[llvm-branch-commits] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Arthur Eubanks via llvm-branch-commits]

@@ -7276,6 +7276,12 @@ void SelectionDAGBuilder::visitIntrinsicCall(const 
CallInst ,
 setValue(, getValue(I.getArgOperand(0)));
 return;
 
+  case Intrinsic::experimental_hot:
+// Default lowering to false. It's intended to be lowered as soon as 
profile
+// is avalible to unblock other optimizations.

aeubanks wrote:

you are still required to run the -O0 "optimization" pipeline to do things like 
lowering (e.g. the examples I gave above) before codegen

https://github.com/llvm/llvm-project/pull/84850
___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Matt Arsenault via llvm-branch-commits]

@@ -7276,6 +7276,12 @@ void SelectionDAGBuilder::visitIntrinsicCall(const 
CallInst ,
 setValue(, getValue(I.getArgOperand(0)));
 return;
 
+  case Intrinsic::experimental_hot:
+// Default lowering to false. It's intended to be lowered as soon as 
profile
+// is avalible to unblock other optimizations.

arsenm wrote:

The optimization pipeline isn't mandatory, -O0 must still work 

https://github.com/llvm/llvm-project/pull/84850
___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Arthur Eubanks via llvm-branch-commits]

aeubanks wrote:

+1, this is a cool addition!

https://github.com/llvm/llvm-project/pull/84850
___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Arthur Eubanks via llvm-branch-commits]

@@ -7276,6 +7276,12 @@ void SelectionDAGBuilder::visitIntrinsicCall(const 
CallInst ,
 setValue(, getValue(I.getArgOperand(0)));
 return;
 
+  case Intrinsic::experimental_hot:
+// Default lowering to false. It's intended to be lowered as soon as 
profile
+// is avalible to unblock other optimizations.

aeubanks wrote:

we shouldn't even get to isel, this should be unconditionally lowered in the 
optimization pipeline and any instance of this intrinsic getting to isel should 
be considered a bug (e.g. coroutines, LowerMatrixIntrinsicsPass)

https://github.com/llvm/llvm-project/pull/84850
___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Matt Arsenault via llvm-branch-commits]

@@ -7276,6 +7276,12 @@ void SelectionDAGBuilder::visitIntrinsicCall(const 
CallInst ,
 setValue(, getValue(I.getArgOperand(0)));
 return;
 
+  case Intrinsic::experimental_hot:
+// Default lowering to false. It's intended to be lowered as soon as 
profile
+// is avalible to unblock other optimizations.

arsenm wrote:

Missing globalisel and LowerIntrinsics handling 

https://github.com/llvm/llvm-project/pull/84850
___
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][Transforms][NFC] Make `rewriterImpl` private in `IRRewrite` (PR #84865)

[via llvm-branch-commits]

llvmbot wrote:



@llvm/pr-subscribers-mlir

@llvm/pr-subscribers-mlir-core

Author: Matthias Springer (matthias-springer)


Changes

This commit makes `rewriterImpl` private in `IRRewrite`. This ensures that only 
the conversion value mapping and the dialect conversion configuration can be 
accessed from an IR rewrite object.


---
Full diff: https://github.com/llvm/llvm-project/pull/84865.diff


1 Files Affected:

- (modified) mlir/lib/Transforms/Utils/DialectConversion.cpp (+23-18) 


``diff
diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp 
b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index e4a022b7a0288b..dbdfaeeeb28d4b 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -232,6 +232,9 @@ class IRRewrite {
 
   const ConversionConfig () const;
 
+  ConversionValueMapping ();
+
+private:
   const Kind kind;
   ConversionPatternRewriterImpl 
 };
@@ -470,7 +473,8 @@ class BlockTypeConversionRewrite : public BlockRewrite {
   /// live users, using the provided `findLiveUser` to search for a user that
   /// survives the conversion process.
   LogicalResult
-  materializeLiveConversions(function_ref findLiveUser);
+  materializeLiveConversions(OpBuilder ,
+ function_ref findLiveUser);
 
   void commit(RewriterBase ) override;
 
@@ -1035,6 +1039,8 @@ const ConversionConfig ::getConfig() const {
   return rewriterImpl.config;
 }
 
+ConversionValueMapping ::getMapping() { return rewriterImpl.mapping; 
}
+
 void BlockTypeConversionRewrite::commit(RewriterBase ) {
   // Inform the listener about all IR modifications that have already taken
   // place: References to the original block have been replaced with the new
@@ -1049,8 +1055,7 @@ void BlockTypeConversionRewrite::commit(RewriterBase 
) {
llvm::zip_equal(origBlock->getArguments(), argInfo)) {
 // Handle the case of a 1->0 value mapping.
 if (!info) {
-  if (Value newArg =
-  rewriterImpl.mapping.lookupOrNull(origArg, origArg.getType()))
+  if (Value newArg = getMapping().lookupOrNull(origArg, origArg.getType()))
 rewriter.replaceAllUsesWith(origArg, newArg);
   continue;
 }
@@ -1061,8 +1066,8 @@ void BlockTypeConversionRewrite::commit(RewriterBase 
) {
 
 // If the argument is still used, replace it with the generated cast.
 if (!origArg.use_empty()) {
-  rewriter.replaceAllUsesWith(origArg, 
rewriterImpl.mapping.lookupOrDefault(
-   castValue, origArg.getType()));
+  rewriter.replaceAllUsesWith(
+  origArg, getMapping().lookupOrDefault(castValue, origArg.getType()));
 }
   }
 }
@@ -1072,23 +1077,23 @@ void BlockTypeConversionRewrite::rollback() {
 }
 
 LogicalResult BlockTypeConversionRewrite::materializeLiveConversions(
-function_ref findLiveUser) {
+OpBuilder , function_ref findLiveUser) {
+  OpBuilder::InsertionGuard g(builder);
+  builder.setInsertionPointToStart(block);
+
   // Process the remapping for each of the original arguments.
   for (auto it : llvm::enumerate(origBlock->getArguments())) {
 BlockArgument origArg = it.value();
-// Note: `block` may be detached, so OpBuilder::atBlockBegin cannot be 
used.
-OpBuilder builder(it.value().getContext(), /*listener=*/);
-builder.setInsertionPointToStart(block);
 
 // If the type of this argument changed and the argument is still live, we
 // need to materialize a conversion.
-if (rewriterImpl.mapping.lookupOrNull(origArg, origArg.getType()))
+if (getMapping().lookupOrNull(origArg, origArg.getType()))
   continue;
 Operation *liveUser = findLiveUser(origArg);
 if (!liveUser)
   continue;
 
-Value replacementValue = rewriterImpl.mapping.lookupOrDefault(origArg);
+Value replacementValue = getMapping().lookupOrDefault(origArg);
 bool isDroppedArg = replacementValue == origArg;
 if (!isDroppedArg)
   builder.setInsertionPointAfterValue(replacementValue);
@@ -1113,13 +1118,13 @@ LogicalResult 
BlockTypeConversionRewrite::materializeLiveConversions(
   << "see existing live user here: " << *liveUser;
   return failure();
 }
-rewriterImpl.mapping.map(origArg, newArg);
+getMapping().map(origArg, newArg);
   }
   return success();
 }
 
 void ReplaceAllUsesRewrite::commit(RewriterBase ) {
-  Value repl = rewriterImpl.mapping.lookupOrNull(value);
+  Value repl = getMapping().lookupOrNull(value);
   assert(repl && "expected that value is mapped");
 
   if (isa(repl)) {
@@ -1138,7 +1143,7 @@ void ReplaceAllUsesRewrite::commit(RewriterBase 
) {
   });
 }
 
-void ReplaceAllUsesRewrite::rollback() { rewriterImpl.mapping.erase(value); }
+void ReplaceAllUsesRewrite::rollback() { getMapping().erase(value); }
 
 void ReplaceOperationRewrite::commit(RewriterBase ) {
   auto *listener = dyn_cast_or_null(
@@ -1147,7 +1152,7 @@ void ReplaceOperationRewrite::commit(RewriterBase 
) {
   // 

[llvm-branch-commits] [mlir] [mlir][Transforms][NFC] Make `rewriterImpl` private in `IRRewrite` (PR #84865)

[Matthias Springer via llvm-branch-commits]

https://github.com/matthias-springer created 
https://github.com/llvm/llvm-project/pull/84865

This commit makes `rewriterImpl` private in `IRRewrite`. This ensures that only 
the conversion value mapping and the dialect conversion configuration can be 
accessed from an IR rewrite object.


>From 71cdb0ef4d9962cc085a1f2182a0bcd1e0c025b5 Mon Sep 17 00:00:00 2001
From: Matthias Springer 
Date: Tue, 12 Mar 2024 02:56:46 +
Subject: [PATCH] [mlir][Transforms][NFC] Make `rewriterImpl` private in
 `IRRewrite`

This commit makes `rewriterImpl` private in `IRRewrite`. This ensures that only 
the conversion value mapping and the dialect conversion configuration can be 
accessed from an IR rewrite object.
---
 .../Transforms/Utils/DialectConversion.cpp| 41 +++
 1 file changed, 23 insertions(+), 18 deletions(-)

diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp 
b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index e4a022b7a0288b..dbdfaeeeb28d4b 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -232,6 +232,9 @@ class IRRewrite {
 
   const ConversionConfig () const;
 
+  ConversionValueMapping ();
+
+private:
   const Kind kind;
   ConversionPatternRewriterImpl 
 };
@@ -470,7 +473,8 @@ class BlockTypeConversionRewrite : public BlockRewrite {
   /// live users, using the provided `findLiveUser` to search for a user that
   /// survives the conversion process.
   LogicalResult
-  materializeLiveConversions(function_ref findLiveUser);
+  materializeLiveConversions(OpBuilder ,
+ function_ref findLiveUser);
 
   void commit(RewriterBase ) override;
 
@@ -1035,6 +1039,8 @@ const ConversionConfig ::getConfig() const {
   return rewriterImpl.config;
 }
 
+ConversionValueMapping ::getMapping() { return rewriterImpl.mapping; 
}
+
 void BlockTypeConversionRewrite::commit(RewriterBase ) {
   // Inform the listener about all IR modifications that have already taken
   // place: References to the original block have been replaced with the new
@@ -1049,8 +1055,7 @@ void BlockTypeConversionRewrite::commit(RewriterBase 
) {
llvm::zip_equal(origBlock->getArguments(), argInfo)) {
 // Handle the case of a 1->0 value mapping.
 if (!info) {
-  if (Value newArg =
-  rewriterImpl.mapping.lookupOrNull(origArg, origArg.getType()))
+  if (Value newArg = getMapping().lookupOrNull(origArg, origArg.getType()))
 rewriter.replaceAllUsesWith(origArg, newArg);
   continue;
 }
@@ -1061,8 +1066,8 @@ void BlockTypeConversionRewrite::commit(RewriterBase 
) {
 
 // If the argument is still used, replace it with the generated cast.
 if (!origArg.use_empty()) {
-  rewriter.replaceAllUsesWith(origArg, 
rewriterImpl.mapping.lookupOrDefault(
-   castValue, origArg.getType()));
+  rewriter.replaceAllUsesWith(
+  origArg, getMapping().lookupOrDefault(castValue, origArg.getType()));
 }
   }
 }
@@ -1072,23 +1077,23 @@ void BlockTypeConversionRewrite::rollback() {
 }
 
 LogicalResult BlockTypeConversionRewrite::materializeLiveConversions(
-function_ref findLiveUser) {
+OpBuilder , function_ref findLiveUser) {
+  OpBuilder::InsertionGuard g(builder);
+  builder.setInsertionPointToStart(block);
+
   // Process the remapping for each of the original arguments.
   for (auto it : llvm::enumerate(origBlock->getArguments())) {
 BlockArgument origArg = it.value();
-// Note: `block` may be detached, so OpBuilder::atBlockBegin cannot be 
used.
-OpBuilder builder(it.value().getContext(), /*listener=*/);
-builder.setInsertionPointToStart(block);
 
 // If the type of this argument changed and the argument is still live, we
 // need to materialize a conversion.
-if (rewriterImpl.mapping.lookupOrNull(origArg, origArg.getType()))
+if (getMapping().lookupOrNull(origArg, origArg.getType()))
   continue;
 Operation *liveUser = findLiveUser(origArg);
 if (!liveUser)
   continue;
 
-Value replacementValue = rewriterImpl.mapping.lookupOrDefault(origArg);
+Value replacementValue = getMapping().lookupOrDefault(origArg);
 bool isDroppedArg = replacementValue == origArg;
 if (!isDroppedArg)
   builder.setInsertionPointAfterValue(replacementValue);
@@ -1113,13 +1118,13 @@ LogicalResult 
BlockTypeConversionRewrite::materializeLiveConversions(
   << "see existing live user here: " << *liveUser;
   return failure();
 }
-rewriterImpl.mapping.map(origArg, newArg);
+getMapping().map(origArg, newArg);
   }
   return success();
 }
 
 void ReplaceAllUsesRewrite::commit(RewriterBase ) {
-  Value repl = rewriterImpl.mapping.lookupOrNull(value);
+  Value repl = getMapping().lookupOrNull(value);
   assert(repl && "expected that value is mapped");
 
   if (isa(repl)) {
@@ -1138,7 +1143,7 @@ void ReplaceAllUsesRewrite::commit(RewriterBase 
) {
   });
 }
 

[llvm-branch-commits] [mlir] [mlir][Transforms] Support `replaceAllUsesWith` in dialect conversion (PR #84725)

[Matthias Springer via llvm-branch-commits]

@@ -751,6 +731,44 @@ class UnresolvedMaterializationRewrite : public 
OperationRewrite {
   /// The original output type. This is only used for argument conversions.
   Type origOutputType;
 };
+
+/// A value rewrite.
+class ValueRewrite : public IRRewrite {
+public:
+  /// Return the operation that this rewrite operates on.
+  Value getValue() const { return value; }
+
+  static bool classof(const IRRewrite *rewrite) {
+return rewrite->getKind() >= Kind::ReplaceAllUses &&
+   rewrite->getKind() <= Kind::ReplaceAllUses;
+  }
+
+protected:
+  ValueRewrite(Kind kind, ConversionPatternRewriterImpl ,
+   Value value)
+  : IRRewrite(kind, rewriterImpl), value(value) {}
+
+  // The value that this rewrite operates on.
+  Value value;
+};
+
+/// Replacing a value. This rewrite is not immediately reflected in the IR. An
+/// internal IR mapping is updated, but the actual replacement is delayed until
+/// the rewrite is committed.
+class ReplaceAllUsesRewrite : public ValueRewrite {
+public:
+  ReplaceAllUsesRewrite(ConversionPatternRewriterImpl ,
+Value value)
+  : ValueRewrite(Kind::ReplaceAllUses, rewriterImpl, value) {}
+
+  static bool classof(const IRRewrite *rewrite) {
+return rewrite->getKind() == Kind::ReplaceAllUses;
+  }
+
+  void commit(RewriterBase ) override;
+
+  void rollback() override;
+};

matthias-springer wrote:

I agree, the dialect conversion is too complicated. I'm working on this for 
over a month now and there are still parts that I do not understand.

Thinking longer term, what would make the design much simpler: no automatic 
rollback, materialize all IR changes immediately (in particular, materialize 
unrealized_conversion_casts eagerly and no more adaptors). Maybe as part of a 
dialect conversion v2 built on top of the existing `RewritePattern`, 
`ConversionPolicy`, `TypeConverter`...


https://github.com/llvm/llvm-project/pull/84725
___
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][Transforms] Support `replaceAllUsesWith` in dialect conversion (PR #84725)

[Mehdi Amini via llvm-branch-commits]

@@ -751,6 +731,44 @@ class UnresolvedMaterializationRewrite : public 
OperationRewrite {
   /// The original output type. This is only used for argument conversions.
   Type origOutputType;
 };
+
+/// A value rewrite.
+class ValueRewrite : public IRRewrite {
+public:
+  /// Return the operation that this rewrite operates on.
+  Value getValue() const { return value; }
+
+  static bool classof(const IRRewrite *rewrite) {
+return rewrite->getKind() >= Kind::ReplaceAllUses &&
+   rewrite->getKind() <= Kind::ReplaceAllUses;
+  }
+
+protected:
+  ValueRewrite(Kind kind, ConversionPatternRewriterImpl ,
+   Value value)
+  : IRRewrite(kind, rewriterImpl), value(value) {}
+
+  // The value that this rewrite operates on.
+  Value value;
+};
+
+/// Replacing a value. This rewrite is not immediately reflected in the IR. An
+/// internal IR mapping is updated, but the actual replacement is delayed until
+/// the rewrite is committed.
+class ReplaceAllUsesRewrite : public ValueRewrite {
+public:
+  ReplaceAllUsesRewrite(ConversionPatternRewriterImpl ,
+Value value)
+  : ValueRewrite(Kind::ReplaceAllUses, rewriterImpl, value) {}
+
+  static bool classof(const IRRewrite *rewrite) {
+return rewrite->getKind() == Kind::ReplaceAllUses;
+  }
+
+  void commit(RewriterBase ) override;
+
+  void rollback() override;
+};

joker-eph wrote:

The whole flow of Dialect conversion tracking of these changes is too 
complicated for me to know whether the commit/rollback logic is safe and 
complete here :(
It's likely that only testing will tell, but that's unfortunate!

https://github.com/llvm/llvm-project/pull/84725
___
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][CodeGen] Remove SimplifyCFGPass preceding RemoveTrapsPass (PR #84852)

[Arthur Eubanks via llvm-branch-commits]

https://github.com/aeubanks approved this pull request.


https://github.com/llvm/llvm-project/pull/84852
___
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][ubsan] Switch UBSAN optimization to `llvm.experimental.hot` (PR #84858)

[Kirill Stoimenov via llvm-branch-commits]

https://github.com/kstoimenov approved this pull request.


https://github.com/llvm/llvm-project/pull/84858
___
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/18.x: [Clang][LoongArch] Fix wrong return value type of __iocsrrd_h (#84100) (PR #84715)

[Lu Weining via llvm-branch-commits]

https://github.com/SixWeining approved this pull request.


https://github.com/llvm/llvm-project/pull/84715
___
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] Rename `remove-traps` to `lower-builtin-hot` (PR #84853)

[Kirill Stoimenov via llvm-branch-commits]

https://github.com/kstoimenov approved this pull request.


https://github.com/llvm/llvm-project/pull/84853
___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Kirill Stoimenov via llvm-branch-commits]

https://github.com/kstoimenov approved this pull request.


https://github.com/llvm/llvm-project/pull/84850
___
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] [compiler-rt] release/18.x: Unbreak *tf builtins for hexfloat (#82208) (PR #82904)

[Alexander Richardson via llvm-branch-commits]

https://github.com/arichardson approved this pull request.


https://github.com/llvm/llvm-project/pull/82904
___
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][ubsan] Switch UBSAN optimization to `llvm.experimental.hot` (PR #84858)

[Jon Roelofs via llvm-branch-commits]

@@ -3805,6 +3812,12 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked,
 SanitizerHandler CheckHandlerID) {
   llvm::BasicBlock *Cont = createBasicBlock("cont");
 
+  if (ClSanitizeExpHot) {
+Checked =
+Builder.CreateOr(Checked, Builder.CreateCall(CGM.getIntrinsic(
+  llvm::Intrinsic::experimental_hot)));

jroelofs wrote:

This would be the moral equivalent of diagnose_if

https://github.com/llvm/llvm-project/pull/84858
___
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][ubsan] Switch UBSAN optimization to `llvm.experimental.hot` (PR #84858)

[Jon Roelofs via llvm-branch-commits]

@@ -3805,6 +3812,12 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked,
 SanitizerHandler CheckHandlerID) {
   llvm::BasicBlock *Cont = createBasicBlock("cont");
 
+  if (ClSanitizeExpHot) {
+Checked =
+Builder.CreateOr(Checked, Builder.CreateCall(CGM.getIntrinsic(
+  llvm::Intrinsic::experimental_hot)));

jroelofs wrote:

It might be helpful if there were a remark that fired when one of these were 
optimized out.

Maybe we could get that with another intrinsic whose semantics are "print the 
message in a remark if this gets optimized out", and then stick one of those on 
the appropriate side of the condition, with an explanation on where it came 
from, and what the threshold was.

https://github.com/llvm/llvm-project/pull/84858
___
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][ubsan] Switch UBSAN optimization to `llvm.experimental.hot` (PR #84858)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka edited 
https://github.com/llvm/llvm-project/pull/84858
___
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][ubsan] Switch UBSAN optimization to `llvm.experimental.hot` (PR #84858)

[via llvm-branch-commits]

llvmbot wrote:




@llvm/pr-subscribers-clang

Author: Vitaly Buka (vitalybuka)


Changes

Intrinsic introduced with #84850. Using intrinsic improves performance
by 3% comparing to removing traps (on
"test-suite/MultiSource/Benchmarks" with PGO+ThinLTO).

The pass will be renamed with #84852.


