[llvm-branch-commits] [llvm] release/18.x: [AArch64] Only apply bool vector bitcast opt if result is scalar (#81256) (PR #81454)
https://github.com/llvmbot milestoned https://github.com/llvm/llvm-project/pull/81454 ___ 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: [AArch64] Only apply bool vector bitcast opt if result is scalar (#81256) (PR #81454)
https://github.com/llvmbot created
https://github.com/llvm/llvm-project/pull/81454
Backport 92d79922051f732560acf3791b543df1e6580689
Requested by: @DianQK
>From c9f2f8538310d6602cfb1cfb0749ef0f0bcbde56 Mon Sep 17 00:00:00 2001
From: Nikita Popov
Date: Mon, 12 Feb 2024 10:00:34 +0100
Subject: [PATCH] [AArch64] Only apply bool vector bitcast opt if result is
scalar (#81256)
This optimization tries to optimize bitcasts from `` to iN, but
currently also triggers for `` to `` bitcasts, if custom
lowering has been requested for these for an unrelated reason. Fix this
by explicitly checking that the result type is scalar.
Fixes https://github.com/llvm/llvm-project/issues/81216.
(cherry picked from commit 92d79922051f732560acf3791b543df1e6580689)
---
.../Target/AArch64/AArch64ISelLowering.cpp| 3 +-
.../AArch64/vec-combine-compare-to-bitmask.ll | 28 +++
2 files changed, 30 insertions(+), 1 deletion(-)
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index e97f5e32201488..2d6b5bc983e739 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -24419,7 +24419,8 @@ void AArch64TargetLowering::ReplaceBITCASTResults(
return;
}
- if (SrcVT.isVector() && SrcVT.getVectorElementType() == MVT::i1)
+ if (SrcVT.isVector() && SrcVT.getVectorElementType() == MVT::i1 &&
+ !VT.isVector())
return replaceBoolVectorBitcast(N, Results, DAG);
if (VT != MVT::i16 || (SrcVT != MVT::f16 && SrcVT != MVT::bf16))
diff --git a/llvm/test/CodeGen/AArch64/vec-combine-compare-to-bitmask.ll
b/llvm/test/CodeGen/AArch64/vec-combine-compare-to-bitmask.ll
index 1b22e2f900ddb7..557aa010b3a7d9 100644
--- a/llvm/test/CodeGen/AArch64/vec-combine-compare-to-bitmask.ll
+++ b/llvm/test/CodeGen/AArch64/vec-combine-compare-to-bitmask.ll
@@ -489,3 +489,31 @@ define i6 @no_combine_illegal_num_elements(<6 x i32> %vec)
{
%bitmask = bitcast <6 x i1> %cmp_result to i6
ret i6 %bitmask
}
+
+; Only apply the combine when casting a vector to a scalar.
+define <2 x i8> @vector_to_vector_cast(<16 x i1> %arg) nounwind {
+; CHECK-LABEL: vector_to_vector_cast:
+; CHECK: ; %bb.0:
+; CHECK-NEXT:sub sp, sp, #16
+; CHECK-NEXT:shl.16b v0, v0, #7
+; CHECK-NEXT: Lloh36:
+; CHECK-NEXT:adrp x8, lCPI20_0@PAGE
+; CHECK-NEXT: Lloh37:
+; CHECK-NEXT:ldr q1, [x8, lCPI20_0@PAGEOFF]
+; CHECK-NEXT:add x8, sp, #14
+; CHECK-NEXT:cmlt.16b v0, v0, #0
+; CHECK-NEXT:and.16b v0, v0, v1
+; CHECK-NEXT:ext.16b v1, v0, v0, #8
+; CHECK-NEXT:zip1.16b v0, v0, v1
+; CHECK-NEXT:addv.8h h0, v0
+; CHECK-NEXT:str h0, [sp, #14]
+; CHECK-NEXT:ld1.b { v0 }[0], [x8]
+; CHECK-NEXT:orr x8, x8, #0x1
+; CHECK-NEXT:ld1.b { v0 }[4], [x8]
+; CHECK-NEXT:; kill: def $d0 killed $d0 killed $q0
+; CHECK-NEXT:add sp, sp, #16
+; CHECK-NEXT:ret
+; CHECK-NEXT:.loh AdrpLdr Lloh36, Lloh37
+ %bc = bitcast <16 x i1> %arg to <2 x i8>
+ ret <2 x i8> %bc
+}
___
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: [AArch64] Only apply bool vector bitcast opt if result is scalar (#81256) (PR #81454)
llvmbot wrote: @lawben What do you think about merging this PR to the release branch? https://github.com/llvm/llvm-project/pull/81454 ___ 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: [AArch64] Only apply bool vector bitcast opt if result is scalar (#81256) (PR #81454)
llvmbot wrote:
@llvm/pr-subscribers-backend-aarch64
Author: None (llvmbot)
Changes
Backport 92d79922051f732560acf3791b543df1e6580689
Requested by: @DianQK
---
Full diff: https://github.com/llvm/llvm-project/pull/81454.diff
2 Files Affected:
- (modified) llvm/lib/Target/AArch64/AArch64ISelLowering.cpp (+2-1)
- (modified) llvm/test/CodeGen/AArch64/vec-combine-compare-to-bitmask.ll (+28)
``diff
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index e97f5e32201488..2d6b5bc983e739 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -24419,7 +24419,8 @@ void AArch64TargetLowering::ReplaceBITCASTResults(
return;
}
- if (SrcVT.isVector() && SrcVT.getVectorElementType() == MVT::i1)
+ if (SrcVT.isVector() && SrcVT.getVectorElementType() == MVT::i1 &&
+ !VT.isVector())
return replaceBoolVectorBitcast(N, Results, DAG);
if (VT != MVT::i16 || (SrcVT != MVT::f16 && SrcVT != MVT::bf16))
diff --git a/llvm/test/CodeGen/AArch64/vec-combine-compare-to-bitmask.ll
b/llvm/test/CodeGen/AArch64/vec-combine-compare-to-bitmask.ll
index 1b22e2f900ddb7..557aa010b3a7d9 100644
--- a/llvm/test/CodeGen/AArch64/vec-combine-compare-to-bitmask.ll
+++ b/llvm/test/CodeGen/AArch64/vec-combine-compare-to-bitmask.ll
@@ -489,3 +489,31 @@ define i6 @no_combine_illegal_num_elements(<6 x i32> %vec)
{
%bitmask = bitcast <6 x i1> %cmp_result to i6
ret i6 %bitmask
}
+
+; Only apply the combine when casting a vector to a scalar.
+define <2 x i8> @vector_to_vector_cast(<16 x i1> %arg) nounwind {
+; CHECK-LABEL: vector_to_vector_cast:
+; CHECK: ; %bb.0:
+; CHECK-NEXT:sub sp, sp, #16
+; CHECK-NEXT:shl.16b v0, v0, #7
+; CHECK-NEXT: Lloh36:
+; CHECK-NEXT:adrp x8, lCPI20_0@PAGE
+; CHECK-NEXT: Lloh37:
+; CHECK-NEXT:ldr q1, [x8, lCPI20_0@PAGEOFF]
+; CHECK-NEXT:add x8, sp, #14
+; CHECK-NEXT:cmlt.16b v0, v0, #0
+; CHECK-NEXT:and.16b v0, v0, v1
+; CHECK-NEXT:ext.16b v1, v0, v0, #8
+; CHECK-NEXT:zip1.16b v0, v0, v1
+; CHECK-NEXT:addv.8h h0, v0
+; CHECK-NEXT:str h0, [sp, #14]
+; CHECK-NEXT:ld1.b { v0 }[0], [x8]
+; CHECK-NEXT:orr x8, x8, #0x1
+; CHECK-NEXT:ld1.b { v0 }[4], [x8]
+; CHECK-NEXT:; kill: def $d0 killed $d0 killed $q0
+; CHECK-NEXT:add sp, sp, #16
+; CHECK-NEXT:ret
+; CHECK-NEXT:.loh AdrpLdr Lloh36, Lloh37
+ %bc = bitcast <16 x i1> %arg to <2 x i8>
+ ret <2 x i8> %bc
+}
``
https://github.com/llvm/llvm-project/pull/81454
___
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 `moveOpBefore`/`After` in dialect conversion (PR #81240)
https://github.com/matthias-springer updated
https://github.com/llvm/llvm-project/pull/81240
>From c60c43bcd2296715ceca83a3f9666433883ec303 Mon Sep 17 00:00:00 2001
From: Matthias Springer
Date: Mon, 12 Feb 2024 09:05:50 +
Subject: [PATCH 1/2] [mlir][Transforms][WIP] RewriteAction
BEGIN_PUBLIC
No public commit message needed for presubmit.
END_PUBLIC
---
.../Transforms/Utils/DialectConversion.cpp| 504 +++---
1 file changed, 306 insertions(+), 198 deletions(-)
diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp
b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index e41231d7cbd390..edca84e5a73f04 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -154,13 +154,12 @@ namespace {
struct RewriterState {
RewriterState(unsigned numCreatedOps, unsigned numUnresolvedMaterializations,
unsigned numReplacements, unsigned numArgReplacements,
-unsigned numBlockActions, unsigned numIgnoredOperations,
+unsigned numRewrites, unsigned numIgnoredOperations,
unsigned numRootUpdates)
: numCreatedOps(numCreatedOps),
numUnresolvedMaterializations(numUnresolvedMaterializations),
numReplacements(numReplacements),
-numArgReplacements(numArgReplacements),
-numBlockActions(numBlockActions),
+numArgReplacements(numArgReplacements), numRewrites(numRewrites),
numIgnoredOperations(numIgnoredOperations),
numRootUpdates(numRootUpdates) {}
@@ -176,8 +175,8 @@ struct RewriterState {
/// The current number of argument replacements queued.
unsigned numArgReplacements;
- /// The current number of block actions performed.
- unsigned numBlockActions;
+ /// The current number of rewrites performed.
+ unsigned numRewrites;
/// The current number of ignored operations.
unsigned numIgnoredOperations;
@@ -235,86 +234,6 @@ struct OpReplacement {
const TypeConverter *converter;
};
-//===--===//
-// BlockAction
-
-/// The kind of the block action performed during the rewrite. Actions can be
-/// undone if the conversion fails.
-enum class BlockActionKind {
- Create,
- Erase,
- Inline,
- Move,
- Split,
- TypeConversion
-};
-
-/// Original position of the given block in its parent region. During undo
-/// actions, the block needs to be placed before `insertBeforeBlock`.
-struct BlockPosition {
- Region *region;
- Block *insertBeforeBlock;
-};
-
-/// Information needed to undo inlining actions.
-/// - the source block
-/// - the first inlined operation (could be null if the source block was empty)
-/// - the last inlined operation (could be null if the source block was empty)
-struct InlineInfo {
- Block *sourceBlock;
- Operation *firstInlinedInst;
- Operation *lastInlinedInst;
-};
-
-/// The storage class for an undoable block action (one of BlockActionKind),
-/// contains the information necessary to undo this action.
-struct BlockAction {
- static BlockAction getCreate(Block *block) {
-return {BlockActionKind::Create, block, {}};
- }
- static BlockAction getErase(Block *block, BlockPosition originalPosition) {
-return {BlockActionKind::Erase, block, {originalPosition}};
- }
- static BlockAction getInline(Block *block, Block *srcBlock,
- Block::iterator before) {
-BlockAction action{BlockActionKind::Inline, block, {}};
-action.inlineInfo = {srcBlock,
- srcBlock->empty() ? nullptr : &srcBlock->front(),
- srcBlock->empty() ? nullptr : &srcBlock->back()};
-return action;
- }
- static BlockAction getMove(Block *block, BlockPosition originalPosition) {
-return {BlockActionKind::Move, block, {originalPosition}};
- }
- static BlockAction getSplit(Block *block, Block *originalBlock) {
-BlockAction action{BlockActionKind::Split, block, {}};
-action.originalBlock = originalBlock;
-return action;
- }
- static BlockAction getTypeConversion(Block *block) {
-return BlockAction{BlockActionKind::TypeConversion, block, {}};
- }
-
- // The action kind.
- BlockActionKind kind;
-
- // A pointer to the block that was created by the action.
- Block *block;
-
- union {
-// In use if kind == BlockActionKind::Inline or BlockActionKind::Erase, and
-// contains a pointer to the region that originally contained the block as
-// well as the position of the block in that region.
-BlockPosition originalPosition;
-// In use if kind == BlockActionKind::Split and contains a pointer to the
-// block that was split into two parts.
-Block *originalBlock;
-// In use if kind == BlockActionKind::Inline, and contains the information
-// needed to undo the inlining.
-InlineInfo inlineInfo;
- };
-};
-
//===--===/
[llvm-branch-commits] [mlir] [mlir][Transforms][NFC] Modularize block actions (PR #81237)
https://github.com/matthias-springer updated
https://github.com/llvm/llvm-project/pull/81237
>From c60c43bcd2296715ceca83a3f9666433883ec303 Mon Sep 17 00:00:00 2001
From: Matthias Springer
Date: Mon, 12 Feb 2024 09:05:50 +
Subject: [PATCH] [mlir][Transforms][WIP] RewriteAction
BEGIN_PUBLIC
No public commit message needed for presubmit.
END_PUBLIC
---
.../Transforms/Utils/DialectConversion.cpp| 504 +++---
1 file changed, 306 insertions(+), 198 deletions(-)
diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp
b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index e41231d7cbd390..edca84e5a73f04 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -154,13 +154,12 @@ namespace {
struct RewriterState {
RewriterState(unsigned numCreatedOps, unsigned numUnresolvedMaterializations,
unsigned numReplacements, unsigned numArgReplacements,
-unsigned numBlockActions, unsigned numIgnoredOperations,
+unsigned numRewrites, unsigned numIgnoredOperations,
unsigned numRootUpdates)
: numCreatedOps(numCreatedOps),
numUnresolvedMaterializations(numUnresolvedMaterializations),
numReplacements(numReplacements),
-numArgReplacements(numArgReplacements),
-numBlockActions(numBlockActions),
+numArgReplacements(numArgReplacements), numRewrites(numRewrites),
numIgnoredOperations(numIgnoredOperations),
numRootUpdates(numRootUpdates) {}
@@ -176,8 +175,8 @@ struct RewriterState {
/// The current number of argument replacements queued.
unsigned numArgReplacements;
- /// The current number of block actions performed.
- unsigned numBlockActions;
+ /// The current number of rewrites performed.
+ unsigned numRewrites;
/// The current number of ignored operations.
unsigned numIgnoredOperations;
@@ -235,86 +234,6 @@ struct OpReplacement {
const TypeConverter *converter;
};
-//===--===//
-// BlockAction
-
-/// The kind of the block action performed during the rewrite. Actions can be
-/// undone if the conversion fails.
-enum class BlockActionKind {
- Create,
- Erase,
- Inline,
- Move,
- Split,
- TypeConversion
-};
-
-/// Original position of the given block in its parent region. During undo
-/// actions, the block needs to be placed before `insertBeforeBlock`.
-struct BlockPosition {
- Region *region;
- Block *insertBeforeBlock;
-};
-
-/// Information needed to undo inlining actions.
-/// - the source block
-/// - the first inlined operation (could be null if the source block was empty)
-/// - the last inlined operation (could be null if the source block was empty)
-struct InlineInfo {
- Block *sourceBlock;
- Operation *firstInlinedInst;
- Operation *lastInlinedInst;
-};
-
-/// The storage class for an undoable block action (one of BlockActionKind),
-/// contains the information necessary to undo this action.
-struct BlockAction {
- static BlockAction getCreate(Block *block) {
-return {BlockActionKind::Create, block, {}};
- }
- static BlockAction getErase(Block *block, BlockPosition originalPosition) {
-return {BlockActionKind::Erase, block, {originalPosition}};
- }
- static BlockAction getInline(Block *block, Block *srcBlock,
- Block::iterator before) {
-BlockAction action{BlockActionKind::Inline, block, {}};
-action.inlineInfo = {srcBlock,
- srcBlock->empty() ? nullptr : &srcBlock->front(),
- srcBlock->empty() ? nullptr : &srcBlock->back()};
-return action;
- }
- static BlockAction getMove(Block *block, BlockPosition originalPosition) {
-return {BlockActionKind::Move, block, {originalPosition}};
- }
- static BlockAction getSplit(Block *block, Block *originalBlock) {
-BlockAction action{BlockActionKind::Split, block, {}};
-action.originalBlock = originalBlock;
-return action;
- }
- static BlockAction getTypeConversion(Block *block) {
-return BlockAction{BlockActionKind::TypeConversion, block, {}};
- }
-
- // The action kind.
- BlockActionKind kind;
-
- // A pointer to the block that was created by the action.
- Block *block;
-
- union {
-// In use if kind == BlockActionKind::Inline or BlockActionKind::Erase, and
-// contains a pointer to the region that originally contained the block as
-// well as the position of the block in that region.
-BlockPosition originalPosition;
-// In use if kind == BlockActionKind::Split and contains a pointer to the
-// block that was split into two parts.
-Block *originalBlock;
-// In use if kind == BlockActionKind::Inline, and contains the information
-// needed to undo the inlining.
