[llvm] 26a0d53 - [CHR] Skip region containing llvm.coro.id
Wei Wang via llvm-commits
llvm-commits at lists.llvm.org
Wed Apr 27 10:27:36 PDT 2022
Author: Wei Wang
Date: 2022-04-27T10:27:25-07:00
New Revision: 26a0d53b15447e011dc1a8de1f035e4aecd1083c
URL: https://github.com/llvm/llvm-project/commit/26a0d53b15447e011dc1a8de1f035e4aecd1083c
DIFF: https://github.com/llvm/llvm-project/commit/26a0d53b15447e011dc1a8de1f035e4aecd1083c.diff
LOG: [CHR] Skip region containing llvm.coro.id
When a block containing llvm.coro.id is cloned during CHR, it inserts an invalid
PHI node with token type to the beginning of the block containing llvm.coro.begin.
To avoid such case, we exclude regions with llvm.coro.id.
Reviewed By: ChuanqiXu
Differential Revision: https://reviews.llvm.org/D124418
Added:
llvm/test/Transforms/PGOProfile/chr_coro.ll
Modified:
llvm/lib/Transforms/Instrumentation/ControlHeightReduction.cpp
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/Instrumentation/ControlHeightReduction.cpp b/llvm/lib/Transforms/Instrumentation/ControlHeightReduction.cpp
index bda7a46db142f..9f3f2d6a07e52 100644
--- a/llvm/lib/Transforms/Instrumentation/ControlHeightReduction.cpp
+++ b/llvm/lib/Transforms/Instrumentation/ControlHeightReduction.cpp
@@ -26,6 +26,7 @@
#include "llvm/IR/CFG.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/IRBuilder.h"
+#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/MDBuilder.h"
#include "llvm/IR/PassManager.h"
#include "llvm/InitializePasses.h"
@@ -769,9 +770,21 @@ CHRScope * CHR::findScope(Region *R) {
return nullptr;
// If any of the basic blocks have address taken, we must skip this region
// because we cannot clone basic blocks that have address taken.
- for (BasicBlock *BB : R->blocks())
+ for (BasicBlock *BB : R->blocks()) {
if (BB->hasAddressTaken())
return nullptr;
+ // If we encounter llvm.coro.id, skip this region because if the basic block
+ // is cloned, we end up inserting a token type PHI node to the block with
+ // llvm.coro.begin.
+ // FIXME: This could lead to less optimal codegen, because the region is
+ // excluded, it can prevent CHR from merging adjacent regions into bigger
+ // scope and hoisting more branches.
+ for (Instruction &I : *BB)
+ if (auto *II = dyn_cast<IntrinsicInst>(&I))
+ if (II->getIntrinsicID() == Intrinsic::coro_id)
+ return nullptr;
+ }
+
if (Exit) {
// Try to find an if-then block (check if R is an if-then).
// if (cond) {
diff --git a/llvm/test/Transforms/PGOProfile/chr_coro.ll b/llvm/test/Transforms/PGOProfile/chr_coro.ll
new file mode 100644
index 0000000000000..d902f3da14ec4
--- /dev/null
+++ b/llvm/test/Transforms/PGOProfile/chr_coro.ll
@@ -0,0 +1,115 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+;RUN: opt < %s -passes='require<profile-summary>,function(chr,instcombine,simplifycfg)' -S | FileCheck %s
+
+declare void @foo()
+declare void @bar()
+
+declare token @llvm.coro.id(i32, i8*, i8*, i8*)
+declare i1 @llvm.coro.alloc(token)
+declare i8* @llvm.coro.begin(token, i8*)
+declare noalias i8* @malloc(i32)
+
+%f.Frame = type { void (%f.Frame*)*, void (%f.Frame*)*, i1 }
+
+; resume part of the coroutine
+define fastcc void @f.resume(%f.Frame* noalias nonnull align 8 dereferenceable(24) %FramePtr) {
+ tail call void @bar()
+ ret void
+}
+
+; destroy part of the coroutine
+define fastcc void @f.destroy(%f.Frame* noalias nonnull align 8 dereferenceable(24) %FramePtr) {
+ tail call void @bar()
+ ret void
+}
+
+; cleanup part of the coroutine
+define fastcc void @f.cleanup(%f.Frame* noalias nonnull align 8 dereferenceable(24) %FramePtr) {
+ tail call void @bar()
+ ret void
+}
+
+ at f.resumers = private constant [3 x void (%f.Frame*)*] [void (%f.Frame*)* @f.resume, void (%f.Frame*)* @f.destroy, void (%f.Frame*)* @f.cleanup]
+
+; Test that chr will skip block containing llvm.coro.id.
