1 : index, %arg2 : 
index,
   // CHECK: omp.

[llvm-branch-commits] [mlir] [MLIR][OpenMP] Update omp.wsloop translation to LLVM IR (4/5) (PR #89214)

[Sergio Afonso via llvm-branch-commits]

https://github.com/skatrak updated 
https://github.com/llvm/llvm-project/pull/89214

>From 25dc3a45645ab2310606b2ab02346eed700c7f97 Mon Sep 17 00:00:00 2001
From: Sergio Afonso 
Date: Thu, 18 Apr 2024 11:07:10 +0100
Subject: [PATCH] [MLIR][OpenMP] Update omp.wsloop translation to LLVM IR (4/5)

This patch introduces minimal changes to the MLIR to LLVM IR translation of
omp.wsloop to support the loop wrapper approach.

There is `omp.loop_nest` related translation code that should be extracted and
shared among all loop operations (e.g. `omp.simd`). This would possibly also
help in the addition of support for compound constructs later on. This first
approach is only intended to keep things running after the transition to loop
wrappers and not to add support for other use cases enabled by that transition.

This PR on its own will not pass premerge tests. All patches in the stack are
needed before it can be compiled and passes tests.
---
 .../OpenMP/OpenMPToLLVMIRTranslation.cpp  |  68 +-
 .../LLVMIR/omptarget-parallel-wsloop.mlir |  11 +-
 .../LLVMIR/omptarget-wsloop-collapsed.mlir|  17 +-
 mlir/test/Target/LLVMIR/omptarget-wsloop.mlir |  18 +-
 .../LLVMIR/openmp-data-target-device.mlir |  31 +-
 mlir/test/Target/LLVMIR/openmp-llvm.mlir  | 741 ++
 mlir/test/Target/LLVMIR/openmp-nested.mlir|  30 +-
 mlir/test/Target/LLVMIR/openmp-reduction.mlir | 113 +--
 .../openmp-wsloop-reduction-cleanup.mlir  |   9 +-
 9 files changed, 572 insertions(+), 466 deletions(-)

diff --git 
a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp 
b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index e89ff9209b034a..22d6462b881dc0 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -916,35 +916,37 @@ static LogicalResult inlineReductionCleanup(
 static LogicalResult
 convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase &builder,
  LLVM::ModuleTranslation &moduleTranslation) {
-  auto loop = cast(opInst);
-  const bool isByRef = loop.getByref();
+  auto wsloopOp = cast(opInst);
+  auto loopOp = cast(wsloopOp.getWrappedLoop());
+  const bool isByRef = wsloopOp.getByref();
+
   // TODO: this should be in the op verifier instead.
-  if (loop.getLowerBound().empty())
+  if (loopOp.getLowerBound().empty())
 return failure();
 
   // Static is the default.
   auto schedule =
-  loop.getScheduleVal().value_or(omp::ClauseScheduleKind::Static);
+  wsloopOp.getScheduleVal().value_or(omp::ClauseScheduleKind::Static);
 
   // Find the loop configuration.
-  llvm::Value *step = moduleTranslation.lookupValue(loop.getStep()[0]);
+  llvm::Value *step = moduleTranslation.lookupValue(loopOp.getStep()[0]);
   llvm::Type *ivType = step->getType();
   llvm::Value *chunk = nullptr;
-  if (loop.getScheduleChunkVar()) {
+  if (wsloopOp.getScheduleChunkVar()) {
 llvm::Value *chunkVar =
-moduleTranslation.lookupValue(loop.getScheduleChunkVar());
+moduleTranslation.lookupValue(wsloopOp.getScheduleChunkVar());
 chunk = builder.CreateSExtOrTrunc(chunkVar, ivType);
   }
 
   SmallVector reductionDecls;
-  collectReductionDecls(loop, reductionDecls);
+  collectReductionDecls(wsloopOp, reductionDecls);
   llvm::OpenMPIRBuilder::InsertPointTy allocaIP =
   findAllocaInsertPoint(builder, moduleTranslation);
 
   SmallVector privateReductionVariables;
   DenseMap reductionVariableMap;
   if (!isByRef) {
-allocByValReductionVars(loop, builder, moduleTranslation, allocaIP,
+allocByValReductionVars(wsloopOp, builder, moduleTranslation, allocaIP,
 reductionDecls, privateReductionVariables,
 reductionVariableMap);
   }
@@ -952,13 +954,12 @@ convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase 
&builder,
   // Before the loop, store the initial values of reductions into reduction
   // variables. Although this could be done after allocas, we don't want to 
mess
   // up with the alloca insertion point.
-  MutableArrayRef reductionArgs =
-  loop.getRegion().getArguments().take_back(loop.getNumReductionVars());
-  for (unsigned i = 0; i < loop.getNumReductionVars(); ++i) {
+  ArrayRef reductionArgs = wsloopOp.getRegion().getArguments();
+  for (unsigned i = 0; i < wsloopOp.getNumReductionVars(); ++i) {
 SmallVector phis;
 