-InlineInfo inlineInfo;
- };
-};
-
//===--===//
/
[llvm-branch-commits] [mlir] [mlir][Transforms][NFC] Turn in-place op modifications into `RewriteAction`s (PR #81245)
https://github.com/matthias-springer updated
https://github.com/llvm/llvm-project/pull/81245
>From f7010ea9f363c011f4171a6329ad6de30880c3e7 Mon Sep 17 00:00:00 2001
From: Matthias Springer
Date: Mon, 12 Feb 2024 09:11:47 +
Subject: [PATCH] [mlir][Transforms][NFC] Turn in-place op modifications into
`RewriteAction`s
This commit simplifies the internal state of the dialect conversion. A separate
field for the previous state of in-place op modifications is no longer needed.
BEGIN_PUBLIC
No public commit message needed for presubmit.
END_PUBLIC
---
.../mlir/Transforms/DialectConversion.h | 4 +-
.../Transforms/Utils/DialectConversion.cpp| 123 --
2 files changed, 56 insertions(+), 71 deletions(-)
diff --git a/mlir/include/mlir/Transforms/DialectConversion.h
b/mlir/include/mlir/Transforms/DialectConversion.h
index b028d2b71b3762..c0c702a7d34821 100644
--- a/mlir/include/mlir/Transforms/DialectConversion.h
+++ b/mlir/include/mlir/Transforms/DialectConversion.h
@@ -721,8 +721,8 @@ class ConversionPatternRewriter final : public
PatternRewriter {
/// PatternRewriter hook for updating the given operation in-place.
/// Note: These methods only track updates to the given operation itself,
- /// and not nested regions. Updates to regions will still require
- /// notification through other more specific hooks above.
+ /// and not nested regions. Updates to regions will still require
notification
+ /// through other more specific hooks above.
void startOpModification(Operation *op) override;
/// PatternRewriter hook for updating the given operation in-place.
diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp
b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index 85b67bb834de7c..489ccd0139c7f2 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -154,14 +154,12 @@ namespace {
struct RewriterState {
RewriterState(unsigned numCreatedOps, unsigned numUnresolvedMaterializations,
unsigned numReplacements, unsigned numArgReplacements,
-unsigned numRewrites, unsigned numIgnoredOperations,
-unsigned numRootUpdates)
+unsigned numRewrites, unsigned numIgnoredOperations)
: numCreatedOps(numCreatedOps),
numUnresolvedMaterializations(numUnresolvedMaterializations),
numReplacements(numReplacements),
numArgReplacements(numArgReplacements), numRewrites(numRewrites),
-numIgnoredOperations(numIgnoredOperations),
-numRootUpdates(numRootUpdates) {}
+numIgnoredOperations(numIgnoredOperations) {}
/// The current number of created operations.
unsigned numCreatedOps;
@@ -180,44 +178,6 @@ struct RewriterState {
/// The current number of ignored operations.
unsigned numIgnoredOperations;
-
- /// The current number of operations that were updated in place.
- unsigned numRootUpdates;
-};
-
-//===--===//
-// OperationTransactionState
-
-/// The state of an operation that was updated by a pattern in-place. This
-/// contains all of the necessary information to reconstruct an operation that
-/// was updated in place.
-class OperationTransactionState {
-public:
- OperationTransactionState() = default;
- OperationTransactionState(Operation *op)
- : op(op), loc(op->getLoc()), attrs(op->getAttrDictionary()),
-operands(op->operand_begin(), op->operand_end()),
-successors(op->successor_begin(), op->successor_end()) {}
-
- /// Discard the transaction state and reset the state of the original
- /// operation.
- void resetOperation() const {
-op->setLoc(loc);
-op->setAttrs(attrs);
-op->setOperands(operands);
-for (const auto &it : llvm::enumerate(successors))
- op->setSuccessor(it.value(), it.index());
- }
-
- /// Return the original operation of this state.
- Operation *getOperation() const { return op; }
-
-private:
- Operation *op;
- LocationAttr loc;
- DictionaryAttr attrs;
- SmallVector operands;
- SmallVector successors;
};
//===--===//
@@ -761,7 +721,8 @@ class IRRewrite {
MoveBlock,
SplitBlock,
BlockTypeConversion,
-MoveOperation
+MoveOperation,
+ModifyOperation
};
virtual ~IRRewrite() = default;
@@ -992,7 +953,7 @@ class OperationRewrite : public IRRewrite {
static bool classof(const IRRewrite *rewrite) {
return rewrite->getKind() >= Kind::MoveOperation &&
- rewrite->getKind() <= Kind::MoveOperation;
+ rewrite->getKind() <= Kind::ModifyOperation;
}
protected:
@@ -1031,6 +992,34 @@ class MoveOperationRewrite : public OperationRewrite {
// this operation was the only operation in the region.
Operation *insertBeforeOp;
};
+
+/// In-place modification of an op. This rewrite is immediately reflected in
+///
[llvm-branch-commits] [mlir] [mlir][Transforms][NFC] Modularize block actions (PR #81237)
https://github.com/matthias-springer edited https://github.com/llvm/llvm-project/pull/81237 ___ 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] Modularize block actions (PR #81237)
@@ -820,6 +740,238 @@ void ArgConverter::insertConversion(Block *newBlock,
conversionInfo.insert({newBlock, std::move(info)});
}
+//===--===//
+// RewriteAction
matthias-springer wrote:
I renamed `RewriteAction` to `IRRewrite`.
https://github.com/llvm/llvm-project/pull/81237
___
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 `moveOpBefore`/`After` in dialect conversion (PR #81240)
https://github.com/matthias-springer edited https://github.com/llvm/llvm-project/pull/81240 ___ 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 `moveOpBefore`/`After` in dialect conversion (PR #81240)
@@ -970,6 +971,54 @@ class BlockTypeConversionAction : public BlockAction {
void rollback() override;
};
+
+/// An operation rewrite.
matthias-springer wrote:
I renamed the class to `OperationRewrite`. I added more documentation to the
superclass `IRRewrite` in the dependent PR.
https://github.com/llvm/llvm-project/pull/81240
___
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] Turn in-place op modification into `IRRewrite` (PR #81245)
https://github.com/matthias-springer edited https://github.com/llvm/llvm-project/pull/81245 ___ 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] [TBAA] Use !tbaa for first accessed field, even if there are others. (PR #81313)
https://github.com/dobbelaj-snps approved this pull request. lgtm Maybe rephrase the commit message to something like: [tbaa] Use !tbaa for first accessed field if it is an exact match in offset and size. https://github.com/llvm/llvm-project/pull/81313 ___ 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] Simplify `ArgConverter` state (PR #81462)
https://github.com/matthias-springer created
https://github.com/llvm/llvm-project/pull/81462
* When converting a block signature, `ArgConverter` creates a new block with
the new signature and moves all operation from the old block to the new block.
The new block is temporarily inserted into a region that is stored in
`regionMapping`. The old block is not yet deleted, so that the conversion can
be rolled back. `regionMapping` is not needed. Instead of moving the old block
to a temporary region, it can just be unlinked. Block erasures are handles in
the same way in the dialect conversion.
* `regionToConverter` is a mapping from regions to type converter. That field
is never accessed within `ArgConverter`. It should be stored in
`ConversionPatternRewriterImpl` instead.
>From 5af1476c5d439c14122ffb50d24923e522a61b32 Mon Sep 17 00:00:00 2001
From: Matthias Springer
Date: Mon, 12 Feb 2024 10:53:59 +
Subject: [PATCH] [mlir][Transforms][NFC] Simplify `ArgConverter` state
* When converting a block signature, `ArgConverter` creates a new block with
the new signature and moves all operation from the old block to the new block.
The new block is temporarily inserted into a region that is stored in
`regionMapping`. The old block is not yet deleted, so that the conversion can
be rolled back. `regionMapping` is not needed. Instead of moving the old block
to a temporary region, it can just be unlinked. Block erasures are handles in
the same way in the dialect conversion.
* `regionToConverter` is a mapping from regions to type converter. That field
is never accessed within `ArgConverter`. It should be stored in
`ConversionPatternRewriterImpl` instead.
---
.../Transforms/Utils/DialectConversion.cpp| 63 ++-
1 file changed, 18 insertions(+), 45 deletions(-)
diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp
b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index 489ccd0139c7f..53717f632621d 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -348,18 +348,6 @@ struct ArgConverter {
return conversionInfo.count(block) || convertedBlocks.count(block);
}
- /// Set the type converter to use for the given region.
- void setConverter(Region *region, const TypeConverter *typeConverter) {
-assert(typeConverter && "expected valid type converter");
-regionToConverter[region] = typeConverter;
- }
-
- /// Return the type converter to use for the given region, or null if there
- /// isn't one.
- const TypeConverter *getConverter(Region *region) {
-return regionToConverter.lookup(region);
- }
-
//======//
// Rewrite Application
//======//
@@ -409,9 +397,6 @@ struct ArgConverter {
ConversionValueMapping &mapping,
SmallVectorImpl &argReplacements);
- /// Insert a new conversion into the cache.
- void insertConversion(Block *newBlock, ConvertedBlockInfo &&info);
-
/// A collection of blocks that have had their arguments converted. This is a
/// map from the new replacement block, back to the original block.
llvm::MapVector conversionInfo;
@@ -419,14 +404,6 @@ struct ArgConverter {
/// The set of original blocks that were converted.
DenseSet convertedBlocks;
- /// A mapping from valid regions, to those containing the original blocks of
a
- /// conversion.
- DenseMap> regionMapping;
-
- /// A mapping of regions to type converters that should be used when
- /// converting the arguments of blocks within that region.
- DenseMap regionToConverter;
-
/// The pattern rewriter to use when materializing conversions.
PatternRewriter &rewriter;
@@ -474,9 +451,10 @@ void ArgConverter::discardRewrites(Block *block) {
block->getArgument(i).dropAllUses();
block->replaceAllUsesWith(origBlock);
- // Move the operations back the original block and the delete the new block.
+ // Move the operations back the original block, move the original block back
+ // into its original location and the delete the new block.
origBlock->getOperations().splice(origBlock->end(), block->getOperations());
- origBlock->moveBefore(block);
+ block->getParent()->getBlocks().insert(Region::iterator(block), origBlock);
block->erase();
convertedBlocks.erase(origBlock);
@@ -510,6 +488,9 @@ void ArgConverter::applyRewrites(ConversionValueMapping
&mapping) {
mapping.lookupOrDefault(castValue, origArg.getType()));
}
}
+
+delete origBlock;
+blockInfo.origBlock = nullptr;
}
}
@@ -603,6 +584,9 @@ Block *ArgConverter::applySignatureConversion(
// signature.
Block *newBlock = block->splitBlock(block->begin());
block->replaceAllUsesWith(newBlock);
+ // Unlink the block, but do not erase it yet, so that the change can be
rolled
+ // back.
+ block->getParent()->getBlocks().remo
[llvm-branch-commits] [mlir] [mlir][Transforms][NFC] Simplify `ArgConverter` state (PR #81462)
llvmbot wrote:
@llvm/pr-subscribers-mlir-core
Author: Matthias Springer (matthias-springer)
Changes
* When converting a block signature, `ArgConverter` creates a new block with
the new signature and moves all operation from the old block to the new block.
The new block is temporarily inserted into a region that is stored in
`regionMapping`. The old block is not yet deleted, so that the conversion can
be rolled back. `regionMapping` is not needed. Instead of moving the old block
to a temporary region, it can just be unlinked. Block erasures are handles in
the same way in the dialect conversion.
* `regionToConverter` is a mapping from regions to type converter. That field
is never accessed within `ArgConverter`. It should be stored in
`ConversionPatternRewriterImpl` instead.
---
Full diff: https://github.com/llvm/llvm-project/pull/81462.diff
1 Files Affected:
- (modified) mlir/lib/Transforms/Utils/DialectConversion.cpp (+18-45)
``diff
diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp
b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index 489ccd0139c7f2..53717f632621dd 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -348,18 +348,6 @@ struct ArgConverter {
return conversionInfo.count(block) || convertedBlocks.count(block);
}
- /// Set the type converter to use for the given region.
- void setConverter(Region *region, const TypeConverter *typeConverter) {
-assert(typeConverter && "expected valid type converter");
-regionToConverter[region] = typeConverter;
- }
-
- /// Return the type converter to use for the given region, or null if there
- /// isn't one.
- const TypeConverter *getConverter(Region *region) {
-return regionToConverter.lookup(region);
- }
-
//======//
// Rewrite Application
//======//
@@ -409,9 +397,6 @@ struct ArgConverter {
ConversionValueMapping &mapping,
SmallVectorImpl &argReplacements);
- /// Insert a new conversion into the cache.
- void insertConversion(Block *newBlock, ConvertedBlockInfo &&info);
-
/// A collection of blocks that have had their arguments converted. This is a
/// map from the new replacement block, back to the original block.
llvm::MapVector conversionInfo;
@@ -419,14 +404,6 @@ struct ArgConverter {
/// The set of original blocks that were converted.
DenseSet convertedBlocks;
- /// A mapping from valid regions, to those containing the original blocks of
a
- /// conversion.
- DenseMap> regionMapping;
-
- /// A mapping of regions to type converters that should be used when
- /// converting the arguments of blocks within that region.
- DenseMap regionToConverter;
-
/// The pattern rewriter to use when materializing conversions.
PatternRewriter &rewriter;
@@ -474,9 +451,10 @@ void ArgConverter::discardRewrites(Block *block) {
block->getArgument(i).dropAllUses();
block->replaceAllUsesWith(origBlock);
- // Move the operations back the original block and the delete the new block.
+ // Move the operations back the original block, move the original block back
+ // into its original location and the delete the new block.
origBlock->getOperations().splice(origBlock->end(), block->getOperations());
- origBlock->moveBefore(block);
+ block->getParent()->getBlocks().insert(Region::iterator(block), origBlock);
block->erase();
convertedBlocks.erase(origBlock);
@@ -510,6 +488,9 @@ void ArgConverter::applyRewrites(ConversionValueMapping
&mapping) {
mapping.lookupOrDefault(castValue, origArg.getType()));
}
}
+
+delete origBlock;
+blockInfo.origBlock = nullptr;
}
}
@@ -603,6 +584,9 @@ Block *ArgConverter::applySignatureConversion(
// signature.
Block *newBlock = block->splitBlock(block->begin());
block->replaceAllUsesWith(newBlock);
+ // Unlink the block, but do not erase it yet, so that the change can be
rolled
+ // back.
+ block->getParent()->getBlocks().remove(block);
// Map all new arguments to the location of the argument they originate from.
SmallVector newLocs(convertedTypes.size(),
@@ -679,24 +663,9 @@ Block *ArgConverter::applySignatureConversion(
ConvertedArgInfo(inputMap->inputNo, inputMap->size, newArg);
}
- // Remove the original block from the region and return the new one.
- insertConversion(newBlock, std::move(info));
- return newBlock;
-}
-
-void ArgConverter::insertConversion(Block *newBlock,
-ConvertedBlockInfo &&info) {
- // Get a region to insert the old block.
- Region *region = newBlock->getParent();
- std::unique_ptr &mappedRegion = regionMapping[region];
- if (!mappedRegion)
-mappedRegion = std::make_unique(region->getParentOp());
-
- // Move the original block to the mapped region and
[llvm-branch-commits] [mlir] [mlir][Transforms][NFC] Simplify `ArgConverter` state (PR #81462)
https://github.com/matthias-springer updated
https://github.com/llvm/llvm-project/pull/81462
>From ba1a8085ab576b56a7d1780d9a3f01cb6769df65 Mon Sep 17 00:00:00 2001
From: Matthias Springer
Date: Mon, 12 Feb 2024 11:24:52 +
Subject: [PATCH] [mlir][Transforms][NFC] Simplify `ArgConverter` state
* When converting a block signature, `ArgConverter` creates a new block with
the new signature and moves all operation from the old block to the new block.
The new block is temporarily inserted into a region that is stored in
`regionMapping`. The old block is not yet deleted, so that the conversion can
be rolled back. `regionMapping` is not needed. Instead of moving the old block
to a temporary region, it can just be unlinked. Block erasures are handles in
the same way in the dialect conversion.
* `regionToConverter` is a mapping from regions to type converter. That field
is never accessed within `ArgConverter`. It should be stored in
`ConversionPatternRewriterImpl` instead.
---
.../Transforms/Utils/DialectConversion.cpp| 79 ++-
1 file changed, 22 insertions(+), 57 deletions(-)
diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp
b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index 489ccd0139c7f2..e1c838ea1e43a4 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -343,23 +343,6 @@ struct ArgConverter {
const TypeConverter *converter;
};
- /// Return if the signature of the given block has already been converted.
- bool hasBeenConverted(Block *block) const {
-return conversionInfo.count(block) || convertedBlocks.count(block);
- }
-
- /// Set the type converter to use for the given region.
- void setConverter(Region *region, const TypeConverter *typeConverter) {
-assert(typeConverter && "expected valid type converter");
-regionToConverter[region] = typeConverter;
- }
-
- /// Return the type converter to use for the given region, or null if there
- /// isn't one.
- const TypeConverter *getConverter(Region *region) {
-return regionToConverter.lookup(region);
- }
-
//======//
// Rewrite Application
//======//
@@ -409,24 +392,10 @@ struct ArgConverter {
ConversionValueMapping &mapping,
SmallVectorImpl &argReplacements);
- /// Insert a new conversion into the cache.