---

Patch is 20.23 KiB, truncated to 20.00 KiB below, full version: 
https://github.com/llvm/llvm-project/pull/84858.diff


5 Files Affected:

- (modified) clang/lib/CodeGen/BackendUtil.cpp (+3-3) 
- (modified) clang/lib/CodeGen/CGExpr.cpp (+14-1) 
- (modified) clang/test/CodeGen/remote-traps.c (+20-3) 
- (modified) llvm/lib/Transforms/Instrumentation/RemoveTrapsPass.cpp (+15-11) 
- (modified) llvm/test/Transforms/RemoveTraps/remove-traps.ll (+64-25) 


``diff
diff --git a/clang/lib/CodeGen/BackendUtil.cpp 
b/clang/lib/CodeGen/BackendUtil.cpp
index 82b30b8d815629..7e53469a48d42c 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -100,9 +100,9 @@ using namespace llvm;
 namespace llvm {
 extern cl::opt PrintPipelinePasses;
 
-cl::opt ClRemoveTraps("clang-remove-traps", cl::Optional,
-cl::desc("Insert remove-traps pass."),
-cl::init(false));
+static cl::opt ClRemoveTraps("clang-remove-traps", cl::Optional,
+   cl::desc("Insert remove-traps pass."),
+   cl::init(false));
 
 // Experiment to move sanitizers earlier.
 static cl::opt ClSanitizeOnOptimizerEarlyEP(
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index 59a7fe8925001c..3a27622f165995 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -56,7 +56,14 @@ using namespace CodeGen;
 // Experiment to make sanitizers easier to debug
 static llvm::cl::opt ClSanitizeDebugDeoptimization(
 "ubsan-unique-traps", llvm::cl::Optional,
-llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check"),
+llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check."),
+llvm::cl::init(false));
+
+// TODO: Introduce frontend options to enabled per sanitizers, similar to
+// `fsanitize-trap`.
+static llvm::cl::opt ClSanitizeExpHot(
+"ubsan-exp-hot", llvm::cl::Optional,
+llvm::cl::desc("Pass UBSAN checks if `llvm.experimental.hot()` is true."),
 llvm::cl::init(false));
 
 //======//
@@ -3805,6 +3812,12 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked,
 SanitizerHandler CheckHandlerID) {
   llvm::BasicBlock *Cont = createBasicBlock("cont");
 
+  if (ClSanitizeExpHot) {
+Checked =
+Builder.CreateOr(Checked, Builder.CreateCall(CGM.getIntrinsic(
+  llvm::Intrinsic::experimental_hot)));
+  }
+
   // If we're optimizing, collapse all calls to trap down to just one per
   // check-type per function to save on code size.
   if ((int)TrapBBs.size() <= CheckHandlerID)
diff --git a/clang/test/CodeGen/remote-traps.c 
b/clang/test/CodeGen/remote-traps.c
index 6751afb96d25f2..934b76fae0d2a5 100644
--- a/clang/test/CodeGen/remote-traps.c
+++ b/clang/test/CodeGen/remote-traps.c
@@ -1,5 +1,7 @@
-// RUN: %clang_cc1 -O1 -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow %s -o - | FileCheck %s 
-// RUN: %clang_cc1 -O1 -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow -mllvm -clang-remove-traps -mllvm 
-remove-traps-random-rate=1 %s -o - | FileCheck %s --implicit-check-not="call 
void @llvm.ubsantrap" --check-prefixes=REMOVE
+// RUN: %clang_cc1 -O1 %s -o - -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow -mllvm -ubsan-exp-hot | FileCheck %s
+// RUN: %clang_cc1 -O1 %s -o - -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow -mllvm -ubsan-exp-hot -mllvm 
-clang-remove-traps -mllvm -remove-traps-random-rate=1 %s -o - | FileCheck %s 
--check-prefixes=REMOVE
+
+#include 
 
 int test(int x) {
   return x + 123;
@@ -12,4 +14,19 @@ int test(int x) {
 // CHECK-NEXT: unreachable
 
 // REMOVE-LABEL: define {{.*}}i32 @test(
-// REMOVE: call { i32, i1 } @llvm.sadd.with.overflow.i32(
+// REMOVE: add i32 %x, 123
+// REMOVE-NEXT: ret i32
+
+
+bool experimental_hot() __asm("llvm.experimental.hot");
+
+bool test_asm() {
+  return experimental_hot();
+}
+
+// CHECK-LABEL: define {{.*}}i1 @test_asm(
+// CHECK: [[R:%.*]] = tail call zeroext i1 @llvm.experimental.hot()
+// CHECK: ret i1 [[R]]
+
+// REMOVE-LABEL: define {{.*}}i1 @test_asm(
+// REMOVE: ret i1 true
diff --git a/llvm/lib/Transforms/Instrumentation/RemoveTrapsPass.cpp 
b/llvm/lib/Transforms/Instrumentation/RemoveTrapsPass.cpp
index d87f7482a21d25..fa4716f2e7a403 100644
--- a/llvm/lib/Transforms/Instrumentation/RemoveTrapsPass.cpp
+++ b/llvm/lib/Transforms/Instrumentation/RemoveTrapsPass.cpp
@@ 

[llvm-branch-commits] Rename `remove-traps` to `lower-builtin-hot` (PR #84853)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka updated 
https://github.com/llvm/llvm-project/pull/84853


___
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] Rename `remove-traps` to `lower-builtin-hot` (PR #84853)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka updated 
https://github.com/llvm/llvm-project/pull/84853


___
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][CodeGen] Remove SimplifyCFGPass preceding RemoveTrapsPass (PR #84852)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka updated 
https://github.com/llvm/llvm-project/pull/84852


___
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][CodeGen] Remove SimplifyCFGPass preceding RemoveTrapsPass (PR #84852)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka updated 
https://github.com/llvm/llvm-project/pull/84852


___
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][ubsan] Switch UBSAN optimization to `llvm.experimental.hot` (PR #84858)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka created 
https://github.com/llvm/llvm-project/pull/84858

Intrinsic introduced with #84850. Using intrinsic improves performance
by 3% comparing to removing traps (on
"test-suite/MultiSource/Benchmarks" with PGO+ThinLTO).

The pass will be renamed with #84852.



___
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] [lld] [llvm] Backport fixes for ARM64EC import libraries (PR #84590)

[Daniel Paoliello via llvm-branch-commits]

dpaoliello wrote:

> So your use-case is basically equivalent to using llvm-dlltool, except not 
> using the text parser?

Yep.

> If this is actually enough to make Rust targets usable, then I guess we could 
> consider it, but the fixes aren't structured in a way to make it obvious this 
> won't impact non-ARM64EC targets.

That's fair enough.

@cjacek you have more experience in this area than I do - what's your 
confidence level with backporting these changes?

https://github.com/llvm/llvm-project/pull/84590
___
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] [InstCombiner] Remove unused `llvm.experimental.hot()` (PR #84851)

[Jon Roelofs via llvm-branch-commits]

@@ -56,7 +56,14 @@ using namespace CodeGen;
 // Experiment to make sanitizers easier to debug
 static llvm::cl::opt ClSanitizeDebugDeoptimization(
 "ubsan-unique-traps", llvm::cl::Optional,
-llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check"),
+llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check."),
+llvm::cl::init(false));
+
+// TODO: Introduce frontend options to enabled per sanitizers, similar to
+// `fsanitize-trap`.
+static llvm::cl::opt ClSanitizeExpHot(
+"ubsan-exp-hot", llvm::cl::Optional,
+llvm::cl::desc("Pass UBSAN checks if `llvm.experimental.hot()` is true."),

jroelofs wrote:

np, I'll move this comment once you get that sorted.

https://github.com/llvm/llvm-project/pull/84851
___
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] [InstCombiner] Remove unused `llvm.experimental.hot()` (PR #84851)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka converted_to_draft 
https://github.com/llvm/llvm-project/pull/84851
___
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] [InstCombiner] Remove unused `llvm.experimental.hot()` (PR #84851)

[Vitaly Buka via llvm-branch-commits]

@@ -56,7 +56,14 @@ using namespace CodeGen;
 // Experiment to make sanitizers easier to debug
 static llvm::cl::opt ClSanitizeDebugDeoptimization(
 "ubsan-unique-traps", llvm::cl::Optional,
-llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check"),
+llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check."),
+llvm::cl::init(false));
+
+// TODO: Introduce frontend options to enabled per sanitizers, similar to
+// `fsanitize-trap`.
+static llvm::cl::opt ClSanitizeExpHot(
+"ubsan-exp-hot", llvm::cl::Optional,
+llvm::cl::desc("Pass UBSAN checks if `llvm.experimental.hot()` is true."),

vitalybuka wrote:

@jroelofs  It should be in a different one, I'll add you into the review

https://github.com/llvm/llvm-project/pull/84851
___
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] [InstCombiner] Remove unused `llvm.experimental.hot()` (PR #84851)

[Vitaly Buka via llvm-branch-commits]

@@ -56,7 +56,14 @@ using namespace CodeGen;
 // Experiment to make sanitizers easier to debug
 static llvm::cl::opt ClSanitizeDebugDeoptimization(
 "ubsan-unique-traps", llvm::cl::Optional,
-llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check"),
+llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check."),
+llvm::cl::init(false));
+
+// TODO: Introduce frontend options to enabled per sanitizers, similar to
+// `fsanitize-trap`.
+static llvm::cl::opt ClSanitizeExpHot(
+"ubsan-exp-hot", llvm::cl::Optional,
+llvm::cl::desc("Pass UBSAN checks if `llvm.experimental.hot()` is true."),

vitalybuka wrote:

Sorry, messed up the spr stack. fixing

https://github.com/llvm/llvm-project/pull/84851
___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Jon Roelofs via llvm-branch-commits]

jroelofs wrote:

Neat idea!

https://github.com/llvm/llvm-project/pull/84850
___
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] [InstCombiner] Remove unused `llvm.experimental.hot()` (PR #84851)

[Jon Roelofs via llvm-branch-commits]

@@ -56,7 +56,14 @@ using namespace CodeGen;
 // Experiment to make sanitizers easier to debug
 static llvm::cl::opt ClSanitizeDebugDeoptimization(
 "ubsan-unique-traps", llvm::cl::Optional,
-llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check"),
+llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check."),
+llvm::cl::init(false));
+
+// TODO: Introduce frontend options to enabled per sanitizers, similar to
+// `fsanitize-trap`.
+static llvm::cl::opt ClSanitizeExpHot(
+"ubsan-exp-hot", llvm::cl::Optional,
+llvm::cl::desc("Pass UBSAN checks if `llvm.experimental.hot()` is true."),

jroelofs wrote:

I'm not sure how best to do it, but it might be useful to have a remark that 
fires when one of these is optimized out.

https://github.com/llvm/llvm-project/pull/84851
___
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] [InstCombiner] Remove unused `llvm.experimental.hot()` (PR #84851)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka edited 
https://github.com/llvm/llvm-project/pull/84851
___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka edited 
https://github.com/llvm/llvm-project/pull/84850
___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka edited 
https://github.com/llvm/llvm-project/pull/84850
___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka edited 
https://github.com/llvm/llvm-project/pull/84850
___
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] Rename `remove-traps` to `lower-builtin-hot` (PR #84853)

[via llvm-branch-commits]

llvmbot wrote:




@llvm/pr-subscribers-llvm-transforms

Author: Vitaly Buka (vitalybuka)


Changes



---
Full diff: https://github.com/llvm/llvm-project/pull/84853.diff


8 Files Affected:

- (modified) clang/lib/CodeGen/BackendUtil.cpp (+7-6) 
- (renamed) clang/test/CodeGen/lower-builtin-hot.c (+1-1) 
- (renamed) llvm/include/llvm/Transforms/Instrumentation/LowerBuiltinHotPass.h 
(+4-4) 
- (modified) llvm/lib/Passes/PassBuilder.cpp (+1-1) 
- (modified) llvm/lib/Passes/PassRegistry.def (+1-1) 
- (modified) llvm/lib/Transforms/Instrumentation/CMakeLists.txt (+1-1) 
- (renamed) llvm/lib/Transforms/Instrumentation/LowerBuiltinHotPass.cpp (+7-7) 
- (renamed) llvm/test/Transforms/LowerBuiltinHotPass/lower-builtin-hot.ll 
(+4-4) 


``diff
diff --git a/clang/lib/CodeGen/BackendUtil.cpp 
b/clang/lib/CodeGen/BackendUtil.cpp
index 7eab1c4ed44b89..e36e08e8bf5102 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -73,10 +73,10 @@
 #include "llvm/Transforms/Instrumentation/HWAddressSanitizer.h"
 #include "llvm/Transforms/Instrumentation/InstrProfiling.h"
 #include "llvm/Transforms/Instrumentation/KCFI.h"
+#include "llvm/Transforms/Instrumentation/LowerBuiltinHotPass.h"
 #include "llvm/Transforms/Instrumentation/MemProfiler.h"
 #include "llvm/Transforms/Instrumentation/MemorySanitizer.h"
 #include "llvm/Transforms/Instrumentation/PGOInstrumentation.h"
-#include "llvm/Transforms/Instrumentation/RemoveTrapsPass.h"
 #include "llvm/Transforms/Instrumentation/SanitizerBinaryMetadata.h"
 #include "llvm/Transforms/Instrumentation/SanitizerCoverage.h"
 #include "llvm/Transforms/Instrumentation/ThreadSanitizer.h"
@@ -99,9 +99,10 @@ using namespace llvm;
 namespace llvm {
 extern cl::opt PrintPipelinePasses;
 
-static cl::opt ClRemoveTraps("clang-remove-traps", cl::Optional,
-   cl::desc("Insert remove-traps pass."),
-   cl::init(false));
+static cl::opt
+ClLowerBuiltinHot("clang-lower-builtin-hot", cl::Optional,
+  cl::desc("Insert lower-builtin-hot pass."),
+  cl::init(false));
 
 // Experiment to move sanitizers earlier.
 static cl::opt ClSanitizeOnOptimizerEarlyEP(
@@ -750,13 +751,13 @@ static void addSanitizers(const Triple ,
 PB.registerOptimizerLastEPCallback(SanitizersCallback);
   }
 
-  if (ClRemoveTraps) {
+  if (ClLowerBuiltinHot) {
 // We can optimize after inliner, and PGO profile matching. The hook below
 // is called at the end `buildFunctionSimplificationPipeline`, which called
 // from `buildInlinerPipeline`, which called after profile matching.
 PB.registerScalarOptimizerLateEPCallback(
 [](FunctionPassManager , OptimizationLevel Level) {
-  FPM.addPass(RemoveTrapsPass());
+  FPM.addPass(LowerBuiltinHotPass());
 });
   }
 }
diff --git a/clang/test/CodeGen/remote-traps.c 
b/clang/test/CodeGen/lower-builtin-hot.c
similarity index 87%
rename from clang/test/CodeGen/remote-traps.c
rename to clang/test/CodeGen/lower-builtin-hot.c
index 16e4ebecb5c326..adcd0130bc73db 100644
--- a/clang/test/CodeGen/remote-traps.c
+++ b/clang/test/CodeGen/lower-builtin-hot.c
@@ -1,5 +1,5 @@
 // RUN: %clang_cc1 -O1 %s -o - -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow -mllvm -ubsan-exp-hot | FileCheck %s 
-// RUN: %clang_cc1 -O1 %s -o - -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow -mllvm -ubsan-exp-hot -mllvm 
-clang-remove-traps -mllvm -remove-traps-random-rate=1 %s -o - | FileCheck %s 
--check-prefixes=REMOVE
+// RUN: %clang_cc1 -O1 %s -o - -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow -mllvm -ubsan-exp-hot -mllvm 
-clang-lower-builtin-hot -mllvm -lower-builtin-hot-random-rate=1 %s -o - | 
FileCheck %s --check-prefixes=REMOVE
 
 #include 
 
diff --git a/llvm/include/llvm/Transforms/Instrumentation/RemoveTrapsPass.h 
b/llvm/include/llvm/Transforms/Instrumentation/LowerBuiltinHotPass.h
similarity index 75%
rename from llvm/include/llvm/Transforms/Instrumentation/RemoveTrapsPass.h
rename to llvm/include/llvm/Transforms/Instrumentation/LowerBuiltinHotPass.h
index 58f6bbcec5dc9d..fa7e9455ce58ea 100644
--- a/llvm/include/llvm/Transforms/Instrumentation/RemoveTrapsPass.h
+++ b/llvm/include/llvm/Transforms/Instrumentation/LowerBuiltinHotPass.h
@@ -1,4 +1,4 @@
-//===- RemoveTrapsPass.h *- C++ 
-*-===//
+//===- LowerBuiltinHotPass.h *- C++ 
-*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM 
Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -11,8 +11,8 @@
 ///
 
//===--===//
 
-#ifndef LLVM_TRANSFORMS_INSTRUMENTATION_UBSANOPTIMIZATIONPASS_H
-#define LLVM_TRANSFORMS_INSTRUMENTATION_UBSANOPTIMIZATIONPASS_H
+#ifndef 

[llvm-branch-commits] [clang][CodeGen] Remove SimplifyCFGPass preceding RemoveTrapsPass (PR #84852)

[via llvm-branch-commits]

llvmbot wrote:




@llvm/pr-subscribers-clang

Author: Vitaly Buka (vitalybuka)


Changes

There is no performance difference after switching to
`llvm.experimental.hot`.


---
Full diff: https://github.com/llvm/llvm-project/pull/84852.diff


1 Files Affected:

- (modified) clang/lib/CodeGen/BackendUtil.cpp (-6) 


``diff
diff --git a/clang/lib/CodeGen/BackendUtil.cpp 
b/clang/lib/CodeGen/BackendUtil.cpp
index 7e53469a48d42c..7eab1c4ed44b89 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -84,7 +84,6 @@
 #include "llvm/Transforms/Scalar/EarlyCSE.h"
 #include "llvm/Transforms/Scalar/GVN.h"
 #include "llvm/Transforms/Scalar/JumpThreading.h"
-#include "llvm/Transforms/Scalar/SimplifyCFG.h"
 #include "llvm/Transforms/Utils/Debugify.h"
 #include "llvm/Transforms/Utils/EntryExitInstrumenter.h"
 #include "llvm/Transforms/Utils/ModuleUtils.h"
@@ -757,11 +756,6 @@ static void addSanitizers(const Triple ,
 // from `buildInlinerPipeline`, which called after profile matching.
 PB.registerScalarOptimizerLateEPCallback(
 [](FunctionPassManager , OptimizationLevel Level) {
-  // RemoveTrapsPass expects trap blocks preceded by conditional
-  // branches, which usually is not the case without SimplifyCFG.
-  // TODO: Remove `SimplifyCFGPass` after switching to dedicated
-  // intrinsic.
-  FPM.addPass(SimplifyCFGPass());
   FPM.addPass(RemoveTrapsPass());
 });
   }

``




https://github.com/llvm/llvm-project/pull/84852
___
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] [InstCombiner] Remove unused `llvm.experimental.hot()` (PR #84851)

[via llvm-branch-commits]

llvmbot wrote:




@llvm/pr-subscribers-clang-codegen

Author: Vitaly Buka (vitalybuka)


Changes

Intrinsic declared to have sideeffects, but it's done only to prevent
moving it. Removing unused ones is OK.


---

Patch is 21.91 KiB, truncated to 20.00 KiB below, full version: 
https://github.com/llvm/llvm-project/pull/84851.diff


7 Files Affected:

- (modified) clang/lib/CodeGen/BackendUtil.cpp (+3-3) 
- (modified) clang/lib/CodeGen/CGExpr.cpp (+14-1) 
- (modified) clang/test/CodeGen/remote-traps.c (+20-3) 
- (modified) llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp (+8) 
- (modified) llvm/lib/Transforms/Instrumentation/RemoveTrapsPass.cpp (+15-11) 
- (added) llvm/test/Transforms/InstCombine/builtin-hot.ll (+25) 
- (modified) llvm/test/Transforms/RemoveTraps/remove-traps.ll (+64-25) 


``diff
diff --git a/clang/lib/CodeGen/BackendUtil.cpp 
b/clang/lib/CodeGen/BackendUtil.cpp
index 82b30b8d815629..7e53469a48d42c 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -100,9 +100,9 @@ using namespace llvm;
 namespace llvm {
 extern cl::opt PrintPipelinePasses;
 
-cl::opt ClRemoveTraps("clang-remove-traps", cl::Optional,
-cl::desc("Insert remove-traps pass."),
-cl::init(false));
+static cl::opt ClRemoveTraps("clang-remove-traps", cl::Optional,
+   cl::desc("Insert remove-traps pass."),
+   cl::init(false));
 
 // Experiment to move sanitizers earlier.
 static cl::opt ClSanitizeOnOptimizerEarlyEP(
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index 59a7fe8925001c..3a27622f165995 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -56,7 +56,14 @@ using namespace CodeGen;
 // Experiment to make sanitizers easier to debug
 static llvm::cl::opt ClSanitizeDebugDeoptimization(
 "ubsan-unique-traps", llvm::cl::Optional,
-llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check"),
+llvm::cl::desc("Deoptimize traps for UBSAN so there is 1 trap per check."),
+llvm::cl::init(false));
+
+// TODO: Introduce frontend options to enabled per sanitizers, similar to
+// `fsanitize-trap`.
+static llvm::cl::opt ClSanitizeExpHot(
+"ubsan-exp-hot", llvm::cl::Optional,
+llvm::cl::desc("Pass UBSAN checks if `llvm.experimental.hot()` is true."),
 llvm::cl::init(false));
 
 //======//
@@ -3805,6 +3812,12 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked,
 SanitizerHandler CheckHandlerID) {
   llvm::BasicBlock *Cont = createBasicBlock("cont");
 
+  if (ClSanitizeExpHot) {
+Checked =
+Builder.CreateOr(Checked, Builder.CreateCall(CGM.getIntrinsic(
+  llvm::Intrinsic::experimental_hot)));
+  }
+
   // If we're optimizing, collapse all calls to trap down to just one per
   // check-type per function to save on code size.
   if ((int)TrapBBs.size() <= CheckHandlerID)
diff --git a/clang/test/CodeGen/remote-traps.c 
b/clang/test/CodeGen/remote-traps.c
index 6751afb96d25f2..16e4ebecb5c326 100644
--- a/clang/test/CodeGen/remote-traps.c
+++ b/clang/test/CodeGen/remote-traps.c
@@ -1,5 +1,7 @@
-// RUN: %clang_cc1 -O1 -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow %s -o - | FileCheck %s 
-// RUN: %clang_cc1 -O1 -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow -mllvm -clang-remove-traps -mllvm 
-remove-traps-random-rate=1 %s -o - | FileCheck %s --implicit-check-not="call 
void @llvm.ubsantrap" --check-prefixes=REMOVE
+// RUN: %clang_cc1 -O1 %s -o - -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow -mllvm -ubsan-exp-hot | FileCheck %s 
+// RUN: %clang_cc1 -O1 %s -o - -emit-llvm -fsanitize=signed-integer-overflow 
-fsanitize-trap=signed-integer-overflow -mllvm -ubsan-exp-hot -mllvm 
-clang-remove-traps -mllvm -remove-traps-random-rate=1 %s -o - | FileCheck %s 
--check-prefixes=REMOVE
+
+#include 
 
 int test(int x) {
   return x + 123;
@@ -12,4 +14,19 @@ int test(int x) {
 // CHECK-NEXT: unreachable
 
 // REMOVE-LABEL: define {{.*}}i32 @test(
-// REMOVE: call { i32, i1 } @llvm.sadd.with.overflow.i32(
+// REMOVE: add i32 %x, 123
+// REMOVE-NEXT: ret i32
+
+
+bool experimental_hot() __asm("llvm.experimental.hot");
+
+bool test_asm() {
+  return experimental_hot();
+}
+
+// CHECK-LABEL: define {{.*}}i1 @test_asm(
+// CHECK: [[R:%.*]] = tail call zeroext i1 @llvm.experimental.hot()
+// CHECK: ret i1 [[R]]
+
+// REMOVE-LABEL: define {{.*}}i1 @test_asm(
+// REMOVE: ret i1 true
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp 
b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index f5f3716d390d77..a7f2b0216bfa00 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ 

[llvm-branch-commits] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[via llvm-branch-commits]

llvmbot wrote:




@llvm/pr-subscribers-llvm-selectiondag

Author: Vitaly Buka (vitalybuka)


Changes



---
Full diff: https://github.com/llvm/llvm-project/pull/84850.diff


4 Files Affected:

- (modified) llvm/docs/LangRef.rst (+48) 
- (modified) llvm/include/llvm/IR/Intrinsics.td (+5) 
- (modified) llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (+6) 
- (added) llvm/test/CodeGen/Generic/builtin-hot.ll (+19) 


``diff
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index d613ceea8654f8..36f4c964ee296c 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -27639,6 +27639,54 @@ constant `true`. However it is always correct to 
replace
 it with any other `i1` value. Any pass can
 freely do it if it can benefit from non-default lowering.
 