+define i8* @test_chr_with_coro_id(i32* %i) !prof !14 {
+; CHECK-LABEL: @test_chr_with_coro_id(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4
+; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 3
+; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 3
+; CHECK-NEXT: br i1 [[TMP2]], label %[[BB0:.*]], label %[[ENTRY_SPLIT_NONCHR:.*]], !prof !15
+; CHECK: [[BB0]]:
+; CHECK-NEXT: call void @foo()
+; CHECK-NEXT: br label %[[BB_CORO_ID:.*]]
+; CHECK: bb1.nonchr:
+; CHECK-NEXT: [[TMP4:%.*]] = and i32 [[TMP0]], 2
+; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[TMP4]], 0
+; CHECK-NEXT: br i1 [[TMP5]], label %[[BB2_NONCHR:.*]], label %[[BB_CORO_ID]], !prof !16
+; CHECK: [[BB2_NONCHR]]:
+; CHECK-NEXT: call void @foo()
+; CHECK-NEXT: br label %[[BB_CORO_ID]]
+; CHECK: [[BB_CORO_ID]]:
+; CHECK-NEXT: [[ID:%.*]] = call token @llvm.coro.id
+; CHECK-NEXT: [[NEED_DYN_ALLOC:%.*]] = call i1 @llvm.coro.alloc(token [[ID]])
+; CHECK-NEXT: br i1 [[NEED_DYN_ALLOC]], label %[[BB_CORO_DYN_ALLOC:.*]], label %[[BB_CORO_BEGIN:.*]]
+; CHECK: [[BB_CORO_BEGIN]]:
+; CHECK-NEXT: [[PHI:%.*]] = phi i8* [ null, %[[BB_CORO_ID]] ], [ %alloc, %[[BB_CORO_DYN_ALLOC]] ]
+; CHECK-NEXT: [[HDL:%.*]] = call noalias nonnull i8* @llvm.coro.begin(token [[ID]], i8* [[PHI]])
+;
+entry:
+ %0 = load i32, i32* %i
+ %1 = and i32 %0, 1
+ %2 = icmp eq i32 %1, 0
+ br i1 %2, label %bb1, label %bb0, !prof !15
+
+bb0:
+ call void @foo()
+ br label %bb1
+
+bb1:
+ %3 = and i32 %0, 2
+ %4 = icmp eq i32 %3, 0
+ br i1 %4, label %bb2, label %bb.coro.id, !prof !15
+
+bb2:
+ call void @foo()
+ br label %bb.coro.id
+
+bb.coro.id:
+ %id = call token @llvm.coro.id(i32 0, i8* null, i8* null, i8* bitcast ([3 x void (%f.Frame*)*]* @f.resumers to i8*))
+ %need.dyn.alloc = call i1 @llvm.coro.alloc(token %id)
+ br i1 %need.dyn.alloc, label %bb.coro.dyn.alloc, label %bb.coro.begin
+
+bb.coro.dyn.alloc:
+ %alloc = call i8* @malloc(i32 24)
+ br label %bb.coro.begin
+
+bb.coro.begin:
+ %phi = phi i8* [ null, %bb.coro.id ], [ %alloc, %bb.coro.dyn.alloc ]
+ %hdl = call noalias nonnull i8* @llvm.coro.begin(token %id, i8* %phi)
+ ret i8* %hdl
+}
+
+!llvm.module.flags = !{!0}
+!0 = !{i32 1, !"ProfileSummary", !1}
+!1 = !{!2, !3, !4, !5, !6, !7, !8, !9}
+!2 = !{!"ProfileFormat", !"InstrProf"}
+!3 = !{!"TotalCount", i64 10000}
+!4 = !{!"MaxCount", i64 10}
+!5 = !{!"MaxInternalCount", i64 1}
+!6 = !{!"MaxFunctionCount", i64 1000}
+!7 = !{!"NumCounts", i64 3}
+!8 = !{!"NumFunctions", i64 3}
+!9 = !{!"DetailedSummary", !10}
+!10 = !{!11, !12, !13}
+!11 = !{i32 10000, i64 100, i32 1}
+!12 = !{i32 999000, i64 100, i32 1}
+!13 = !{i32 999999, i64 1, i32 2}
+
+!14 = !{!"function_entry_count", i64 100}
+!15 = !{!"branch_weights", i32 0, i32 1}
+!16 = !{!"branch_weights", i32 1, i32 1}
+!17 = !{!"branch_weights", i32 0, i32 0}
+; CHECK: !15 = !{!"branch_weights", i32 1000, i32 0}
+; CHECK: !16 = !{!"branch_weights", i32 0, i32 1}
More information about the llvm-commits
mailing list