- void insertConversion(Block *newBlock, ConvertedBlockInfo &&info);
-
/// A collection of blocks that have had their arguments converted. This is a
/// map from the new replacement block, back to the original block.
llvm::MapVector conversionInfo;
- /// The set of original blocks that were converted.
- DenseSet convertedBlocks;
-
- /// A mapping from valid regions, to those containing the original blocks of
a
- /// conversion.
- DenseMap> regionMapping;
-
- /// A mapping of regions to type converters that should be used when
- /// converting the arguments of blocks within that region.
- DenseMap regionToConverter;
-
/// The pattern rewriter to use when materializing conversions.
PatternRewriter &rewriter;
@@ -474,12 +443,12 @@ void ArgConverter::discardRewrites(Block *block) {
block->getArgument(i).dropAllUses();
block->replaceAllUsesWith(origBlock);
- // Move the operations back the original block and the delete the new block.
+ // Move the operations back the original block, move the original block back
+ // into its original location and the delete the new block.
origBlock->getOperations().splice(origBlock->end(), block->getOperations());
- origBlock->moveBefore(block);
+ block->getParent()->getBlocks().insert(Region::iterator(block), origBlock);
block->erase();
- convertedBlocks.erase(origBlock);
conversionInfo.erase(it);
}
@@ -510,6 +479,9 @@ void ArgConverter::applyRewrites(ConversionValueMapping
&mapping) {
mapping.lookupOrDefault(castValue, origArg.getType()));
}
}
+
+delete origBlock;
+blockInfo.origBlock = nullptr;
}
}
@@ -572,9 +544,11 @@ FailureOr ArgConverter::convertSignature(
Block *block, const TypeConverter *converter,
ConversionValueMapping &mapping,
SmallVectorImpl &argReplacements) {
- // Check if the block was already converted. If the block is detached,
- // conservatively assume it is going to be deleted.
- if (hasBeenConverted(block) || !block->getParent())
+ // Check if the block was already converted.
+ // * If the block is mapped in `conversionInfo`, it is a converted block.
+ // * If the block is detached, conservatively assume that it is going to be
+ // deleted; it is likely the old block (before it was converted).
+ if (conversionInfo.count(block) || !block->getParent())
return block;
// If a converter wasn't provided, and the block wasn't already converted,
// there is nothing we can do
[llvm-branch-commits] [mlir] [mlir][Transforms][NFC] Simplify `ArgConverter` state (PR #81462)
https://github.com/matthias-springer edited https://github.com/llvm/llvm-project/pull/81462 ___ 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] [test] Avoid printing timestamps past INT32_MAX with llvm-readobj (#81463) (PR #81468)
https://github.com/llvmbot milestoned https://github.com/llvm/llvm-project/pull/81468 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] [AArch64] Improve cost model for legal subvec insert/extract (PR #81135)
https://github.com/huntergr-arm edited https://github.com/llvm/llvm-project/pull/81135 ___ 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] [test] Avoid printing timestamps past INT32_MAX with llvm-readobj (#81463) (PR #81468)
https://github.com/llvmbot updated
https://github.com/llvm/llvm-project/pull/81468
>From 996b6d7194c6bc6dc4b9aa87bcca3525dafcaa03 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?=
Date: Mon, 12 Feb 2024 13:22:45 +0200
Subject: [PATCH] [LLD] [test] Avoid printing timestamps past INT32_MAX with
llvm-readobj (#81463)
If llvm-readobj is built with a 32 bit time_t, it can't print such
timestamps correctly.
(cherry picked from commit 0bf4ff29816c0eead99ba576a2df2e3c4d214b1f)
---
lld/test/COFF/timestamp.test | 29 +++--
1 file changed, 19 insertions(+), 10 deletions(-)
diff --git a/lld/test/COFF/timestamp.test b/lld/test/COFF/timestamp.test
index c0658d6109811b..cc73af13c38ca6 100644
--- a/lld/test/COFF/timestamp.test
+++ b/lld/test/COFF/timestamp.test
@@ -4,19 +4,28 @@ RUN: lld-link %t.obj /debug /Brepro /entry:main /nodefaultlib
/out:%t.1.exe
RUN: lld-link %t.obj /debug /Brepro /entry:main /nodefaultlib /out:%t.2.exe
RUN: lld-link %t.obj /debug /timestamp:0 /entry:main /nodefaultlib
/out:%t.3.exe
RUN: env SOURCE_DATE_EPOCH=0 lld-link %t.obj /debug /entry:main /nodefaultlib
/out:%t.4.exe
-RUN: lld-link %t.obj /debug /timestamp:4294967295 /entry:main /nodefaultlib
/out:%t.5.exe
-RUN: env SOURCE_DATE_EPOCH=4294967295 lld-link %t.obj /debug /entry:main
/nodefaultlib /out:%t.6.exe
+# Test timestamps corresponding to INT32_TMAX
+RUN: lld-link %t.obj /debug /timestamp:2147483647 /entry:main /nodefaultlib
/out:%t.5.exe
+RUN: env SOURCE_DATE_EPOCH=2147483647 lld-link %t.obj /debug /entry:main
/nodefaultlib /out:%t.6.exe
+# Test that the command line option /timestamp has precedence over
SOURCE_DATE_EPOCH
RUN: env SOURCE_DATE_EPOCH=12345 lld-link %t.obj /debug /timestamp:0
/entry:main /nodefaultlib /out:%t.7.exe
-RUN: env LLD_IN_TEST=1 not lld-link %t.obj /debug /timestamp:4294967296
/entry:main /nodefaultlib /out:%t.8.exe 2>&1 | FileCheck %s --check-prefix=ERROR
-RUN: env SOURCE_DATE_EPOCH=4294967296 env LLD_IN_TEST=1 not lld-link %t.obj
/debug /entry:main /nodefaultlib /out:%t.9.exe 2>&1 | FileCheck %s
--check-prefix=ERROR2
+# Test timestamps corresponding to UINT32_TMAX
+RUN: lld-link %t.obj /debug /timestamp:4294967295 /entry:main /nodefaultlib
/out:%t.8.exe
+RUN: env SOURCE_DATE_EPOCH=4294967295 lld-link %t.obj /debug /entry:main
/nodefaultlib /out:%t.9.exe
+# Test that setting UINT32_MAX+1 as timestamp fails.
+RUN: env LLD_IN_TEST=1 not lld-link %t.obj /debug /timestamp:4294967296
/entry:main /nodefaultlib /out:%t.10.exe 2>&1 | FileCheck %s
--check-prefix=ERROR
+RUN: env SOURCE_DATE_EPOCH=4294967296 env LLD_IN_TEST=1 not lld-link %t.obj
/debug /entry:main /nodefaultlib /out:%t.11.exe 2>&1 | FileCheck %s
--check-prefix=ERROR2
RUN: llvm-readobj --file-headers --coff-debug-directory %t.1.exe | FileCheck
%s --check-prefix=HASH
RUN: llvm-readobj --file-headers --coff-debug-directory %t.2.exe | FileCheck
%s --check-prefix=HASH
RUN: llvm-readobj --file-headers --coff-debug-directory %t.3.exe | FileCheck
%s --check-prefix=ZERO
RUN: llvm-readobj --file-headers --coff-debug-directory %t.4.exe | FileCheck
%s --check-prefix=ZERO
-RUN: llvm-readobj --file-headers --coff-debug-directory %t.5.exe | FileCheck
%s --check-prefix=MAX
-RUN: llvm-readobj --file-headers --coff-debug-directory %t.6.exe | FileCheck
%s --check-prefix=MAX
+RUN: llvm-readobj --file-headers --coff-debug-directory %t.5.exe | FileCheck
%s --check-prefix=LARGE
+RUN: llvm-readobj --file-headers --coff-debug-directory %t.6.exe | FileCheck
%s --check-prefix=LARGE
RUN: llvm-readobj --file-headers --coff-debug-directory %t.7.exe | FileCheck
%s --check-prefix=ZERO
+# Not inspecting %t.8.exe and %t.9.exe; llvm-readobj with a 32 bit time_t
fails to print dates
+# past INT32_MAX correctly.
+
HASH: ImageFileHeader {
HASH: TimeDateStamp: [[STAMP:.*]]
HASH: DebugDirectory [
@@ -27,10 +36,10 @@ ZERO: TimeDateStamp: 1970-01-01 00:00:00 (0x0)
ZERO: DebugDirectory [
ZERO: TimeDateStamp: 1970-01-01 00:00:00 (0x0)
-MAX: ImageFileHeader {
-MAX: TimeDateStamp: 2106-02-07 06:28:15 (0x)
-MAX: DebugDirectory [
-MAX: TimeDateStamp: 2106-02-07 06:28:15 (0x)
+LARGE: ImageFileHeader {
+LARGE: TimeDateStamp: 2038-01-19 03:14:07 (0x7FFF)
+LARGE: DebugDirectory [
+LARGE: TimeDateStamp: 2038-01-19 03:14:07 (0x7FFF)
ERROR: error: invalid timestamp: 4294967296. Expected 32-bit integer
ERROR2: error: invalid SOURCE_DATE_EPOCH timestamp: 4294967296. Expected
32-bit integer
___
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] [test] Avoid printing timestamps past INT32_MAX with llvm-readobj (#81463) (PR #81468)
llvmbot wrote: @DavidSpickett What do you think about merging this PR to the release branch? https://github.com/llvm/llvm-project/pull/81468 ___ 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] [test] Avoid printing timestamps past INT32_MAX with llvm-readobj (#81463) (PR #81468)
llvmbot wrote:
@llvm/pr-subscribers-lld
Author: None (llvmbot)
Changes
Backport 0bf4ff29816c0eead99ba576a2df2e3c4d214b1f
Requested by: @mstorsjo
---
Full diff: https://github.com/llvm/llvm-project/pull/81468.diff
1 Files Affected:
- (modified) lld/test/COFF/timestamp.test (+19-10)
``diff
diff --git a/lld/test/COFF/timestamp.test b/lld/test/COFF/timestamp.test
index c0658d6109811b..cc73af13c38ca6 100644
--- a/lld/test/COFF/timestamp.test
+++ b/lld/test/COFF/timestamp.test
@@ -4,19 +4,28 @@ RUN: lld-link %t.obj /debug /Brepro /entry:main /nodefaultlib
/out:%t.1.exe
RUN: lld-link %t.obj /debug /Brepro /entry:main /nodefaultlib /out:%t.2.exe
RUN: lld-link %t.obj /debug /timestamp:0 /entry:main /nodefaultlib
/out:%t.3.exe
RUN: env SOURCE_DATE_EPOCH=0 lld-link %t.obj /debug /entry:main /nodefaultlib
/out:%t.4.exe
-RUN: lld-link %t.obj /debug /timestamp:4294967295 /entry:main /nodefaultlib
/out:%t.5.exe
-RUN: env SOURCE_DATE_EPOCH=4294967295 lld-link %t.obj /debug /entry:main
/nodefaultlib /out:%t.6.exe
+# Test timestamps corresponding to INT32_TMAX
+RUN: lld-link %t.obj /debug /timestamp:2147483647 /entry:main /nodefaultlib
/out:%t.5.exe
+RUN: env SOURCE_DATE_EPOCH=2147483647 lld-link %t.obj /debug /entry:main
/nodefaultlib /out:%t.6.exe
+# Test that the command line option /timestamp has precedence over
SOURCE_DATE_EPOCH
RUN: env SOURCE_DATE_EPOCH=12345 lld-link %t.obj /debug /timestamp:0
/entry:main /nodefaultlib /out:%t.7.exe
-RUN: env LLD_IN_TEST=1 not lld-link %t.obj /debug /timestamp:4294967296
/entry:main /nodefaultlib /out:%t.8.exe 2>&1 | FileCheck %s --check-prefix=ERROR
-RUN: env SOURCE_DATE_EPOCH=4294967296 env LLD_IN_TEST=1 not lld-link %t.obj
/debug /entry:main /nodefaultlib /out:%t.9.exe 2>&1 | FileCheck %s
--check-prefix=ERROR2
+# Test timestamps corresponding to UINT32_TMAX
+RUN: lld-link %t.obj /debug /timestamp:4294967295 /entry:main /nodefaultlib
/out:%t.8.exe
+RUN: env SOURCE_DATE_EPOCH=4294967295 lld-link %t.obj /debug /entry:main
/nodefaultlib /out:%t.9.exe
+# Test that setting UINT32_MAX+1 as timestamp fails.
+RUN: env LLD_IN_TEST=1 not lld-link %t.obj /debug /timestamp:4294967296
/entry:main /nodefaultlib /out:%t.10.exe 2>&1 | FileCheck %s
--check-prefix=ERROR
+RUN: env SOURCE_DATE_EPOCH=4294967296 env LLD_IN_TEST=1 not lld-link %t.obj
/debug /entry:main /nodefaultlib /out:%t.11.exe 2>&1 | FileCheck %s
--check-prefix=ERROR2
RUN: llvm-readobj --file-headers --coff-debug-directory %t.1.exe | FileCheck
%s --check-prefix=HASH
RUN: llvm-readobj --file-headers --coff-debug-directory %t.2.exe | FileCheck
%s --check-prefix=HASH
RUN: llvm-readobj --file-headers --coff-debug-directory %t.3.exe | FileCheck
%s --check-prefix=ZERO
RUN: llvm-readobj --file-headers --coff-debug-directory %t.4.exe | FileCheck
%s --check-prefix=ZERO
-RUN: llvm-readobj --file-headers --coff-debug-directory %t.5.exe | FileCheck
%s --check-prefix=MAX
-RUN: llvm-readobj --file-headers --coff-debug-directory %t.6.exe | FileCheck
%s --check-prefix=MAX
+RUN: llvm-readobj --file-headers --coff-debug-directory %t.5.exe | FileCheck
%s --check-prefix=LARGE
+RUN: llvm-readobj --file-headers --coff-debug-directory %t.6.exe | FileCheck
%s --check-prefix=LARGE
RUN: llvm-readobj --file-headers --coff-debug-directory %t.7.exe | FileCheck
%s --check-prefix=ZERO
+# Not inspecting %t.8.exe and %t.9.exe; llvm-readobj with a 32 bit time_t
fails to print dates
+# past INT32_MAX correctly.
+
HASH: ImageFileHeader {
HASH: TimeDateStamp: [[STAMP:.*]]
HASH: DebugDirectory [
@@ -27,10 +36,10 @@ ZERO: TimeDateStamp: 1970-01-01 00:00:00 (0x0)
ZERO: DebugDirectory [
ZERO: TimeDateStamp: 1970-01-01 00:00:00 (0x0)
-MAX: ImageFileHeader {
-MAX: TimeDateStamp: 2106-02-07 06:28:15 (0x)
-MAX: DebugDirectory [
-MAX: TimeDateStamp: 2106-02-07 06:28:15 (0x)
+LARGE: ImageFileHeader {
+LARGE: TimeDateStamp: 2038-01-19 03:14:07 (0x7FFF)
+LARGE: DebugDirectory [
+LARGE: TimeDateStamp: 2038-01-19 03:14:07 (0x7FFF)
ERROR: error: invalid timestamp: 4294967296. Expected 32-bit integer
ERROR2: error: invalid SOURCE_DATE_EPOCH timestamp: 4294967296. Expected
32-bit integer
``
https://github.com/llvm/llvm-project/pull/81468
___
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] [test] Avoid printing timestamps past INT32_MAX with llvm-readobj (#81463) (PR #81468)
DavidSpickett wrote: I've confirmed locally that this fixes the test as claimed, let's merge it. https://github.com/llvm/llvm-project/pull/81468 ___ 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] [test] Avoid printing timestamps past INT32_MAX with llvm-readobj (#81463) (PR #81468)
https://github.com/DavidSpickett approved this pull request. https://github.com/llvm/llvm-project/pull/81468 ___ 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] [TBAA] Use !tbaa for first accessed field if it is an exact match in offset and size. (PR #81313)
https://github.com/fhahn edited https://github.com/llvm/llvm-project/pull/81313 ___ 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] [TBAA] Use !tbaa for first accessed field if it is an exact match in offset and size. (PR #81313)
fhahn wrote: > lgtm > > Maybe rephrase the commit message to something like: > > ``` > [tbaa] Use !tbaa for first accessed field if it is an exact match in offset > and size. > ``` Updated, thanks! It would be great if you could take another look at https://github.com/llvm/llvm-project/pull/81289 in case you missed my response to your comment, as this PR depends on https://github.com/llvm/llvm-project/pull/81289 https://github.com/llvm/llvm-project/pull/81313 ___ 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] [SROA] Use !tbaa instead of !tbaa.struct if op matches field. (PR #81289)
dobbelaj-snps wrote: > > Hmm. 10 changes + 1 new usage of setAAMetaData, But only 4 relevant changes > > in tests.. That part of SROA seems to lack some testing ? > > Yes, will add the missing coverage, just wanted to make sure this makes sense > in general beforehand I think it makes sense. For this one to go through, I think we need at least a testcase that triggers the new 'PStore->setAAMetaData'... case. (line 4582) https://github.com/llvm/llvm-project/pull/81289 ___ 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++][print] Moves is_terminal to the dylib. (#80464) (PR #81410)
ldionne wrote: Approved. https://github.com/llvm/llvm-project/pull/81410 ___ 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] [NFC][CallPromotionUtils]Extract a helper function versionCallSiteWithCond from versionCallSite (PR #81181)
https://github.com/minglotus-6 ready_for_review https://github.com/llvm/llvm-project/pull/81181 ___ 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] [NFC][CallPromotionUtils]Extract a helper function versionCallSiteWithCond from versionCallSite (PR #81181)
llvmbot wrote:
@llvm/pr-subscribers-llvm-transforms
Author: Mingming Liu (minglotus-6)
Changes
The parent patch is https://github.com/llvm/llvm-project/pull/81051
---
Full diff: https://github.com/llvm/llvm-project/pull/81181.diff
1 Files Affected:
- (modified) llvm/lib/Transforms/Utils/CallPromotionUtils.cpp (+21-15)
``diff
diff --git a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
index 4e84927f1cfc90..d0cf0792eface0 100644
--- a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
+++ b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
@@ -188,10 +188,9 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Predicate and clone the given call site.