+'``llvm.experimental.hot``' Intrinsic
+^
+
+Syntax:
+"""
+
+::
+
+  declare i1 @llvm.experimental.hot()
+
+Overview:
+"
+
+This intrinsic returns true iff it's known that containing basic block is hot 
in
+profile.
+
+When used with profile based optimization allows to change program behaviour
+deppending on the code hotness.
+
+Arguments:
+""
+
+None.
+
+Semantics:
+""
+
+The intrinsic ``@llvm.experimental.hot()`` returns either `true` or `false`,
+deppending on profile used. Expresion is evaluated as `true` iff profile and
+summary are availible and profile counter for the block reach hotness 
threshold.
+For each evaluation of a call to this intrinsic, the program must be valid and
+correct both if it returns `true` and if it returns `false`.
+
+When used in a branch condition, it allows us to choose between
+two alternative correct solutions for the same problem, like
+in example below:
+
+.. code-block:: text
+
+%cond = call i1 @llvm.experimental.hot()
+br i1 %cond, label %fast_path, label %slow_path
+
+  label %fast_path:
+; Omit diagnostics.
+
+  label %slow_path:
+; Additional diagnostics.
+
 
 '``llvm.load.relative``' Intrinsic
 ^^
diff --git a/llvm/include/llvm/IR/Intrinsics.td 
b/llvm/include/llvm/IR/Intrinsics.td
index 144298fd7c0162..96e9cdf6627a75 100644
--- a/llvm/include/llvm/IR/Intrinsics.td
+++ b/llvm/include/llvm/IR/Intrinsics.td
@@ -1722,6 +1722,11 @@ def int_debugtrap : Intrinsic<[]>,
 def int_ubsantrap : Intrinsic<[], [llvm_i8_ty],
   [IntrNoReturn, IntrCold, ImmArg>]>;
 
+// Return true if profile counter for containing block is hot.
+def int_experimental_hot : Intrinsic<[llvm_i1_ty], [],
+  [IntrInaccessibleMemOnly, IntrWriteMem,
+   IntrWillReturn, NoUndef]>;
+
 // Support for dynamic deoptimization (or de-specialization)
 def int_experimental_deoptimize : Intrinsic<[llvm_any_ty], [llvm_vararg_ty],
 [Throws]>;
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp 
b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index 22e57d0d99e9b1..8e73433ce82ea5 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -7276,6 +7276,12 @@ void SelectionDAGBuilder::visitIntrinsicCall(const 
CallInst ,
 setValue(, getValue(I.getArgOperand(0)));
 return;
 
+  case Intrinsic::experimental_hot:
+// Default lowering to false. It's intended to be lowered as soon as 
profile
+// is avalible to unblock other optimizations.
+setValue(, DAG.getConstant(0, sdl, MVT::i1));
+return;
+
   case Intrinsic::ubsantrap:
   case Intrinsic::debugtrap:
   case Intrinsic::trap: {
diff --git a/llvm/test/CodeGen/Generic/builtin-hot.ll 
b/llvm/test/CodeGen/Generic/builtin-hot.ll
new file mode 100644
index 00..449f58d3c00675
--- /dev/null
+++ b/llvm/test/CodeGen/Generic/builtin-hot.ll
@@ -0,0 +1,19 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 
UTC_ARGS: --version 4
+; RUN: llc -o - %s | FileCheck %s
+
+; REQUIRES: aarch64-registered-target
+
+target triple = "aarch64-linux"
+
+define i1 @test()  {
+; CHECK-LABEL: test:
+; CHECK:   // %bb.0: // %entry
+; CHECK-NEXT:mov w0, wzr
+; CHECK-NEXT:ret
+entry:
+  %hot = call i1 @llvm.experimental.hot()
+  ret i1 %hot
+}
+
+declare i1 @llvm.expect.hot() nounwind
+

``




https://github.com/llvm/llvm-project/pull/84850
___
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] Rename `remove-traps` to `lower-builtin-hot` (PR #84853)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka created 
https://github.com/llvm/llvm-project/pull/84853

None


___
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][CodeGen] Remove SimplifyCFGPass preceding RemoveTrapsPass (PR #84852)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka created 
https://github.com/llvm/llvm-project/pull/84852

There is no performance difference after switching to
`llvm.experimental.hot`.



___
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] [InstCombiner] Remove unused `llvm.experimental.hot()` (PR #84851)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka created 
https://github.com/llvm/llvm-project/pull/84851

Intrinsic declared to have sideeffects, but it's done only to prevent
moving it. Removing unused ones is OK.



___
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] [IR] Introduce `llvm.experimental.hot()` (PR #84850)

[Vitaly Buka via llvm-branch-commits]

https://github.com/vitalybuka created 
https://github.com/llvm/llvm-project/pull/84850

None


___
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] [lld] [llvm][lld][RISCV] Support x3_reg_usage (PR #84598)

[Paul Kirth via llvm-branch-commits]

@@ -520,3 +520,8 @@ define i8 @atomic_load_i8_seq_cst(ptr %a) nounwind {
 ; A6S: .attribute 14, 2
 ; A6C: .attribute 14, 1
 }
+

ilovepi wrote:

Sorry, I'm a little stuck on how to infer the attribute. This function will get 
a test once I have that figured out.

https://github.com/llvm/llvm-project/pull/84598
___
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] [lld] [llvm][lld][RISCV] Support x3_reg_usage (PR #84598)

[Paul Kirth via llvm-branch-commits]

https://github.com/ilovepi updated 
https://github.com/llvm/llvm-project/pull/84598

>From 7c9298eea6d8239f9afedc3d6aabb1ec0f71e273 Mon Sep 17 00:00:00 2001
From: Paul Kirth 
Date: Mon, 11 Mar 2024 15:35:59 -0700
Subject: [PATCH] Update callsite parameter

Created using spr 1.3.4
---
 lld/ELF/Arch/RISCV.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lld/ELF/Arch/RISCV.cpp b/lld/ELF/Arch/RISCV.cpp
index b2e0ba17e4efe5..5eb2ce3d64513a 100644
--- a/lld/ELF/Arch/RISCV.cpp
+++ b/lld/ELF/Arch/RISCV.cpp
@@ -1228,8 +1228,8 @@ mergeAttributesSection(const SmallVector ) {
   if (r.second) {
 firstX3RegUse = sec;
   } else {
-mergeX3RegUse(merged.intAttr, firstX3RegUse, sec,
-  r.first->getSecond(), *i);
+mergeX3RegUse(r.first, firstX3RegUse, sec, r.first->getSecond(),
+  *i);
   }
 }
 continue;

___
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] [lld] [llvm][lld][RISCV] Support x3_reg_usage (PR #84598)

[Paul Kirth via llvm-branch-commits]

https://github.com/ilovepi updated 
https://github.com/llvm/llvm-project/pull/84598

>From 7c9298eea6d8239f9afedc3d6aabb1ec0f71e273 Mon Sep 17 00:00:00 2001
From: Paul Kirth 
Date: Mon, 11 Mar 2024 15:35:59 -0700
Subject: [PATCH] Update callsite parameter

Created using spr 1.3.4
---
 lld/ELF/Arch/RISCV.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lld/ELF/Arch/RISCV.cpp b/lld/ELF/Arch/RISCV.cpp
index b2e0ba17e4efe5..5eb2ce3d64513a 100644
--- a/lld/ELF/Arch/RISCV.cpp
+++ b/lld/ELF/Arch/RISCV.cpp
@@ -1228,8 +1228,8 @@ mergeAttributesSection(const SmallVector ) {
   if (r.second) {
 firstX3RegUse = sec;
   } else {
-mergeX3RegUse(merged.intAttr, firstX3RegUse, sec,
-  r.first->getSecond(), *i);
+mergeX3RegUse(r.first, firstX3RegUse, sec, r.first->getSecond(),
+  *i);
   }
 }
 continue;

___
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] [lld] [llvm] Backport fixes for ARM64EC import libraries (PR #84590)

[Eli Friedman via llvm-branch-commits]

efriedma-quic wrote:

So your use-case is basically equivalent to using llvm-dlltool, except not 
using the text parser?

If this is actually enough to make Rust targets usable, then I guess we could 
consider it, but the fixes aren't structured in a way to make it obvious this 
won't impact non-ARM64EC targets.

https://github.com/llvm/llvm-project/pull/84590
___
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] [lld] [llvm][lld][RISCV] Support x3_reg_usage (PR #84598)

[Paul Kirth via llvm-branch-commits]

https://github.com/ilovepi updated 
https://github.com/llvm/llvm-project/pull/84598

>From 7c9298eea6d8239f9afedc3d6aabb1ec0f71e273 Mon Sep 17 00:00:00 2001
From: Paul Kirth 
Date: Mon, 11 Mar 2024 15:35:59 -0700
Subject: [PATCH] Update callsite parameter

Created using spr 1.3.4
---
 lld/ELF/Arch/RISCV.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lld/ELF/Arch/RISCV.cpp b/lld/ELF/Arch/RISCV.cpp
index b2e0ba17e4efe5..5eb2ce3d64513a 100644
--- a/lld/ELF/Arch/RISCV.cpp
+++ b/lld/ELF/Arch/RISCV.cpp
@@ -1228,8 +1228,8 @@ mergeAttributesSection(const SmallVector ) {
   if (r.second) {
 firstX3RegUse = sec;
   } else {
-mergeX3RegUse(merged.intAttr, firstX3RegUse, sec,
-  r.first->getSecond(), *i);
+mergeX3RegUse(r.first, firstX3RegUse, sec, r.first->getSecond(),
+  *i);
   }
 }
 continue;

___
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] [lld] release/18.x: [LLD] [COFF] Set the right alignment for DelayDirectoryChunk (#84697) (PR #84844)

[via llvm-branch-commits]

llvmbot wrote:




@llvm/pr-subscribers-lld

Author: None (llvmbot)


Changes

Backport c93c76b562784926b22a69d3f82a5032dcb4a274

Requested by: @mstorsjo

---
Full diff: https://github.com/llvm/llvm-project/pull/84844.diff


2 Files Affected:

- (modified) lld/COFF/DLL.cpp (+1-1) 
- (modified) lld/test/COFF/delayimports-armnt.yaml (+19-6) 


``diff
diff --git a/lld/COFF/DLL.cpp b/lld/COFF/DLL.cpp
index 6b516d8c6d5ef8..c4388ba9e40d0b 100644
--- a/lld/COFF/DLL.cpp
+++ b/lld/COFF/DLL.cpp
@@ -172,7 +172,7 @@ binImports(COFFLinkerContext ,
 // A chunk for the delay import descriptor table etnry.
 class DelayDirectoryChunk : public NonSectionChunk {
 public:
-  explicit DelayDirectoryChunk(Chunk *n) : dllName(n) {}
+  explicit DelayDirectoryChunk(Chunk *n) : dllName(n) { setAlignment(4); }
 
   size_t getSize() const override {
 return sizeof(delay_import_directory_table_entry);
diff --git a/lld/test/COFF/delayimports-armnt.yaml 
b/lld/test/COFF/delayimports-armnt.yaml
index 7d9bc38c5c3606..ea96d864ef53d5 100644
--- a/lld/test/COFF/delayimports-armnt.yaml
+++ b/lld/test/COFF/delayimports-armnt.yaml
@@ -6,6 +6,7 @@
 # RUN: llvm-readobj --coff-imports %t.exe | FileCheck -check-prefix=IMPORT %s
 # RUN: llvm-readobj --coff-basereloc %t.exe | FileCheck -check-prefix=BASEREL 
%s
 # RUN: llvm-objdump --no-print-imm-hex -d %t.exe | FileCheck 
--check-prefix=DISASM %s
+# RUN: llvm-readobj --file-headers %t.exe | FileCheck -check-prefix=DIR %s
 
 # IMPORT:  Format: COFF-ARM
 # IMPORT-NEXT: Arch: thumb
@@ -13,9 +14,9 @@
 # IMPORT-NEXT: DelayImport {
 # IMPORT-NEXT:   Name: library.dll
 # IMPORT-NEXT:   Attributes: 0x1
-# IMPORT-NEXT:   ModuleHandle: 0x3000
-# IMPORT-NEXT:   ImportAddressTable: 0x3008
-# IMPORT-NEXT:   ImportNameTable: 0x2040
+# IMPORT-NEXT:   ModuleHandle: 0x3008
+# IMPORT-NEXT:   ImportAddressTable: 0x3010
+# IMPORT-NEXT:   ImportNameTable: 0x2044
 # IMPORT-NEXT:   BoundDelayImportTable: 0x0
 # IMPORT-NEXT:   UnloadDelayImportTable: 0x0
 # IMPORT-NEXT:   Import {
@@ -43,7 +44,7 @@
 # BASEREL-NEXT:   }
 # BASEREL-NEXT:   Entry {
 # BASEREL-NEXT: Type: HIGHLOW
-# BASEREL-NEXT: Address: 0x3008
+# BASEREL-NEXT: Address: 0x3010
 # BASEREL-NEXT:   }
 # BASEREL-NEXT:   Entry {
 # BASEREL-NEXT: Type: ABSOLUTE
@@ -52,20 +53,24 @@
 # BASEREL-NEXT: ]
 #
 # DISASM:00401000 <.text>:
-# DISASM:  40100c:   f243 0c08   movw r12, #12296
+# DISASM:  40100c:   f243 0c10   movw r12, #12304
 # DISASM-NEXT:   f2c0 0c40   movtr12, #64
 # DISASM-NEXT:   f000 b800   b.w {{.+}} @ imm = #0
 # DISASM-NEXT:   e92d 480f   push.w  {r0, r1, r2, r3, r11, lr}
 # DISASM-NEXT:   f20d 0b10   addwr11, sp, #16
 # DISASM-NEXT:   ed2d 0b10   vpush   {d0, d1, d2, d3, d4, d5, 
d6, d7}
 # DISASM-NEXT:   4661mov r1, r12
-# DISASM-NEXT:   f242    movw r0, #8192
+# DISASM-NEXT:   f242 0004   movw r0, #8196
 # DISASM-NEXT:   f2c0 0040   movtr0, #64
 # DISASM-NEXT:   f7ff ffe7   bl  0x401000 <.text>
 # DISASM-NEXT:   4684mov r12, r0
 # DISASM-NEXT:   ecbd 0b10   vpop{d0, d1, d2, d3, d4, d5, 
d6, d7}
 # DISASM-NEXT:   e8bd 480f   pop.w   {r0, r1, r2, r3, r11, lr}
 # DISASM-NEXT:   4760bx  r12
+#
+# DIR: DelayImportDescriptorRVA: 0x2004
+# DIR-NEXT:DelayImportDescriptorSize: 0x40
+
 
 --- !COFF
 header:
@@ -80,6 +85,14 @@ sections:
   - VirtualAddress:  0
 SymbolName:  __imp_function
 Type:IMAGE_REL_ARM_MOV32T
+  - Name:.rdata
+Characteristics: [ IMAGE_SCN_CNT_INITIALIZED_DATA, IMAGE_SCN_MEM_READ ]
+Alignment:   1
+SectionData: 01
+  - Name:.data
+Characteristics: [ IMAGE_SCN_CNT_INITIALIZED_DATA, IMAGE_SCN_MEM_READ, 
IMAGE_SCN_MEM_WRITE ]
+Alignment:   1
+SectionData: 02
 symbols:
   - Name:.text
 Value:   0

``




https://github.com/llvm/llvm-project/pull/84844
___
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] [lld] release/18.x: [LLD] [COFF] Set the right alignment for DelayDirectoryChunk (#84697) (PR #84844)

[via llvm-branch-commits]

llvmbot wrote:

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

https://github.com/llvm/llvm-project/pull/84844
___
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] [lld] release/18.x: [LLD] [COFF] Set the right alignment for DelayDirectoryChunk (#84697) (PR #84844)

[via llvm-branch-commits]

https://github.com/llvmbot created 
https://github.com/llvm/llvm-project/pull/84844

Backport c93c76b562784926b22a69d3f82a5032dcb4a274

Requested by: @mstorsjo

>From 021ae6a2ceb1c9067f16ca8a6ae5c1bad3aba235 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= 
Date: Tue, 12 Mar 2024 00:03:26 +0200
Subject: [PATCH] [LLD] [COFF] Set the right alignment for DelayDirectoryChunk
 (#84697)

This makes a difference when linking executables with delay loaded
libraries for arm32; the delay loader implementation can load data from
the registry with instructions that assume alignment.

This issue does not show up when linking in MinGW mode, because a
PseudoRelocTableChunk gets injected, which also sets alignment, even if
the chunk itself is empty.

(cherry picked from commit c93c76b562784926b22a69d3f82a5032dcb4a274)
---
 lld/COFF/DLL.cpp  |  2 +-
 lld/test/COFF/delayimports-armnt.yaml | 25 +++--
 2 files changed, 20 insertions(+), 7 deletions(-)

diff --git a/lld/COFF/DLL.cpp b/lld/COFF/DLL.cpp
index 6b516d8c6d5ef8..c4388ba9e40d0b 100644
--- a/lld/COFF/DLL.cpp
+++ b/lld/COFF/DLL.cpp
@@ -172,7 +172,7 @@ binImports(COFFLinkerContext ,
 // A chunk for the delay import descriptor table etnry.
 class DelayDirectoryChunk : public NonSectionChunk {
 public:
-  explicit DelayDirectoryChunk(Chunk *n) : dllName(n) {}
+  explicit DelayDirectoryChunk(Chunk *n) : dllName(n) { setAlignment(4); }
 
   size_t getSize() const override {
 return sizeof(delay_import_directory_table_entry);
diff --git a/lld/test/COFF/delayimports-armnt.yaml 
b/lld/test/COFF/delayimports-armnt.yaml
index 7d9bc38c5c3606..ea96d864ef53d5 100644
--- a/lld/test/COFF/delayimports-armnt.yaml
+++ b/lld/test/COFF/delayimports-armnt.yaml
@@ -6,6 +6,7 @@
 # RUN: llvm-readobj --coff-imports %t.exe | FileCheck -check-prefix=IMPORT %s
 # RUN: llvm-readobj --coff-basereloc %t.exe | FileCheck -check-prefix=BASEREL 
%s
 # RUN: llvm-objdump --no-print-imm-hex -d %t.exe | FileCheck 
--check-prefix=DISASM %s
+# RUN: llvm-readobj --file-headers %t.exe | FileCheck -check-prefix=DIR %s
 
 # IMPORT:  Format: COFF-ARM
 # IMPORT-NEXT: Arch: thumb
@@ -13,9 +14,9 @@
 # IMPORT-NEXT: DelayImport {
 # IMPORT-NEXT:   Name: library.dll
 # IMPORT-NEXT:   Attributes: 0x1
-# IMPORT-NEXT:   ModuleHandle: 0x3000
-# IMPORT-NEXT:   ImportAddressTable: 0x3008
-# IMPORT-NEXT:   ImportNameTable: 0x2040
+# IMPORT-NEXT:   ModuleHandle: 0x3008
+# IMPORT-NEXT:   ImportAddressTable: 0x3010
+# IMPORT-NEXT:   ImportNameTable: 0x2044
 # IMPORT-NEXT:   BoundDelayImportTable: 0x0
 # IMPORT-NEXT:   UnloadDelayImportTable: 0x0
 # IMPORT-NEXT:   Import {
@@ -43,7 +44,7 @@
 # BASEREL-NEXT:   }
 # BASEREL-NEXT:   Entry {
 # BASEREL-NEXT: Type: HIGHLOW
-# BASEREL-NEXT: Address: 0x3008
+# BASEREL-NEXT: Address: 0x3010
 # BASEREL-NEXT:   }
 # BASEREL-NEXT:   Entry {
 # BASEREL-NEXT: Type: ABSOLUTE
@@ -52,20 +53,24 @@
 # BASEREL-NEXT: ]
 #
 # DISASM:00401000 <.text>:
-# DISASM:  40100c:   f243 0c08   movw r12, #12296
+# DISASM:  40100c:   f243 0c10   movw r12, #12304
 # DISASM-NEXT:   f2c0 0c40   movtr12, #64
 # DISASM-NEXT:   f000 b800   b.w {{.+}} @ imm = #0
 # DISASM-NEXT:   e92d 480f   push.w  {r0, r1, r2, r3, r11, lr}
 # DISASM-NEXT:   f20d 0b10   addwr11, sp, #16
 # DISASM-NEXT:   ed2d 0b10   vpush   {d0, d1, d2, d3, d4, d5, 
d6, d7}
 # DISASM-NEXT:   4661mov r1, r12
-# DISASM-NEXT:   f242    movw r0, #8192
+# DISASM-NEXT:   f242 0004   movw r0, #8196
 # DISASM-NEXT:   f2c0 0040   movtr0, #64
 # DISASM-NEXT:   f7ff ffe7   bl  0x401000 <.text>
 # DISASM-NEXT:   4684mov r12, r0
 # DISASM-NEXT:   ecbd 0b10   vpop{d0, d1, d2, d3, d4, d5, 
d6, d7}
 # DISASM-NEXT:   e8bd 480f   pop.w   {r0, r1, r2, r3, r11, lr}
 # DISASM-NEXT:   4760bx  r12
+#
+# DIR: DelayImportDescriptorRVA: 0x2004
+# DIR-NEXT:DelayImportDescriptorSize: 0x40
+
 
 --- !COFF
 header:
@@ -80,6 +85,14 @@ sections:
   - VirtualAddress:  0
 SymbolName:  __imp_function
 Type:IMAGE_REL_ARM_MOV32T
+  - Name:.rdata
+Characteristics: [ IMAGE_SCN_CNT_INITIALIZED_DATA, IMAGE_SCN_MEM_READ ]
+Alignment:   1
+SectionData: 01
+  - Name:.data
+Characteristics: [ IMAGE_SCN_CNT_INITIALIZED_DATA, IMAGE_SCN_MEM_READ, 
IMAGE_SCN_MEM_WRITE ]
+Alignment:   1
+SectionData: 02
 symbols:
   - Name:.text
 Value:   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] [lld] release/18.x: [LLD] [COFF] Set the right alignment for DelayDirectoryChunk (#84697) (PR #84844)

[via llvm-branch-commits]

https://github.com/llvmbot milestoned 
https://github.com/llvm/llvm-project/pull/84844
___
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][lld][RISCV] Support x3_reg_usage (PR #84598)

[Paul Kirth via llvm-branch-commits]

https://github.com/ilovepi updated 
https://github.com/llvm/llvm-project/pull/84598


___
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][lld][RISCV] Support x3_reg_usage (PR #84598)

[Paul Kirth via llvm-branch-commits]

https://github.com/ilovepi updated 
https://github.com/llvm/llvm-project/pull/84598


___
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] Backport ARM64EC variadic args fixes to LLVM 18 (PR #81800)

[Daniel Paoliello via llvm-branch-commits]

https://github.com/dpaoliello updated 
https://github.com/llvm/llvm-project/pull/81800

>From d7a9810f9c14e6598265ab41519be9b861228450 Mon Sep 17 00:00:00 2001
From: Billy Laws 
Date: Wed, 31 Jan 2024 02:32:15 +
Subject: [PATCH 1/3] [AArch64] Fix variadic tail-calls on ARM64EC (#79774)

ARM64EC varargs calls expect that x4 = sp at entry, special handling is
needed to ensure this with tail calls since they occur after the
epilogue and the x4 write happens before.

