[Pipelines] Perform mergefunc after constmerge #92498
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
yamt
force-pushed
the
mergefunc-after-constmerge
branch
from
ebc89c7
to
91eadcf
Compare
yamt
changed the title
nikic
reviewed
Jun 26, 2024
Constmerge can fold switch jump tables, possibly making functions identical again. It can help mergefunc. On the otherhand, the opposite seems unlikely. Fixes llvm#92201
yamt
force-pushed
the
mergefunc-after-constmerge
branch
from
91eadcf
to
a6d30c8
Compare
i added a few tests. |
|
@llvm/pr-subscribers-llvm-transforms Author: YAMAMOTO Takashi (yamt) ChangesConstmerge can fold switch jump tables, possibly making functions identical again. It can help mergefunc. Fixes #92201 Full diff: https://github.com/llvm/llvm-project/pull/92498.diff 4 Files Affected:
diff --git a/llvm/lib/Passes/PassBuilderPipelines.cpp b/llvm/lib/Passes/PassBuilderPipelines.cpp
index 926515c9508a97..4fd5ee1946bb77 100644
--- a/llvm/lib/Passes/PassBuilderPipelines.cpp
+++ b/llvm/lib/Passes/PassBuilderPipelines.cpp
@@ -1527,10 +1527,6 @@ PassBuilder::buildModuleOptimizationPipeline(OptimizationLevel Level,
if (EnableIROutliner)
MPM.addPass(IROutlinerPass());
- // Merge functions if requested.
- if (PTO.MergeFunctions)
- MPM.addPass(MergeFunctionsPass());
-
// Now we need to do some global optimization transforms.
// FIXME: It would seem like these should come first in the optimization
// pipeline and maybe be the bottom of the canonicalization pipeline? Weird
@@ -1538,6 +1534,11 @@ PassBuilder::buildModuleOptimizationPipeline(OptimizationLevel Level,
MPM.addPass(GlobalDCEPass());
MPM.addPass(ConstantMergePass());
+ // Merge functions if requested. It has a better chance to merge functions
+ // after ConstantMerge folded jump tables.
+ if (PTO.MergeFunctions)
+ MPM.addPass(MergeFunctionsPass());
+
if (PTO.CallGraphProfile && !LTOPreLink)
MPM.addPass(CGProfilePass(LTOPhase == ThinOrFullLTOPhase::FullLTOPostLink ||
LTOPhase == ThinOrFullLTOPhase::ThinLTOPostLink));
diff --git a/llvm/test/Other/new-pm-defaults.ll b/llvm/test/Other/new-pm-defaults.ll
index 489aed40c190b4..588337c15625e6 100644
--- a/llvm/test/Other/new-pm-defaults.ll
+++ b/llvm/test/Other/new-pm-defaults.ll
@@ -281,9 +281,9 @@
; CHECK-HOT-COLD-SPLIT-NEXT: Running pass: HotColdSplittingPass
; CHECK-IR-OUTLINER-NEXT: Running pass: IROutlinerPass
; CHECK-IR-OUTLINER-NEXT: Running analysis: IRSimilarityAnalysis
-; CHECK-MERGE-FUNCS-NEXT: Running pass: MergeFunctionsPass
; CHECK-O-NEXT: Running pass: GlobalDCEPass
; CHECK-O-NEXT: Running pass: ConstantMergePass
+; CHECK-MERGE-FUNCS-NEXT: Running pass: MergeFunctionsPass
; CHECK-DEFAULT-NEXT: Running pass: CGProfilePass
; CHECK-DEFAULT-NEXT: Running pass: RelLookupTableConverterPass
; CHECK-LTO-NOT: Running pass: RelLookupTableConverterPass
diff --git a/llvm/test/Transforms/PhaseOrdering/X86/merge-functions2.ll b/llvm/test/Transforms/PhaseOrdering/X86/merge-functions2.ll
new file mode 100644
index 00000000000000..5d650d5f080bac
--- /dev/null
+++ b/llvm/test/Transforms/PhaseOrdering/X86/merge-functions2.ll
@@ -0,0 +1,67 @@
+; RUN: opt -passes="default<O3>" -enable-merge-functions -S < %s | FileCheck %s
+
+target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx12.0.0"
+
+; Function Attrs: noinline nounwind optsize ssp uwtable
+define i32 @f(i32 noundef %x) #0 {
+; CHECK-LABEL: @f(
+entry:
+ %x.addr = alloca i32, align 4
+ store i32 %x, ptr %x.addr, align 4, !tbaa !5
+ %0 = load i32, ptr %x.addr, align 4, !tbaa !5
+ switch i32 %0, label %sw.default [
+ i32 0, label %sw.bb
+ i32 2, label %sw.bb
+ i32 4, label %sw.bb
+ i32 6, label %sw.bb
+ i32 7, label %sw.bb
+ ]
+
+sw.bb: ; preds = %entry, %entry, %entry, %entry, %entry
+ store i32 1, ptr %x.addr, align 4, !tbaa !5
+ br label %sw.epilog
+
+sw.default: ; preds = %entry
+ store i32 0, ptr %x.addr, align 4, !tbaa !5
+ br label %sw.epilog
+
+sw.epilog: ; preds = %sw.default, %sw.bb
+ %1 = load i32, ptr %x.addr, align 4, !tbaa !5
+ ret i32 %1
+}
+
+; Function Attrs: noinline nounwind optsize ssp uwtable
+define i32 @g(i32 noundef %x) #0 {
+; CHECK-LABEL: @g(
+; CHECK-NEXT: [[TMP2:%.*]] = tail call range(i32 0, 2) i32 @f(i32 noundef [[TMP0:%.*]]) #[[ATTR0:[0-9]+]]