///
/// This function creates an if-then-else structure at the location of the call
-/// site. The "if" condition compares the call site's called value to the given
-/// callee. The original call site is moved into the "else" block, and a clone
-/// of the call site is placed in the "then" block. The cloned instruction is
-/// returned.
+/// site. The "if" condition is specified by `Cond`.
+/// The original call site is moved into the "else" block, and a clone of the
+/// call site is placed in the "then" block. The cloned instruction is
returned.
///
/// For example, the call instruction below:
///
@@ -202,7 +201,6 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Is replace by the following:
///
/// orig_bb:
-/// %cond = icmp eq i32 ()* %ptr, @func
/// br i1 %cond, %then_bb, %else_bb
///
/// then_bb:
@@ -232,7 +230,6 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Is replace by the following:
///
/// orig_bb:
-/// %cond = icmp eq i32 ()* %ptr, @func
/// br i1 %cond, %then_bb, %else_bb
///
/// then_bb:
@@ -267,7 +264,6 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Is replaced by the following:
///
/// cond_bb:
-/// %cond = icmp eq i32 ()* %ptr, @func
/// br i1 %cond, %then_bb, %orig_bb
///
/// then_bb:
@@ -280,19 +276,13 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// ; The original call instruction stays in its original block.
/// %t0 = musttail call i32 %ptr()
/// ret %t0
-CallBase &llvm::versionCallSite(CallBase &CB, Value *Callee,
-MDNode *BranchWeights) {
+static CallBase &versionCallSiteWithCond(CallBase &CB, Value *Cond,
+ MDNode *BranchWeights) {
IRBuilder<> Builder(&CB);
CallBase *OrigInst = &CB;
BasicBlock *OrigBlock = OrigInst->getParent();
- // Create the compare. The called value and callee must have the same type to
- // be compared.
- if (CB.getCalledOperand()->getType() != Callee->getType())
-Callee = Builder.CreateBitCast(Callee, CB.getCalledOperand()->getType());
- auto *Cond = Builder.CreateICmpEQ(CB.getCalledOperand(), Callee);
-
if (OrigInst->isMustTailCall()) {
// Create an if-then structure. The original instruction stays in its
block,
// and a clone of the original instruction is placed in the "then" block.
@@ -380,6 +370,22 @@ CallBase &llvm::versionCallSite(CallBase &CB, Value
*Callee,
return *NewInst;
}
+// Predicate and clone the given call site usingc condition `CB.callee ==
+// Callee`. See the comment `versionCallSiteWithCond` for the transformation.
+CallBase &llvm::versionCallSite(CallBase &CB, Value *Callee,
+MDNode *BranchWeights) {
+
+ IRBuilder<> Builder(&CB);
+
+ // Create the compare. The called value and callee must have the same type to
+ // be compared.
+ if (CB.getCalledOperand()->getType() != Callee->getType())
+Callee = Builder.CreateBitCast(Callee, CB.getCalledOperand()->getType());
+ auto *Cond = Builder.CreateICmpEQ(CB.getCalledOperand(), Callee);
+
+ return versionCallSiteWithCond(CB, Cond, BranchWeights);
+}
+
bool llvm::isLegalToPromote(const CallBase &CB, Function *Callee,
const char **FailureReason) {
assert(!CB.getCalledFunction() && "Only indirect call sites can be
promoted");
``
https://github.com/llvm/llvm-project/pull/81181
___
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] [NFC][CallPromotionUtils]Extract a helper function versionCallSiteWithCond from versionCallSite (PR #81181)
https://github.com/minglotus-6 edited https://github.com/llvm/llvm-project/pull/81181 ___ 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] [CallPromotionUtils]Implement conditional indirect call promotion with vtable-based comparison (PR #81378)
https://github.com/minglotus-6 updated
https://github.com/llvm/llvm-project/pull/81378
>From ac5dc1bf77b67cbf0aa5e2c8fb6a7ce0080fb501 Mon Sep 17 00:00:00 2001
From: mingmingl
Date: Sat, 10 Feb 2024 12:03:25 -0800
Subject: [PATCH 1/2] [CallPromotionUtils]Implement conditional indirect call
promotion with vtable-based comparison
---
.../Transforms/Utils/CallPromotionUtils.h | 50 ++-
.../Transforms/Utils/CallPromotionUtils.cpp | 64 -
.../Utils/CallPromotionUtilsTest.cpp | 127 ++
3 files changed, 233 insertions(+), 8 deletions(-)
diff --git a/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
b/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
index fcb384ec361339..5f3a71206876c6 100644
--- a/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
+++ b/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
@@ -14,10 +14,17 @@
#ifndef LLVM_TRANSFORMS_UTILS_CALLPROMOTIONUTILS_H
#define LLVM_TRANSFORMS_UTILS_CALLPROMOTIONUTILS_H
+#include
+
+#include "llvm/ADT/ArrayRef.h"
+
namespace llvm {
+class Constant;
class CallBase;
class CastInst;
class Function;
+class GlobalVariable;
+class Instruction;
class MDNode;
class Value;
@@ -41,7 +48,9 @@ bool isLegalToPromote(const CallBase &CB, Function *Callee,
CallBase &promoteCall(CallBase &CB, Function *Callee,
CastInst **RetBitCast = nullptr);
-/// Promote the given indirect call site to conditionally call \p Callee.
+/// Promote the given indirect call site to conditionally call \p Callee. The
+/// promoted direct call instruction is predicated on `CB.getCalledOperand() ==
+/// Callee`.
///
/// This function creates an if-then-else structure at the location of the call
/// site. The original call site is moved into the "else" block. A clone of the
@@ -51,6 +60,31 @@ CallBase &promoteCall(CallBase &CB, Function *Callee,
CallBase &promoteCallWithIfThenElse(CallBase &CB, Function *Callee,
MDNode *BranchWeights = nullptr);
+/// This is similar to `promoteCallWithIfThenElse` except that the condition to
+/// promote a virtual call is that \p VPtr is the same as any of \p
+/// AddressPoints.
+///
+/// This function is expected to be used on virtual calls (a subset of indirect
+/// calls). \p VPtr is the virtual table address stored in the objects, and
+/// \p AddressPoints contains address points of vtables to be compared with.
+///
+/// It's the responsibility of caller to guarantee the transformation
+/// correctness (by specifying \p VPtr and \p AddressPoints properly).
+///
+/// This function doesn't sink the address-calculation instructions of indirect
+/// callee to the indirect call fallback. The subsequent passes (e.g.
+/// inst-combine) should sink them if possible and handle the sink of debug
+/// intrinsics together.
+CallBase &promoteCallWithVTableCmp(CallBase &CB, Instruction *VPtr,
+ Function *Callee,
+ ArrayRef AddressPoints,
+ MDNode *BranchWeights);
+
+/// Returns a constant representing the vtable's address point specified by the
+/// offset. Caller should ensure \p AddressPointOffset is valid.
+Constant *getVTableAddressPointOffset(GlobalVariable *VTable,
+ uint32_t AddressPointOffset);
+
/// Try to promote (devirtualize) a virtual call on an Alloca. Return true on
/// success.
///
@@ -74,13 +108,17 @@ CallBase &promoteCallWithIfThenElse(CallBase &CB, Function
*Callee,
///
bool tryPromoteCall(CallBase &CB);
+/// Predicate and clone the given call site using the given condition.
+CallBase &versionCallSiteWithCond(CallBase &CB, Value *Cond,
+ MDNode *BranchWeights);
+
/// Predicate and clone the given call site.
///
-/// This function creates an if-then-else structure at the location of the call
-/// site. The "if" condition compares the call site's called value to the given
-/// callee. The original call site is moved into the "else" block, and a clone
-/// of the call site is placed in the "then" block. The cloned instruction is
-/// returned.
+/// This function creates an if-then-else structure at the location of the
+/// call site. The "if" condition compares the call site's called value to
+/// the given callee. The original call site is moved into the "else" block,
+/// and a clone of the call site is placed in the "then" block. The cloned
+/// instruction is returned.
CallBase &versionCallSite(CallBase &CB, Value *Callee, MDNode *BranchWeights);
} // end namespace llvm
diff --git a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
index d0cf0792eface0..ea855b9a4d8416 100644
--- a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
+++ b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
@@ -12,9 +12,11 @@
//===
[llvm-branch-commits] [llvm] [CallPromotionUtils]Implement conditional indirect call promotion with vtable-based comparison (PR #81378)
https://github.com/minglotus-6 updated
https://github.com/llvm/llvm-project/pull/81378
>From ac5dc1bf77b67cbf0aa5e2c8fb6a7ce0080fb501 Mon Sep 17 00:00:00 2001
From: mingmingl
Date: Sat, 10 Feb 2024 12:03:25 -0800
Subject: [PATCH 1/3] [CallPromotionUtils]Implement conditional indirect call
promotion with vtable-based comparison
---
.../Transforms/Utils/CallPromotionUtils.h | 50 ++-
.../Transforms/Utils/CallPromotionUtils.cpp | 64 -
.../Utils/CallPromotionUtilsTest.cpp | 127 ++
3 files changed, 233 insertions(+), 8 deletions(-)
diff --git a/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
b/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
index fcb384ec361339..5f3a71206876c6 100644
--- a/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
+++ b/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
@@ -14,10 +14,17 @@
#ifndef LLVM_TRANSFORMS_UTILS_CALLPROMOTIONUTILS_H
#define LLVM_TRANSFORMS_UTILS_CALLPROMOTIONUTILS_H
+#include
+
+#include "llvm/ADT/ArrayRef.h"
+
namespace llvm {
+class Constant;
class CallBase;
class CastInst;
class Function;
+class GlobalVariable;
+class Instruction;
class MDNode;
class Value;
@@ -41,7 +48,9 @@ bool isLegalToPromote(const CallBase &CB, Function *Callee,
CallBase &promoteCall(CallBase &CB, Function *Callee,
CastInst **RetBitCast = nullptr);
-/// Promote the given indirect call site to conditionally call \p Callee.
+/// Promote the given indirect call site to conditionally call \p Callee. The
+/// promoted direct call instruction is predicated on `CB.getCalledOperand() ==
+/// Callee`.
///
/// This function creates an if-then-else structure at the location of the call
/// site. The original call site is moved into the "else" block. A clone of the
@@ -51,6 +60,31 @@ CallBase &promoteCall(CallBase &CB, Function *Callee,
CallBase &promoteCallWithIfThenElse(CallBase &CB, Function *Callee,
MDNode *BranchWeights = nullptr);
+/// This is similar to `promoteCallWithIfThenElse` except that the condition to
+/// promote a virtual call is that \p VPtr is the same as any of \p
+/// AddressPoints.
+///
+/// This function is expected to be used on virtual calls (a subset of indirect
+/// calls). \p VPtr is the virtual table address stored in the objects, and
+/// \p AddressPoints contains address points of vtables to be compared with.
+///
+/// It's the responsibility of caller to guarantee the transformation
+/// correctness (by specifying \p VPtr and \p AddressPoints properly).
+///
+/// This function doesn't sink the address-calculation instructions of indirect
+/// callee to the indirect call fallback. The subsequent passes (e.g.
+/// inst-combine) should sink them if possible and handle the sink of debug
+/// intrinsics together.
+CallBase &promoteCallWithVTableCmp(CallBase &CB, Instruction *VPtr,
+ Function *Callee,
+ ArrayRef AddressPoints,
+ MDNode *BranchWeights);
+
+/// Returns a constant representing the vtable's address point specified by the
+/// offset. Caller should ensure \p AddressPointOffset is valid.
+Constant *getVTableAddressPointOffset(GlobalVariable *VTable,
+ uint32_t AddressPointOffset);
+
/// Try to promote (devirtualize) a virtual call on an Alloca. Return true on
/// success.
///
@@ -74,13 +108,17 @@ CallBase &promoteCallWithIfThenElse(CallBase &CB, Function
*Callee,
///
bool tryPromoteCall(CallBase &CB);
+/// Predicate and clone the given call site using the given condition.
+CallBase &versionCallSiteWithCond(CallBase &CB, Value *Cond,
+ MDNode *BranchWeights);
+
/// Predicate and clone the given call site.
///
-/// This function creates an if-then-else structure at the location of the call
-/// site. The "if" condition compares the call site's called value to the given
-/// callee. The original call site is moved into the "else" block, and a clone
-/// of the call site is placed in the "then" block. The cloned instruction is
-/// returned.
+/// This function creates an if-then-else structure at the location of the
+/// call site. The "if" condition compares the call site's called value to
+/// the given callee. The original call site is moved into the "else" block,
+/// and a clone of the call site is placed in the "then" block. The cloned
+/// instruction is returned.
CallBase &versionCallSite(CallBase &CB, Value *Callee, MDNode *BranchWeights);
} // end namespace llvm
diff --git a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
index d0cf0792eface0..ea855b9a4d8416 100644
--- a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
+++ b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
@@ -12,9 +12,11 @@
//===
[llvm-branch-commits] [llvm] [CallPromotionUtils]Implement conditional indirect call promotion with vtable-based comparison (PR #81378)
https://github.com/minglotus-6 edited https://github.com/llvm/llvm-project/pull/81378 ___ 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] [RISCV] Support select optimization (PR #80124)
@@ -445,6 +450,9 @@ void RISCVPassConfig::addIRPasses() {
if (EnableLoopDataPrefetch)
addPass(createLoopDataPrefetchPass());
+if (EnableSelectOpt && getOptLevel() == CodeGenOptLevel::Aggressive)
topperc wrote:
Shoudl this be after addIRPasses like it is for AArch64?
https://github.com/llvm/llvm-project/pull/80124
___
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] [RISCV] Support select optimization (PR #80124)
https://github.com/topperc edited https://github.com/llvm/llvm-project/pull/80124 ___ 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] [CallPromotionUtils]Implement conditional indirect call promotion with vtable-based comparison (PR #81378)
https://github.com/minglotus-6 ready_for_review https://github.com/llvm/llvm-project/pull/81378 ___ 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] [CallPromotionUtils]Implement conditional indirect call promotion with vtable-based comparison (PR #81378)
llvmbot wrote:
@llvm/pr-subscribers-llvm-transforms
Author: Mingming Liu (minglotus-6)
Changes
* Given the code sequence
```
bb:
%vtable = load ptr, ptr %d, !prof !8
%vfn = getelementptr inbounds ptr, ptr %vtable, i64 1
%1 = load ptr, ptr %vfn
%call = tail call i32 %1(ptr %d), !prof !9
```
The transformation looks like
```
bb:
%vtable = load ptr, ptr %d, align 8
%vfn = getelementptr inbounds i8, ptr %vtable, i64 8 <-- Inst 1
%func-addr = load ptr, ptr %vfn, align 8 <-- Inst 2
# compare loaded pointers with address point of vtables
%1 = icmp eq ptr %vtable, getelementptr inbounds ({ [4 x ptr] }, ptr @_ZTV, i64 0, i32 0, i64 2)
br i1 %1, label %if.true.direct_targ, label %if.false.orig_indirect, !prof
!18
if.true.direct_targ: ; preds = %entry
%2 = tail call i32 @(ptr nonnull %d)
br label %if.end.icp
if.false.orig_indirect: ; preds = %entry
%call = tail call i32 %func-addr(ptr nonnull %d)
br label %if.end.icp
if.end.icp: ; preds =
%if.false.orig_indirect, %if.true.direct_targ
%4 = phi i32 [ %call, %if.false.orig_indirect ], [ %2, %if.true.direct_targ
]
```
It's intentional that `Inst 1` and `Inst2` remains in `bb` (not in
`if.false.orig_indirect`). Subsequent pass (notably `instcombine` would sink
them (and handles debug intrinsics properly) if possible.