 // map block argument to initializer region
-mapInitializationArg(loop, moduleTranslation, reductionDecls, i);
+mapInitializationArg(wsloopOp, moduleTranslation, reductionDecls, i);
 
 if 
(failed(inlineConvertOmpRegions(reductionDecls[i].getInitializerRegion(),
"omp.reduction.neutral", builder,
@@ -977,7 +978,7 @@ convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase 
&builder,
 
   privateReductionVariables.push_back(var);
   moduleTranslation.mapValue(reductionArgs[i], phi

[llvm-branch-commits] [mlir] [MLIR][OpenMP] Update op verifiers dependent on omp.wsloop (2/5) (PR #89211)

[Sergio Afonso via llvm-branch-commits]

https://github.com/skatrak updated 
https://github.com/llvm/llvm-project/pull/89211

>From f9b14e37a6f437768c405291c064f541f0655b1c Mon Sep 17 00:00:00 2001
From: Sergio Afonso 
Date: Wed, 17 Apr 2024 16:40:03 +0100
Subject: [PATCH 1/2] [MLIR][OpenMP] Update op verifiers dependent on
 omp.wsloop (2/5)

This patch updates verifiers for `omp.ordered.region`, `omp.cancel` and
`omp.cancellation_point`, which check for a parent `omp.wsloop`.

After transitioning to a loop wrapper-based approach, the expected direct
parent will become `omp.loop_nest` instead, so verifiers need to take this into
account.

This PR on its own will not pass premerge tests. All patches in the stack are
needed before it can be compiled and passes tests.
---
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp | 23 
 1 file changed, 14 insertions(+), 9 deletions(-)

diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp 
b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index d66186effc31d6..d014c27e1ee157 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1977,9 +1977,10 @@ LogicalResult OrderedRegionOp::verify() {
   if (getSimd())
 return failure();
 
-  if (auto container = (*this)->getParentOfType()) {
-if (!container.getOrderedValAttr() ||
-container.getOrderedValAttr().getInt() != 0)
+  if (auto loopOp = dyn_cast((*this)->getParentOp())) {
+auto wsloopOp = llvm::dyn_cast_if_present(loopOp->getParentOp());
+if (!wsloopOp || !wsloopOp.getOrderedValAttr() ||
+wsloopOp.getOrderedValAttr().getInt() != 0)
   return emitOpError() << "ordered region must be closely nested inside "
<< "a worksharing-loop region with an ordered "
<< "clause without parameter present";
@@ -2130,15 +2131,19 @@ LogicalResult CancelOp::verify() {
  << "inside a parallel region";
   }
   if (cct == ClauseCancellationConstructType::Loop) {
-if (!isa(parentOp)) {
-  return emitOpError() << "cancel loop must appear "
-   << "inside a worksharing-loop region";
+auto loopOp = dyn_cast(parentOp);
+auto wsloopOp = llvm::dyn_cast_if_present(
+loopOp ? loopOp->getParentOp() : nullptr);
+
+if (!wsloopOp) {
+  return emitOpError()
+ << "cancel loop must appear inside a worksharing-loop region";
 }
-if (cast(parentOp).getNowaitAttr()) {
+if (wsloopOp.getNowaitAttr()) {
   return emitError() << "A worksharing construct that is canceled "
  << "must not have a nowait clause";
 }
-if (cast(parentOp).getOrderedValAttr()) {
+if (wsloopOp.getOrderedValAttr()) {
   return emitError() << "A worksharing construct that is canceled "
  << "must not have an ordered clause";
 }
@@ -2176,7 +2181,7 @@ LogicalResult CancellationPointOp::verify() {
  << "inside a parallel region";
   }
   if ((cct == ClauseCancellationConstructType::Loop) &&
-  !isa(parentOp)) {
+  (!isa(parentOp) || !isa(parentOp->getParentOp( 
{
 return emitOpError() << "cancellation point loop must appear "
  << "inside a worksharing-loop region";
   }

>From 18c8bda112d60824c74904d8c27a16f7f016c020 Mon Sep 17 00:00:00 2001
From: Sergio Afonso 
Date: Fri, 19 Apr 2024 11:37:21 +0100
Subject: [PATCH 2/2] Address review comment, improve tests