+; CHECK-NEXT: ret i32 [[TMP2]]
+entry:
+ %x.addr = alloca i32, align 4
+ store i32 %x, ptr %x.addr, align 4, !tbaa !5
+ %0 = load i32, ptr %x.addr, align 4, !tbaa !5
+ switch i32 %0, label %sw.default [
+ i32 0, label %sw.bb
+ i32 2, label %sw.bb
+ i32 4, label %sw.bb
+ i32 6, label %sw.bb
+ i32 7, label %sw.bb
+ ]
+
+sw.bb: ; preds = %entry, %entry, %entry, %entry, %entry
+ store i32 1, ptr %x.addr, align 4, !tbaa !5
+ br label %sw.epilog
+
+sw.default: ; preds = %entry
+ store i32 0, ptr %x.addr, align 4, !tbaa !5
+ br label %sw.epilog
+
+sw.epilog: ; preds = %sw.default, %sw.bb
+ %1 = load i32, ptr %x.addr, align 4, !tbaa !5
+ ret i32 %1
+}
+
+!5 = !{!6, !6, i64 0}
+!6 = !{!"int", !7, i64 0}
+!7 = !{!"omnipotent char", !8, i64 0}
+!8 = !{!"Simple C/C++ TBAA"}
diff --git a/llvm/test/Transforms/PhaseOrdering/X86/merge-functions3.ll b/llvm/test/Transforms/PhaseOrdering/X86/merge-functions3.ll
new file mode 100644
index 00000000000000..bce8f08ceda5e0
--- /dev/null
+++ b/llvm/test/Transforms/PhaseOrdering/X86/merge-functions3.ll
@@ -0,0 +1,47 @@
+; RUN: opt -passes="default<O3>" -enable-merge-functions -S < %s | FileCheck %s
+
+target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx12.0.0"
+
+@switch.table.f = private unnamed_addr constant [8 x i32] [i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 1], align 4
+@switch.table.g = private unnamed_addr constant [8 x i32] [i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 1], align 4
+
+; Function Attrs: mustprogress nofree noinline norecurse nosync nounwind optsize ssp willreturn memory(none) uwtable
+define range(i32 0, 2) i32 @f(i32 noundef %x) local_unnamed_addr #0 {
+; CHECK-LABEL: @f(
+entry:
+ %0 = icmp ult i32 %x, 8
+ br i1 %0, label %switch.lookup, label %sw.epilog
+
+switch.lookup: ; preds = %entry
+ %1 = zext nneg i32 %x to i64
+ %switch.gep = getelementptr inbounds [8 x i32], ptr @switch.table.f, i64 0, i64 %1
+ %switch.load = load i32, ptr %switch.gep, align 4
+ br label %sw.epilog
+
+sw.epilog: ; preds = %entry, %switch.lookup
+ %x.addr.0 = phi i32 [ %switch.load, %switch.lookup ], [ 0, %entry ]
+ ret i32 %x.addr.0
+}
+
+; Function Attrs: mustprogress nofree noinline norecurse nosync nounwind optsize ssp willreturn memory(none) uwtable
+define range(i32 0, 2) i32 @g(i32 noundef %x) local_unnamed_addr #0 {
+; CHECK-LABEL: @g(
+; CHECK-NEXT: [[TMP2:%.*]] = tail call range(i32 0, 2) i32 @f(i32 noundef [[TMP0:%.*]]) #[[ATTR0:[0-9]+]]
+; CHECK-NEXT: ret i32 [[TMP2]]
+entry:
+ %0 = icmp ult i32 %x, 8
+ br i1 %0, label %switch.lookup, label %sw.epilog
+
+switch.lookup: ; preds = %entry
+ %1 = zext nneg i32 %x to i64
+ %switch.gep = getelementptr inbounds [8 x i32], ptr @switch.table.g, i64 0, i64 %1
+ %switch.load = load i32, ptr %switch.gep, align 4
+ br label %sw.epilog
+
+sw.epilog: ; preds = %entry, %switch.lookup
+ %x.addr.0 = phi i32 [ %switch.load, %switch.lookup ], [ 0, %entry ]
+ ret i32 %x.addr.0
+}
+
+attributes #0 = { mustprogress nofree noinline norecurse nosync nounwind optsize ssp willreturn memory(none) uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cmov,+cx16,+cx8,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "tune-cpu"="generic" }
|
nikic
reviewed
Jun 26, 2024
nikic
approved these changes
Jun 26, 2024
Comment on lines
7
to
8
| ; Function Attrs: noinline nounwind optsize ssp uwtable | ||
| define i32 @f(i32 noundef %x) #0 { |
There was a problem hiding this comment.
Suggested change
| ; Function Attrs: noinline nounwind optsize ssp uwtable | |
| define i32 @f(i32 noundef %x) #0 { | |
| define i32 @f(i32 noundef %x) { |
Here and elsewhere.
|
can this land? |
kbluck
pushed a commit
to kbluck/llvm-project
that referenced
this pull request
Jul 6, 2024
Constmerge can fold switch jump tables, possibly making functions identical again. It can help mergefunc. On the other hand, the opposite seems unlikely. Fixes llvm#92201.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Constmerge can fold switch jump tables, possibly making functions identical again. It can help mergefunc.
On the otherhand, the opposite seems unlikely.
Fixes #92201