* The parent patch is https://github.com/llvm/llvm-project/pull/81181
---
Full diff: https://github.com/llvm/llvm-project/pull/81378.diff
3 Files Affected:
- (modified) llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h (+43-6)
- (modified) llvm/lib/Transforms/Utils/CallPromotionUtils.cpp (+62-2)
- (modified) llvm/unittests/Transforms/Utils/CallPromotionUtilsTest.cpp (+119)
``diff
diff --git a/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
b/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
index fcb384ec361339..32b252d132c04c 100644
--- a/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
+++ b/llvm/include/llvm/Transforms/Utils/CallPromotionUtils.h
@@ -14,10 +14,16 @@
#ifndef LLVM_TRANSFORMS_UTILS_CALLPROMOTIONUTILS_H
#define LLVM_TRANSFORMS_UTILS_CALLPROMOTIONUTILS_H
+#include
+
namespace llvm {
+template class ArrayRef;
+class Constant;
class CallBase;
class CastInst;
class Function;
+class GlobalVariable;
+class Instruction;
class MDNode;
class Value;
@@ -41,7 +47,9 @@ bool isLegalToPromote(const CallBase &CB, Function *Callee,
CallBase &promoteCall(CallBase &CB, Function *Callee,
CastInst **RetBitCast = nullptr);
-/// Promote the given indirect call site to conditionally call \p Callee.
+/// Promote the given indirect call site to conditionally call \p Callee. The
+/// promoted direct call instruction is predicated on `CB.getCalledOperand() ==
+/// Callee`.
///
/// This function creates an if-then-else structure at the location of the call
/// site. The original call site is moved into the "else" block. A clone of the
@@ -51,6 +59,31 @@ CallBase &promoteCall(CallBase &CB, Function *Callee,
CallBase &promoteCallWithIfThenElse(CallBase &CB, Function *Callee,
MDNode *BranchWeights = nullptr);
+/// This is similar to `promoteCallWithIfThenElse` except that the condition to
+/// promote a virtual call is that \p VPtr is the same as any of \p
+/// AddressPoints.
+///
+/// This function is expected to be used on virtual calls (a subset of indirect
+/// calls). \p VPtr is the virtual table address stored in the objects, and
+/// \p AddressPoints contains address points of vtables to be compared with.
+///
+/// It's the responsibility of caller to guarantee the transformation
+/// correctness (by specifying \p VPtr and \p AddressPoints properly).
+///
+/// This function doesn't sink the address-calculation instructions of indirect
+/// callee to the indirect call fallback. The subsequent passes (e.g.
+/// inst-combine) should sink them if possible and handle the sink of debug
+/// intrinsics together.
+CallBase &promoteCallWithVTableCmp(CallBase &CB, Instruction *VPtr,
+ Function *Callee,
+ ArrayRef AddressPoints,
+ MDNode *BranchWeights);
+
+/// Returns a constant representing the vtable's address point specified by the
+/// offset. Caller should ensure \p AddressPointOffset is valid.
+Constant *getVTableAddressPointOffset(GlobalVariable *VTable,
+ uint32_t AddressPointOffset);
+
/// Try to promote (devirtualize) a virtual call on an Alloca. Return true on
/// success.
///
@@ -74,13 +107,17 @@ CallBase &promoteCallWithIfThenElse(CallBase &CB, Function
*Callee,
///
bool tryPromoteCall(CallBase &CB);
+/// Predicate and clone the given call site using the given condition.
[llvm-branch-commits] [OpenMP][MLIR] Add new partial_map and members_index arguments to omp.map_info operations to help support record type mapping (PR #81509)
https://github.com/agozillon created https://github.com/llvm/llvm-project/pull/81509 This PR adds two new fields to omp.map_info, one BoolAttr and one I64ArrayAttr. The BoolAttr is named partial_map, and is a flag that indicates if the record type captured by the map_info operation is a partial map, or if it is mapped in its entirety, this currently helps the later lowering determine the type of map entries that need to be generated. The I64ArrayAttr named members_index is intended to track the placement of each member map_info operations (and by extension mapped member variable) placement in the parent record type. This may need to be extended to an N-D array for nested member mapping. ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [OpenMP][MLIR] Extend explicit derived type member mapping support for OpenMP dialects lowering to LLVM-IR (PR #81510)
https://github.com/agozillon created https://github.com/llvm/llvm-project/pull/81510 This patch seeks to refactor slightly and extend the current record type map support that was put in place for Fortran's descriptor types to handle explicit member mapping for record types at a single level of depth (the case of explicit mapping of nested record types is currently unsupported). This patch seeks to support this by extending the OpenMPToLLVMIRTranslation phase to more generally support record types, building on the prior groundwork in the Fortran allocatables/pointers patch. It now supports different kinds of record type mapping, in this case full record type mapping and then explicit member mapping in which there is a special case for certain types when mapped individually to not require any parent map link in the kernel argument structure. To facilitate this required: * The movement of the setting of the map flag type "ptr_and_obj" to respective frontends, now supporting it as a possible flag that can be read and printed in mlir form. Some minor changes to declare target map type setting was neccesary for this. * The addition of a member index array operand, which tracks the position of the member in the parent, required for caclulating the appropriate size to offload to the target, alongside the parents offload pointer (always the first member currently being mapped). * A partial mapping attribute operand, to indicate if the entire record type is being mapped or just member components, aiding the ability to lower record types in the different manners that are possible. * Refactoring bounds calculation for record types and general arrays to one location (as well as load/store generation prior to assigning to the kernel argument structure), as a side affect enter/exit/update/data mapping should now be more correct and fully support bounds mapping, previously this would have only worked for target. ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [Flang][OpenMP][MLIR] Extend derived (record) type map support in Flang OpenMP by adding some initial support for explicit member mapping (PR #81511)
https://github.com/agozillon created https://github.com/llvm/llvm-project/pull/81511 This patch is one in a series of four patches that seeks to refactor slightly and extend the current record type map support that was put in place for Fortran's descriptor types to handle explicit member mapping for record types at a single level of depth. For example, the below case where two members of a Fortran derived type are mapped explicitly: type :: scalar_and_array real(4) :: real integer(4) :: array(10) integer(4) :: int end type scalar_and_array type(scalar_and_array) :: scalar_arr !$omp target map(tofrom: scalar_arr%int, scalar_arr%real) Current cases of derived type mapping left for future work are: > explicit member mapping of nested members (e.g. two layers of record types where we explicitly map a member from the internal record type) > Fortran's automagical mapping of all elements and nested elements of a derived type > explicit member mapping of a derived type and then constituient members (redundant in Fortran due to former case but still legal as far as I am aware) > explicit member mapping of a record type (may be handled reasonably, just not fully tested in this iteration) > explicit member mapping for Fortran allocatable types (a variation of nested record types) This patch seeks to support this by extending the Flang-new OpenMP lowering to support generation of this newly required information, creating the neccessary parent <-to-> member map_info links, calculating the member indices and setting if it's a partial map. The OMPDescriptorMapInfoGen pass has also been generalized into a map finalization phase, now named OMPMapInfoFinalization. This pass was extended to support the insertion of member maps into the BlockArg and MapOperands of relevant map carrying operations. Similar to the method in which descriptor types are expanded and constituient members inserted. ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [OpenMP][MLIR] Extend explicit derived type member mapping support for OpenMP dialects lowering to LLVM-IR (PR #81510)
llvmbot wrote:
@llvm/pr-subscribers-mlir-openmp
@llvm/pr-subscribers-mlir
Author: None (agozillon)
Changes
This patch seeks to refactor slightly and extend the current record type map
support that was put in place for Fortran's descriptor types to handle explicit
member mapping for record types at a single level of depth (the case of explicit
mapping of nested record types is currently unsupported).
This patch seeks to support this by extending the OpenMPToLLVMIRTranslation
phase
to more generally support record types, building on the prior groundwork in the
Fortran allocatables/pointers patch. It now supports different kinds of record
type
mapping, in this case full record type mapping and then explicit member mapping
in which there is a special case for certain types when mapped individually to
not
require any parent map link in the kernel argument structure. To facilitate this
required:
* The movement of the setting of the map flag type "ptr_and_obj" to
respective
frontends, now supporting it as a possible flag that can be read and
printed
in mlir form. Some minor changes to declare target map type setting was
neccesary for this.
* The addition of a member index array operand, which tracks the position
of the member in the parent, required for caclulating the appropriate
size
to offload to the target, alongside the parents offload pointer (always
the
first member currently being mapped).
* A partial mapping attribute operand, to indicate if the entire record type
is
being mapped or just member components, aiding the ability to lower
record types in the different manners that are possible.
* Refactoring bounds calculation for record types and general arrays to one
location (as well as load/store generation prior to assigning to the
kernel
argument structure), as a side affect enter/exit/update/data mapping
should now be more correct and fully support bounds mapping, previously
this would have only worked for target.
---
Patch is 49.34 KiB, truncated to 20.00 KiB below, full version:
https://github.com/llvm/llvm-project/pull/81510.diff
4 Files Affected:
- (modified)
mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp (+300-204)
- (modified)
mlir/test/Target/LLVMIR/omptarget-fortran-allocatable-types-host.mlir (+14-16)
- (modified) mlir/test/Target/LLVMIR/omptarget-llvm.mlir (+27-22)
- (added) mlir/test/Target/LLVMIR/omptarget-record-type-mapping-host.mlir (+63)
``diff
diff --git
a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index 79956f82ed141a..1ba8099135c35f 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -1750,9 +1750,9 @@ void collectMapDataFromMapOperands(MapInfoData &mapData,
mapData.BaseType.push_back(
moduleTranslation.convertType(mapOp.getVarType()));
- mapData.Sizes.push_back(getSizeInBytes(
- dl, mapOp.getVarType(), mapOp, mapData.BasePointers.back(),
- mapData.BaseType.back(), builder, moduleTranslation));
+ mapData.Sizes.push_back(
+ getSizeInBytes(dl, mapOp.getVarType(), mapOp,
mapData.Pointers.back(),
+ mapData.BaseType.back(), builder, moduleTranslation));
mapData.MapClause.push_back(mapOp.getOperation());
mapData.Types.push_back(
llvm::omp::OpenMPOffloadMappingFlags(mapOp.getMapType().value()));
@@ -1783,6 +1783,98 @@ void collectMapDataFromMapOperands(MapInfoData &mapData,
}
}
+static int getMapDataMemberIdx(MapInfoData &mapData,
+ mlir::omp::MapInfoOp memberOp) {
+ int memberDataIdx = -1;
+ for (size_t i = 0; i < mapData.MapClause.size(); ++i) {
+if (mapData.MapClause[i] == memberOp)
+ memberDataIdx = i;
+ }
+ return memberDataIdx;
+}
+
+static mlir::omp::MapInfoOp
+getFirstOrLastMappedMemberPtr(mlir::omp::MapInfoOp mapInfo, bool first) {
+ // Only 1 member has been mapped, we can return it.
+ if (mapInfo.getMembersIndex()->size() == 1)
+if (auto mapOp = mlir::dyn_cast(
+mapInfo.getMembers()[0].getDefiningOp()))
+ return mapOp;
+
+ int64_t curPos =
+ mapInfo.getMembersIndex()->begin()->cast().getInt();
+
+ int64_t idx = 1, curIdx = 0, memberPlacement = 0;
+ for (const auto *iter = std::next(mapInfo.getMembersIndex()->begin());
+ iter != mapInfo.getMembersIndex()->end(); iter++) {
+memberPlacement = iter->cast().getInt();
+if (first) {
+ if (memberPlacement < curPos) {
+curIdx = idx;
+curPos = memberPlacement;
+ }
+} else {
+ if (memberPlacement > curPos) {
+curIdx = idx;
+curPos = memberPlacement;
+ }
+}
+idx++;
+ }
+
+ if (auto mapOp = mlir::dyn_cast(
+ m
[llvm-branch-commits] [OpenMP][MLIR] Add new partial_map and members_index arguments to omp.map_info operations to help support record type mapping (PR #81509)
llvmbot wrote:
@llvm/pr-subscribers-flang-openmp
@llvm/pr-subscribers-mlir
Author: None (agozillon)
Changes
This PR adds two new fields to omp.map_info, one BoolAttr and one I64ArrayAttr.
The BoolAttr is named partial_map, and is a flag that indicates if the record
type captured by
the map_info operation is a partial map, or if it is mapped in its entirety,
this currently helps
the later lowering determine the type of map entries that need to be generated.
The I64ArrayAttr named members_index is intended to track the placement of each
member
map_info operations (and by extension mapped member variable) placement in the
parent
record type. This may need to be extended to an N-D array for nested member
mapping.
---
Full diff: https://github.com/llvm/llvm-project/pull/81509.diff
3 Files Affected:
- (modified) mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td (+12-4)
- (modified) mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp (+7)
- (modified) mlir/test/Dialect/OpenMP/ops.mlir (+22-6)
``diff
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index ca36350548577..cdfe6a8f8edf4 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -1195,10 +1195,12 @@ def MapInfoOp : OpenMP_Op<"map_info",
[AttrSizedOperandSegments]> {
TypeAttr:$var_type,
Optional:$var_ptr_ptr,
Variadic:$members,
+ OptionalAttr:$members_index,
Variadic:$bounds, /* rank-0 to
rank-{n-1} */
OptionalAttr:$map_type,
OptionalAttr:$map_capture_type,
- OptionalAttr:$name);
+ OptionalAttr:$name,
+ DefaultValuedAttr:$partial_map);
let results = (outs OpenMP_PointerLikeType:$omp_ptr);
let description = [{
@@ -1234,10 +1236,14 @@ def MapInfoOp : OpenMP_Op<"map_info",
[AttrSizedOperandSegments]> {
- `var_type`: The type of the variable to copy.
- `var_ptr_ptr`: Used when the variable copied is a member of a class,
structure
or derived type and refers to the originating struct.
-- `members`: Used to indicate mapped child members for the current
MapInfoOp,
+- `members`: Used to indicate mapped child members for the current
MapInfoOp,
represented as other MapInfoOp's, utilised in cases where a parent
structure
type and members of the structure type are being mapped at the same
time.
- For example: map(to: parent, parent->member, parent->member2[:10])
+ For example: map(to: parent, parent->member, parent->member2[:10])
+- `members_index`: Used to indicate the ordering of members within the
containing
+ parent (generally a record type such as a structure, class or derived
type),
+ e.g. struct {int x, float y, double z}, x would be 0, y would be 1, and
z
+ would be 2. This aids the mapping
- `bounds`: Used when copying slices of array's, pointers or pointer
members of
objects (e.g. derived types or classes), indicates the bounds to be
copied
of the variable. When it's an array slice it is in rank order where
rank 0
@@ -1248,6 +1254,8 @@ def MapInfoOp : OpenMP_Op<"map_info",
[AttrSizedOperandSegments]> {
- 'map_capture_type': Capture type for the variable e.g. this, byref,
byvalue, byvla
this can affect how the variable is lowered.
- `name`: Holds the name of variable as specified in user clause
(including bounds).
+- `partial_map`: The record type being mapped will not be mapped in its
entirety,
+ it may be used however, in a mapping to bind it's mapped components
together.