---
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp | 57 +++--
 mlir/test/Dialect/OpenMP/invalid.mlir| 67 
 2 files changed, 94 insertions(+), 30 deletions(-)

diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp 
b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index d014c27e1ee157..c546d3a9044de1 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1945,6 +1945,39 @@ LogicalResult 
CriticalOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
 // Ordered construct
 
//===--===//
 
+static LogicalResult verifyOrderedParent(Operation &op) {
+  bool hasRegion = op.getNumRegions() > 0;
+  auto loopOp = op.getParentOfType();
+  if (!loopOp) {
+if (hasRegion)
+  return success();
+
+// TODO: Consider if this needs to be the case only for the standalone
+// variant of the ordered construct.
+return op.emitOpError() << "must be nested inside of a loop";
+  }
+
+  Operation *wrapper = loopOp->getParentOp();
+  if (auto wsloopOp = dyn_cast(wrapper)) {
+IntegerAttr orderedAttr = wsloopOp.getOrderedValAttr();
+if (!orderedAttr)
+  return op.emitOpError() << "the enclosing worksharing-loop region must "
+ "have an ordered clause";
+
+if (hasRegion && orderedAttr.getInt() != 0)
+  return op.emitOpError() << "the enclosing loop's ordered clause must not 

[llvm-branch-commits] [mlir] [MLIR][SCF] Update scf.parallel lowering to OpenMP (3/5) (PR #89212)

[Sergio Afonso via llvm-branch-commits]

https://github.com/skatrak updated 
https://github.com/llvm/llvm-project/pull/89212

>From fdee8cf17cd7d2dbdb6cf872574776f02e70be7c Mon Sep 17 00:00:00 2001
From: Sergio Afonso 
Date: Thu, 18 Apr 2024 10:55:16 +0100
Subject: [PATCH 1/2] [MLIR][SCF] Update scf.parallel lowering to OpenMP (3/5)

This patch makes changes to the `scf.parallel` to `omp.parallel` + `omp.wsloop`
lowering pass in order to introduce a nested `omp.loop_nest` as well, and to
follow the new loop wrapper role for `omp.wsloop`.

This PR on its own will not pass premerge tests. All patches in the stack are
needed before it can be compiled and passes tests.
---
 .../Conversion/SCFToOpenMP/SCFToOpenMP.cpp| 52 ++-
 .../Conversion/SCFToOpenMP/reductions.mlir|  5 ++
 .../Conversion/SCFToOpenMP/scf-to-openmp.mlir | 31 ---
 3 files changed, 68 insertions(+), 20 deletions(-)

diff --git a/mlir/lib/Conversion/SCFToOpenMP/SCFToOpenMP.cpp 
b/mlir/lib/Conversion/SCFToOpenMP/SCFToOpenMP.cpp
index 7f91367ad427a2..f0b8d6c5309357 100644
--- a/mlir/lib/Conversion/SCFToOpenMP/SCFToOpenMP.cpp
+++ b/mlir/lib/Conversion/SCFToOpenMP/SCFToOpenMP.cpp
@@ -461,18 +461,51 @@ struct ParallelOpLowering : public 
OpRewritePattern {
   // Replace the loop.
   {
 OpBuilder::InsertionGuard allocaGuard(rewriter);
-auto loop = rewriter.create(
+// Create worksharing loop wrapper.
+auto wsloopOp = rewriter.create(parallelOp.getLoc());
+if (!reductionVariables.empty()) {
+  wsloopOp.setReductionsAttr(
+  ArrayAttr::get(rewriter.getContext(), reductionDeclSymbols));
+  wsloopOp.getReductionVarsMutable().append(reductionVariables);
+}
+rewriter.create(loc); // omp.parallel terminator.
+
+// The wrapper's entry block arguments will define the reduction
+// variables.
+llvm::SmallVector reductionTypes;
+reductionTypes.reserve(reductionVariables.size());
+llvm::transform(reductionVariables, std::back_inserter(reductionTypes),
+[](mlir::Value v) { return v.getType(); });
+rewriter.createBlock(
+&wsloopOp.getRegion(), {}, reductionTypes,
+llvm::SmallVector(reductionVariables.size(),
+  parallelOp.getLoc()));
+
+rewriter.setInsertionPoint(
+rewriter.create(parallelOp.getLoc()));
+
+// Create loop nest and populate region with contents of scf.parallel.
+auto loopOp = rewriter.create(
 parallelOp.getLoc(), parallelOp.getLowerBound(),
 parallelOp.getUpperBound(), parallelOp.getStep());
-rewriter.create(loc);
 