I tried going through AArch64MachineFrameLowering for this, hoping to
avoid creating the dummy object but this was the best I could do since
the stack info that uses isn't populated at this stage,
CreateFixedObject also explicitly forbids 0 sized objects.
---
 .../Target/AArch64/AArch64ISelLowering.cpp| 10 -
 llvm/test/CodeGen/AArch64/arm64ec-varargs.ll  | 37 +++
 llvm/test/CodeGen/AArch64/vararg-tallcall.ll  |  8 
 3 files changed, 54 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp 
b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 0287856560e91a..196aa50cf4060b 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -8007,11 +8007,19 @@ AArch64TargetLowering::LowerCall(CallLoweringInfo ,
   }
 
   if (IsVarArg && Subtarget->isWindowsArm64EC()) {
+SDValue ParamPtr = StackPtr;
+if (IsTailCall) {
+  // Create a dummy object at the top of the stack that can be used to get
+  // the SP after the epilogue
+  int FI = MF.getFrameInfo().CreateFixedObject(1, FPDiff, true);
+  ParamPtr = DAG.getFrameIndex(FI, PtrVT);
+}
+
 // For vararg calls, the Arm64EC ABI requires values in x4 and x5
 // describing the argument list.  x4 contains the address of the
 // first stack parameter. x5 contains the size in bytes of all parameters
 // passed on the stack.
-RegsToPass.emplace_back(AArch64::X4, StackPtr);
+RegsToPass.emplace_back(AArch64::X4, ParamPtr);
 RegsToPass.emplace_back(AArch64::X5,
 DAG.getConstant(NumBytes, DL, MVT::i64));
   }
diff --git a/llvm/test/CodeGen/AArch64/arm64ec-varargs.ll 
b/llvm/test/CodeGen/AArch64/arm64ec-varargs.ll
index dc16b3a1a0f270..844fc52ddade63 100644
--- a/llvm/test/CodeGen/AArch64/arm64ec-varargs.ll
+++ b/llvm/test/CodeGen/AArch64/arm64ec-varargs.ll
@@ -100,5 +100,42 @@ define void @varargs_many_argscalleer() nounwind {
   ret void
 }
 
+define void @varargs_caller_tail() nounwind {
+; CHECK-LABEL: varargs_caller_tail:
+; CHECK:// %bb.0:
+; CHECK-NEXT:sub sp, sp, #48
+; CHECK-NEXT:mov x4, sp
+; CHECK-NEXT:add x8, sp, #16
+; CHECK-NEXT:mov x9, #4617315517961601024// 
=0x4014
+; CHECK-NEXT:mov x0, #4607182418800017408// 
=0x3ff0
+; CHECK-NEXT:mov w1, #2  // =0x2
+; CHECK-NEXT:mov x2, #4613937818241073152// 
=0x4008
+; CHECK-NEXT:mov w3, #4  // =0x4
+; CHECK-NEXT:mov w5, #16 // =0x10
+; CHECK-NEXT:stp xzr, x30, [sp, #24] // 8-byte Folded 
Spill
+; CHECK-NEXT:stp x9, x8, [sp]
+; CHECK-NEXT:str xzr, [sp, #16]
+; CHECK-NEXT:.weak_anti_dep  varargs_callee
+; CHECK-NEXT:.set varargs_callee, "#varargs_callee"@WEAKREF
+; CHECK-NEXT:.weak_anti_dep  "#varargs_callee"
+; CHECK-NEXT:.set "#varargs_callee", varargs_callee@WEAKREF
+; CHECK-NEXT:bl  "#varargs_callee"
+; CHECK-NEXT:ldr x30, [sp, #32]  // 8-byte Folded 
Reload
+; CHECK-NEXT:add x4, sp, #48
+; CHECK-NEXT:mov x0, #4607182418800017408// 
=0x3ff0
+; CHECK-NEXT:mov w1, #4  // =0x4
+; CHECK-NEXT:mov w2, #3  // =0x3
+; CHECK-NEXT:mov w3, #2  // =0x2
+; CHECK-NEXT:mov x5, xzr
+; CHECK-NEXT:add sp, sp, #48
+; CHECK-NEXT:.weak_anti_dep  varargs_callee
+; CHECK-NEXT:.set varargs_callee, "#varargs_callee"@WEAKREF
+; CHECK-NEXT:.weak_anti_dep  "#varargs_callee"
+; CHECK-NEXT:.set "#varargs_callee", varargs_callee@WEAKREF
+; CHECK-NEXT:b   "#varargs_callee"
+  call void (double, ...) @varargs_callee(double 1.0, i32 2, double 3.0, i32 
4, double 5.0, <2 x double> )
+  tail call void (double, ...) @varargs_callee(double 1.0, i32 4, i32 3, i32 2)
+  ret void
+}
 
 declare void @llvm.va_start(ptr)
diff --git a/llvm/test/CodeGen/AArch64/vararg-tallcall.ll 
b/llvm/test/CodeGen/AArch64/vararg-tallcall.ll
index 2d6db1642247d7..812837639196e6 100644
--- a/llvm/test/CodeGen/AArch64/vararg-tallcall.ll
+++ b/llvm/test/CodeGen/AArch64/vararg-tallcall.ll
@@ -1,5 +1,6 @@
 ; RUN: llc -mtriple=aarch64-windows-msvc %s -o - | FileCheck %s
 ; 

[llvm-branch-commits] [lld] [llvm] Backport fixes for ARM64EC import libraries (PR #84590)

[Daniel Paoliello via llvm-branch-commits]

dpaoliello wrote:

Sorry, I've got to disagree with this: this is blocking Rust's use of Arm64EC 
since it can generate import libraries on-the-fly using this functionality 
(rather than requiring that users install the Windows SDK).

I've tested Arm64EC with Rust, and this is the only blocking issue that I've 
found.

https://github.com/llvm/llvm-project/pull/84590
___
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: [ARM] Switch to LiveRegUnits to fix r7 register allocation bug (PR #84475)

[via llvm-branch-commits]

AtariDreams wrote:

> This is, as far as I can tell, not a miscompile; probably not worth taking on 
> the 18.x branch.
> 
> (Also, it's usually not a good idea to open a PR for a cherry-pick before the 
> original patch is merged.)

On that last part, it was supposed to be merged but then split off in the event 
other changes had to be reverted. 

https://github.com/llvm/llvm-project/pull/84475
___
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: [LoongArch] Override LoongArchTargetLowering::getExtendForAtomicCmpSwapArg (#83656) (PR #83750)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/83750
___
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] ea6c457 - [LoongArch] Override LoongArchTargetLowering::getExtendForAtomicCmpSwapArg (#83656)

[Tom Stellard via llvm-branch-commits]

Author: Lu Weining
Date: 2024-03-11T14:19:24-07:00
New Revision: ea6c457b8dd2d0e6a7f05b4a5bdd2686085e1ec0

URL: 
https://github.com/llvm/llvm-project/commit/ea6c457b8dd2d0e6a7f05b4a5bdd2686085e1ec0
DIFF: 
https://github.com/llvm/llvm-project/commit/ea6c457b8dd2d0e6a7f05b4a5bdd2686085e1ec0.diff

LOG: [LoongArch] Override LoongArchTargetLowering::getExtendForAtomicCmpSwapArg 
(#83656)

This patch aims to solve Firefox issue:
https://bugzilla.mozilla.org/show_bug.cgi?id=1882301

Similar to 616289ed2922. Currently LoongArch uses an ll.[wd]/sc.[wd]
loop for ATOMIC_CMP_XCHG. Because the comparison in the loop is
full-width (i.e. the `bne` instruction), we must sign extend the input
comparsion argument.

Note that LoongArch ISA manual V1.1 has introduced compare-and-swap
instructions. We would change the implementation (return `ANY_EXTEND`)
when we support them.

(cherry picked from commit 5f058aa211995d2f0df2a0e063532832569cb7a8)

Added: 


Modified: 
llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
llvm/lib/Target/LoongArch/LoongArchISelLowering.h
llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
llvm/test/CodeGen/LoongArch/ir-instruction/atomic-cmpxchg.ll
llvm/test/CodeGen/LoongArch/ir-instruction/atomicrmw-fp.ll

Removed: 




diff  --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp 
b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index 76c1a14fe0156c..b161c5434ca13e 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -4940,3 +4940,8 @@ bool LoongArchTargetLowering::hasAndNotCompare(SDValue Y) 
const {
 
   return !isa(Y);
 }
+
+ISD::NodeType LoongArchTargetLowering::getExtendForAtomicCmpSwapArg() const {
+  // TODO: LAMCAS will use amcas{_DB,}.[bhwd] which does not require extension.
+  return ISD::SIGN_EXTEND;
+}

diff  --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h 
b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
index 72182623b2c3dd..9e9ac0b8269291 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
@@ -206,6 +206,8 @@ class LoongArchTargetLowering : public TargetLowering {
 return ISD::SIGN_EXTEND;
   }
 
+  ISD::NodeType getExtendForAtomicCmpSwapArg() const override;
+
   Register getRegisterByName(const char *RegName, LLT VT,
  const MachineFunction ) const override;
   bool mayBeEmittedAsTailCall(const CallInst *CI) const override;

diff  --git a/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll 
b/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
index b0f29ee790885d..b84c1093eb75f2 100644
--- a/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
+++ b/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
@@ -25,15 +25,16 @@ define i8 @atomicrmw_uinc_wrap_i8(ptr %ptr, i8 %val) {
 ; LA64-NEXT:andi $a5, $a5, 255
 ; LA64-NEXT:sll.w $a5, $a5, $a3
 ; LA64-NEXT:and $a6, $a2, $a4
-; LA64-NEXT:or $a6, $a6, $a5
+; LA64-NEXT:or $a5, $a6, $a5
+; LA64-NEXT:addi.w $a6, $a2, 0
 ; LA64-NEXT:  .LBB0_3: # %atomicrmw.start
 ; LA64-NEXT:# Parent Loop BB0_1 Depth=1
 ; LA64-NEXT:# => This Inner Loop Header: Depth=2
-; LA64-NEXT:ll.w $a5, $a0, 0
-; LA64-NEXT:bne $a5, $a2, .LBB0_5
+; LA64-NEXT:ll.w $a2, $a0, 0
+; LA64-NEXT:bne $a2, $a6, .LBB0_5
 ; LA64-NEXT:  # %bb.4: # %atomicrmw.start
 ; LA64-NEXT:# in Loop: Header=BB0_3 Depth=2
-; LA64-NEXT:move $a7, $a6
+; LA64-NEXT:move $a7, $a5
 ; LA64-NEXT:sc.w $a7, $a0, 0
 ; LA64-NEXT:beqz $a7, .LBB0_3
 ; LA64-NEXT:b .LBB0_6
@@ -42,11 +43,9 @@ define i8 @atomicrmw_uinc_wrap_i8(ptr %ptr, i8 %val) {
 ; LA64-NEXT:dbar 20
 ; LA64-NEXT:  .LBB0_6: # %atomicrmw.start
 ; LA64-NEXT:# in Loop: Header=BB0_1 Depth=1
-; LA64-NEXT:addi.w $a6, $a2, 0
-; LA64-NEXT:move $a2, $a5
-; LA64-NEXT:bne $a5, $a6, .LBB0_1
+; LA64-NEXT:bne $a2, $a6, .LBB0_1
 ; LA64-NEXT:  # %bb.2: # %atomicrmw.end
-; LA64-NEXT:srl.w $a0, $a5, $a3
+; LA64-NEXT:srl.w $a0, $a2, $a3
 ; LA64-NEXT:ret
   %result = atomicrmw uinc_wrap ptr %ptr, i8 %val seq_cst
   ret i8 %result
@@ -77,15 +76,16 @@ define i16 @atomicrmw_uinc_wrap_i16(ptr %ptr, i16 %val) {
 ; LA64-NEXT:bstrpick.d $a5, $a5, 15, 0
 ; LA64-NEXT:sll.w $a5, $a5, $a3
 ; LA64-NEXT:and $a6, $a2, $a4
-; LA64-NEXT:or $a6, $a6, $a5
+; LA64-NEXT:or $a5, $a6, $a5
+; LA64-NEXT:addi.w $a6, $a2, 0
 ; LA64-NEXT:  .LBB1_3: # %atomicrmw.start
 ; LA64-NEXT:# Parent Loop BB1_1 Depth=1
 ; LA64-NEXT:# => This Inner Loop Header: Depth=2
-; LA64-NEXT:ll.w $a5, $a0, 0
-; LA64-NEXT:bne $a5, $a2, .LBB1_5
+; LA64-NEXT:ll.w $a2, $a0, 0
+; LA64-NEXT:bne $a2, $a6, .LBB1_5
 ; LA64-NEXT:  # %bb.4: # %atomicrmw.start
 ; LA64-NEXT:# in Loop: Header=BB1_3 Depth=2
-; LA64-NEXT:move $a7, $a6
+; LA64-NEXT:

[llvm-branch-commits] [llvm] release/18.x: [LoongArch] Override LoongArchTargetLowering::getExtendForAtomicCmpSwapArg (#83656) (PR #83750)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/83750

>From ea6c457b8dd2d0e6a7f05b4a5bdd2686085e1ec0 Mon Sep 17 00:00:00 2001
From: Lu Weining 
Date: Mon, 4 Mar 2024 08:38:52 +0800
Subject: [PATCH] [LoongArch] Override
 LoongArchTargetLowering::getExtendForAtomicCmpSwapArg (#83656)

This patch aims to solve Firefox issue:
https://bugzilla.mozilla.org/show_bug.cgi?id=1882301

Similar to 616289ed2922. Currently LoongArch uses an ll.[wd]/sc.[wd]
loop for ATOMIC_CMP_XCHG. Because the comparison in the loop is
full-width (i.e. the `bne` instruction), we must sign extend the input
comparsion argument.

Note that LoongArch ISA manual V1.1 has introduced compare-and-swap
instructions. We would change the implementation (return `ANY_EXTEND`)
when we support them.

(cherry picked from commit 5f058aa211995d2f0df2a0e063532832569cb7a8)
---
 .../LoongArch/LoongArchISelLowering.cpp   |   5 +
 .../Target/LoongArch/LoongArchISelLowering.h  |   2 +
 .../LoongArch/atomicrmw-uinc-udec-wrap.ll | 120 +++--
 .../ir-instruction/atomic-cmpxchg.ll  |  25 +--
 .../LoongArch/ir-instruction/atomicrmw-fp.ll  | 160 +-
 5 files changed, 159 insertions(+), 153 deletions(-)

diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp 
b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index 76c1a14fe0156c..b161c5434ca13e 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -4940,3 +4940,8 @@ bool LoongArchTargetLowering::hasAndNotCompare(SDValue Y) 
const {
 
   return !isa(Y);
 }
+
+ISD::NodeType LoongArchTargetLowering::getExtendForAtomicCmpSwapArg() const {
+  // TODO: LAMCAS will use amcas{_DB,}.[bhwd] which does not require extension.
+  return ISD::SIGN_EXTEND;
+}
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h 
b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
index 72182623b2c3dd..9e9ac0b8269291 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
@@ -206,6 +206,8 @@ class LoongArchTargetLowering : public TargetLowering {
 return ISD::SIGN_EXTEND;
   }
 
+  ISD::NodeType getExtendForAtomicCmpSwapArg() const override;
+
   Register getRegisterByName(const char *RegName, LLT VT,
  const MachineFunction ) const override;
   bool mayBeEmittedAsTailCall(const CallInst *CI) const override;
diff --git a/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll 
b/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
index b0f29ee790885d..b84c1093eb75f2 100644
--- a/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
+++ b/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
@@ -25,15 +25,16 @@ define i8 @atomicrmw_uinc_wrap_i8(ptr %ptr, i8 %val) {
 ; LA64-NEXT:andi $a5, $a5, 255
 ; LA64-NEXT:sll.w $a5, $a5, $a3
 ; LA64-NEXT:and $a6, $a2, $a4
-; LA64-NEXT:or $a6, $a6, $a5
+; LA64-NEXT:or $a5, $a6, $a5
+; LA64-NEXT:addi.w $a6, $a2, 0
 ; LA64-NEXT:  .LBB0_3: # %atomicrmw.start
 ; LA64-NEXT:# Parent Loop BB0_1 Depth=1
 ; LA64-NEXT:# => This Inner Loop Header: Depth=2
-; LA64-NEXT:ll.w $a5, $a0, 0
-; LA64-NEXT:bne $a5, $a2, .LBB0_5
+; LA64-NEXT:ll.w $a2, $a0, 0
+; LA64-NEXT:bne $a2, $a6, .LBB0_5
 ; LA64-NEXT:  # %bb.4: # %atomicrmw.start
 ; LA64-NEXT:# in Loop: Header=BB0_3 Depth=2
-; LA64-NEXT:move $a7, $a6
+; LA64-NEXT:move $a7, $a5
 ; LA64-NEXT:sc.w $a7, $a0, 0
 ; LA64-NEXT:beqz $a7, .LBB0_3
 ; LA64-NEXT:b .LBB0_6
@@ -42,11 +43,9 @@ define i8 @atomicrmw_uinc_wrap_i8(ptr %ptr, i8 %val) {
 ; LA64-NEXT:dbar 20
 ; LA64-NEXT:  .LBB0_6: # %atomicrmw.start
 ; LA64-NEXT:# in Loop: Header=BB0_1 Depth=1
-; LA64-NEXT:addi.w $a6, $a2, 0
-; LA64-NEXT:move $a2, $a5
-; LA64-NEXT:bne $a5, $a6, .LBB0_1
+; LA64-NEXT:bne $a2, $a6, .LBB0_1
 ; LA64-NEXT:  # %bb.2: # %atomicrmw.end
-; LA64-NEXT:srl.w $a0, $a5, $a3
+; LA64-NEXT:srl.w $a0, $a2, $a3
 ; LA64-NEXT:ret
   %result = atomicrmw uinc_wrap ptr %ptr, i8 %val seq_cst
   ret i8 %result
@@ -77,15 +76,16 @@ define i16 @atomicrmw_uinc_wrap_i16(ptr %ptr, i16 %val) {
 ; LA64-NEXT:bstrpick.d $a5, $a5, 15, 0
 ; LA64-NEXT:sll.w $a5, $a5, $a3
 ; LA64-NEXT:and $a6, $a2, $a4
-; LA64-NEXT:or $a6, $a6, $a5
+; LA64-NEXT:or $a5, $a6, $a5
+; LA64-NEXT:addi.w $a6, $a2, 0
 ; LA64-NEXT:  .LBB1_3: # %atomicrmw.start
 ; LA64-NEXT:# Parent Loop BB1_1 Depth=1
 ; LA64-NEXT:# => This Inner Loop Header: Depth=2
-; LA64-NEXT:ll.w $a5, $a0, 0
-; LA64-NEXT:bne $a5, $a2, .LBB1_5
+; LA64-NEXT:ll.w $a2, $a0, 0
+; LA64-NEXT:bne $a2, $a6, .LBB1_5
 ; LA64-NEXT:  # %bb.4: # %atomicrmw.start
 ; LA64-NEXT:# in Loop: Header=BB1_3 Depth=2
-; LA64-NEXT:move $a7, $a6
+; LA64-NEXT:move $a7, $a5
 ; LA64-NEXT:sc.w $a7, $a0, 0
 ; LA64-NEXT:beqz $a7, .LBB1_3
 ; LA64-NEXT:b .LBB1_6
@@ -94,11 +94,9 @@ define 

[llvm-branch-commits] [llvm] release/18.x: [InstCombine] Fix miscompilation in PR83947 (#83993) (PR #84021)

[Tom Stellard via llvm-branch-commits]

tstellar wrote:

This is fixed in #84298.

https://github.com/llvm/llvm-project/pull/84021
___
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] [lld] [llvm] Backport fixes for ARM64EC import libraries (PR #84590)

[Jacek Caban via llvm-branch-commits]

cjacek wrote:

I have similar feeling to @efriedma-quic . There is probably not much use of 
those fixes on their own. They are a step towards stand-alone toolchain and are 
required for the linker, both for its libs part and to generate test cases, but 
they alone are not enough. They will be more useful when paired with linker 
support, which I hope will be in a useful shape for LLVM 19. Until then MSVC is 
required anyway.

https://github.com/llvm/llvm-project/pull/84590
___
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] Backport ARM64EC variadic args fixes to LLVM 18 (PR #81800)

[Daniel Paoliello via llvm-branch-commits]

https://github.com/dpaoliello updated 
https://github.com/llvm/llvm-project/pull/81800

>From a276993409f867760dbc1993acb54063fd8a0fbb Mon Sep 17 00:00:00 2001
From: Billy Laws 
Date: Wed, 31 Jan 2024 02:32:15 +
Subject: [PATCH 1/3] [AArch64] Fix variadic tail-calls on ARM64EC (#79774)

ARM64EC varargs calls expect that x4 = sp at entry, special handling is
needed to ensure this with tail calls since they occur after the
epilogue and the x4 write happens before.