}];
let assemblyFormat = [{
@@ -1256,7 +1264,7 @@ def MapInfoOp : OpenMP_Op<"map_info",
[AttrSizedOperandSegments]> {
`var_ptr_ptr` `(` $var_ptr_ptr `:` type($var_ptr_ptr) `)`
| `map_clauses` `(` custom($map_type) `)`
| `capture` `(` custom($map_capture_type) `)`
- | `members` `(` $members `:` type($members) `)`
+ | `members` `(` $members `:` type($members) `:` $members_index `)`
| `bounds` `(` $bounds `)`
) `->` type($omp_ptr) attr-dict
}];
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index 381f17d080419..7d166b1f78bff 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -709,6 +709,9 @@ static ParseResult parseMapClause(OpAsmParser &parser,
IntegerAttr &mapType) {
if (mapTypeMod == "delete")
mapTypeBits |= llvm::omp::OpenMPOffloadMappingFlags::OMP_MAP_DELETE;
+if (mapTypeMod == "ptr_and_obj")
+ mapTypeBits |= llvm::omp::OpenMPOffloadMappingFlags::OMP_MAP_PTR_AND_OBJ;
+
return success();
};
@@ -745,6 +748,10 @@ static void printMapClause(OpAsmPrinter &p, Oper
[llvm-branch-commits] [Flang][OpenMP][MLIR] Extend derived (record) type map support in Flang OpenMP by adding some initial support for explicit member mapping (PR #81511)
llvmbot wrote: @llvm/pr-subscribers-flang-fir-hlfir Author: None (agozillon) Changes This patch is one in a series of four patches that seeks to refactor slightly and extend the current record type map support that was put in place for Fortran's descriptor types to handle explicit member mapping for record types at a single level of depth. For example, the below case where two members of a Fortran derived type are mapped explicitly: type :: scalar_and_array real(4) :: real integer(4) :: array(10) integer(4) :: int end type scalar_and_array type(scalar_and_array) :: scalar_arr !$omp target map(tofrom: scalar_arr%int, scalar_arr%real) Current cases of derived type mapping left for future work are: > explicit member mapping of nested members (e.g. two layers of record types where we explicitly map a member from the internal record type) > Fortran's automagical mapping of all elements and nested elements of a derived type > explicit member mapping of a derived type and then constituient members (redundant in Fortran due to former case but still legal as far as I am aware) > explicit member mapping of a record type (may be handled reasonably, just not fully tested in this iteration) > explicit member mapping for Fortran allocatable types (a variation of nested record types) This patch seeks to support this by extending the Flang-new OpenMP lowering to support generation of this newly required information, creating the neccessary parent <-to-> member map_info links, calculating the member indices and setting if it's a partial map. The OMPDescriptorMapInfoGen pass has also been generalized into a map finalization phase, now named OMPMapInfoFinalization. This pass was extended to support the insertion of member maps into the BlockArg and MapOperands of relevant map carrying operations. Similar to the method in which descriptor types are expanded and constituient members inserted. --- Patch is 98.14 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/81511.diff 20 Files Affected: - (modified) flang/docs/OpenMP-descriptor-management.md (+2-2) - (modified) flang/include/flang/Optimizer/Transforms/Passes.h (+1-1) - (modified) flang/include/flang/Optimizer/Transforms/Passes.td (+3-3) - (modified) flang/include/flang/Tools/CLOptions.inc (+1-1) - (modified) flang/lib/Lower/OpenMP.cpp (+210-25) - (modified) flang/lib/Optimizer/Transforms/CMakeLists.txt (+1-1) - (removed) flang/lib/Optimizer/Transforms/OMPDescriptorMapInfoGen.cpp (-168) - (added) flang/lib/Optimizer/Transforms/OMPMapInfoFinalization.cpp (+262) - (modified) flang/test/Fir/convert-to-llvm-openmp-and-fir.fir (+31-4) - (modified) flang/test/Integration/OpenMP/map-types-and-sizes.f90 (+153-1) - (modified) flang/test/Lower/OpenMP/FIR/array-bounds.f90 (+2-2) - (modified) flang/test/Lower/OpenMP/FIR/map-component-ref.f90 (+2-2) - (modified) flang/test/Lower/OpenMP/FIR/target.f90 (+2-2) - (modified) flang/test/Lower/OpenMP/allocatable-array-bounds.f90 (+6-6) - (modified) flang/test/Lower/OpenMP/allocatable-map.f90 (+4-4) - (modified) flang/test/Lower/OpenMP/array-bounds.f90 (+4-4) - (added) flang/test/Lower/OpenMP/derived-type-map.f90 (+105) - (modified) flang/test/Lower/OpenMP/map-component-ref.f90 (+1-1) - (modified) flang/test/Lower/OpenMP/target.f90 (+2-2) - (renamed) flang/test/Transforms/omp-map-info-finalization.fir (+29-5) ``diff diff --git a/flang/docs/OpenMP-descriptor-management.md b/flang/docs/OpenMP-descriptor-management.md index 90a20282e05126..af02b3a99cb07d 100644 --- a/flang/docs/OpenMP-descriptor-management.md +++ b/flang/docs/OpenMP-descriptor-management.md @@ -44,7 +44,7 @@ Currently, Flang will lower these descriptor types in the OpenMP lowering (lower to all other map types, generating an omp.MapInfoOp containing relevant information required for lowering the OpenMP dialect to LLVM-IR during the final stages of the MLIR lowering. However, after the lowering to FIR/HLFIR has been performed an OpenMP dialect specific pass for Fortran, -`OMPDescriptorMapInfoGenPass` (Optimizer/OMPDescriptorMapInfoGen.cpp) will expand the +`OMPMapInfoFinalizationPass` (Optimizer/OMPMapInfoFinalization.cpp) will expand the `omp.MapInfoOp`'s containing descriptors (which currently will be a `BoxType` or `BoxAddrOp`) into multiple mappings, with one extra per pointer member in the descriptor that is supported on top of the original descriptor map operation. These pointers members are linked to the parent descriptor by adding them to @@ -52,7 +52,7 @@ the member field of the original descriptor map operation, they are then inserte owning operation's (`omp.TargetOp`, `omp.DataOp` etc.) map operand list and in cases where the owning operation is `IsolatedFromAbove`, it also inserts them as `BlockArgs` to canonicalize the mappings and simplify lowering. -An example transformation b
[llvm-branch-commits] [asan] isInterestingAlloca: remove the isAllocaPromotable condition (PR #77221)
MaskRay wrote: Ping:) https://github.com/llvm/llvm-project/pull/77221 ___ 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] release/18.x: [mlir] Skip invalid test on big endian platform (s390x) (#80246) (PR #81373)
https://github.com/jpienaar approved this pull request. https://github.com/llvm/llvm-project/pull/81373 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [Flang][OpenMP][MLIR] Extend derived (record) type map support in Flang OpenMP by adding some initial support for explicit member mapping (PR #81511)
agozillon wrote: I believe this is the top of the PR stack that should pass... however, it's my first time using SPR so we'll see how it goes. If there's anything odd with the PR (outside of the usual code review) please don't hesitate to mention it so I can address it! I believe I fragmented the stack appropriately into: MLIR Dialect change Flang change MLIR -> LLVM IR changes Set of offload runtime tests However, if you spot a test or file that isn't in the right stack, please do point it out and I can move it, there's a lot of tests across the patch series so there's a chance I've misplaced one or two. https://github.com/llvm/llvm-project/pull/81511 ___ 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] [BOLT][NFC] Add BOLTError and return it from passes (1/n) (PR #81522)
https://github.com/aaupov created https://github.com/llvm/llvm-project/pull/81522 Co-authored-by: Rafael Auler Test Plan: NFC ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/n) (PR #81523)
https://github.com/aaupov created https://github.com/llvm/llvm-project/pull/81523 Test Plan: NFC Co-authored-by: Rafael Auler ___ 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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/aaupov created https://github.com/llvm/llvm-project/pull/81524 Co-authored-by: Rafael Auler Test Plan: NFC ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/n) (PR #81523)
https://github.com/aaupov converted_to_draft https://github.com/llvm/llvm-project/pull/81523 ___ 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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/aaupov converted_to_draft https://github.com/llvm/llvm-project/pull/81524 ___ 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] [CallPromotionUtils]Implement conditional indirect call promotion with vtable-based comparison (PR #81378)
https://github.com/minglotus-6 edited https://github.com/llvm/llvm-project/pull/81378 ___ 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] [CallPromotionUtils]Implement conditional indirect call promotion with vtable-based comparison (PR #81378)
https://github.com/minglotus-6 edited https://github.com/llvm/llvm-project/pull/81378 ___ 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] [BOLT][NFC] Add BOLTError and return it from passes (1/n) (PR #81529)
https://github.com/aaupov created https://github.com/llvm/llvm-project/pull/81529 Co-authored-by: Rafael Auler Test Plan: NFC ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/n) (PR #81530)
https://github.com/aaupov created https://github.com/llvm/llvm-project/pull/81530 Test Plan: NFC Co-authored-by: Rafael Auler ___ 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] [BOLT] Log through JournalingStreams (PR #81531)
https://github.com/aaupov created https://github.com/llvm/llvm-project/pull/81531 Co-authored-by: Rafael Auler Test Plan: NFC ___ 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] [InstrProf] Add vtables with type metadata into symtab to look it up with GUID (PR #81051)
teresajohnson wrote: Can you pull out the part that does the refactoring into getCanonicalName and adds comments etc into a separate nfc patch, so that this one is just about adding vtables? https://github.com/llvm/llvm-project/pull/81051 ___ 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] [BOLT] Log through JournalingStreams (PR #81531)
https://github.com/aaupov closed https://github.com/llvm/llvm-project/pull/81531 ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/n) (PR #81530)
https://github.com/aaupov closed https://github.com/llvm/llvm-project/pull/81530 ___ 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] [BOLT][NFC] Add BOLTError and return it from passes (1/n) (PR #81529)
https://github.com/aaupov closed https://github.com/llvm/llvm-project/pull/81529 ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/n) (PR #81523)
https://github.com/aaupov updated
https://github.com/llvm/llvm-project/pull/81523
>From 17ab8e592a01d4402e88de21c8f66742134b64f3 Mon Sep 17 00:00:00 2001
From: Amir Ayupov
Date: Mon, 12 Feb 2024 12:42:25 -0800
Subject: [PATCH] Fix build
Created using spr 1.3.4
---
bolt/lib/Rewrite/MachORewriteInstance.cpp | 4 ++--
bolt/lib/Rewrite/RewriteInstance.cpp | 11 ++-
2 files changed, 8 insertions(+), 7 deletions(-)
diff --git a/bolt/lib/Rewrite/MachORewriteInstance.cpp
b/bolt/lib/Rewrite/MachORewriteInstance.cpp
index 8be8257f15c1ce..c2a6c1f88ab070 100644
--- a/bolt/lib/Rewrite/MachORewriteInstance.cpp
+++ b/bolt/lib/Rewrite/MachORewriteInstance.cpp
@@ -337,7 +337,7 @@ void MachORewriteInstance::disassembleFunctions() {
BinaryFunction &Function = BFI.second;
if (!Function.isSimple())
continue;
-Function.disassemble();
+cantFail(Function.disassemble());
if (opts::PrintDisasm)
Function.print(outs(), "after disassembly");
}
@@ -387,7 +387,7 @@ void MachORewriteInstance::runOptimizationPasses() {
Manager.registerPass(
std::make_unique(opts::PrintFinalized));
- Manager.runPasses();
+ cantFail(Manager.runPasses());
}
void MachORewriteInstance::mapInstrumentationSection(
diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp
b/bolt/lib/Rewrite/RewriteInstance.cpp
index cc994c213482d4..06aed246de0cf2 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -1283,7 +1283,7 @@ void RewriteInstance::discoverFileObjects() {
/*UseMaxSize*/ true);
if (BF) {
assert(Rel.isRelative() && "Expected relative relocation for
island");
- BF->markIslandDynamicRelocationAtAddress(RelAddress);
+ cantFail(BF->markIslandDynamicRelocationAtAddress(RelAddress));
}
}
}
@@ -2859,8 +2859,9 @@ void RewriteInstance::selectFunctionsToProcess() {
StringSet<> ReorderFunctionsUserSet;
StringSet<> ReorderFunctionsLTOCommonSet;
if (opts::ReorderFunctions == ReorderFunctions::RT_USER) {
-for (const std::string &Function :
- ReorderFunctions::readFunctionOrderFile()) {
+std::vector FunctionNames;
+cantFail(ReorderFunctions::readFunctionOrderFile(FunctionNames));
+for (const std::string &Function : FunctionNames) {
ReorderFunctionsUserSet.insert(Function);
if (std::optional LTOCommonName = getLTOCommonName(Function))
ReorderFunctionsLTOCommonSet.insert(*LTOCommonName);
@@ -3207,7 +3208,7 @@ void RewriteInstance::disassembleFunctions() {
check_error(LSDASection.getError(), "failed to get LSDA section");
ArrayRef LSDAData = ArrayRef(
LSDASection->getData(), LSDASection->getContents().size());
- Function.parseLSDA(LSDAData, LSDASection->getAddress());
+ cantFail(Function.parseLSDA(LSDAData, LSDASection->getAddress()));
}
}
}
@@ -3298,7 +3299,7 @@ void RewriteInstance::postProcessFunctions() {
void RewriteInstance::runOptimizationPasses() {
NamedRegionTimer T("runOptimizationPasses", "run optimization passes",
TimerGroupName, TimerGroupDesc, opts::TimeRewrite);
- BinaryFunctionPassManager::runAllPasses(*BC);
+ cantFail(BinaryFunctionPassManager::runAllPasses(*BC));
}
void RewriteInstance::preregisterSections() {
___
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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/aaupov updated https://github.com/llvm/llvm-project/pull/81524 ___ 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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/aaupov updated https://github.com/llvm/llvm-project/pull/81524 ___ 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] [NFC][CallPromotionUtils]Extract a helper function versionCallSiteWithCond from versionCallSite (PR #81181)
https://github.com/teresajohnson approved this pull request. lgtm with a suggested change to the comments https://github.com/llvm/llvm-project/pull/81181 ___ 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] [NFC][CallPromotionUtils]Extract a helper function versionCallSiteWithCond from versionCallSite (PR #81181)
@@ -267,7 +264,6 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Is replaced by the following:
///
/// cond_bb:
-/// %cond = icmp eq i32 ()* %ptr, @func
teresajohnson wrote:
I'm not sure these conditions should be completely removed. Perhaps just show
something like `%cond = Cond`.
https://github.com/llvm/llvm-project/pull/81181
___
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] [NFC][CallPromotionUtils]Extract a helper function versionCallSiteWithCond from versionCallSite (PR #81181)
https://github.com/teresajohnson edited https://github.com/llvm/llvm-project/pull/81181 ___ 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] [BOLT][NFC] Add BOLTError and return it from passes (1/2) (PR #81522)
https://github.com/aaupov edited https://github.com/llvm/llvm-project/pull/81522 ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/2) (PR #81523)
https://github.com/aaupov edited https://github.com/llvm/llvm-project/pull/81523 ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/2) (PR #81523)
https://github.com/aaupov ready_for_review https://github.com/llvm/llvm-project/pull/81523 ___ 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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/aaupov ready_for_review https://github.com/llvm/llvm-project/pull/81524 ___ 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] [InstrProf] Add vtables with type metadata into symtab to look it up with GUID (PR #81051)
minglotus-6 wrote: > Can you pull out the part that does the refactoring into getCanonicalName and > adds comments etc into a separate nfc patch, so that this one is just about > adding vtables? Sure! I'm thinking about doing the nfc patch into LLVM main branch given `getCanonicalName` simplifies `addFuncWithName` a little bit. But it should be a quick [click](https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/proposing-changes-to-your-work-with-pull-requests/changing-the-base-branch-of-a-pull-request) to switch the base branch to merge a PR into at any time. https://github.com/llvm/llvm-project/pull/81051 ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/2) (PR #81523)
https://github.com/aaupov updated
https://github.com/llvm/llvm-project/pull/81523
>From 17ab8e592a01d4402e88de21c8f66742134b64f3 Mon Sep 17 00:00:00 2001
From: Amir Ayupov
Date: Mon, 12 Feb 2024 12:42:25 -0800
Subject: [PATCH 1/2] Fix build
Created using spr 1.3.4
---
bolt/lib/Rewrite/MachORewriteInstance.cpp | 4 ++--
bolt/lib/Rewrite/RewriteInstance.cpp | 11 ++-
2 files changed, 8 insertions(+), 7 deletions(-)
diff --git a/bolt/lib/Rewrite/MachORewriteInstance.cpp
b/bolt/lib/Rewrite/MachORewriteInstance.cpp
index 8be8257f15c1ce..c2a6c1f88ab070 100644
--- a/bolt/lib/Rewrite/MachORewriteInstance.cpp
+++ b/bolt/lib/Rewrite/MachORewriteInstance.cpp
@@ -337,7 +337,7 @@ void MachORewriteInstance::disassembleFunctions() {
BinaryFunction &Function = BFI.second;
if (!Function.isSimple())
continue;
-Function.disassemble();
+cantFail(Function.disassemble());
if (opts::PrintDisasm)
Function.print(outs(), "after disassembly");
}
@@ -387,7 +387,7 @@ void MachORewriteInstance::runOptimizationPasses() {
Manager.registerPass(
std::make_unique(opts::PrintFinalized));
- Manager.runPasses();
+ cantFail(Manager.runPasses());
}
void MachORewriteInstance::mapInstrumentationSection(
diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp
b/bolt/lib/Rewrite/RewriteInstance.cpp
index cc994c213482d4..06aed246de0cf2 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -1283,7 +1283,7 @@ void RewriteInstance::discoverFileObjects() {
/*UseMaxSize*/ true);
if (BF) {
assert(Rel.isRelative() && "Expected relative relocation for