-rewriter.inlineRegionBefore(parallelOp.getRegion(), loop.getRegion(),
-loop.getRegion().begin());
+rewriter.inlineRegionBefore(parallelOp.getRegion(), loopOp.getRegion(),
+loopOp.getRegion().begin());
+
+// Remove reduction-related block arguments from omp.loop_nest and
+// redirect uses to the corresponding omp.wsloop block argument.
+mlir::Block &loopOpEntryBlock = loopOp.getRegion().front();
+unsigned numLoops = parallelOp.getNumLoops();
+rewriter.replaceAllUsesWith(
+loopOpEntryBlock.getArguments().drop_front(numLoops),
+wsloopOp.getRegion().getArguments());
+loopOpEntryBlock.eraseArguments(
+numLoops, loopOpEntryBlock.getNumArguments() - numLoops);
 
-Block *ops = rewriter.splitBlock(&*loop.getRegion().begin(),
- loop.getRegion().begin()->begin());
+Block *ops = rewriter.splitBlock(&*loopOp.getRegion().begin(),
+ loopOp.getRegion().begin()->begin());
 
-rewriter.setInsertionPointToStart(&*loop.getRegion().begin());
+rewriter.setInsertionPointToStart(&*loopOp.getRegion().begin());
 
 auto scope = 
rewriter.create(parallelOp.getLoc(),
 TypeRange());
@@ -481,11 +514,6 @@ struct ParallelOpLowering : public 
OpRewritePattern {
 rewriter.mergeBlocks(ops, scopeBlock);
 rewriter.setInsertionPointToEnd(&*scope.getBodyRegion().begin());
 rewriter.create(loc, ValueRange());
-if (!reductionVariables.empty()) {
-  loop.setReductionsAttr(
-  ArrayAttr::get(rewriter.getContext(), reductionDeclSymbols));
-  loop.getReductionVarsMutable().append(reductionVariables);
-}
   }
 }
 
diff --git a/mlir/test/Conversion/SCFToOpenMP/reductions.mlir 
b/mlir/test/Conversion/SCFToOpenMP/reductions.mlir
index 3b6c145d62f1a8..fc6d56559c2618 100644
--- a/mlir/test/Conversion/SCFToOpenMP/reductions.mlir
+++ b/mlir/test/Conversion/SCFToOpenMP/reductions.mlir
@@ -28,6 +28,7 @@ func.func @reduction1(%arg0 : index, %arg1 : index, %arg2 : 
index,
   // CHECK: omp.

[llvm-branch-commits] [mlir] [MLIR][OpenMP] Update omp.wsloop translation to LLVM IR (4/5) (PR #89214)

[Sergio Afonso via llvm-branch-commits]

https://github.com/skatrak updated 
https://github.com/llvm/llvm-project/pull/89214

>From 25dc3a45645ab2310606b2ab02346eed700c7f97 Mon Sep 17 00:00:00 2001
From: Sergio Afonso 
Date: Thu, 18 Apr 2024 11:07:10 +0100
Subject: [PATCH] [MLIR][OpenMP] Update omp.wsloop translation to LLVM IR (4/5)

This patch introduces minimal changes to the MLIR to LLVM IR translation of
omp.wsloop to support the loop wrapper approach.

There is `omp.loop_nest` related translation code that should be extracted and
shared among all loop operations (e.g. `omp.simd`). This would possibly also
help in the addition of support for compound constructs later on. This first
approach is only intended to keep things running after the transition to loop
wrappers and not to add support for other use cases enabled by that transition.

This PR on its own will not pass premerge tests. All patches in the stack are
needed before it can be compiled and passes tests.
---
 .../OpenMP/OpenMPToLLVMIRTranslation.cpp  |  68 +-
 .../LLVMIR/omptarget-parallel-wsloop.mlir |  11 +-
 .../LLVMIR/omptarget-wsloop-collapsed.mlir|  17 +-
 mlir/test/Target/LLVMIR/omptarget-wsloop.mlir |  18 +-
 .../LLVMIR/openmp-data-target-device.mlir |  31 +-
 mlir/test/Target/LLVMIR/openmp-llvm.mlir  | 741 ++
 mlir/test/Target/LLVMIR/openmp-nested.mlir|  30 +-
 mlir/test/Target/LLVMIR/openmp-reduction.mlir | 113 +--
 .../openmp-wsloop-reduction-cleanup.mlir  |   9 +-
 9 files changed, 572 insertions(+), 466 deletions(-)