I tried going through AArch64MachineFrameLowering for this, hoping to
avoid creating the dummy object but this was the best I could do since
the stack info that uses isn't populated at this stage,
CreateFixedObject also explicitly forbids 0 sized objects.
---
 .../Target/AArch64/AArch64ISelLowering.cpp| 10 -
 llvm/test/CodeGen/AArch64/arm64ec-varargs.ll  | 37 +++
 llvm/test/CodeGen/AArch64/vararg-tallcall.ll  |  8 
 3 files changed, 54 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp 
b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 0287856560e91a..196aa50cf4060b 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -8007,11 +8007,19 @@ AArch64TargetLowering::LowerCall(CallLoweringInfo ,
   }
 
   if (IsVarArg && Subtarget->isWindowsArm64EC()) {
+SDValue ParamPtr = StackPtr;
+if (IsTailCall) {
+  // Create a dummy object at the top of the stack that can be used to get
+  // the SP after the epilogue
+  int FI = MF.getFrameInfo().CreateFixedObject(1, FPDiff, true);
+  ParamPtr = DAG.getFrameIndex(FI, PtrVT);
+}
+
 // For vararg calls, the Arm64EC ABI requires values in x4 and x5
 // describing the argument list.  x4 contains the address of the
 // first stack parameter. x5 contains the size in bytes of all parameters
 // passed on the stack.
-RegsToPass.emplace_back(AArch64::X4, StackPtr);
+RegsToPass.emplace_back(AArch64::X4, ParamPtr);
 RegsToPass.emplace_back(AArch64::X5,
 DAG.getConstant(NumBytes, DL, MVT::i64));
   }
diff --git a/llvm/test/CodeGen/AArch64/arm64ec-varargs.ll 
b/llvm/test/CodeGen/AArch64/arm64ec-varargs.ll
index dc16b3a1a0f270..844fc52ddade63 100644
--- a/llvm/test/CodeGen/AArch64/arm64ec-varargs.ll
+++ b/llvm/test/CodeGen/AArch64/arm64ec-varargs.ll
@@ -100,5 +100,42 @@ define void @varargs_many_argscalleer() nounwind {
   ret void
 }
 
+define void @varargs_caller_tail() nounwind {
+; CHECK-LABEL: varargs_caller_tail:
+; CHECK:// %bb.0:
+; CHECK-NEXT:sub sp, sp, #48
+; CHECK-NEXT:mov x4, sp
+; CHECK-NEXT:add x8, sp, #16
+; CHECK-NEXT:mov x9, #4617315517961601024// 
=0x4014
+; CHECK-NEXT:mov x0, #4607182418800017408// 
=0x3ff0
+; CHECK-NEXT:mov w1, #2  // =0x2
+; CHECK-NEXT:mov x2, #4613937818241073152// 
=0x4008
+; CHECK-NEXT:mov w3, #4  // =0x4
+; CHECK-NEXT:mov w5, #16 // =0x10
+; CHECK-NEXT:stp xzr, x30, [sp, #24] // 8-byte Folded 
Spill
+; CHECK-NEXT:stp x9, x8, [sp]
+; CHECK-NEXT:str xzr, [sp, #16]
+; CHECK-NEXT:.weak_anti_dep  varargs_callee
+; CHECK-NEXT:.set varargs_callee, "#varargs_callee"@WEAKREF
+; CHECK-NEXT:.weak_anti_dep  "#varargs_callee"
+; CHECK-NEXT:.set "#varargs_callee", varargs_callee@WEAKREF
+; CHECK-NEXT:bl  "#varargs_callee"
+; CHECK-NEXT:ldr x30, [sp, #32]  // 8-byte Folded 
Reload
+; CHECK-NEXT:add x4, sp, #48
+; CHECK-NEXT:mov x0, #4607182418800017408// 
=0x3ff0
+; CHECK-NEXT:mov w1, #4  // =0x4
+; CHECK-NEXT:mov w2, #3  // =0x3
+; CHECK-NEXT:mov w3, #2  // =0x2
+; CHECK-NEXT:mov x5, xzr
+; CHECK-NEXT:add sp, sp, #48
+; CHECK-NEXT:.weak_anti_dep  varargs_callee
+; CHECK-NEXT:.set varargs_callee, "#varargs_callee"@WEAKREF
+; CHECK-NEXT:.weak_anti_dep  "#varargs_callee"
+; CHECK-NEXT:.set "#varargs_callee", varargs_callee@WEAKREF
+; CHECK-NEXT:b   "#varargs_callee"
+  call void (double, ...) @varargs_callee(double 1.0, i32 2, double 3.0, i32 
4, double 5.0, <2 x double> )
+  tail call void (double, ...) @varargs_callee(double 1.0, i32 4, i32 3, i32 2)
+  ret void
+}
 
 declare void @llvm.va_start(ptr)
diff --git a/llvm/test/CodeGen/AArch64/vararg-tallcall.ll 
b/llvm/test/CodeGen/AArch64/vararg-tallcall.ll
index 2d6db1642247d7..812837639196e6 100644
--- a/llvm/test/CodeGen/AArch64/vararg-tallcall.ll
+++ b/llvm/test/CodeGen/AArch64/vararg-tallcall.ll
@@ -1,5 +1,6 @@
 ; RUN: llc -mtriple=aarch64-windows-msvc %s -o - | FileCheck %s
 ; 

[llvm-branch-commits] [lld] [llvm] Backport fixes for ARM64EC import libraries (PR #84590)

[Daniel Paoliello via llvm-branch-commits]

dpaoliello wrote:

Blocking to wait for 

https://github.com/llvm/llvm-project/pull/84590
___
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] Backport PR83993 to 18.x (PR #84298)

[Nikita Popov via llvm-branch-commits]

https://github.com/nikic approved this pull request.


https://github.com/llvm/llvm-project/pull/84298
___
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 wrong codegen of BothFusionPredicateWithMCInstPredicate (#83990) (PR #83999)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/83999
___
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] 69d9b15 - [TableGen] Fix wrong codegen of BothFusionPredicateWithMCInstPredicate (#83990)

[Tom Stellard via llvm-branch-commits]

Author: Wang Pengcheng
Date: 2024-03-11T13:57:41-07:00
New Revision: 69d9b15fe872bb188474b0ad9e36c8506bdf9cc3

URL: 
https://github.com/llvm/llvm-project/commit/69d9b15fe872bb188474b0ad9e36c8506bdf9cc3
DIFF: 
https://github.com/llvm/llvm-project/commit/69d9b15fe872bb188474b0ad9e36c8506bdf9cc3.diff

LOG: [TableGen] Fix wrong codegen of BothFusionPredicateWithMCInstPredicate 
(#83990)

We should generate the `MCInstPredicate` twice, one with `FirstMI`
and another with `SecondMI`.

(cherry picked from commit de1f33873beff93063577195e1214a9509e229e0)

Added: 


Modified: 
llvm/include/llvm/Target/TargetSchedule.td
llvm/test/TableGen/MacroFusion.td
llvm/utils/TableGen/MacroFusionPredicatorEmitter.cpp

Removed: 




diff  --git a/llvm/include/llvm/Target/TargetSchedule.td 
b/llvm/include/llvm/Target/TargetSchedule.td
index 032de728517827..40c2cce8c6effe 100644
--- a/llvm/include/llvm/Target/TargetSchedule.td
+++ b/llvm/include/llvm/Target/TargetSchedule.td
@@ -620,7 +620,7 @@ class 
SecondFusionPredicateWithMCInstPredicate
   : FusionPredicateWithMCInstPredicate;
 // The pred will be applied on both firstMI and secondMI.
 class BothFusionPredicateWithMCInstPredicate
-  : FusionPredicateWithMCInstPredicate;
+  : FusionPredicateWithMCInstPredicate;
 
 // Tie firstOpIdx and secondOpIdx. The operand of `FirstMI` at position
 // `firstOpIdx` should be the same as the operand of `SecondMI` at position

diff  --git a/llvm/test/TableGen/MacroFusion.td 
b/llvm/test/TableGen/MacroFusion.td
index 4aa6c8d9acb273..ce76e7f0f7fa64 100644
--- a/llvm/test/TableGen/MacroFusion.td
+++ b/llvm/test/TableGen/MacroFusion.td
@@ -34,6 +34,11 @@ let Namespace = "Test" in {
 def Inst0 : TestInst<0>;
 def Inst1 : TestInst<1>;
 
+def BothFusionPredicate: 
BothFusionPredicateWithMCInstPredicate>;
+def TestBothFusionPredicate: Fusion<"test-both-fusion-predicate", 
"HasBothFusionPredicate",
+"Test BothFusionPredicate",
+[BothFusionPredicate]>;
+
 def TestFusion: SimpleFusion<"test-fusion", "HasTestFusion", "Test Fusion",
  CheckOpcode<[Inst0]>,
  CheckAll<[
@@ -45,6 +50,7 @@ def TestFusion: SimpleFusion<"test-fusion", "HasTestFusion", 
"Test Fusion",
 // CHECK-PREDICATOR-NEXT:  #undef GET_Test_MACRO_FUSION_PRED_DECL
 // CHECK-PREDICATOR-EMPTY:
 // CHECK-PREDICATOR-NEXT:  namespace llvm {
+// CHECK-PREDICATOR-NEXT:  bool isTestBothFusionPredicate(const 
TargetInstrInfo &, const TargetSubtargetInfo &, const MachineInstr *, const 
MachineInstr &);
 // CHECK-PREDICATOR-NEXT:  bool isTestFusion(const TargetInstrInfo &, const 
TargetSubtargetInfo &, const MachineInstr *, const MachineInstr &);
 // CHECK-PREDICATOR-NEXT:  } // end namespace llvm
 // CHECK-PREDICATOR-EMPTY:
@@ -54,6 +60,24 @@ def TestFusion: SimpleFusion<"test-fusion", "HasTestFusion", 
"Test Fusion",
 // CHECK-PREDICATOR-NEXT:  #undef GET_Test_MACRO_FUSION_PRED_IMPL
 // CHECK-PREDICATOR-EMPTY:
 // CHECK-PREDICATOR-NEXT:  namespace llvm {
+// CHECK-PREDICATOR-NEXT:  bool isTestBothFusionPredicate(
+// CHECK-PREDICATOR-NEXT:  const TargetInstrInfo ,
+// CHECK-PREDICATOR-NEXT:  const TargetSubtargetInfo ,
+// CHECK-PREDICATOR-NEXT:  const MachineInstr *FirstMI,
+// CHECK-PREDICATOR-NEXT:  const MachineInstr ) {
+// CHECK-PREDICATOR-NEXT:auto  = SecondMI.getMF()->getRegInfo();
+// CHECK-PREDICATOR-NEXT:{
+// CHECK-PREDICATOR-NEXT:  const MachineInstr *MI = FirstMI;
+// CHECK-PREDICATOR-NEXT:  if (MI->getOperand(0).getReg() != Test::X0)
+// CHECK-PREDICATOR-NEXT:return false;
+// CHECK-PREDICATOR-NEXT:}
+// CHECK-PREDICATOR-NEXT:{
+// CHECK-PREDICATOR-NEXT:  const MachineInstr *MI = 
+// CHECK-PREDICATOR-NEXT:  if (MI->getOperand(0).getReg() != Test::X0)
+// CHECK-PREDICATOR-NEXT:return false;
+// CHECK-PREDICATOR-NEXT:}
+// CHECK-PREDICATOR-NEXT:return true;
+// CHECK-PREDICATOR-NEXT:  }
 // CHECK-PREDICATOR-NEXT:  bool isTestFusion(
 // CHECK-PREDICATOR-NEXT:  const TargetInstrInfo ,
 // CHECK-PREDICATOR-NEXT:  const TargetSubtargetInfo ,
@@ -106,6 +130,7 @@ def TestFusion: SimpleFusion<"test-fusion", 
"HasTestFusion", "Test Fusion",
 
 // CHECK-SUBTARGET:  std::vector 
TestGenSubtargetInfo::getMacroFusions() const {
 // CHECK-SUBTARGET-NEXT:   std::vector Fusions;
+// CHECK-SUBTARGET-NEXT:   if (hasFeature(Test::TestBothFusionPredicate)) 
Fusions.push_back(llvm::isTestBothFusionPredicate); 
 // CHECK-SUBTARGET-NEXT:   if (hasFeature(Test::TestFusion)) 
Fusions.push_back(llvm::isTestFusion);
 // CHECK-SUBTARGET-NEXT:   return Fusions;
 // CHECK-SUBTARGET-NEXT: }

diff  --git a/llvm/utils/TableGen/MacroFusionPredicatorEmitter.cpp 
b/llvm/utils/TableGen/MacroFusionPredicatorEmitter.cpp
index 78dcd4471ae747..7f494e532b1f44 100644
--- 

[llvm-branch-commits] [llvm] release/18.x: [TableGen] Fix wrong codegen of BothFusionPredicateWithMCInstPredicate (#83990) (PR #83999)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/83999

>From 69d9b15fe872bb188474b0ad9e36c8506bdf9cc3 Mon Sep 17 00:00:00 2001
From: Wang Pengcheng 
Date: Tue, 5 Mar 2024 19:54:02 +0800
Subject: [PATCH] [TableGen] Fix wrong codegen of
 BothFusionPredicateWithMCInstPredicate (#83990)

We should generate the `MCInstPredicate` twice, one with `FirstMI`
and another with `SecondMI`.

(cherry picked from commit de1f33873beff93063577195e1214a9509e229e0)
---
 llvm/include/llvm/Target/TargetSchedule.td|  2 +-
 llvm/test/TableGen/MacroFusion.td | 25 +++
 .../TableGen/MacroFusionPredicatorEmitter.cpp | 12 -
 3 files changed, 31 insertions(+), 8 deletions(-)

diff --git a/llvm/include/llvm/Target/TargetSchedule.td 
b/llvm/include/llvm/Target/TargetSchedule.td
index 032de728517827..40c2cce8c6effe 100644
--- a/llvm/include/llvm/Target/TargetSchedule.td
+++ b/llvm/include/llvm/Target/TargetSchedule.td
@@ -620,7 +620,7 @@ class 
SecondFusionPredicateWithMCInstPredicate
   : FusionPredicateWithMCInstPredicate;
 // The pred will be applied on both firstMI and secondMI.
 class BothFusionPredicateWithMCInstPredicate
-  : FusionPredicateWithMCInstPredicate;
+  : FusionPredicateWithMCInstPredicate;
 
 // Tie firstOpIdx and secondOpIdx. The operand of `FirstMI` at position
 // `firstOpIdx` should be the same as the operand of `SecondMI` at position
diff --git a/llvm/test/TableGen/MacroFusion.td 
b/llvm/test/TableGen/MacroFusion.td
index 4aa6c8d9acb273..ce76e7f0f7fa64 100644
--- a/llvm/test/TableGen/MacroFusion.td
+++ b/llvm/test/TableGen/MacroFusion.td
@@ -34,6 +34,11 @@ let Namespace = "Test" in {
 def Inst0 : TestInst<0>;
 def Inst1 : TestInst<1>;
 
+def BothFusionPredicate: 
BothFusionPredicateWithMCInstPredicate>;
+def TestBothFusionPredicate: Fusion<"test-both-fusion-predicate", 
"HasBothFusionPredicate",
+"Test BothFusionPredicate",
+[BothFusionPredicate]>;
+
 def TestFusion: SimpleFusion<"test-fusion", "HasTestFusion", "Test Fusion",
  CheckOpcode<[Inst0]>,
  CheckAll<[
@@ -45,6 +50,7 @@ def TestFusion: SimpleFusion<"test-fusion", "HasTestFusion", 
"Test Fusion",
 // CHECK-PREDICATOR-NEXT:  #undef GET_Test_MACRO_FUSION_PRED_DECL
 // CHECK-PREDICATOR-EMPTY:
 // CHECK-PREDICATOR-NEXT:  namespace llvm {
+// CHECK-PREDICATOR-NEXT:  bool isTestBothFusionPredicate(const 
TargetInstrInfo &, const TargetSubtargetInfo &, const MachineInstr *, const 
MachineInstr &);
 // CHECK-PREDICATOR-NEXT:  bool isTestFusion(const TargetInstrInfo &, const 
TargetSubtargetInfo &, const MachineInstr *, const MachineInstr &);
 // CHECK-PREDICATOR-NEXT:  } // end namespace llvm
 // CHECK-PREDICATOR-EMPTY:
@@ -54,6 +60,24 @@ def TestFusion: SimpleFusion<"test-fusion", "HasTestFusion", 
"Test Fusion",
 // CHECK-PREDICATOR-NEXT:  #undef GET_Test_MACRO_FUSION_PRED_IMPL
 // CHECK-PREDICATOR-EMPTY:
 // CHECK-PREDICATOR-NEXT:  namespace llvm {
+// CHECK-PREDICATOR-NEXT:  bool isTestBothFusionPredicate(
+// CHECK-PREDICATOR-NEXT:  const TargetInstrInfo ,
+// CHECK-PREDICATOR-NEXT:  const TargetSubtargetInfo ,
+// CHECK-PREDICATOR-NEXT:  const MachineInstr *FirstMI,
+// CHECK-PREDICATOR-NEXT:  const MachineInstr ) {
+// CHECK-PREDICATOR-NEXT:auto  = SecondMI.getMF()->getRegInfo();
+// CHECK-PREDICATOR-NEXT:{
+// CHECK-PREDICATOR-NEXT:  const MachineInstr *MI = FirstMI;
+// CHECK-PREDICATOR-NEXT:  if (MI->getOperand(0).getReg() != Test::X0)
+// CHECK-PREDICATOR-NEXT:return false;
+// CHECK-PREDICATOR-NEXT:}
+// CHECK-PREDICATOR-NEXT:{
+// CHECK-PREDICATOR-NEXT:  const MachineInstr *MI = 
+// CHECK-PREDICATOR-NEXT:  if (MI->getOperand(0).getReg() != Test::X0)
+// CHECK-PREDICATOR-NEXT:return false;
+// CHECK-PREDICATOR-NEXT:}
+// CHECK-PREDICATOR-NEXT:return true;
+// CHECK-PREDICATOR-NEXT:  }
 // CHECK-PREDICATOR-NEXT:  bool isTestFusion(
 // CHECK-PREDICATOR-NEXT:  const TargetInstrInfo ,
 // CHECK-PREDICATOR-NEXT:  const TargetSubtargetInfo ,
@@ -106,6 +130,7 @@ def TestFusion: SimpleFusion<"test-fusion", 
"HasTestFusion", "Test Fusion",
 
 // CHECK-SUBTARGET:  std::vector 
TestGenSubtargetInfo::getMacroFusions() const {
 // CHECK-SUBTARGET-NEXT:   std::vector Fusions;
+// CHECK-SUBTARGET-NEXT:   if (hasFeature(Test::TestBothFusionPredicate)) 
Fusions.push_back(llvm::isTestBothFusionPredicate); 
 // CHECK-SUBTARGET-NEXT:   if (hasFeature(Test::TestFusion)) 
Fusions.push_back(llvm::isTestFusion);
 // CHECK-SUBTARGET-NEXT:   return Fusions;
 // CHECK-SUBTARGET-NEXT: }
diff --git a/llvm/utils/TableGen/MacroFusionPredicatorEmitter.cpp 
b/llvm/utils/TableGen/MacroFusionPredicatorEmitter.cpp
index 78dcd4471ae747..7f494e532b1f44 100644
--- a/llvm/utils/TableGen/MacroFusionPredicatorEmitter.cpp
+++ b/llvm/utils/TableGen/MacroFusionPredicatorEmitter.cpp
@@ -152,8 +152,7 @@ void 

[llvm-branch-commits] [lld] [llvm] Backport fixes for ARM64EC import libraries (PR #84590)

[Eli Friedman via llvm-branch-commits]

efriedma-quic wrote:

This seems like a pretty big change to backport... how useful is it in 
practice?  I was under the impression that arm64ec lld support is still 
immature... and if you're using the MSVC linker, you might as well use the MSVC 
lib/dlltool.

https://github.com/llvm/llvm-project/pull/84590
___
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] release/18.x: [OpenMP] fix endianness dependent definitions in OMP headers for MSVC (#84540) (PR #84668)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/84668
___
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] a91b9bd - [OpenMP] fix endianness dependent definitions in OMP headers for MSVC (#84540)

[Tom Stellard via llvm-branch-commits]

Author: Vadim Paretsky
Date: 2024-03-11T13:42:47-07:00
New Revision: a91b9bd9c7507b378a7f318db52484c8bacc12eb

URL: 
https://github.com/llvm/llvm-project/commit/a91b9bd9c7507b378a7f318db52484c8bacc12eb
DIFF: 
https://github.com/llvm/llvm-project/commit/a91b9bd9c7507b378a7f318db52484c8bacc12eb.diff

LOG: [OpenMP] fix endianness dependent definitions in OMP headers for MSVC 
(#84540)

MSVC does not define __BYTE_ORDER__ making the check for BigEndian
erroneously evaluate to true and breaking the struct definitions in MSVC
compiled builds correspondingly. The fix adds an additional check for
whether __BYTE_ORDER__ is defined by the compiler to fix these.

-

Co-authored-by: Vadim Paretsky 
(cherry picked from commit 110141b37813dc48af33de5e1407231e56acdfc5)

Added: 


Modified: 
openmp/runtime/src/kmp.h
openmp/runtime/src/kmp_lock.h
openmp/runtime/test/tasking/bug_nested_proxy_task.c
openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
openmp/runtime/test/tasking/hidden_helper_task/common.h

Removed: 




diff  --git a/openmp/runtime/src/kmp.h b/openmp/runtime/src/kmp.h
index 259c57b5afbca5..e3a1e20731bbe0 100644
--- a/openmp/runtime/src/kmp.h
+++ b/openmp/runtime/src/kmp.h
@@ -2506,7 +2506,7 @@ typedef struct kmp_depend_info {
   union {
 kmp_uint8 flag; // flag as an unsigned char
 struct { // flag as a set of 8 bits
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   /* Same fields as in the #else branch, but in reverse order */
   unsigned all : 1;
   unsigned unused : 3;
@@ -2671,7 +2671,7 @@ typedef struct kmp_task_stack {
 #endif // BUILD_TIED_TASK_STACK
 
 typedef struct kmp_tasking_flags { /* Total struct must be exactly 32 bits */
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   /* Same fields as in the #else branch, but in reverse order */
 #if OMPX_TASKGRAPH
   unsigned reserved31 : 6;

diff  --git a/openmp/runtime/src/kmp_lock.h b/openmp/runtime/src/kmp_lock.h
index e2a0cda01a9718..6202f3d617cc59 100644
--- a/openmp/runtime/src/kmp_lock.h
+++ b/openmp/runtime/src/kmp_lock.h
@@ -120,7 +120,8 @@ extern void __kmp_validate_locks(void);
 
 struct kmp_base_tas_lock {
   // KMP_LOCK_FREE(tas) => unlocked; locked: (gtid+1) of owning thread
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ && __LP64__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) && 
\
+__LP64__
   // Flip the ordering of the high and low 32-bit member to be consistent
   // with the memory layout of the address in 64-bit big-endian.
   kmp_int32 depth_locked; // depth locked, for nested locks only

diff  --git a/openmp/runtime/test/tasking/bug_nested_proxy_task.c 
b/openmp/runtime/test/tasking/bug_nested_proxy_task.c
index 24fe1f3fe7607c..9e0b412efce609 100644
--- a/openmp/runtime/test/tasking/bug_nested_proxy_task.c
+++ b/openmp/runtime/test/tasking/bug_nested_proxy_task.c
@@ -50,7 +50,7 @@ typedef struct kmp_depend_info {
  union {
 kmp_uint8 flag; // flag as an unsigned char
 struct { // flag as a set of 8 bits
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   unsigned all : 1;
   unsigned unused : 3;
   unsigned set : 1;

diff  --git a/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c 
b/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
index 688860c035728f..1e86d574f4f6a8 100644
--- a/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
+++ b/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
@@ -47,7 +47,7 @@ typedef struct kmp_depend_info {
  union {
 kmp_uint8 flag; // flag as an unsigned char
 struct { // flag as a set of 8 bits
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   unsigned all : 1;
   unsigned unused : 3;
   unsigned set : 1;

diff  --git a/openmp/runtime/test/tasking/hidden_helper_task/common.h 
b/openmp/runtime/test/tasking/hidden_helper_task/common.h
index ba57656cbac41d..68e2b584c87739 100644
--- a/openmp/runtime/test/tasking/hidden_helper_task/common.h
+++ b/openmp/runtime/test/tasking/hidden_helper_task/common.h
@@ -17,7 +17,7 @@ typedef struct kmp_depend_info {
   union {
 unsigned char flag;
 struct {
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   unsigned all : 1;
   unsigned unused : 3;
   unsigned set : 1;



___
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] release/18.x: [OpenMP] fix endianness dependent definitions in OMP headers for MSVC (#84540) (PR #84668)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/84668

>From a91b9bd9c7507b378a7f318db52484c8bacc12eb Mon Sep 17 00:00:00 2001
From: Vadim Paretsky 
Date: Sat, 9 Mar 2024 10:47:31 -0800
Subject: [PATCH] [OpenMP] fix endianness dependent definitions in OMP headers
 for MSVC (#84540)

MSVC does not define __BYTE_ORDER__ making the check for BigEndian
erroneously evaluate to true and breaking the struct definitions in MSVC
compiled builds correspondingly. The fix adds an additional check for
whether __BYTE_ORDER__ is defined by the compiler to fix these.