island");
- BF->markIslandDynamicRelocationAtAddress(RelAddress);
+ cantFail(BF->markIslandDynamicRelocationAtAddress(RelAddress));
}
}
}
@@ -2859,8 +2859,9 @@ void RewriteInstance::selectFunctionsToProcess() {
StringSet<> ReorderFunctionsUserSet;
StringSet<> ReorderFunctionsLTOCommonSet;
if (opts::ReorderFunctions == ReorderFunctions::RT_USER) {
-for (const std::string &Function :
- ReorderFunctions::readFunctionOrderFile()) {
+std::vector FunctionNames;
+cantFail(ReorderFunctions::readFunctionOrderFile(FunctionNames));
+for (const std::string &Function : FunctionNames) {
ReorderFunctionsUserSet.insert(Function);
if (std::optional LTOCommonName = getLTOCommonName(Function))
ReorderFunctionsLTOCommonSet.insert(*LTOCommonName);
@@ -3207,7 +3208,7 @@ void RewriteInstance::disassembleFunctions() {
check_error(LSDASection.getError(), "failed to get LSDA section");
ArrayRef LSDAData = ArrayRef(
LSDASection->getData(), LSDASection->getContents().size());
- Function.parseLSDA(LSDAData, LSDASection->getAddress());
+ cantFail(Function.parseLSDA(LSDAData, LSDASection->getAddress()));
}
}
}
@@ -3298,7 +3299,7 @@ void RewriteInstance::postProcessFunctions() {
void RewriteInstance::runOptimizationPasses() {
NamedRegionTimer T("runOptimizationPasses", "run optimization passes",
TimerGroupName, TimerGroupDesc, opts::TimeRewrite);
- BinaryFunctionPassManager::runAllPasses(*BC);
+ cantFail(BinaryFunctionPassManager::runAllPasses(*BC));
}
void RewriteInstance::preregisterSections() {
>From cb5a6939b09a88e1f1442d8ea761a04b2fcad332 Mon Sep 17 00:00:00 2001
From: Amir Ayupov
Date: Mon, 12 Feb 2024 13:52:48 -0800
Subject: [PATCH 2/2] Handle BOLTErrors in runOptimizationPasses
Created using spr 1.3.4
---
bolt/lib/Rewrite/RewriteInstance.cpp | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp
b/bolt/lib/Rewrite/RewriteInstance.cpp
index 06aed246de0cf2..829568cefff01f 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -3299,7 +3299,11 @@ void RewriteInstance::postProcessFunctions() {
void RewriteInstance::runOptimizationPasses() {
NamedRegionTimer T("runOptimizationPasses", "run optimization passes",
TimerGroupName, TimerGroupDesc, opts::TimeRewrite);
- cantFail(BinaryFunctionPassManager::runAllPasses(*BC));
+ handleAllErrors(BinaryFunctionPassManager::runAllPasses(*BC),
+ [](const BOLTError &E) {
+E.log(errs());
+exit(1);
+ });
}
void RewriteInstance::preregisterSections() {
___
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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/aaupov updated https://github.com/llvm/llvm-project/pull/81524 ___ 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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/aaupov updated https://github.com/llvm/llvm-project/pull/81524 ___ 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] [BOLT][NFC] Add BOLTError and return it from passes (1/2) (PR #81522)
https://github.com/rafaelauler approved this pull request. Thanks, as commit message we can put something like: As part of the effort to refactor old error handling code that would directly call exit(1), in this patch we add a new class BOLTError and auxiliary functions createFatalBOLTError() and createNonFatalBOLTError() that allow BOLT code to bubble up the problem to the caller by using the Error class as a return type (or Expected). Also change passes to use these. https://github.com/llvm/llvm-project/pull/81522 ___ 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] [BOLT][NFC] Add BOLTError and return it from passes (1/2) (PR #81522)
https://github.com/rafaelauler edited https://github.com/llvm/llvm-project/pull/81522 ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/2) (PR #81523)
https://github.com/rafaelauler approved this pull request. Thanks, as commit message: As part of the effort to refactor old error handling code that would directly call exit(1), in this patch continue the migration on libCore, libRewrite and libPasses to use the new BOLTError class whenever a failure occurs. https://github.com/llvm/llvm-project/pull/81523 ___ 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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/rafaelauler approved this pull request. Thanks, here's what I think is a suitable commit message for this one: Make core BOLT functionality more friendly to being used as a library instead of in our standalone driver llvm-bolt. To accomplish this, we augment BinaryContext with journaling streams that are to be used by most BOLT code whenever something needs to be logged to the screen. Users of the library can decide if logs should be printed to a file, no file or to the screen, as before. To illustrate this, this patch adds a new option --log-file that allows the user to redirect BOLT logging to a file on disk or completely hide it by using --log-file=/dev/null. Future BOLT code should now use BinaryContext::outs() for printing important messages instead of llvm::outs(). A new test log.test enforces this by verifying that no strings are print to screen once the --log-file option is used. In previous patches we also added a new BOLTError class to report common and fatal errors, so code shouldn't call exit(1) now. To easily handle problems as before (by quitting with exit(1)), callers can now use BinaryContext::logBOLTErrorsAndQuitOnFatal(Error) whenever code needs to deal with BOLT errors. To test this, we have fatal.s that checks we are correctly quitting and printing a fatal error to the screen. Because this is a significant change by itself, not all code was yet ported. Code from Profiler libs (DataAggregator and friends) still print errors directly to screen. https://github.com/llvm/llvm-project/pull/81524 ___ 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] [test] Avoid printing timestamps past INT32_MAX with llvm-readobj (#81463) (PR #81468)
https://github.com/MaskRay approved this pull request. https://github.com/llvm/llvm-project/pull/81468 ___ 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] [BOLT][NFC] Add BOLTError and return it from passes (1/2) (PR #81522)
https://github.com/aaupov edited https://github.com/llvm/llvm-project/pull/81522 ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/2) (PR #81523)
https://github.com/aaupov updated
https://github.com/llvm/llvm-project/pull/81523
>From 17ab8e592a01d4402e88de21c8f66742134b64f3 Mon Sep 17 00:00:00 2001
From: Amir Ayupov
Date: Mon, 12 Feb 2024 12:42:25 -0800
Subject: [PATCH 1/2] Fix build
Created using spr 1.3.4
---
bolt/lib/Rewrite/MachORewriteInstance.cpp | 4 ++--
bolt/lib/Rewrite/RewriteInstance.cpp | 11 ++-
2 files changed, 8 insertions(+), 7 deletions(-)
diff --git a/bolt/lib/Rewrite/MachORewriteInstance.cpp
b/bolt/lib/Rewrite/MachORewriteInstance.cpp
index 8be8257f15c1ce..c2a6c1f88ab070 100644
--- a/bolt/lib/Rewrite/MachORewriteInstance.cpp
+++ b/bolt/lib/Rewrite/MachORewriteInstance.cpp
@@ -337,7 +337,7 @@ void MachORewriteInstance::disassembleFunctions() {
BinaryFunction &Function = BFI.second;
if (!Function.isSimple())
continue;
-Function.disassemble();
+cantFail(Function.disassemble());
if (opts::PrintDisasm)
Function.print(outs(), "after disassembly");
}
@@ -387,7 +387,7 @@ void MachORewriteInstance::runOptimizationPasses() {
Manager.registerPass(
std::make_unique(opts::PrintFinalized));
- Manager.runPasses();
+ cantFail(Manager.runPasses());
}
void MachORewriteInstance::mapInstrumentationSection(
diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp
b/bolt/lib/Rewrite/RewriteInstance.cpp
index cc994c213482d4..06aed246de0cf2 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -1283,7 +1283,7 @@ void RewriteInstance::discoverFileObjects() {
/*UseMaxSize*/ true);
if (BF) {
assert(Rel.isRelative() && "Expected relative relocation for
island");
- BF->markIslandDynamicRelocationAtAddress(RelAddress);
+ cantFail(BF->markIslandDynamicRelocationAtAddress(RelAddress));
}
}
}
@@ -2859,8 +2859,9 @@ void RewriteInstance::selectFunctionsToProcess() {
StringSet<> ReorderFunctionsUserSet;
StringSet<> ReorderFunctionsLTOCommonSet;
if (opts::ReorderFunctions == ReorderFunctions::RT_USER) {
-for (const std::string &Function :
- ReorderFunctions::readFunctionOrderFile()) {
+std::vector FunctionNames;
+cantFail(ReorderFunctions::readFunctionOrderFile(FunctionNames));
+for (const std::string &Function : FunctionNames) {
ReorderFunctionsUserSet.insert(Function);
if (std::optional LTOCommonName = getLTOCommonName(Function))
ReorderFunctionsLTOCommonSet.insert(*LTOCommonName);
@@ -3207,7 +3208,7 @@ void RewriteInstance::disassembleFunctions() {
check_error(LSDASection.getError(), "failed to get LSDA section");
ArrayRef LSDAData = ArrayRef(
LSDASection->getData(), LSDASection->getContents().size());
- Function.parseLSDA(LSDAData, LSDASection->getAddress());
+ cantFail(Function.parseLSDA(LSDAData, LSDASection->getAddress()));
}
}
}
@@ -3298,7 +3299,7 @@ void RewriteInstance::postProcessFunctions() {
void RewriteInstance::runOptimizationPasses() {
NamedRegionTimer T("runOptimizationPasses", "run optimization passes",
TimerGroupName, TimerGroupDesc, opts::TimeRewrite);
- BinaryFunctionPassManager::runAllPasses(*BC);
+ cantFail(BinaryFunctionPassManager::runAllPasses(*BC));
}
void RewriteInstance::preregisterSections() {
>From cb5a6939b09a88e1f1442d8ea761a04b2fcad332 Mon Sep 17 00:00:00 2001
From: Amir Ayupov
Date: Mon, 12 Feb 2024 13:52:48 -0800
Subject: [PATCH 2/2] Handle BOLTErrors in runOptimizationPasses
Created using spr 1.3.4
---
bolt/lib/Rewrite/RewriteInstance.cpp | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp
b/bolt/lib/Rewrite/RewriteInstance.cpp
index 06aed246de0cf2..829568cefff01f 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -3299,7 +3299,11 @@ void RewriteInstance::postProcessFunctions() {
void RewriteInstance::runOptimizationPasses() {
NamedRegionTimer T("runOptimizationPasses", "run optimization passes",
TimerGroupName, TimerGroupDesc, opts::TimeRewrite);
- cantFail(BinaryFunctionPassManager::runAllPasses(*BC));
+ handleAllErrors(BinaryFunctionPassManager::runAllPasses(*BC),
+ [](const BOLTError &E) {
+E.log(errs());
+exit(1);
+ });
}
void RewriteInstance::preregisterSections() {
___
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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/2) (PR #81523)
https://github.com/aaupov updated
https://github.com/llvm/llvm-project/pull/81523
>From 17ab8e592a01d4402e88de21c8f66742134b64f3 Mon Sep 17 00:00:00 2001
From: Amir Ayupov
Date: Mon, 12 Feb 2024 12:42:25 -0800
Subject: [PATCH 1/2] Fix build
Created using spr 1.3.4
---
bolt/lib/Rewrite/MachORewriteInstance.cpp | 4 ++--
bolt/lib/Rewrite/RewriteInstance.cpp | 11 ++-
2 files changed, 8 insertions(+), 7 deletions(-)
diff --git a/bolt/lib/Rewrite/MachORewriteInstance.cpp
b/bolt/lib/Rewrite/MachORewriteInstance.cpp
index 8be8257f15c1ce..c2a6c1f88ab070 100644
--- a/bolt/lib/Rewrite/MachORewriteInstance.cpp
+++ b/bolt/lib/Rewrite/MachORewriteInstance.cpp
@@ -337,7 +337,7 @@ void MachORewriteInstance::disassembleFunctions() {
BinaryFunction &Function = BFI.second;
if (!Function.isSimple())
continue;
-Function.disassemble();
+cantFail(Function.disassemble());
if (opts::PrintDisasm)
Function.print(outs(), "after disassembly");
}
@@ -387,7 +387,7 @@ void MachORewriteInstance::runOptimizationPasses() {
Manager.registerPass(
std::make_unique(opts::PrintFinalized));
- Manager.runPasses();
+ cantFail(Manager.runPasses());
}
void MachORewriteInstance::mapInstrumentationSection(
diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp
b/bolt/lib/Rewrite/RewriteInstance.cpp
index cc994c213482d4..06aed246de0cf2 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -1283,7 +1283,7 @@ void RewriteInstance::discoverFileObjects() {
/*UseMaxSize*/ true);
if (BF) {
assert(Rel.isRelative() && "Expected relative relocation for
island");
- BF->markIslandDynamicRelocationAtAddress(RelAddress);
+ cantFail(BF->markIslandDynamicRelocationAtAddress(RelAddress));
}
}
}
@@ -2859,8 +2859,9 @@ void RewriteInstance::selectFunctionsToProcess() {
StringSet<> ReorderFunctionsUserSet;
StringSet<> ReorderFunctionsLTOCommonSet;
if (opts::ReorderFunctions == ReorderFunctions::RT_USER) {
-for (const std::string &Function :
- ReorderFunctions::readFunctionOrderFile()) {
+std::vector FunctionNames;
+cantFail(ReorderFunctions::readFunctionOrderFile(FunctionNames));
+for (const std::string &Function : FunctionNames) {
ReorderFunctionsUserSet.insert(Function);
if (std::optional LTOCommonName = getLTOCommonName(Function))
ReorderFunctionsLTOCommonSet.insert(*LTOCommonName);
@@ -3207,7 +3208,7 @@ void RewriteInstance::disassembleFunctions() {
check_error(LSDASection.getError(), "failed to get LSDA section");
ArrayRef LSDAData = ArrayRef(
LSDASection->getData(), LSDASection->getContents().size());
- Function.parseLSDA(LSDAData, LSDASection->getAddress());
+ cantFail(Function.parseLSDA(LSDAData, LSDASection->getAddress()));
}
}
}
@@ -3298,7 +3299,7 @@ void RewriteInstance::postProcessFunctions() {
void RewriteInstance::runOptimizationPasses() {
NamedRegionTimer T("runOptimizationPasses", "run optimization passes",
TimerGroupName, TimerGroupDesc, opts::TimeRewrite);
- BinaryFunctionPassManager::runAllPasses(*BC);
+ cantFail(BinaryFunctionPassManager::runAllPasses(*BC));
}
void RewriteInstance::preregisterSections() {
>From cb5a6939b09a88e1f1442d8ea761a04b2fcad332 Mon Sep 17 00:00:00 2001
From: Amir Ayupov
Date: Mon, 12 Feb 2024 13:52:48 -0800
Subject: [PATCH 2/2] Handle BOLTErrors in runOptimizationPasses
Created using spr 1.3.4
---
bolt/lib/Rewrite/RewriteInstance.cpp | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp
b/bolt/lib/Rewrite/RewriteInstance.cpp
index 06aed246de0cf2..829568cefff01f 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -3299,7 +3299,11 @@ void RewriteInstance::postProcessFunctions() {
void RewriteInstance::runOptimizationPasses() {
NamedRegionTimer T("runOptimizationPasses", "run optimization passes",
TimerGroupName, TimerGroupDesc, opts::TimeRewrite);
- cantFail(BinaryFunctionPassManager::runAllPasses(*BC));
+ handleAllErrors(BinaryFunctionPassManager::runAllPasses(*BC),
+ [](const BOLTError &E) {
+E.log(errs());
+exit(1);
+ });
}
void RewriteInstance::preregisterSections() {
___
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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/aaupov updated https://github.com/llvm/llvm-project/pull/81524 ___ 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] [BOLT] Log through JournalingStreams (PR #81524)
https://github.com/aaupov updated https://github.com/llvm/llvm-project/pull/81524 ___ 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] [BOLT] Propagate BOLTErrors from Core, RewriteInstance, and passes (2/2) (PR #81523)
https://github.com/aaupov edited https://github.com/llvm/llvm-project/pull/81523 ___ 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] [NFC][CallPromotionUtils]Extract a helper function versionCallSiteWithCond from versionCallSite (PR #81181)
https://github.com/minglotus-6 updated
https://github.com/llvm/llvm-project/pull/81181
>From 7ebae253ab1808bca328453f68af2b595d07176e Mon Sep 17 00:00:00 2001
From: mingmingl
Date: Thu, 8 Feb 2024 11:32:50 -0800
Subject: [PATCH 1/2] [NFC][CallPromotionUtils]Extract a helper function
versionCallSiteWithCond from versionCallSite
---
.../Transforms/Utils/CallPromotionUtils.cpp | 36 +++
1 file changed, 21 insertions(+), 15 deletions(-)
diff --git a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
index 4e84927f1cfc90..d0cf0792eface0 100644
--- a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
+++ b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
@@ -188,10 +188,9 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Predicate and clone the given call site.
///
/// This function creates an if-then-else structure at the location of the call
-/// site. The "if" condition compares the call site's called value to the given
-/// callee. The original call site is moved into the "else" block, and a clone
-/// of the call site is placed in the "then" block. The cloned instruction is
-/// returned.
+/// site. The "if" condition is specified by `Cond`.
+/// The original call site is moved into the "else" block, and a clone of the
+/// call site is placed in the "then" block. The cloned instruction is
returned.
///
/// For example, the call instruction below:
///
@@ -202,7 +201,6 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Is replace by the following:
///
/// orig_bb:
-/// %cond = icmp eq i32 ()* %ptr, @func
/// br i1 %cond, %then_bb, %else_bb
///
/// then_bb:
@@ -232,7 +230,6 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Is replace by the following:
///
/// orig_bb:
-/// %cond = icmp eq i32 ()* %ptr, @func
/// br i1 %cond, %then_bb, %else_bb
///
/// then_bb:
@@ -267,7 +264,6 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Is replaced by the following:
///
/// cond_bb:
-/// %cond = icmp eq i32 ()* %ptr, @func
/// br i1 %cond, %then_bb, %orig_bb
///
/// then_bb:
@@ -280,19 +276,13 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// ; The original call instruction stays in its original block.
/// %t0 = musttail call i32 %ptr()
/// ret %t0
-CallBase &llvm::versionCallSite(CallBase &CB, Value *Callee,
-MDNode *BranchWeights) {
+static CallBase &versionCallSiteWithCond(CallBase &CB, Value *Cond,
+ MDNode *BranchWeights) {
IRBuilder<> Builder(&CB);
CallBase *OrigInst = &CB;
BasicBlock *OrigBlock = OrigInst->getParent();
- // Create the compare. The called value and callee must have the same type to
- // be compared.
- if (CB.getCalledOperand()->getType() != Callee->getType())
-Callee = Builder.CreateBitCast(Callee, CB.getCalledOperand()->getType());
- auto *Cond = Builder.CreateICmpEQ(CB.getCalledOperand(), Callee);
-
if (OrigInst->isMustTailCall()) {
// Create an if-then structure. The original instruction stays in its
block,
// and a clone of the original instruction is placed in the "then" block.
