[llvm] 153a0b8 - [PGO][PGSO] Add profile guided size optimization to the X86 LEA fixup.
Hiroshi Yamauchi via llvm-commits
llvm-commits at lists.llvm.org
Mon Jul 13 10:06:53 PDT 2020
Author: Hiroshi Yamauchi
Date: 2020-07-13T09:46:22-07:00
New Revision: 153a0b8906d57f07f6711c2cfd10361616987038
URL: https://github.com/llvm/llvm-project/commit/153a0b8906d57f07f6711c2cfd10361616987038
DIFF: https://github.com/llvm/llvm-project/commit/153a0b8906d57f07f6711c2cfd10361616987038.diff
LOG: [PGO][PGSO] Add profile guided size optimization to the X86 LEA fixup.
Differential Revision: https://reviews.llvm.org/D83330
Added:
Modified:
llvm/lib/Target/X86/X86FixupLEAs.cpp
llvm/lib/Target/X86/X86PadShortFunction.cpp
llvm/test/CodeGen/X86/fixup-lea.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86FixupLEAs.cpp b/llvm/lib/Target/X86/X86FixupLEAs.cpp
index 9ac401bb0253..424279038921 100644
--- a/llvm/lib/Target/X86/X86FixupLEAs.cpp
+++ b/llvm/lib/Target/X86/X86FixupLEAs.cpp
@@ -16,8 +16,11 @@
#include "X86InstrInfo.h"
#include "X86Subtarget.h"
#include "llvm/ADT/Statistic.h"
+#include "llvm/Analysis/ProfileSummaryInfo.h"
+#include "llvm/CodeGen/LazyMachineBlockFrequencyInfo.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
+#include "llvm/CodeGen/MachineSizeOpts.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/CodeGen/TargetSchedule.h"
#include "llvm/Support/Debug.h"
@@ -111,6 +114,12 @@ class FixupLEAPass : public MachineFunctionPass {
MachineFunctionProperties::Property::NoVRegs);
}
+ void getAnalysisUsage(AnalysisUsage &AU) const override {
+ AU.addRequired<ProfileSummaryInfoWrapperPass>();
+ AU.addRequired<LazyMachineBlockFrequencyInfoPass>();
+ MachineFunctionPass::getAnalysisUsage(AU);
+ }
+
private:
TargetSchedModel TSM;
const X86InstrInfo *TII = nullptr;
@@ -205,21 +214,27 @@ bool FixupLEAPass::runOnMachineFunction(MachineFunction &MF) {
TSM.init(&ST);
TII = ST.getInstrInfo();
TRI = ST.getRegisterInfo();
+ auto *PSI = &getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI();
+ auto *MBFI = (PSI && PSI->hasProfileSummary())
+ ? &getAnalysis<LazyMachineBlockFrequencyInfoPass>().getBFI()
+ : nullptr;
LLVM_DEBUG(dbgs() << "Start X86FixupLEAs\n";);
for (MachineBasicBlock &MBB : MF) {
// First pass. Try to remove or optimize existing LEAs.
+ bool OptIncDecPerBB =
+ OptIncDec || llvm::shouldOptimizeForSize(&MBB, PSI, MBFI);
for (MachineBasicBlock::iterator I = MBB.begin(); I != MBB.end(); ++I) {
if (!isLEA(I->getOpcode()))
continue;
- if (optTwoAddrLEA(I, MBB, OptIncDec, UseLEAForSP))
+ if (optTwoAddrLEA(I, MBB, OptIncDecPerBB, UseLEAForSP))
continue;
if (IsSlowLEA)
processInstructionForSlowLEA(I, MBB);
else if (IsSlow3OpsLEA)
- processInstrForSlow3OpLEA(I, MBB, OptIncDec);
+ processInstrForSlow3OpLEA(I, MBB, OptIncDecPerBB);
}
// Second pass for creating LEAs. This may reverse some of the
diff --git a/llvm/lib/Target/X86/X86PadShortFunction.cpp b/llvm/lib/Target/X86/X86PadShortFunction.cpp
index 4c6bd0ccc2cd..ec81b07f9e5f 100644