diff --git 
a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp 
b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index e89ff9209b034a..22d6462b881dc0 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -916,35 +916,37 @@ static LogicalResult inlineReductionCleanup(
 static LogicalResult
 convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase &builder,
  LLVM::ModuleTranslation &moduleTranslation) {
-  auto loop = cast(opInst);
-  const bool isByRef = loop.getByref();
+  auto wsloopOp = cast(opInst);
+  auto loopOp = cast(wsloopOp.getWrappedLoop());
+  const bool isByRef = wsloopOp.getByref();
+
   // TODO: this should be in the op verifier instead.
-  if (loop.getLowerBound().empty())
+  if (loopOp.getLowerBound().empty())
 return failure();
 
   // Static is the default.
   auto schedule =
-  loop.getScheduleVal().value_or(omp::ClauseScheduleKind::Static);
+  wsloopOp.getScheduleVal().value_or(omp::ClauseScheduleKind::Static);
 
   // Find the loop configuration.
-  llvm::Value *step = moduleTranslation.lookupValue(loop.getStep()[0]);
+  llvm::Value *step = moduleTranslation.lookupValue(loopOp.getStep()[0]);
   llvm::Type *ivType = step->getType();
   llvm::Value *chunk = nullptr;
-  if (loop.getScheduleChunkVar()) {
+  if (wsloopOp.getScheduleChunkVar()) {
 llvm::Value *chunkVar =
-moduleTranslation.lookupValue(loop.getScheduleChunkVar());
+moduleTranslation.lookupValue(wsloopOp.getScheduleChunkVar());
 chunk = builder.CreateSExtOrTrunc(chunkVar, ivType);
   }
 
   SmallVector reductionDecls;
-  collectReductionDecls(loop, reductionDecls);
+  collectReductionDecls(wsloopOp, reductionDecls);
   llvm::OpenMPIRBuilder::InsertPointTy allocaIP =
   findAllocaInsertPoint(builder, moduleTranslation);
 
   SmallVector privateReductionVariables;
   DenseMap reductionVariableMap;
   if (!isByRef) {
-allocByValReductionVars(loop, builder, moduleTranslation, allocaIP,
+allocByValReductionVars(wsloopOp, builder, moduleTranslation, allocaIP,
 reductionDecls, privateReductionVariables,
 reductionVariableMap);
   }
@@ -952,13 +954,12 @@ convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase 
&builder,
   // Before the loop, store the initial values of reductions into reduction
   // variables. Although this could be done after allocas, we don't want to 
mess
   // up with the alloca insertion point.
-  MutableArrayRef reductionArgs =
-  loop.getRegion().getArguments().take_back(loop.getNumReductionVars());
-  for (unsigned i = 0; i < loop.getNumReductionVars(); ++i) {
+  ArrayRef reductionArgs = wsloopOp.getRegion().getArguments();
+  for (unsigned i = 0; i < wsloopOp.getNumReductionVars(); ++i) {
 SmallVector phis;
 
 // map block argument to initializer region
-mapInitializationArg(loop, moduleTranslation, reductionDecls, i);
+mapInitializationArg(wsloopOp, moduleTranslation, reductionDecls, i);
 
 if 
(failed(inlineConvertOmpRegions(reductionDecls[i].getInitializerRegion(),
"omp.reduction.neutral", builder,
@@ -977,7 +978,7 @@ convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase 
&builder,
 
   privateReductionVariables.push_back(var);
   moduleTranslation.mapValue(reductionArgs[i], phi

[llvm-branch-commits] [llvm] [bolt] Revise IDE folder structure (PR #89742)

[Michael Kruse via llvm-branch-commits]

https://github.com/Meinersbur ready_for_review 
https://github.com/llvm/llvm-project/pull/89742
___
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] [clang] Revise IDE folder structure (PR #89743)

[Michael Kruse via llvm-branch-commits]

https://github.com/Meinersbur ready_for_review 
https://github.com/llvm/llvm-project/pull/89743
___
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] [clang] Revise IDE folder structure (PR #89743)

[via llvm-branch-commits]

llvmbot wrote:




@llvm/pr-subscribers-clang-analysis

Author: Michael Kruse (Meinersbur)


Changes

Reviewers of #89153 suggested to break up the patch into per-subproject 
patches. This is the Clang part. See #89153 for the entire series and 
motivation.

Update the folder titles for targets in the monorepository that have not seen 
taken care of for some time. These are the folders that targets are organized 
in Visual Studio and XCode (`set_property(TARGET  PROPERTY FOLDER 
"