@@ -380,6 +370,22 @@ CallBase &llvm::versionCallSite(CallBase &CB, Value
*Callee,
return *NewInst;
}
+// Predicate and clone the given call site usingc condition `CB.callee ==
+// Callee`. See the comment `versionCallSiteWithCond` for the transformation.
+CallBase &llvm::versionCallSite(CallBase &CB, Value *Callee,
+MDNode *BranchWeights) {
+
+ IRBuilder<> Builder(&CB);
+
+ // Create the compare. The called value and callee must have the same type to
+ // be compared.
+ if (CB.getCalledOperand()->getType() != Callee->getType())
+Callee = Builder.CreateBitCast(Callee, CB.getCalledOperand()->getType());
+ auto *Cond = Builder.CreateICmpEQ(CB.getCalledOperand(), Callee);
+
+ return versionCallSiteWithCond(CB, Cond, BranchWeights);
+}
+
bool llvm::isLegalToPromote(const CallBase &CB, Function *Callee,
const char **FailureReason) {
assert(!CB.getCalledFunction() && "Only indirect call sites can be
promoted");
>From e774f69e9e960966399aa5569664c254ff0040bd Mon Sep 17 00:00:00 2001
From: mingmingl
Date: Mon, 12 Feb 2024 15:38:07 -0800
Subject: [PATCH 2/2] resolve review feedback
---
llvm/lib/Transforms/Utils/CallPromotionUtils.cpp | 9 ++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
index d0cf0792eface0..47e6d299ae8607 100644
--- a/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
+++ b/llvm/lib/Transforms/Utils/CallPromotionUtils.cpp
@@ -188,9 +188,9 @@ static void createRetB
[llvm-branch-commits] [llvm] [NFC][CallPromotionUtils]Extract a helper function versionCallSiteWithCond from versionCallSite (PR #81181)
@@ -267,7 +264,6 @@ static void createRetBitCast(CallBase &CB, Type *RetTy,
CastInst **RetBitCast) {
/// Is replaced by the following:
///
/// cond_bb:
-/// %cond = icmp eq i32 ()* %ptr, @func
minglotus-6 wrote:
done.
https://github.com/llvm/llvm-project/pull/81181
___
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] [BOLT] Add binary introspection/JIT manager (PR #81346)
@@ -0,0 +1,99 @@
+#include "bolt/Rewrite/JITRewriteInstance.h"
+#include "bolt/Core/BinaryContext.h"
+#include "llvm/BinaryFormat/ELF.h"
+#include "llvm/Support/TargetSelect.h"
+#include "gtest/gtest.h"
+
+using namespace llvm;
+using namespace llvm::object;
+using namespace llvm::ELF;
+using namespace bolt;
+
+namespace {
+struct JITRewriteInstanceTester
+: public testing::TestWithParam {
+ void SetUp() override {
+initalizeLLVM();
+initializeBOLT();
+ }
+
+protected:
+ void initalizeLLVM() {
+llvm::InitializeAllTargetInfos();
+llvm::InitializeAllTargetMCs();
+llvm::InitializeAllAsmParsers();
+llvm::InitializeAllDisassemblers();
+llvm::InitializeAllTargets();
+llvm::InitializeAllAsmPrinters();
+ }
+
+ void initializeBOLT() {
+BOLTJIT = cantFail(bolt::JITRewriteInstance::createJITRewriteInstance(
+{llvm::outs(), llvm::errs()}, /*IsPIC*/ false));
+ASSERT_FALSE(!BOLTJIT);
+ }
+
+ std::unique_ptr BOLTJIT;
+};
+} // namespace
+
+#ifdef X86_AVAILABLE
+
+// clang-format off
+extern "C" __attribute((naked)) int fib(int n)
+{
+ __asm__ __volatile__(
+"pushq %%r14\n"
+"pushq %%rbx\n"
+"pushq %%rax\n"
+"movl%%edi, %%r14d\n"
+"xorl%%ebx, %%ebx\n"
+"cmpl$0x2, %%edi\n"
+"jge .Ltmp0\n"
+"movl%%r14d, %%ecx\n"
+"jmp .Ltmp1\n"
+".Ltmp0:\n"
+"xorl%%ebx, %%ebx\n"
+"nopw%%cs:(%%rax,%%rax)\n"
+".Ltmp2:\n"
+"leal-0x1(%%r14), %%edi\n"
+"callq fib\n"
+"leal-0x2(%%r14), %%ecx\n"
+"addl%%eax, %%ebx\n"
+"cmpl$0x3, %%r14d\n"
+"movl%%ecx, %%r14d\n"
+"ja .Ltmp2\n"
+".Ltmp1:\n"
+"addl%%ecx, %%ebx\n"
+"movl%%ebx, %%eax\n"
+"addq$0x8, %%rsp\n"
+"popq%%rbx\n"
+"popq%%r14\n"
+"retq\n"
+:::);
+}
+// clang-format on
+
+INSTANTIATE_TEST_SUITE_P(X86, JITRewriteInstanceTester,
+ ::testing::Values(Triple::x86_64));
+
+TEST_P(JITRewriteInstanceTester, DisassembleFib) {
+ EXPECT_EQ(fib(7), 13);
+
+ // BOLT JIT test/example
+ // Analyze fib function in this binary
+ // Disassemble 63 bytes
+ uint64_t Address = reinterpret_cast(&fib);
+ StringRef Data = StringRef(reinterpret_cast(&fib), 63);
aaupov wrote:
Should JITRewriter expose symbol size so users use it directly?
https://github.com/llvm/llvm-project/pull/81346
___
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] [BOLT] Add binary introspection/JIT manager (PR #81346)
@@ -0,0 +1,99 @@
+#include "bolt/Rewrite/JITRewriteInstance.h"
+#include "bolt/Core/BinaryContext.h"
+#include "llvm/BinaryFormat/ELF.h"
+#include "llvm/Support/TargetSelect.h"
+#include "gtest/gtest.h"
+
+using namespace llvm;
+using namespace llvm::object;
+using namespace llvm::ELF;
+using namespace bolt;
+
+namespace {
+struct JITRewriteInstanceTester
+: public testing::TestWithParam {
+ void SetUp() override {
+initalizeLLVM();
+initializeBOLT();
+ }
+
+protected:
+ void initalizeLLVM() {
+llvm::InitializeAllTargetInfos();
+llvm::InitializeAllTargetMCs();
+llvm::InitializeAllAsmParsers();
+llvm::InitializeAllDisassemblers();
+llvm::InitializeAllTargets();
+llvm::InitializeAllAsmPrinters();
+ }
+
+ void initializeBOLT() {
+BOLTJIT = cantFail(bolt::JITRewriteInstance::createJITRewriteInstance(
+{llvm::outs(), llvm::errs()}, /*IsPIC*/ false));
+ASSERT_FALSE(!BOLTJIT);
+ }
+
+ std::unique_ptr BOLTJIT;
+};
+} // namespace
+
+#ifdef X86_AVAILABLE
+
+// clang-format off
+extern "C" __attribute((naked)) int fib(int n)
+{
+ __asm__ __volatile__(
+"pushq %%r14\n"
+"pushq %%rbx\n"
+"pushq %%rax\n"
+"movl%%edi, %%r14d\n"
+"xorl%%ebx, %%ebx\n"
+"cmpl$0x2, %%edi\n"
+"jge .Ltmp0\n"
+"movl%%r14d, %%ecx\n"
+"jmp .Ltmp1\n"
+".Ltmp0:\n"
+"xorl%%ebx, %%ebx\n"
+"nopw%%cs:(%%rax,%%rax)\n"
+".Ltmp2:\n"
+"leal-0x1(%%r14), %%edi\n"
+"callq fib\n"
+"leal-0x2(%%r14), %%ecx\n"
+"addl%%eax, %%ebx\n"
+"cmpl$0x3, %%r14d\n"
+"movl%%ecx, %%r14d\n"
+"ja .Ltmp2\n"
+".Ltmp1:\n"
+"addl%%ecx, %%ebx\n"
+"movl%%ebx, %%eax\n"
+"addq$0x8, %%rsp\n"
+"popq%%rbx\n"
+"popq%%r14\n"
+"retq\n"
+:::);
+}
+// clang-format on
+
+INSTANTIATE_TEST_SUITE_P(X86, JITRewriteInstanceTester,
+ ::testing::Values(Triple::x86_64));
+
+TEST_P(JITRewriteInstanceTester, DisassembleFib) {
+ EXPECT_EQ(fib(7), 13);
+
+ // BOLT JIT test/example
+ // Analyze fib function in this binary
+ // Disassemble 63 bytes
+ uint64_t Address = reinterpret_cast(&fib);
+ StringRef Data = StringRef(reinterpret_cast(&fib), 63);
rafaelauler wrote:
The rewriter object will only get knowledge about the symbol size after a call
to to registerJITFunction. Not sure if I answered the question, but does it
make sense to you? The user calling registerJITSection function will first make
a section visible to the Rewriter, and then the user can partition this section
in different functions using registerJITFunction, so the user passes the size
down to the rewriter instead of the opposite.
https://github.com/llvm/llvm-project/pull/81346
___
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] [ELF] Place .lbss/.lrodata/.ldata after .bss (PR #81224)
https://github.com/MaskRay updated https://github.com/llvm/llvm-project/pull/81224 ___ 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] [ELF] Place .lbss/.lrodata/.ldata after .bss (PR #81224)
https://github.com/MaskRay updated https://github.com/llvm/llvm-project/pull/81224 ___ 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] [ELF] Place .lbss/.lrodata/.ldata after .bss (PR #81224)
https://github.com/MaskRay edited https://github.com/llvm/llvm-project/pull/81224 ___ 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] [ELF] Place .lbss/.lrodata/.ldata after .bss (PR #81224)
jyknight wrote: ISTM that making the layout conditional in order to not regress the most-common case (PIE) would be worth a minor amount of code in lld, even though the benefit is also relatively small. Do you think INSERT scripts are going to be a real-world issue there? It seems to me like it wouldn't much matter to someone doing INSERT BEFORE/AFTER .lrodata, if lrodata goes near the beginning or near the end? https://github.com/llvm/llvm-project/pull/81224 ___ 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-format] Don't remove parentheses in macro definitions (#81444) (PR #81566)
https://github.com/llvmbot created
https://github.com/llvm/llvm-project/pull/81566
Backport 4af24d4ab765
Requested by: @owenca
>From 0acacbbad0160dc697375ab657b4231012e4e4d0 Mon Sep 17 00:00:00 2001
From: Owen Pan
Date: Mon, 12 Feb 2024 19:20:26 -0800
Subject: [PATCH] [clang-format] Don't remove parentheses in macro definitions
(#81444)
Closes #81399.
(cherry picked from commit 4af24d4ab76539706bfbceec4b3923426fb1b9e7)
---
clang/lib/Format/UnwrappedLineParser.cpp | 2 +-
clang/unittests/Format/FormatTest.cpp| 2 ++
2 files changed, 3 insertions(+), 1 deletion(-)
diff --git a/clang/lib/Format/UnwrappedLineParser.cpp
b/clang/lib/Format/UnwrappedLineParser.cpp
index b904e0e56d9eb3..57391979887078 100644
--- a/clang/lib/Format/UnwrappedLineParser.cpp
+++ b/clang/lib/Format/UnwrappedLineParser.cpp
@@ -2515,7 +2515,7 @@ bool UnwrappedLineParser::parseParens(TokenType
AmpAmpTokenType) {
parseChildBlock();
break;
case tok::r_paren:
- if (!MightBeStmtExpr &&
+ if (!MightBeStmtExpr && !Line->InMacroBody &&
Style.RemoveParentheses > FormatStyle::RPS_Leave) {
const auto *Prev = LeftParen->Previous;
const auto *Next = Tokens->peekNextToken();
diff --git a/clang/unittests/Format/FormatTest.cpp
b/clang/unittests/Format/FormatTest.cpp
index a471e36f8d6825..0beba12dda62ae 100644
--- a/clang/unittests/Format/FormatTest.cpp
+++ b/clang/unittests/Format/FormatTest.cpp
@@ -26856,6 +26856,7 @@ TEST_F(FormatTest, RemoveParentheses) {
EXPECT_EQ(Style.RemoveParentheses, FormatStyle::RPS_Leave);
Style.RemoveParentheses = FormatStyle::RPS_MultipleParentheses;
+ verifyFormat("#define Foo(...) foo((__VA_ARGS__))", Style);
verifyFormat("int x __attribute__((aligned(16))) = 0;", Style);
verifyFormat("decltype((foo->bar)) baz;", Style);
verifyFormat("class __declspec(dllimport) X {};",
@@ -26890,6 +26891,7 @@ TEST_F(FormatTest, RemoveParentheses) {
verifyFormat("return (({ 0; }));", "return ((({ 0; })));", Style);
Style.RemoveParentheses = FormatStyle::RPS_ReturnStatement;
+ verifyFormat("#define Return0 return (0);", Style);
verifyFormat("return 0;", "return (0);", Style);
verifyFormat("co_return 0;", "co_return ((0));", Style);
verifyFormat("return 0;", "return (((0)));", Style);
___
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-format] Don't remove parentheses in macro definitions (#81444) (PR #81566)
https://github.com/llvmbot milestoned https://github.com/llvm/llvm-project/pull/81566 ___ 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-format] Don't remove parentheses in macro definitions (#81444) (PR #81566)
llvmbot wrote: @HazardyKnusperkeks What do you think about merging this PR to the release branch? https://github.com/llvm/llvm-project/pull/81566 ___ 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-format] Don't remove parentheses in macro definitions (#81444) (PR #81566)
llvmbot wrote:
@llvm/pr-subscribers-clang-format
Author: None (llvmbot)
Changes
Backport 4af24d4ab765
Requested by: @owenca
---
Full diff: https://github.com/llvm/llvm-project/pull/81566.diff
2 Files Affected:
- (modified) clang/lib/Format/UnwrappedLineParser.cpp (+1-1)
- (modified) clang/unittests/Format/FormatTest.cpp (+2)
``diff
diff --git a/clang/lib/Format/UnwrappedLineParser.cpp
b/clang/lib/Format/UnwrappedLineParser.cpp
index b904e0e56d9eb3..57391979887078 100644
--- a/clang/lib/Format/UnwrappedLineParser.cpp
+++ b/clang/lib/Format/UnwrappedLineParser.cpp
@@ -2515,7 +2515,7 @@ bool UnwrappedLineParser::parseParens(TokenType
AmpAmpTokenType) {
parseChildBlock();
break;
case tok::r_paren:
- if (!MightBeStmtExpr &&
+ if (!MightBeStmtExpr && !Line->InMacroBody &&
Style.RemoveParentheses > FormatStyle::RPS_Leave) {
const auto *Prev = LeftParen->Previous;
const auto *Next = Tokens->peekNextToken();
diff --git a/clang/unittests/Format/FormatTest.cpp
b/clang/unittests/Format/FormatTest.cpp
index a471e36f8d6825..0beba12dda62ae 100644
--- a/clang/unittests/Format/FormatTest.cpp
+++ b/clang/unittests/Format/FormatTest.cpp
@@ -26856,6 +26856,7 @@ TEST_F(FormatTest, RemoveParentheses) {
EXPECT_EQ(Style.RemoveParentheses, FormatStyle::RPS_Leave);
Style.RemoveParentheses = FormatStyle::RPS_MultipleParentheses;
+ verifyFormat("#define Foo(...) foo((__VA_ARGS__))", Style);
verifyFormat("int x __attribute__((aligned(16))) = 0;", Style);
verifyFormat("decltype((foo->bar)) baz;", Style);
verifyFormat("class __declspec(dllimport) X {};",
@@ -26890,6 +26891,7 @@ TEST_F(FormatTest, RemoveParentheses) {
verifyFormat("return (({ 0; }));", "return ((({ 0; })));", Style);
Style.RemoveParentheses = FormatStyle::RPS_ReturnStatement;
+ verifyFormat("#define Return0 return (0);", Style);
verifyFormat("return 0;", "return (0);", Style);
verifyFormat("co_return 0;", "co_return ((0));", Style);
verifyFormat("return 0;", "return (((0)));", Style);
``
https://github.com/llvm/llvm-project/pull/81566
___
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] [ELF] Place .lbss/.lrodata/.ldata after .bss (PR #81224)
MaskRay wrote:
> ISTM that making the layout conditional in order to not regress the
> most-common case (PIE) would be worth a minor amount of code in lld, even
> though the benefit is also relatively small.
>
> Do you think INSERT scripts are going to be a real-world issue there? It
> seems to me like it wouldn't much matter to someone doing INSERT BEFORE/AFTER
> .lrodata, if lrodata goes near the beginning or near the end?
This has been my objection to making this layout change at all.
If we consider `-no-pie` an uncommon case that has to be supported, ISTM we can
require the user to use a linker script fragment
```
SECTIONS { .lrodata : { *(.lrodata .lrodata.*) } } INSERT AFTER .bss;
```
https://github.com/llvm/llvm-project/pull/81224
___
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] [ELF] Place .lbss/.lrodata/.ldata after .bss (PR #81224)
jyknight wrote: I don't follow. no-pie is _uncommon_ but not _unsupported_. Making users add a linker script when they wouldn't otherwise have one, in order to have a working no-pie build seems unjustified. https://github.com/llvm/llvm-project/pull/81224 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