-

Co-authored-by: Vadim Paretsky 
(cherry picked from commit 110141b37813dc48af33de5e1407231e56acdfc5)
---
 openmp/runtime/src/kmp.h | 4 ++--
 openmp/runtime/src/kmp_lock.h| 3 ++-
 openmp/runtime/test/tasking/bug_nested_proxy_task.c  | 2 +-
 openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c | 2 +-
 openmp/runtime/test/tasking/hidden_helper_task/common.h  | 2 +-
 5 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/openmp/runtime/src/kmp.h b/openmp/runtime/src/kmp.h
index 259c57b5afbca5..e3a1e20731bbe0 100644
--- a/openmp/runtime/src/kmp.h
+++ b/openmp/runtime/src/kmp.h
@@ -2506,7 +2506,7 @@ typedef struct kmp_depend_info {
   union {
 kmp_uint8 flag; // flag as an unsigned char
 struct { // flag as a set of 8 bits
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   /* Same fields as in the #else branch, but in reverse order */
   unsigned all : 1;
   unsigned unused : 3;
@@ -2671,7 +2671,7 @@ typedef struct kmp_task_stack {
 #endif // BUILD_TIED_TASK_STACK
 
 typedef struct kmp_tasking_flags { /* Total struct must be exactly 32 bits */
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   /* Same fields as in the #else branch, but in reverse order */
 #if OMPX_TASKGRAPH
   unsigned reserved31 : 6;
diff --git a/openmp/runtime/src/kmp_lock.h b/openmp/runtime/src/kmp_lock.h
index e2a0cda01a9718..6202f3d617cc59 100644
--- a/openmp/runtime/src/kmp_lock.h
+++ b/openmp/runtime/src/kmp_lock.h
@@ -120,7 +120,8 @@ extern void __kmp_validate_locks(void);
 
 struct kmp_base_tas_lock {
   // KMP_LOCK_FREE(tas) => unlocked; locked: (gtid+1) of owning thread
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ && __LP64__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) && 
\
+__LP64__
   // Flip the ordering of the high and low 32-bit member to be consistent
   // with the memory layout of the address in 64-bit big-endian.
   kmp_int32 depth_locked; // depth locked, for nested locks only
diff --git a/openmp/runtime/test/tasking/bug_nested_proxy_task.c 
b/openmp/runtime/test/tasking/bug_nested_proxy_task.c
index 24fe1f3fe7607c..9e0b412efce609 100644
--- a/openmp/runtime/test/tasking/bug_nested_proxy_task.c
+++ b/openmp/runtime/test/tasking/bug_nested_proxy_task.c
@@ -50,7 +50,7 @@ typedef struct kmp_depend_info {
  union {
 kmp_uint8 flag; // flag as an unsigned char
 struct { // flag as a set of 8 bits
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   unsigned all : 1;
   unsigned unused : 3;
   unsigned set : 1;
diff --git a/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c 
b/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
index 688860c035728f..1e86d574f4f6a8 100644
--- a/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
+++ b/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
@@ -47,7 +47,7 @@ typedef struct kmp_depend_info {
  union {
 kmp_uint8 flag; // flag as an unsigned char
 struct { // flag as a set of 8 bits
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   unsigned all : 1;
   unsigned unused : 3;
   unsigned set : 1;
diff --git a/openmp/runtime/test/tasking/hidden_helper_task/common.h 
b/openmp/runtime/test/tasking/hidden_helper_task/common.h
index ba57656cbac41d..68e2b584c87739 100644
--- a/openmp/runtime/test/tasking/hidden_helper_task/common.h
+++ b/openmp/runtime/test/tasking/hidden_helper_task/common.h
@@ -17,7 +17,7 @@ typedef struct kmp_depend_info {
   union {
 unsigned char flag;
 struct {
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   unsigned all : 1;
   unsigned unused : 3;
   unsigned set : 1;

___
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] release/18.x: [OpenMP] fix endianness dependent definitions in OMP headers for MSVC (#84540) (PR #84668)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/84668

>From 39c1f0de2011dec072fdf64ee25b110eab945fc2 Mon Sep 17 00:00:00 2001
From: Vadim Paretsky 
Date: Sat, 9 Mar 2024 10:47:31 -0800
Subject: [PATCH] [OpenMP] fix endianness dependent definitions in OMP headers
 for MSVC (#84540)

MSVC does not define __BYTE_ORDER__ making the check for BigEndian
erroneously evaluate to true and breaking the struct definitions in MSVC
compiled builds correspondingly. The fix adds an additional check for
whether __BYTE_ORDER__ is defined by the compiler to fix these.

-

Co-authored-by: Vadim Paretsky 
(cherry picked from commit 110141b37813dc48af33de5e1407231e56acdfc5)
---
 openmp/runtime/src/kmp.h | 4 ++--
 openmp/runtime/src/kmp_lock.h| 3 ++-
 openmp/runtime/test/tasking/bug_nested_proxy_task.c  | 2 +-
 openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c | 2 +-
 openmp/runtime/test/tasking/hidden_helper_task/common.h  | 2 +-
 5 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/openmp/runtime/src/kmp.h b/openmp/runtime/src/kmp.h
index 259c57b5afbca5..e3a1e20731bbe0 100644
--- a/openmp/runtime/src/kmp.h
+++ b/openmp/runtime/src/kmp.h
@@ -2506,7 +2506,7 @@ typedef struct kmp_depend_info {
   union {
 kmp_uint8 flag; // flag as an unsigned char
 struct { // flag as a set of 8 bits
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   /* Same fields as in the #else branch, but in reverse order */
   unsigned all : 1;
   unsigned unused : 3;
@@ -2671,7 +2671,7 @@ typedef struct kmp_task_stack {
 #endif // BUILD_TIED_TASK_STACK
 
 typedef struct kmp_tasking_flags { /* Total struct must be exactly 32 bits */
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   /* Same fields as in the #else branch, but in reverse order */
 #if OMPX_TASKGRAPH
   unsigned reserved31 : 6;
diff --git a/openmp/runtime/src/kmp_lock.h b/openmp/runtime/src/kmp_lock.h
index e2a0cda01a9718..6202f3d617cc59 100644
--- a/openmp/runtime/src/kmp_lock.h
+++ b/openmp/runtime/src/kmp_lock.h
@@ -120,7 +120,8 @@ extern void __kmp_validate_locks(void);
 
 struct kmp_base_tas_lock {
   // KMP_LOCK_FREE(tas) => unlocked; locked: (gtid+1) of owning thread
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ && __LP64__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) && 
\
+__LP64__
   // Flip the ordering of the high and low 32-bit member to be consistent
   // with the memory layout of the address in 64-bit big-endian.
   kmp_int32 depth_locked; // depth locked, for nested locks only
diff --git a/openmp/runtime/test/tasking/bug_nested_proxy_task.c 
b/openmp/runtime/test/tasking/bug_nested_proxy_task.c
index 24fe1f3fe7607c..9e0b412efce609 100644
--- a/openmp/runtime/test/tasking/bug_nested_proxy_task.c
+++ b/openmp/runtime/test/tasking/bug_nested_proxy_task.c
@@ -50,7 +50,7 @@ typedef struct kmp_depend_info {
  union {
 kmp_uint8 flag; // flag as an unsigned char
 struct { // flag as a set of 8 bits
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   unsigned all : 1;
   unsigned unused : 3;
   unsigned set : 1;
diff --git a/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c 
b/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
index 688860c035728f..1e86d574f4f6a8 100644
--- a/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
+++ b/openmp/runtime/test/tasking/bug_proxy_task_dep_waiting.c
@@ -47,7 +47,7 @@ typedef struct kmp_depend_info {
  union {
 kmp_uint8 flag; // flag as an unsigned char
 struct { // flag as a set of 8 bits
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   unsigned all : 1;
   unsigned unused : 3;
   unsigned set : 1;
diff --git a/openmp/runtime/test/tasking/hidden_helper_task/common.h 
b/openmp/runtime/test/tasking/hidden_helper_task/common.h
index ba57656cbac41d..68e2b584c87739 100644
--- a/openmp/runtime/test/tasking/hidden_helper_task/common.h
+++ b/openmp/runtime/test/tasking/hidden_helper_task/common.h
@@ -17,7 +17,7 @@ typedef struct kmp_depend_info {
   union {
 unsigned char flag;
 struct {
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   unsigned all : 1;
   unsigned unused : 3;
   unsigned set : 1;

___
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] ReleaseNotes for LLVM binary utilities (PR #83751)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/83751
___
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] 7cb6753 - ReleaseNotes for LLVM binary utilities (#83751)

[via llvm-branch-commits]

Author: Fangrui Song
Date: 2024-03-11T13:40:01-07:00
New Revision: 7cb67530d2e9b7364e81f0d35da99a2855e391ac

URL: 
https://github.com/llvm/llvm-project/commit/7cb67530d2e9b7364e81f0d35da99a2855e391ac
DIFF: 
https://github.com/llvm/llvm-project/commit/7cb67530d2e9b7364e81f0d35da99a2855e391ac.diff

LOG: ReleaseNotes for LLVM binary utilities (#83751)

Added: 


Modified: 
llvm/docs/ReleaseNotes.rst

Removed: 




diff  --git a/llvm/docs/ReleaseNotes.rst b/llvm/docs/ReleaseNotes.rst
index 5b3210138f2f89..bfa8e93da05cb8 100644
--- a/llvm/docs/ReleaseNotes.rst
+++ b/llvm/docs/ReleaseNotes.rst
@@ -339,36 +339,41 @@ Changes to the Debug Info
 Changes to the LLVM tools
 -
 
-* llvm-symbolizer now treats invalid input as an address for which source
+* ``llvm-symbolizer`` now treats invalid input as an address for which source
   information is not found.
-* Fixed big-endian support in llvm-symbolizer's DWARF location parser.
-* llvm-readelf now supports ``--extra-sym-info`` (``-X``) to display extra
+* Fixed big-endian support in ``llvm-symbolizer``'s DWARF location parser.
+* ``llvm-readelf`` now supports ``--extra-sym-info`` (``-X``) to display extra
   information (section name) when showing symbols.
+* ``llvm-readobj``/``llvm-readelf`` now supports ``--decompress``/``-z`` with
+  string and hex dump for ELF object files.
 
-* ``llvm-nm`` now supports the ``--line-numbers`` (``-l``) option to use
-  debugging information to print symbols' filenames and line numbers.
-
-* llvm-symbolizer and llvm-addr2line now support addresses specified as symbol 
names.
+* ``llvm-symbolizer`` and ``llvm-addr2line`` now support addresses specified 
as symbol names.
 
-* llvm-objcopy now supports ``--gap-fill`` and ``--pad-to`` options, for
+* ``llvm-objcopy`` now supports ``--gap-fill`` and ``--pad-to`` options, for
   ELF input and binary output files only.
+* ``llvm-objcopy`` now supports ``-O elf64-s390`` for SystemZ.
 
-* Supported parsing XCOFF auxiliary symbols in obj2yaml.
+* Supported parsing XCOFF auxiliary symbols in ``obj2yaml``.
 
 * ``llvm-ranlib`` now supports ``-X`` on AIX to specify the type of object file
   ranlib should examine.
 
+* ``llvm-cxxfilt`` now supports ``--no-params``/``-p`` to skip function
+  parameters.
+
 * ``llvm-nm`` now supports ``--export-symbol`` to ignore the import symbol 
file.
+* ``llvm-nm`` now supports the ``--line-numbers`` (``-l``) option to use
+  debugging information to print symbols' filenames and line numbers.
 
-* llvm-rc and llvm-windres now accept file path references in ``.rc`` files
+* ``llvm-rc`` and ``llvm-windres`` now accept file path references in ``.rc`` 
files
   concatenated from multiple string literals.
 
-* The llvm-windres option ``--preprocessor`` now resolves its argument
-  in the PATH environment variable as expected, and options passed with
+* The ``llvm-windres`` option ``--preprocessor`` now resolves its argument
+  in the ``PATH`` environment variable as expected, and options passed with
   ``--preprocessor-arg`` are placed before the input file as they should
   be.
 
-* The llvm-windres option ``--preprocessor`` has been updated with the
+* The ``llvm-windres`` option ``--preprocessor`` has been updated with the
   breaking behaviour change from GNU windres from binutils 2.36, where
   the whole argument is considered as one path, not considered as a
   sequence of tool name and parameters.



___
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: [X86] Add missing subvector_subreg_lowering for BF16 (#83720) (PR #84491)

[Tom Stellard via llvm-branch-commits]

tstellar wrote:

The new test is failing on the release/18.x branch.

https://github.com/llvm/llvm-project/pull/84491
___
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: [InstCombine] Fix infinite loop in select equivalence fold (#84036) (PR #84141)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/84141
___
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] 94d8f15 - [InstCombine] Fix infinite loop in select equivalence fold (#84036)

[Tom Stellard via llvm-branch-commits]

Author: Nikita Popov
Date: 2024-03-11T13:36:40-07:00
New Revision: 94d8f150ed8b6afb83ba47f68116a0247a23cb52

URL: 
https://github.com/llvm/llvm-project/commit/94d8f150ed8b6afb83ba47f68116a0247a23cb52
DIFF: 
https://github.com/llvm/llvm-project/commit/94d8f150ed8b6afb83ba47f68116a0247a23cb52.diff

LOG: [InstCombine] Fix infinite loop in select equivalence fold (#84036)

When replacing with a non-constant, it's possible that the result of the
simplification is actually more complicated than the original, and may
result in an infinite combine loop.

Mitigate the issue by requiring that either the replacement or
simplification result is constant, which should ensure that it's
simpler. While this check is crude, it does not appear to cause
optimization regressions in real-world code in practice.

Fixes https://github.com/llvm/llvm-project/issues/83127.

(cherry picked from commit 9f45c5e1a65a1abf4920b617d36ed05e73c04bea)

Added: 


Modified: 
llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
llvm/test/Transforms/InstCombine/select.ll

Removed: 




diff  --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp 
b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index 21bfc91148bfeb..9f220ec003ec33 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -1284,7 +1284,11 @@ Instruction 
*InstCombinerImpl::foldSelectValueEquivalence(SelectInst ,
   isGuaranteedNotToBeUndefOrPoison(CmpRHS, SQ.AC, , )) {
 if (Value *V = simplifyWithOpReplaced(TrueVal, CmpLHS, CmpRHS, SQ,
   /* AllowRefinement */ true))
-  return replaceOperand(Sel, Swapped ? 2 : 1, V);
+  // Require either the replacement or the simplification result to be a
+  // constant to avoid infinite loops.
+  // FIXME: Make this check more precise.
+  if (isa(CmpRHS) || isa(V))
+return replaceOperand(Sel, Swapped ? 2 : 1, V);
 
 // Even if TrueVal does not simplify, we can directly replace a use of
 // CmpLHS with CmpRHS, as long as the instruction is not used anywhere
@@ -1302,7 +1306,8 @@ Instruction 
*InstCombinerImpl::foldSelectValueEquivalence(SelectInst ,
   isGuaranteedNotToBeUndefOrPoison(CmpLHS, SQ.AC, , ))
 if (Value *V = simplifyWithOpReplaced(TrueVal, CmpRHS, CmpLHS, SQ,
   /* AllowRefinement */ true))
-  return replaceOperand(Sel, Swapped ? 2 : 1, V);
+  if (isa(CmpLHS) || isa(V))
+return replaceOperand(Sel, Swapped ? 2 : 1, V);
 
   auto *FalseInst = dyn_cast(FalseVal);
   if (!FalseInst)

diff  --git a/llvm/test/Transforms/InstCombine/select.ll 
b/llvm/test/Transforms/InstCombine/select.ll
index c5f1b77c6d7404..b7e743c14a52ca 100644
--- a/llvm/test/Transforms/InstCombine/select.ll
+++ b/llvm/test/Transforms/InstCombine/select.ll
@@ -2849,12 +2849,14 @@ define i8 @select_replacement_sub(i8 %x, i8 %y, i8 %z) {
   ret i8 %sel
 }
 
+; FIXME: This is safe to fold.
 define i8 @select_replacement_shift_noundef(i8 %x, i8 %y, i8 %z) {
 ; CHECK-LABEL: @select_replacement_shift_noundef(
 ; CHECK-NEXT:[[SHR:%.*]] = lshr exact i8 [[X:%.*]], 1
 ; CHECK-NEXT:call void @use_i8(i8 noundef [[SHR]])
 ; CHECK-NEXT:[[CMP:%.*]] = icmp eq i8 [[SHR]], [[Y:%.*]]
-; CHECK-NEXT:[[SEL:%.*]] = select i1 [[CMP]], i8 [[X]], i8 [[Z:%.*]]
+; CHECK-NEXT:[[SHL:%.*]] = shl i8 [[Y]], 1
+; CHECK-NEXT:[[SEL:%.*]] = select i1 [[CMP]], i8 [[SHL]], i8 [[Z:%.*]]
 ; CHECK-NEXT:ret i8 [[SEL]]
 ;
   %shr = lshr exact i8 %x, 1
@@ -2904,6 +2906,40 @@ define i32 @select_replacement_loop2(i32 %arg, i32 
%arg2) {
   ret i32 %sel
 }
 
+define i8 @select_replacement_loop3(i32 noundef %x) {
+; CHECK-LABEL: @select_replacement_loop3(
+; CHECK-NEXT:[[TRUNC:%.*]] = trunc i32 [[X:%.*]] to i8
+; CHECK-NEXT:[[REV:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[TRUNC]])
+; CHECK-NEXT:[[EXT:%.*]] = zext i8 [[REV]] to i32
+; CHECK-NEXT:[[CMP:%.*]] = icmp eq i32 [[EXT]], [[X]]
+; CHECK-NEXT:[[SEL:%.*]] = select i1 [[CMP]], i8 [[TRUNC]], i8 0
+; CHECK-NEXT:ret i8 [[SEL]]
+;
+  %trunc = trunc i32 %x to i8
+  %rev = call i8 @llvm.bitreverse.i8(i8 %trunc)
+  %ext = zext i8 %rev to i32
+  %cmp = icmp eq i32 %ext, %x
+  %sel = select i1 %cmp, i8 %trunc, i8 0
+  ret i8 %sel
+}
+
+define i16 @select_replacement_loop4(i16 noundef %p_12) {
+; CHECK-LABEL: @select_replacement_loop4(
+; CHECK-NEXT:[[CMP1:%.*]] = icmp ult i16 [[P_12:%.*]], 2
+; CHECK-NEXT:[[AND1:%.*]] = and i16 [[P_12]], 1
+; CHECK-NEXT:[[AND2:%.*]] = select i1 [[CMP1]], i16 [[AND1]], i16 0
+; CHECK-NEXT:[[CMP2:%.*]] = icmp eq i16 [[AND2]], [[P_12]]
+; CHECK-NEXT:[[AND3:%.*]] = select i1 [[CMP2]], i16 [[AND1]], i16 0
+; CHECK-NEXT:ret i16 [[AND3]]
+;
+  %cmp1 = icmp ult i16 %p_12, 2
+  %and1 = and i16 %p_12, 1
+  %and2 = select i1 %cmp1, i16 %and1, i16 0
+  

[llvm-branch-commits] [llvm] release/18.x: [InstCombine] Fix infinite loop in select equivalence fold (#84036) (PR #84141)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/84141

>From 94d8f150ed8b6afb83ba47f68116a0247a23cb52 Mon Sep 17 00:00:00 2001
From: Nikita Popov 
Date: Wed, 6 Mar 2024 09:33:51 +0100
Subject: [PATCH] [InstCombine] Fix infinite loop in select equivalence fold
 (#84036)

When replacing with a non-constant, it's possible that the result of the
simplification is actually more complicated than the original, and may
result in an infinite combine loop.

Mitigate the issue by requiring that either the replacement or
simplification result is constant, which should ensure that it's
simpler. While this check is crude, it does not appear to cause
optimization regressions in real-world code in practice.

Fixes https://github.com/llvm/llvm-project/issues/83127.

(cherry picked from commit 9f45c5e1a65a1abf4920b617d36ed05e73c04bea)
---
 .../InstCombine/InstCombineSelect.cpp |  9 -
 llvm/test/Transforms/InstCombine/select.ll| 38 ++-
 2 files changed, 44 insertions(+), 3 deletions(-)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp 
b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index 21bfc91148bfeb..9f220ec003ec33 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -1284,7 +1284,11 @@ Instruction 
*InstCombinerImpl::foldSelectValueEquivalence(SelectInst ,
   isGuaranteedNotToBeUndefOrPoison(CmpRHS, SQ.AC, , )) {
 if (Value *V = simplifyWithOpReplaced(TrueVal, CmpLHS, CmpRHS, SQ,
   /* AllowRefinement */ true))
-  return replaceOperand(Sel, Swapped ? 2 : 1, V);
+  // Require either the replacement or the simplification result to be a
+  // constant to avoid infinite loops.
+  // FIXME: Make this check more precise.
+  if (isa(CmpRHS) || isa(V))
+return replaceOperand(Sel, Swapped ? 2 : 1, V);
 
 // Even if TrueVal does not simplify, we can directly replace a use of
 // CmpLHS with CmpRHS, as long as the instruction is not used anywhere
@@ -1302,7 +1306,8 @@ Instruction 
*InstCombinerImpl::foldSelectValueEquivalence(SelectInst ,
   isGuaranteedNotToBeUndefOrPoison(CmpLHS, SQ.AC, , ))
 if (Value *V = simplifyWithOpReplaced(TrueVal, CmpRHS, CmpLHS, SQ,
   /* AllowRefinement */ true))
-  return replaceOperand(Sel, Swapped ? 2 : 1, V);
+  if (isa(CmpLHS) || isa(V))
+return replaceOperand(Sel, Swapped ? 2 : 1, V);
 
   auto *FalseInst = dyn_cast(FalseVal);
   if (!FalseInst)
diff --git a/llvm/test/Transforms/InstCombine/select.ll 
b/llvm/test/Transforms/InstCombine/select.ll
index c5f1b77c6d7404..b7e743c14a52ca 100644
--- a/llvm/test/Transforms/InstCombine/select.ll
+++ b/llvm/test/Transforms/InstCombine/select.ll
@@ -2849,12 +2849,14 @@ define i8 @select_replacement_sub(i8 %x, i8 %y, i8 %z) {
   ret i8 %sel
 }
 
+; FIXME: This is safe to fold.
 define i8 @select_replacement_shift_noundef(i8 %x, i8 %y, i8 %z) {
 ; CHECK-LABEL: @select_replacement_shift_noundef(
 ; CHECK-NEXT:[[SHR:%.*]] = lshr exact i8 [[X:%.*]], 1
 ; CHECK-NEXT:call void @use_i8(i8 noundef [[SHR]])
 ; CHECK-NEXT:[[CMP:%.*]] = icmp eq i8 [[SHR]], [[Y:%.*]]
-; CHECK-NEXT:[[SEL:%.*]] = select i1 [[CMP]], i8 [[X]], i8 [[Z:%.*]]
+; CHECK-NEXT:[[SHL:%.*]] = shl i8 [[Y]], 1
+; CHECK-NEXT:[[SEL:%.*]] = select i1 [[CMP]], i8 [[SHL]], i8 [[Z:%.*]]
 ; CHECK-NEXT:ret i8 [[SEL]]
 ;
   %shr = lshr exact i8 %x, 1
@@ -2904,6 +2906,40 @@ define i32 @select_replacement_loop2(i32 %arg, i32 
%arg2) {
   ret i32 %sel
 }
 
+define i8 @select_replacement_loop3(i32 noundef %x) {
+; CHECK-LABEL: @select_replacement_loop3(
+; CHECK-NEXT:[[TRUNC:%.*]] = trunc i32 [[X:%.*]] to i8
+; CHECK-NEXT:[[REV:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[TRUNC]])
+; CHECK-NEXT:[[EXT:%.*]] = zext i8 [[REV]] to i32
+; CHECK-NEXT:[[CMP:%.*]] = icmp eq i32 [[EXT]], [[X]]
+; CHECK-NEXT:[[SEL:%.*]] = select i1 [[CMP]], i8 [[TRUNC]], i8 0
+; CHECK-NEXT:ret i8 [[SEL]]
+;
+  %trunc = trunc i32 %x to i8
+  %rev = call i8 @llvm.bitreverse.i8(i8 %trunc)
+  %ext = zext i8 %rev to i32
+  %cmp = icmp eq i32 %ext, %x
+  %sel = select i1 %cmp, i8 %trunc, i8 0
+  ret i8 %sel
+}
+
+define i16 @select_replacement_loop4(i16 noundef %p_12) {
+; CHECK-LABEL: @select_replacement_loop4(
+; CHECK-NEXT:[[CMP1:%.*]] = icmp ult i16 [[P_12:%.*]], 2
+; CHECK-NEXT:[[AND1:%.*]] = and i16 [[P_12]], 1
+; CHECK-NEXT:[[AND2:%.*]] = select i1 [[CMP1]], i16 [[AND1]], i16 0
+; CHECK-NEXT:[[CMP2:%.*]] = icmp eq i16 [[AND2]], [[P_12]]
+; CHECK-NEXT:[[AND3:%.*]] = select i1 [[CMP2]], i16 [[AND1]], i16 0
+; CHECK-NEXT:ret i16 [[AND3]]
+;
+  %cmp1 = icmp ult i16 %p_12, 2
+  %and1 = and i16 %p_12, 1
+  %and2 = select i1 %cmp1, i16 %and1, i16 0
+  %cmp2 = icmp eq i16 %and2, %p_12
+  %and3 = select i1 %cmp2, i16 %and1, i16 0
+  ret i16 %and3
+}
+
 define ptr 

[llvm-branch-commits] [llvm] release/18.x: [InstCombine] Fix shift calculation in InstCombineCasts (#84027) (PR #84080)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/84080
___
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] 4c36ecb - [InstCombine] Fix shift calculation in InstCombineCasts (#84027)

[Tom Stellard via llvm-branch-commits]

Author: Quentin Dian
Date: 2024-03-11T13:33:20-07:00
New Revision: 4c36ecbe0e162155e6032aec75323dcdd7c81b90

URL: 
https://github.com/llvm/llvm-project/commit/4c36ecbe0e162155e6032aec75323dcdd7c81b90
DIFF: 
https://github.com/llvm/llvm-project/commit/4c36ecbe0e162155e6032aec75323dcdd7c81b90.diff

LOG:  [InstCombine] Fix shift calculation in InstCombineCasts (#84027)

Fixes #84025.