--- a/llvm/lib/Target/X86/X86PadShortFunction.cpp
+++ b/llvm/lib/Target/X86/X86PadShortFunction.cpp
@@ -58,6 +58,7 @@ namespace {
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.addRequired<ProfileSummaryInfoWrapperPass>();
AU.addRequired<LazyMachineBlockFrequencyInfoPass>();
+ AU.addPreserved<LazyMachineBlockFrequencyInfoPass>();
MachineFunctionPass::getAnalysisUsage(AU);
}
diff --git a/llvm/test/CodeGen/X86/fixup-lea.ll b/llvm/test/CodeGen/X86/fixup-lea.ll
index 3f661a8e991e..35c3976408f6 100644
--- a/llvm/test/CodeGen/X86/fixup-lea.ll
+++ b/llvm/test/CodeGen/X86/fixup-lea.ll
@@ -109,31 +109,18 @@ for.end:
}
define void @foo_pgso(i32 inreg %dns) !prof !14 {
-; SLOW-LABEL: foo_pgso:
-; SLOW: # %bb.0: # %entry
-; SLOW-NEXT: xorl %ecx, %ecx
-; SLOW-NEXT: decl %ecx
-; SLOW-NEXT: .LBB4_1: # %for.body
-; SLOW-NEXT: # =>This Inner Loop Header: Depth=1
-; SLOW-NEXT: movzwl %cx, %edx
-; SLOW-NEXT: decl %ecx
-; SLOW-NEXT: cmpl %eax, %edx
-; SLOW-NEXT: jl .LBB4_1
-; SLOW-NEXT: # %bb.2: # %for.end
-; SLOW-NEXT: retl
-;
-; FAST-LABEL: foo_pgso:
-; FAST: # %bb.0: # %entry
-; FAST-NEXT: xorl %ecx, %ecx
-; FAST-NEXT: decl %ecx
-; FAST-NEXT: .LBB4_1: # %for.body
-; FAST-NEXT: # =>This Inner Loop Header: Depth=1
-; FAST-NEXT: movzwl %cx, %edx
-; FAST-NEXT: addl $-1, %ecx
-; FAST-NEXT: cmpl %eax, %edx
-; FAST-NEXT: jl .LBB4_1
-; FAST-NEXT: # %bb.2: # %for.end
-; FAST-NEXT: retl
+; CHECK-LABEL: foo_pgso:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %ecx, %ecx
+; CHECK-NEXT: decl %ecx
+; CHECK-NEXT: .LBB4_1: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: movzwl %cx, %edx
+; CHECK-NEXT: decl %ecx
+; CHECK-NEXT: cmpl %eax, %edx
+; CHECK-NEXT: jl .LBB4_1
+; CHECK-NEXT: # %bb.2: # %for.end
+; CHECK-NEXT: retl
entry:
br label %for.body
@@ -149,31 +136,18 @@ for.end:
}
define void @bar_pgso(i32 inreg %dns) !prof !14 {
-; SLOW-LABEL: bar_pgso:
-; SLOW: # %bb.0: # %entry
-; SLOW-NEXT: xorl %ecx, %ecx
-; SLOW-NEXT: incl %ecx
-; SLOW-NEXT: .LBB5_1: # %for.body
-; SLOW-NEXT: # =>This Inner Loop Header: Depth=1
-; SLOW-NEXT: movzwl %cx, %edx
-; SLOW-NEXT: incl %ecx
-; SLOW-NEXT: cmpl %eax, %edx
-; SLOW-NEXT: jl .LBB5_1
-; SLOW-NEXT: # %bb.2: # %for.end
-; SLOW-NEXT: retl
-;
-; FAST-LABEL: bar_pgso:
-; FAST: # %bb.0: # %entry
-; FAST-NEXT: xorl %ecx, %ecx
-; FAST-NEXT: incl %ecx
-; FAST-NEXT: .LBB5_1: # %for.body
-; FAST-NEXT: # =>This Inner Loop Header: Depth=1
-; FAST-NEXT: movzwl %cx, %edx
-; FAST-NEXT: addl $1, %ecx
-; FAST-NEXT: cmpl %eax, %edx
-; FAST-NEXT: jl .LBB5_1
-; FAST-NEXT: # %bb.2: # %for.end
-; FAST-NEXT: retl
+; CHECK-LABEL: bar_pgso:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %ecx, %ecx
+; CHECK-NEXT: incl %ecx
+; CHECK-NEXT: .LBB5_1: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: movzwl %cx, %edx
+; CHECK-NEXT: incl %ecx
+; CHECK-NEXT: cmpl %eax, %edx
+; CHECK-NEXT: jl .LBB5_1
+; CHECK-NEXT: # %bb.2: # %for.end
+; CHECK-NEXT: retl
entry:
br label %for.body
More information about the llvm-commits
mailing list