[PGO][PGSO] Add profile guided size optimization to the X86 LEA fixup.

Differential Revision: https://reviews.llvm.org/D83330

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 @@ public:
         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

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);
     }
 

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:
+; CHECK-LABEL: foo_pgso:
-; SLOW:       # %bb.0: # %entry
+; CHECK:       # %bb.0: # %entry
-; SLOW-NEXT:    xorl %ecx, %ecx
+; CHECK-NEXT:    xorl %ecx, %ecx
-; SLOW-NEXT:    decl %ecx
+; CHECK-NEXT:    decl %ecx
-; SLOW-NEXT:  .LBB4_1: # %for.body
+; CHECK-NEXT:  .LBB4_1: # %for.body
-; SLOW-NEXT:    # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
-; SLOW-NEXT:    movzwl %cx, %edx
+; CHECK-NEXT:    movzwl %cx, %edx
-; SLOW-NEXT:    decl %ecx
+; CHECK-NEXT:    decl %ecx
-; SLOW-NEXT:    cmpl %eax, %edx
+; CHECK-NEXT:    cmpl %eax, %edx
-; SLOW-NEXT:    jl .LBB4_1
+; CHECK-NEXT:    jl .LBB4_1
-; SLOW-NEXT:  # %bb.2: # %for.end
+; CHECK-NEXT:  # %bb.2: # %for.end
-; SLOW-NEXT:    retl
+; CHECK-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
 entry:
   br label %for.body
 
@@ -149,31 +136,18 @@ for.end:
 }
 
 define void @bar_pgso(i32 inreg %dns) !prof !14 {
-; SLOW-LABEL: bar_pgso:
+; CHECK-LABEL: bar_pgso:
-; SLOW:       # %bb.0: # %entry
+; CHECK:       # %bb.0: # %entry
-; SLOW-NEXT:    xorl %ecx, %ecx
+; CHECK-NEXT:    xorl %ecx, %ecx
-; SLOW-NEXT:    incl %ecx
+; CHECK-NEXT:    incl %ecx
-; SLOW-NEXT:  .LBB5_1: # %for.body
+; CHECK-NEXT:  .LBB5_1: # %for.body
-; SLOW-NEXT:    # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
-; SLOW-NEXT:    movzwl %cx, %edx
+; CHECK-NEXT:    movzwl %cx, %edx
-; SLOW-NEXT:    incl %ecx
+; CHECK-NEXT:    incl %ecx
-; SLOW-NEXT:    cmpl %eax, %edx
+; CHECK-NEXT:    cmpl %eax, %edx
-; SLOW-NEXT:    jl .LBB5_1
+; CHECK-NEXT:    jl .LBB5_1
-; SLOW-NEXT:  # %bb.2: # %for.end
+; CHECK-NEXT:  # %bb.2: # %for.end
-; SLOW-NEXT:    retl
+; CHECK-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
 entry:
   br label %for.body