(cherry picked from commit e96c0c1d5e0a9916098b1a31acb006ea6c1108fb)

Added: 


Modified: 
llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
llvm/test/Transforms/InstCombine/bitcast.ll

Removed: 




diff  --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp 
b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 58f0763bb0c0cd..c5d3f60176a826 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -2156,14 +2156,14 @@ static bool collectInsertionElements(Value *V, unsigned 
Shift,
 Type *ElementIntTy = IntegerType::get(C->getContext(), ElementSize);
 
 for (unsigned i = 0; i != NumElts; ++i) {
-  unsigned ShiftI = Shift + i * ElementSize;
+  unsigned ShiftI = i * ElementSize;
   Constant *Piece = ConstantFoldBinaryInstruction(
   Instruction::LShr, C, ConstantInt::get(C->getType(), ShiftI));
   if (!Piece)
 return false;
 
   Piece = ConstantExpr::getTrunc(Piece, ElementIntTy);
-  if (!collectInsertionElements(Piece, ShiftI, Elements, VecEltTy,
+  if (!collectInsertionElements(Piece, ShiftI + Shift, Elements, VecEltTy,
 isBigEndian))
 return false;
 }

diff  --git a/llvm/test/Transforms/InstCombine/bitcast.ll 
b/llvm/test/Transforms/InstCombine/bitcast.ll
index 58bd81297b0dd9..5ace1039c37825 100644
--- a/llvm/test/Transforms/InstCombine/bitcast.ll
+++ b/llvm/test/Transforms/InstCombine/bitcast.ll
@@ -686,6 +686,21 @@ define ptr 
@bitcast_from_single_element_pointer_vector_to_pointer(<1 x ptr> %ptr
   ret ptr %ptr
 }
 
+; Sure that we calculate the correct shift.
+define <4 x i32> @bitcast_shl(i32 %arg) {
+; CHECK-LABEL: @bitcast_shl(
+; CHECK-NEXT:[[I5:%.*]] = insertelement <4 x i32> , i32 [[ARG:%.*]], i64 3
+; CHECK-NEXT:ret <4 x i32> [[I5]]
+;
+  %i = zext i32 %arg to i64
+  %i1 = shl i64 %i, 32
+  %i2 = or i64 %i1, 65
+  %i3 = zext i64 %i2 to i128
+  %i4 = shl i128 %i3, 64
+  %i5 = bitcast i128 %i4 to <4 x i32>
+  ret <4 x i32> %i5
+}
+
 declare void @f1()
 declare void @f2()
 define ptr @select_bitcast_unsized_pointer(i1 %c) {



___
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: [InstCombine] Fix shift calculation in InstCombineCasts (#84027) (PR #84080)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/84080

>From 4c36ecbe0e162155e6032aec75323dcdd7c81b90 Mon Sep 17 00:00:00 2001
From: Quentin Dian 
Date: Wed, 6 Mar 2024 06:16:28 +0800
Subject: [PATCH]  [InstCombine] Fix shift calculation in InstCombineCasts
 (#84027)

Fixes #84025.

(cherry picked from commit e96c0c1d5e0a9916098b1a31acb006ea6c1108fb)
---
 .../Transforms/InstCombine/InstCombineCasts.cpp   |  4 ++--
 llvm/test/Transforms/InstCombine/bitcast.ll   | 15 +++
 2 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp 
b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 58f0763bb0c0cd..c5d3f60176a826 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -2156,14 +2156,14 @@ static bool collectInsertionElements(Value *V, unsigned 
Shift,
 Type *ElementIntTy = IntegerType::get(C->getContext(), ElementSize);
 
 for (unsigned i = 0; i != NumElts; ++i) {
-  unsigned ShiftI = Shift + i * ElementSize;
+  unsigned ShiftI = i * ElementSize;
   Constant *Piece = ConstantFoldBinaryInstruction(
   Instruction::LShr, C, ConstantInt::get(C->getType(), ShiftI));
   if (!Piece)
 return false;
 
   Piece = ConstantExpr::getTrunc(Piece, ElementIntTy);
-  if (!collectInsertionElements(Piece, ShiftI, Elements, VecEltTy,
+  if (!collectInsertionElements(Piece, ShiftI + Shift, Elements, VecEltTy,
 isBigEndian))
 return false;
 }
diff --git a/llvm/test/Transforms/InstCombine/bitcast.ll 
b/llvm/test/Transforms/InstCombine/bitcast.ll
index 58bd81297b0dd9..5ace1039c37825 100644
--- a/llvm/test/Transforms/InstCombine/bitcast.ll
+++ b/llvm/test/Transforms/InstCombine/bitcast.ll
@@ -686,6 +686,21 @@ define ptr 
@bitcast_from_single_element_pointer_vector_to_pointer(<1 x ptr> %ptr
   ret ptr %ptr
 }
 
+; Sure that we calculate the correct shift.
+define <4 x i32> @bitcast_shl(i32 %arg) {
+; CHECK-LABEL: @bitcast_shl(
+; CHECK-NEXT:[[I5:%.*]] = insertelement <4 x i32> , i32 [[ARG:%.*]], i64 3
+; CHECK-NEXT:ret <4 x i32> [[I5]]
+;
+  %i = zext i32 %arg to i64
+  %i1 = shl i64 %i, 32
+  %i2 = or i64 %i1, 65
+  %i3 = zext i64 %i2 to i128
+  %i4 = shl i128 %i3, 64
+  %i5 = bitcast i128 %i4 to <4 x i32>
+  ret <4 x i32> %i5
+}
+
 declare void @f1()
 declare void @f2()
 define ptr @select_bitcast_unsized_pointer(i1 %c) {

___
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] Backport ARM64EC variadic args fixes to LLVM 18 (PR #81800)

[Eli Friedman via llvm-branch-commits]

https://github.com/efriedma-quic approved this pull request.

LGTM

https://github.com/llvm/llvm-project/pull/81800
___
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: [libc++] Enable availability based on the compiler instead of __has_extension (#84065) (PR #84374)

[Louis Dionne via llvm-branch-commits]

ldionne wrote:

No, it's not legitimate. Our CI is rather unstable at the moment but this looks 
as good as it'll get.

https://github.com/llvm/llvm-project/pull/84374
___
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] [compiler-rt] release/18.x: [test] Make two sanitize-coverage tests pass with glibc 2.39+ (PR #84239)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/84239
___
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] [compiler-rt] e90bfdb - [test] Make two sanitize-coverage tests pass with glibc 2.39+

[Tom Stellard via llvm-branch-commits]

Author: Fangrui Song
Date: 2024-03-11T13:07:50-07:00
New Revision: e90bfdb4ddced8dff215672ffeceece8ebe60426

URL: 
https://github.com/llvm/llvm-project/commit/e90bfdb4ddced8dff215672ffeceece8ebe60426
DIFF: 
https://github.com/llvm/llvm-project/commit/e90bfdb4ddced8dff215672ffeceece8ebe60426.diff

LOG: [test] Make two sanitize-coverage tests pass with glibc 2.39+

glibc 2.39 added `nonnull` attribute to most libio functions accepting a
`FILE*` parameter, including fprintf[1]. The -fsanitize=undefined mode
checks the argument to fprintf and has extra counters, not expected by
two tests. Specify -fno-sanitize=nonnull-attribute to make the two tests
pass.

Fix #82883

[1]: 
https://sourceware.org/git/?p=glibc.git;a=commit;h=64b1a44183a3094672ed304532bedb9acc707554

Pull Request: https://github.com/llvm/llvm-project/pull/84231

(cherry picked from commit c3acbf6bb06f9039f9850e18e0ae2f2adef63905)

Added: 


Modified: 

compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter_default_impl.cpp
compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_symbolize.cpp

Removed: 




diff  --git 
a/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter_default_impl.cpp
 
b/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter_default_impl.cpp
index 1ac04b53491e14..1d1fbf7299e8b4 100644
--- 
a/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter_default_impl.cpp
+++ 
b/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter_default_impl.cpp
@@ -3,7 +3,9 @@
 
 // REQUIRES: has_sancovcc,stable-runtime,linux,x86_64-target-arch
 
-// RUN: %clangxx -O0 %s -fsanitize-coverage=inline-8bit-counters,pc-table -o %t
+/// In glibc 2.39+, fprintf has a nonnull attribute. Disable nonnull-attribute,
+/// which would increase counters for ubsan.
+// RUN: %clangxx -O0 %s -fsanitize-coverage=inline-8bit-counters,pc-table 
-fno-sanitize=nonnull-attribute -o %t
 // RUN: rm -f %t-counters %t-pcs
 // RUN: env %tool_options="cov_8bit_counters_out=%t-counters 
cov_pcs_out=%t-pcs verbosity=1" %run %t 2>&1 | FileCheck %s
 

diff  --git 
a/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_symbolize.cpp 
b/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_symbolize.cpp
index daa994c8116251..b168954a1c92cf 100644
--- 
a/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_symbolize.cpp
+++ 
b/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_symbolize.cpp
@@ -7,7 +7,9 @@
 // RUN: rm -rf $DIR
 // RUN: mkdir -p $DIR
 // RUN: cd $DIR
-// RUN: %clangxx -O0 -fsanitize-coverage=trace-pc-guard %s -o %t
+/// In glibc 2.39+, fprintf has a nonnull attribute. Disable nonnull-attribute,
+/// which would increase counters for ubsan.
+// RUN: %clangxx -O0 -fsanitize-coverage=trace-pc-guard 
-fno-sanitize=nonnull-attribute %s -o %t
 // RUN: %env_tool_opts=coverage=1 %t 2>&1 | FileCheck %s
 // RUN: rm -rf $DIR
 



___
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] [compiler-rt] release/18.x: [test] Make two sanitize-coverage tests pass with glibc 2.39+ (PR #84239)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/84239

>From e90bfdb4ddced8dff215672ffeceece8ebe60426 Mon Sep 17 00:00:00 2001
From: Fangrui Song 
Date: Wed, 6 Mar 2024 13:17:43 -0800
Subject: [PATCH] [test] Make two sanitize-coverage tests pass with glibc 2.39+

glibc 2.39 added `nonnull` attribute to most libio functions accepting a
`FILE*` parameter, including fprintf[1]. The -fsanitize=undefined mode
checks the argument to fprintf and has extra counters, not expected by
two tests. Specify -fno-sanitize=nonnull-attribute to make the two tests
pass.

Fix #82883

[1]: 
https://sourceware.org/git/?p=glibc.git;a=commit;h=64b1a44183a3094672ed304532bedb9acc707554

Pull Request: https://github.com/llvm/llvm-project/pull/84231

(cherry picked from commit c3acbf6bb06f9039f9850e18e0ae2f2adef63905)
---
 .../sanitizer_coverage_inline8bit_counter_default_impl.cpp| 4 +++-
 .../TestCases/sanitizer_coverage_symbolize.cpp| 4 +++-
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git 
a/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter_default_impl.cpp
 
b/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter_default_impl.cpp
index 1ac04b53491e14..1d1fbf7299e8b4 100644
--- 
a/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter_default_impl.cpp
+++ 
b/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter_default_impl.cpp
@@ -3,7 +3,9 @@
 
 // REQUIRES: has_sancovcc,stable-runtime,linux,x86_64-target-arch
 
-// RUN: %clangxx -O0 %s -fsanitize-coverage=inline-8bit-counters,pc-table -o %t
+/// In glibc 2.39+, fprintf has a nonnull attribute. Disable nonnull-attribute,
+/// which would increase counters for ubsan.
+// RUN: %clangxx -O0 %s -fsanitize-coverage=inline-8bit-counters,pc-table 
-fno-sanitize=nonnull-attribute -o %t
 // RUN: rm -f %t-counters %t-pcs
 // RUN: env %tool_options="cov_8bit_counters_out=%t-counters 
cov_pcs_out=%t-pcs verbosity=1" %run %t 2>&1 | FileCheck %s
 
diff --git 
a/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_symbolize.cpp 
b/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_symbolize.cpp
index daa994c8116251..b168954a1c92cf 100644
--- 
a/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_symbolize.cpp
+++ 
b/compiler-rt/test/sanitizer_common/TestCases/sanitizer_coverage_symbolize.cpp
@@ -7,7 +7,9 @@
 // RUN: rm -rf $DIR
 // RUN: mkdir -p $DIR
 // RUN: cd $DIR
-// RUN: %clangxx -O0 -fsanitize-coverage=trace-pc-guard %s -o %t
+/// In glibc 2.39+, fprintf has a nonnull attribute. Disable nonnull-attribute,
+/// which would increase counters for ubsan.
+// RUN: %clangxx -O0 -fsanitize-coverage=trace-pc-guard 
-fno-sanitize=nonnull-attribute %s -o %t
 // RUN: %env_tool_opts=coverage=1 %t 2>&1 | FileCheck %s
 // RUN: rm -rf $DIR
 

___
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/18.x: [ObjC] Check entire chain of superclasses to determine class layout (PR #84093)

[Eli Friedman via llvm-branch-commits]

https://github.com/efriedma-quic requested changes to this pull request.

We usually only take bugfixes on release branches (miscompiles/crashes/etc.).

https://github.com/llvm/llvm-project/pull/84093
___
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: [ARM] Switch to LiveRegUnits to fix r7 register allocation bug (PR #84475)

[Eli Friedman via llvm-branch-commits]

https://github.com/efriedma-quic requested changes to this pull request.

This is, as far as I can tell, not a miscompile; probably not worth taking on 
the 18.x branch.

(Also, it's usually not a good idea to open a PR for a cherry-pick before the 
original patch is merged.)

https://github.com/llvm/llvm-project/pull/84475
___
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] bf45c3a - [DSE] Delay deleting non-memory-defs until end of DSE. (#83411)

[Tom Stellard via llvm-branch-commits]

Author: Florian Hahn
Date: 2024-03-11T12:53:04-07:00
New Revision: bf45c3a07918c14577ef7a829f16ec339b9ed610

URL: 
https://github.com/llvm/llvm-project/commit/bf45c3a07918c14577ef7a829f16ec339b9ed610
DIFF: 
https://github.com/llvm/llvm-project/commit/bf45c3a07918c14577ef7a829f16ec339b9ed610.diff

LOG: [DSE] Delay deleting non-memory-defs until end of DSE. (#83411)

DSE uses BatchAA, which caches queries using pairs of MemoryLocations.
At the moment, DSE may remove instructions that are used as pointers in
cached MemoryLocations. If a new instruction used by a new MemoryLoation
and this instruction gets allocated at the same address as a previosuly
cached and then removed instruction, we may access an incorrect entry in
the cache.

To avoid this delay removing all instructions except MemoryDefs until
the end of DSE. This should avoid removing any values used in BatchAA's
cache.

Test case by @vporpo from
https://github.com/llvm/llvm-project/pull/83181.
(Test not precommitted because the results are non-determinstic - memset
only sometimes gets removed)

PR: https://github.com/llvm/llvm-project/pull/83411
(cherry picked from commit 10f5e983a9e3162a569cbebeb32168716e391340)

Added: 
llvm/test/Transforms/DeadStoreElimination/batchaa-caching-new-pointers.ll

Modified: 
llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp

Removed: 




diff  --git a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp 
b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
index 11a91bfbe5baff..340fba4fb9c5a2 100644
--- a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
+++ b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
@@ -857,6 +857,9 @@ struct DSEState {
   // no longer be captured.
   bool ShouldIterateEndOfFunctionDSE;
 
+  /// Dead instructions to be removed at the end of DSE.
+  SmallVector ToRemove;
+
   // Class contains self-reference, make sure it's not copied/moved.
   DSEState(const DSEState &) = delete;
   DSEState =(const DSEState &) = delete;
@@ -1692,7 +1695,8 @@ struct DSEState {
 return {MaybeDeadAccess};
   }
 
-  // Delete dead memory defs
+  /// Delete dead memory defs and recursively add their operands to ToRemove if
+  /// they became dead.
   void deleteDeadInstruction(Instruction *SI) {
 MemorySSAUpdater Updater();
 SmallVector NowDeadInsts;
@@ -1708,8 +1712,11 @@ struct DSEState {
   salvageKnowledge(DeadInst);
 
   // Remove the Instruction from MSSA.
-  if (MemoryAccess *MA = MSSA.getMemoryAccess(DeadInst)) {
-if (MemoryDef *MD = dyn_cast(MA)) {
+  MemoryAccess *MA = MSSA.getMemoryAccess(DeadInst);
+  bool IsMemDef = MA && isa(MA);
+  if (MA) {
+if (IsMemDef) {
+  auto *MD = cast(MA);
   SkipStores.insert(MD);
   if (auto *SI = dyn_cast(MD->getMemoryInst())) {
 if (SI->getValueOperand()->getType()->isPointerTy()) {
@@ -1730,13 +1737,21 @@ struct DSEState {
   // Remove its operands
   for (Use  : DeadInst->operands())
 if (Instruction *OpI = dyn_cast(O)) {
-  O = nullptr;
+  O.set(PoisonValue::get(O->getType()));
   if (isInstructionTriviallyDead(OpI, ))
 NowDeadInsts.push_back(OpI);
 }
 
   EI.removeInstruction(DeadInst);
-  DeadInst->eraseFromParent();
+  // Remove memory defs directly if they don't produce results, but only
+  // queue other dead instructions for later removal. They may have been
+  // used as memory locations that have been cached by BatchAA. Removing
+  // them here may lead to newly created instructions to be allocated at 
the
+  // same address, yielding stale cache entries.
+  if (IsMemDef && DeadInst->getType()->isVoidTy())
+DeadInst->eraseFromParent();
+  else
+ToRemove.push_back(DeadInst);
 }
   }
 
@@ -2233,6 +2248,12 @@ static bool eliminateDeadStores(Function , 
AliasAnalysis , MemorySSA ,
 
   MadeChange |= State.eliminateRedundantStoresOfExistingValues();
   MadeChange |= State.eliminateDeadWritesAtEndOfFunction();
+
+  while (!State.ToRemove.empty()) {
+Instruction *DeadInst = State.ToRemove.pop_back_val();
+DeadInst->eraseFromParent();
+  }
+
   return MadeChange;
 }
 } // end anonymous namespace

diff  --git 
a/llvm/test/Transforms/DeadStoreElimination/batchaa-caching-new-pointers.ll 
b/llvm/test/Transforms/DeadStoreElimination/batchaa-caching-new-pointers.ll
new file mode 100644
index 00..ee9bd6912e2ae4
--- /dev/null
+++ b/llvm/test/Transforms/DeadStoreElimination/batchaa-caching-new-pointers.ll
@@ -0,0 +1,189 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 
UTC_ARGS: --version 4
+; RUN: opt -S -passes=dse < %s | FileCheck %s
+;
+; DSE kills `store i32 44, ptr %struct.byte.4, align 4` but should not kill
+; `call void @llvm.memset.p0.i64(...)`  because it has a clobber read:
+; `%ret = load ptr, ptr %struct.byte.8`

[llvm-branch-commits] [llvm] release/18.x: [DSE] Delay deleting non-memory-defs until end of DSE. (#83411) (PR #84227)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/84227
___
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: [DSE] Delay deleting non-memory-defs until end of DSE. (#83411) (PR #84227)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/84227

>From bf45c3a07918c14577ef7a829f16ec339b9ed610 Mon Sep 17 00:00:00 2001
From: Florian Hahn 
Date: Sat, 2 Mar 2024 12:34:36 +
Subject: [PATCH] [DSE] Delay deleting non-memory-defs until end of DSE.
 (#83411)

DSE uses BatchAA, which caches queries using pairs of MemoryLocations.
At the moment, DSE may remove instructions that are used as pointers in
cached MemoryLocations. If a new instruction used by a new MemoryLoation
and this instruction gets allocated at the same address as a previosuly
cached and then removed instruction, we may access an incorrect entry in
the cache.

To avoid this delay removing all instructions except MemoryDefs until
the end of DSE. This should avoid removing any values used in BatchAA's
cache.

Test case by @vporpo from
https://github.com/llvm/llvm-project/pull/83181.
(Test not precommitted because the results are non-determinstic - memset
only sometimes gets removed)

PR: https://github.com/llvm/llvm-project/pull/83411
(cherry picked from commit 10f5e983a9e3162a569cbebeb32168716e391340)
---
 .../Scalar/DeadStoreElimination.cpp   |  31 ++-
 .../batchaa-caching-new-pointers.ll   | 189 ++
 2 files changed, 215 insertions(+), 5 deletions(-)
 create mode 100644 
llvm/test/Transforms/DeadStoreElimination/batchaa-caching-new-pointers.ll

diff --git a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp 
b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
index 11a91bfbe5baff..340fba4fb9c5a2 100644
--- a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
+++ b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
@@ -857,6 +857,9 @@ struct DSEState {
   // no longer be captured.
   bool ShouldIterateEndOfFunctionDSE;
 
+  /// Dead instructions to be removed at the end of DSE.
+  SmallVector ToRemove;
+
   // Class contains self-reference, make sure it's not copied/moved.
   DSEState(const DSEState &) = delete;
   DSEState =(const DSEState &) = delete;
@@ -1692,7 +1695,8 @@ struct DSEState {
 return {MaybeDeadAccess};
   }
 
-  // Delete dead memory defs
+  /// Delete dead memory defs and recursively add their operands to ToRemove if
+  /// they became dead.
   void deleteDeadInstruction(Instruction *SI) {
 MemorySSAUpdater Updater();
 SmallVector NowDeadInsts;
@@ -1708,8 +1712,11 @@ struct DSEState {
   salvageKnowledge(DeadInst);
 
   // Remove the Instruction from MSSA.
-  if (MemoryAccess *MA = MSSA.getMemoryAccess(DeadInst)) {
-if (MemoryDef *MD = dyn_cast(MA)) {
+  MemoryAccess *MA = MSSA.getMemoryAccess(DeadInst);
+  bool IsMemDef = MA && isa(MA);
+  if (MA) {
+if (IsMemDef) {
+  auto *MD = cast(MA);
   SkipStores.insert(MD);
   if (auto *SI = dyn_cast(MD->getMemoryInst())) {
 if (SI->getValueOperand()->getType()->isPointerTy()) {
@@ -1730,13 +1737,21 @@ struct DSEState {
   // Remove its operands
   for (Use  : DeadInst->operands())
 if (Instruction *OpI = dyn_cast(O)) {
-  O = nullptr;
+  O.set(PoisonValue::get(O->getType()));
   if (isInstructionTriviallyDead(OpI, ))
 NowDeadInsts.push_back(OpI);
 }
 
   EI.removeInstruction(DeadInst);
-  DeadInst->eraseFromParent();
+  // Remove memory defs directly if they don't produce results, but only
+  // queue other dead instructions for later removal. They may have been
+  // used as memory locations that have been cached by BatchAA. Removing
+  // them here may lead to newly created instructions to be allocated at 
the
+  // same address, yielding stale cache entries.
+  if (IsMemDef && DeadInst->getType()->isVoidTy())
+DeadInst->eraseFromParent();
+  else
+ToRemove.push_back(DeadInst);
 }
   }
 
@@ -2233,6 +2248,12 @@ static bool eliminateDeadStores(Function , 
AliasAnalysis , MemorySSA ,
 
   MadeChange |= State.eliminateRedundantStoresOfExistingValues();
   MadeChange |= State.eliminateDeadWritesAtEndOfFunction();
+
+  while (!State.ToRemove.empty()) {
+Instruction *DeadInst = State.ToRemove.pop_back_val();
+DeadInst->eraseFromParent();
+  }
+
   return MadeChange;
 }
 } // end anonymous namespace
diff --git 
a/llvm/test/Transforms/DeadStoreElimination/batchaa-caching-new-pointers.ll 
b/llvm/test/Transforms/DeadStoreElimination/batchaa-caching-new-pointers.ll
new file mode 100644
index 00..ee9bd6912e2ae4
--- /dev/null
+++ b/llvm/test/Transforms/DeadStoreElimination/batchaa-caching-new-pointers.ll
@@ -0,0 +1,189 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 
UTC_ARGS: --version 4
+; RUN: opt -S -passes=dse < %s | FileCheck %s
+;
+; DSE kills `store i32 44, ptr %struct.byte.4, align 4` but should not kill
+; `call void @llvm.memset.p0.i64(...)`  because it has a clobber read:
+; `%ret = load ptr, ptr %struct.byte.8`
+
+
+%struct.type = type { ptr, ptr }
+

[llvm-branch-commits] [libcxx] release/18.x: [libc++] Enable availability based on the compiler instead of __has_extension (#84065) (PR #84374)

[Tom Stellard via llvm-branch-commits]

tstellar wrote:

@ldionne Is the test failure legitimate?

https://github.com/llvm/llvm-project/pull/84374
___
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/18.x: [clang][fat-lto-objects] Make module flags match non-FatLTO pipelines (#83159) (PR #84290)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/84290
___
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] 16ab081 - [clang][fat-lto-objects] Make module flags match non-FatLTO pipelines (#83159)

[Tom Stellard via llvm-branch-commits]

Author: Paul Kirth
Date: 2024-03-11T12:47:52-07:00
New Revision: 16ab0812d2010dad76f87d4d50da8e79e0e75e71

URL: 
https://github.com/llvm/llvm-project/commit/16ab0812d2010dad76f87d4d50da8e79e0e75e71
DIFF: 
https://github.com/llvm/llvm-project/commit/16ab0812d2010dad76f87d4d50da8e79e0e75e71.diff

LOG: [clang][fat-lto-objects] Make module flags match non-FatLTO pipelines 
(#83159)

In addition to being rather hard to follow, there isn't a good reason
why FatLTO shouldn't just share the same code for setting module flags
for (Thin)LTO. This patch simplifies the logic and makes sure we use set
these flags in a consistent way, independent of FatLTO.

Additionally, we now test that output in the .llvm.lto section actually
matches the output from Full and Thin LTO compilation.

(cherry picked from commit 7d8b50aaab8e0f935e3cb1f3f397e98b9e3ee241)

Added: 


Modified: 
clang/lib/CodeGen/BackendUtil.cpp
clang/test/CodeGen/fat-lto-objects.c

Removed: 




diff  --git a/clang/lib/CodeGen/BackendUtil.cpp 
b/clang/lib/CodeGen/BackendUtil.cpp
index 7877e20d77f772..4f22d35f9d3a94 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -186,6 +186,14 @@ class EmitAssemblyHelper {
TargetTriple.getVendor() != llvm::Triple::Apple;
   }
 
+  /// Check whether we should emit a flag for UnifiedLTO.
+  /// The UnifiedLTO module flag should be set when UnifiedLTO is enabled for
+  /// ThinLTO or Full LTO with module summaries.
+  bool shouldEmitUnifiedLTOModueFlag() const {
+return CodeGenOpts.UnifiedLTO &&
+   (CodeGenOpts.PrepareForThinLTO || shouldEmitRegularLTOSummary());
+  }
+
 public:
   EmitAssemblyHelper(DiagnosticsEngine &_Diags,
  const HeaderSearchOptions ,
@@ -1029,7 +1037,8 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
   if (!actionRequiresCodeGen(Action) && CodeGenOpts.VerifyModule)
 MPM.addPass(VerifierPass());
 
-  if (Action == Backend_EmitBC || Action == Backend_EmitLL) {
+  if (Action == Backend_EmitBC || Action == Backend_EmitLL ||
+  CodeGenOpts.FatLTO) {
 if (CodeGenOpts.PrepareForThinLTO && !CodeGenOpts.DisableLLVMPasses) {
   if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
 TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
@@ -1040,11 +1049,9 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
   if (!ThinLinkOS)
 return;
 }
-if (CodeGenOpts.UnifiedLTO)
-  TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", 
uint32_t(1));
 MPM.addPass(ThinLTOBitcodeWriterPass(
 *OS, ThinLinkOS ? >os() : nullptr));
-  } else {
+  } else if (Action == Backend_EmitLL) {
 MPM.addPass(PrintModulePass(*OS, "", CodeGenOpts.EmitLLVMUseLists,
 /*EmitLTOSummary=*/true));
   }
@@ -1058,24 +1065,17 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
 if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
   TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
uint32_t(1));
-if (CodeGenOpts.UnifiedLTO)
-  TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", 
uint32_t(1));
   }
-  if (Action == Backend_EmitBC)
+  if (Action == Backend_EmitBC) {
 MPM.addPass(BitcodeWriterPass(*OS, CodeGenOpts.EmitLLVMUseLists,
   EmitLTOSummary));
-  else
+  } else if (Action == Backend_EmitLL) {
 MPM.addPass(PrintModulePass(*OS, "", CodeGenOpts.EmitLLVMUseLists,
 EmitLTOSummary));
+  }
 }
-  }
-  if (CodeGenOpts.FatLTO) {
-// Set the EnableSplitLTOUnit and UnifiedLTO module flags, since FatLTO
-// uses a 
diff erent action than Backend_EmitBC or Backend_EmitLL.
-if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
-  TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
-   uint32_t(CodeGenOpts.EnableSplitLTOUnit));
-if (CodeGenOpts.UnifiedLTO && !TheModule->getModuleFlag("UnifiedLTO"))
+
+if (shouldEmitUnifiedLTOModueFlag())
   TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", uint32_t(1));
   }
 

diff  --git a/clang/test/CodeGen/fat-lto-objects.c 
b/clang/test/CodeGen/fat-lto-objects.c
index afce798c5c8194..b50567c024fc8c 100644
--- a/clang/test/CodeGen/fat-lto-objects.c
+++ b/clang/test/CodeGen/fat-lto-objects.c
@@ -11,10 +11,11 @@
 // RUN: llvm-objcopy --dump-section=.llvm.lto=%t.full.split.bc %t.full.split.o
 // RUN: llvm-dis %t.full.split.bc -o - | FileCheck %s 
--check-prefixes=FULL,SPLIT,NOUNIFIED
 
+/// Full LTO always sets EnableSplitLTOUnit when the summary is used.
 // RUN: %clang -cc1 -triple x86_64-unknown-linux-gnu -flto=full 
-ffat-lto-objects -emit-obj < %s -o %t.full.nosplit.o
 // RUN: llvm-readelf -S 

[llvm-branch-commits] [clang] release/18.x: [clang][fat-lto-objects] Make module flags match non-FatLTO pipelines (#83159) (PR #84290)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/84290

>From 16ab0812d2010dad76f87d4d50da8e79e0e75e71 Mon Sep 17 00:00:00 2001
From: Paul Kirth 
Date: Wed, 28 Feb 2024 19:11:55 -0800
Subject: [PATCH] [clang][fat-lto-objects] Make module flags match non-FatLTO
 pipelines (#83159)

In addition to being rather hard to follow, there isn't a good reason
why FatLTO shouldn't just share the same code for setting module flags
for (Thin)LTO. This patch simplifies the logic and makes sure we use set
these flags in a consistent way, independent of FatLTO.

Additionally, we now test that output in the .llvm.lto section actually
matches the output from Full and Thin LTO compilation.

(cherry picked from commit 7d8b50aaab8e0f935e3cb1f3f397e98b9e3ee241)
---
 clang/lib/CodeGen/BackendUtil.cpp| 32 ++--
 clang/test/CodeGen/fat-lto-objects.c | 21 +-
 2 files changed, 36 insertions(+), 17 deletions(-)

diff --git a/clang/lib/CodeGen/BackendUtil.cpp 
b/clang/lib/CodeGen/BackendUtil.cpp
index 7877e20d77f772..4f22d35f9d3a94 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -186,6 +186,14 @@ class EmitAssemblyHelper {
TargetTriple.getVendor() != llvm::Triple::Apple;
   }
 
+  /// Check whether we should emit a flag for UnifiedLTO.
+  /// The UnifiedLTO module flag should be set when UnifiedLTO is enabled for
+  /// ThinLTO or Full LTO with module summaries.
+  bool shouldEmitUnifiedLTOModueFlag() const {
+return CodeGenOpts.UnifiedLTO &&
+   (CodeGenOpts.PrepareForThinLTO || shouldEmitRegularLTOSummary());
+  }
+
 public:
   EmitAssemblyHelper(DiagnosticsEngine &_Diags,
  const HeaderSearchOptions ,
@@ -1029,7 +1037,8 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
   if (!actionRequiresCodeGen(Action) && CodeGenOpts.VerifyModule)
 MPM.addPass(VerifierPass());
 
-  if (Action == Backend_EmitBC || Action == Backend_EmitLL) {
+  if (Action == Backend_EmitBC || Action == Backend_EmitLL ||
+  CodeGenOpts.FatLTO) {
 if (CodeGenOpts.PrepareForThinLTO && !CodeGenOpts.DisableLLVMPasses) {
   if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
 TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
@@ -1040,11 +1049,9 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
   if (!ThinLinkOS)
 return;
 }
-if (CodeGenOpts.UnifiedLTO)
-  TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", 
uint32_t(1));
 MPM.addPass(ThinLTOBitcodeWriterPass(
 *OS, ThinLinkOS ? >os() : nullptr));
-  } else {
+  } else if (Action == Backend_EmitLL) {
 MPM.addPass(PrintModulePass(*OS, "", CodeGenOpts.EmitLLVMUseLists,
 /*EmitLTOSummary=*/true));
   }
@@ -1058,24 +1065,17 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
 if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
   TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
uint32_t(1));
-if (CodeGenOpts.UnifiedLTO)
-  TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", 
uint32_t(1));
   }
-  if (Action == Backend_EmitBC)
+  if (Action == Backend_EmitBC) {
 MPM.addPass(BitcodeWriterPass(*OS, CodeGenOpts.EmitLLVMUseLists,
   EmitLTOSummary));
-  else
+  } else if (Action == Backend_EmitLL) {
 MPM.addPass(PrintModulePass(*OS, "", CodeGenOpts.EmitLLVMUseLists,
 EmitLTOSummary));
+  }
 }
-  }
-  if (CodeGenOpts.FatLTO) {
-// Set the EnableSplitLTOUnit and UnifiedLTO module flags, since FatLTO
-// uses a different action than Backend_EmitBC or Backend_EmitLL.
-if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
-  TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
-   uint32_t(CodeGenOpts.EnableSplitLTOUnit));
-if (CodeGenOpts.UnifiedLTO && !TheModule->getModuleFlag("UnifiedLTO"))
+
+if (shouldEmitUnifiedLTOModueFlag())
   TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", uint32_t(1));
   }
 
diff --git a/clang/test/CodeGen/fat-lto-objects.c 
b/clang/test/CodeGen/fat-lto-objects.c
index afce798c5c8194..b50567c024fc8c 100644
--- a/clang/test/CodeGen/fat-lto-objects.c
+++ b/clang/test/CodeGen/fat-lto-objects.c
@@ -11,10 +11,11 @@
 // RUN: llvm-objcopy --dump-section=.llvm.lto=%t.full.split.bc %t.full.split.o
 // RUN: llvm-dis %t.full.split.bc -o - | FileCheck %s 
--check-prefixes=FULL,SPLIT,NOUNIFIED
 
+/// Full LTO always sets EnableSplitLTOUnit when the summary is used.
 // RUN: %clang -cc1 -triple x86_64-unknown-linux-gnu -flto=full 
-ffat-lto-objects -emit-obj < %s -o %t.full.nosplit.o
 // RUN: llvm-readelf -S %t.full.nosplit.o | FileCheck %s --check-prefixes=ELF
 // RUN: llvm-objcopy 

[llvm-branch-commits] [llvm] Lld mac fix (PR #84764)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar updated 
https://github.com/llvm/llvm-project/pull/84764

>From 2ddb32b50752ca91ca9946cb9c9ea3d92c8616d0 Mon Sep 17 00:00:00 2001
From: Tom Stellard 
Date: Sat, 9 Mar 2024 14:26:47 -0800
Subject: [PATCH 1/2] Bump version to 18.1.2

---
 llvm/CMakeLists.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/CMakeLists.txt b/llvm/CMakeLists.txt
index ddf95cbc6c5175..c5fa66390bba8d 100644
--- a/llvm/CMakeLists.txt
+++ b/llvm/CMakeLists.txt
@@ -22,7 +22,7 @@ if(NOT DEFINED LLVM_VERSION_MINOR)
   set(LLVM_VERSION_MINOR 1)
 endif()
 if(NOT DEFINED LLVM_VERSION_PATCH)
-  set(LLVM_VERSION_PATCH 1)
+  set(LLVM_VERSION_PATCH 2)
 endif()
 if(NOT DEFINED LLVM_VERSION_SUFFIX)
   set(LLVM_VERSION_SUFFIX)

>From 9edcbcbb219c221521894b04e668b6f9f450497f Mon Sep 17 00:00:00 2001
From: Tom Stellard 
Date: Mon, 11 Mar 2024 07:28:58 -0700
Subject: [PATCH 2/2] workflows: Add workaround for lld failures on MacOS

See #81967
---
 .github/workflows/llvm-project-tests.yml | 5 +
 1 file changed, 5 insertions(+)

diff --git a/.github/workflows/llvm-project-tests.yml 
b/.github/workflows/llvm-project-tests.yml
index 43b90193406fc9..64788457fb6d60 100644
--- a/.github/workflows/llvm-project-tests.yml
+++ b/.github/workflows/llvm-project-tests.yml
@@ -118,6 +118,11 @@ jobs:
   else
 builddir="$(pwd)"/build
   fi
+  if [ "${{ runner.os }}" == "MacOS" ]; then
+# Workaround test failure on some lld tests on MacOS
+# https://github.com/llvm/llvm-project/issues/81967
+extra_cmake_args="-DLLVM_DISABLE_ASSEMBLY_FILES=ON"
+  fi
   echo "llvm-builddir=$builddir" >> "$GITHUB_OUTPUT"
   cmake -G Ninja \
 -B "$builddir" \

___
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: Allow .alt_entry symbols to pass the .cfi nesting check (#82268) (PR #83336)

[Tom Stellard via llvm-branch-commits]

https://github.com/tstellar closed 
https://github.com/llvm/llvm-project/pull/83336
___
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] 267d9b1 - Allow .alt_entry symbols to pass the .cfi nesting check (#82268)

[Tom Stellard via llvm-branch-commits]

Author: Jon Roelofs
Date: 2024-03-11T12:35:50-07:00
New Revision: 267d9b1a74c466fff6bde255a211732aa3bcd7e8

URL: 
https://github.com/llvm/llvm-project/commit/267d9b1a74c466fff6bde255a211732aa3bcd7e8
DIFF: 
https://github.com/llvm/llvm-project/commit/267d9b1a74c466fff6bde255a211732aa3bcd7e8.diff

LOG: Allow .alt_entry symbols to pass the .cfi nesting check (#82268)

A symbol with an `N_ALT_ENTRY` attribute may be defined in the middle of
a subsection, so it is reasonable to opt them out of the
`.cfi_{start,end}proc` nesting check.

Fixes: https://github.com/llvm/llvm-project/issues/82261
(cherry picked from commit 5b91647e3f82c9747c42c3239b7d7f3ade4542a7)

Added: 


Modified: 
llvm/lib/MC/MCParser/AsmParser.cpp
llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s

Removed: 




diff  --git a/llvm/lib/MC/MCParser/AsmParser.cpp 
b/llvm/lib/MC/MCParser/AsmParser.cpp
index 8e508dbdb1c69b..026d252ec5bcd7 100644
--- a/llvm/lib/MC/MCParser/AsmParser.cpp
+++ b/llvm/lib/MC/MCParser/AsmParser.cpp
@@ -44,6 +44,7 @@
 #include "llvm/MC/MCSection.h"
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCSymbol.h"
+#include "llvm/MC/MCSymbolMachO.h"
 #include "llvm/MC/MCTargetOptions.h"
 #include "llvm/MC/MCValue.h"
 #include "llvm/Support/Casting.h"
@@ -1950,7 +1951,8 @@ bool AsmParser::parseStatement(ParseStatementInfo ,
   Lex();
 }
 
-if (MAI.hasSubsectionsViaSymbols() && CFIStartProcLoc && Sym->isExternal())
+if (MAI.hasSubsectionsViaSymbols() && CFIStartProcLoc &&
+Sym->isExternal() && !cast(Sym)->isAltEntry())
   return Error(StartTokLoc, "non-private labels cannot appear between "
 ".cfi_startproc / .cfi_endproc pairs") &&
  Error(*CFIStartProcLoc, "previous .cfi_startproc was here");

diff  --git a/llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s 
b/llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s
index 235b7d44809929..3a5af86defc592 100644
--- a/llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s
+++ b/llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s
@@ -8,6 +8,10 @@
.p2align2
 _locomotive:
.cfi_startproc
+   ; An N_ALT_ENTRY symbol can be defined in the middle of a subsection, so
+   ; these are opted out of the .cfi_{start,end}proc nesting check.
+   .alt_entry _engineer
+_engineer:
ret
 
; It is invalid to have a non-private label between .cfi_startproc and
@@ -17,7 +21,7 @@ _locomotive:
.p2align2
 _caboose:
 ; DARWIN: [[#@LINE-1]]:1: error: non-private labels cannot appear between 
.cfi_startproc / .cfi_endproc pairs
-; DARWIN: [[#@LINE-10]]:2: error: previous .cfi_startproc was here
+; DARWIN: [[#@LINE-14]]:2: error: previous .cfi_startproc was here
ret
.cfi_endproc
 



___
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: Allow .alt_entry symbols to pass the .cfi nesting check (#82268) (PR #83336)

[via llvm-branch-commits]

https://github.com/llvmbot updated 
https://github.com/llvm/llvm-project/pull/83336

>From 267d9b1a74c466fff6bde255a211732aa3bcd7e8 Mon Sep 17 00:00:00 2001
From: Jon Roelofs 
Date: Wed, 28 Feb 2024 13:03:35 -0800
Subject: [PATCH] Allow .alt_entry symbols to pass the .cfi nesting check
 (#82268)

A symbol with an `N_ALT_ENTRY` attribute may be defined in the middle of
a subsection, so it is reasonable to opt them out of the
`.cfi_{start,end}proc` nesting check.

Fixes: https://github.com/llvm/llvm-project/issues/82261
(cherry picked from commit 5b91647e3f82c9747c42c3239b7d7f3ade4542a7)
---
 llvm/lib/MC/MCParser/AsmParser.cpp| 4 +++-
 llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s | 6 +-
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/MC/MCParser/AsmParser.cpp 
b/llvm/lib/MC/MCParser/AsmParser.cpp
index 8e508dbdb1c69b..026d252ec5bcd7 100644
--- a/llvm/lib/MC/MCParser/AsmParser.cpp
+++ b/llvm/lib/MC/MCParser/AsmParser.cpp
@@ -44,6 +44,7 @@
 #include "llvm/MC/MCSection.h"
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCSymbol.h"
+#include "llvm/MC/MCSymbolMachO.h"
 #include "llvm/MC/MCTargetOptions.h"
 #include "llvm/MC/MCValue.h"
 #include "llvm/Support/Casting.h"
@@ -1950,7 +1951,8 @@ bool AsmParser::parseStatement(ParseStatementInfo ,
   Lex();
 }
 
-if (MAI.hasSubsectionsViaSymbols() && CFIStartProcLoc && Sym->isExternal())
+if (MAI.hasSubsectionsViaSymbols() && CFIStartProcLoc &&
+Sym->isExternal() && !cast(Sym)->isAltEntry())
   return Error(StartTokLoc, "non-private labels cannot appear between "
 ".cfi_startproc / .cfi_endproc pairs") &&
  Error(*CFIStartProcLoc, "previous .cfi_startproc was here");
diff --git a/llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s 
b/llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s
index 235b7d44809929..3a5af86defc592 100644
--- a/llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s
+++ b/llvm/test/MC/AArch64/cfi-bad-nesting-darwin.s
@@ -8,6 +8,10 @@
.p2align2
 _locomotive:
.cfi_startproc
+   ; An N_ALT_ENTRY symbol can be defined in the middle of a subsection, so
+   ; these are opted out of the .cfi_{start,end}proc nesting check.
+   .alt_entry _engineer
+_engineer:
ret
 
; It is invalid to have a non-private label between .cfi_startproc and
@@ -17,7 +21,7 @@ _locomotive:
.p2align2
 _caboose:
 ; DARWIN: [[#@LINE-1]]:1: error: non-private labels cannot appear between 
.cfi_startproc / .cfi_endproc pairs
-; DARWIN: [[#@LINE-10]]:2: error: previous .cfi_startproc was here
+; DARWIN: [[#@LINE-14]]:2: error: previous .cfi_startproc was here
ret
.cfi_endproc
 

___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits