[profcheck] Add heuristical profile metadata for lowering table-based cttz. (#161898)

When lowering a `table-based cttz` calculation to the `llvm.cttz`
intrinsic, `AggressiveInstCombine` was not attaching profile metadata to
the newly generated `select` instruction.

This PR adds heuristic branch weights to the `select`. It uses a strong
100-to-1 probability favoring the `cttz` path over the zero-input case.
This allows later passes to optimize code layout and branch prediction.

Author: jinhuang1102

llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp

@@ -28,8 +28,12 @@
 #include "llvm/IR/Dominators.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/IRBuilder.h"
+#include "llvm/IR/Instruction.h"
+#include "llvm/IR/MDBuilder.h"
 #include "llvm/IR/PatternMatch.h"
 #include "llvm/IR/ProfDataUtils.h"
+#include "llvm/Support/Casting.h"
+#include "llvm/Support/CommandLine.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include "llvm/Transforms/Utils/BuildLibCalls.h"
 #include "llvm/Transforms/Utils/Local.h"
@@ -39,6 +43,10 @@ using namespace PatternMatch;
 
 #define DEBUG_TYPE "aggressive-instcombine"
 
+namespace llvm {
+extern cl::opt<bool> ProfcheckDisableMetadataFixes;
+}
+
 STATISTIC(NumAnyOrAllBitsSet, "Number of any/all-bits-set patterns folded");
 STATISTIC(NumGuardedRotates,
           "Number of guarded rotates transformed into funnel shifts");
@@ -599,6 +607,14 @@ static bool tryToRecognizeTableBasedCttz(Instruction &I, const DataLayout &DL) {
     auto Cmp = B.CreateICmpEQ(X1, ConstantInt::get(XType, 0));
     auto Select = B.CreateSelect(Cmp, B.CreateZExt(ZeroTableElem, XType), Cttz);
 
+    // The true branch of select handles the cttz(0) case, which is rare.
+    if (!ProfcheckDisableMetadataFixes) {
+      if (Instruction *SelectI = dyn_cast<Instruction>(Select))
+        SelectI->setMetadata(
+            LLVMContext::MD_prof,
+            MDBuilder(SelectI->getContext()).createUnlikelyBranchWeights());
+    }
+
     // NOTE: If the table[0] is 0, but the cttz(0) is defined by the Target
     // it should be handled as: `cttz(x) & (typeSize - 1)`.
 

llvm/test/Transforms/AggressiveInstCombine/lower-table-based-cttz-basics.ll

@@ -91,12 +91,13 @@
 
 @ctz7.table = internal unnamed_addr constant [32 x i8] c"\00\01\1C\02\1D\0E\18\03\1E\16\14\0F\19\11\04\08\1F\1B\0D\17\15\13\10\07\1A\0C\12\06\0B\05\0A\09", align 1
 
-define i32 @ctz1(i32 %x) {
+define i32 @ctz1(i32 %x) !prof !0 {
 ; CHECK-LABEL: @ctz1(
+; CHECK: !prof [[PROF_0:![0-9]+]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.cttz.i32(i32 [[X:%.*]], i1 true)
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[X]], 0
-; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[TMP0]], !prof [[PROF_1:![0-9]+]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = trunc i32 [[TMP2]] to i8
 ; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP3]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV]]
@@ -498,3 +499,7 @@ entry:
   %conv = zext i8 %0 to i32
   ret i32 %conv
 }
+
+!0 = !{!"function_entry_count", i64 1000}
+; CHECK: [[PROF_0]] = !{!"function_entry_count", i64 1000}
+; CHECK: [[PROF_1]] = !{!"branch_weights", i32 1, i32 1048575}

llvm/test/Transforms/AggressiveInstCombine/lower-table-based-cttz-dereferencing-pointer.ll

@@ -20,13 +20,14 @@
 
 @table = internal unnamed_addr constant [64 x i32] [i32 0, i32 1, i32 12, i32 2, i32 13, i32 22, i32 17, i32 3, i32 14, i32 33, i32 23, i32 36, i32 18, i32 58, i32 28, i32 4, i32 62, i32 15, i32 34, i32 26, i32 24, i32 48, i32 50, i32 37, i32 19, i32 55, i32 59, i32 52, i32 29, i32 44, i32 39, i32 5, i32 63, i32 11, i32 21, i32 16, i32 32, i32 35, i32 57, i32 27, i32 61, i32 25, i32 47, i32 49, i32 54, i32 51, i32 43, i32 38, i32 10, i32 20, i32 31, i32 56, i32 60, i32 46, i32 53, i32 42, i32 9, i32 30, i32 45, i32 41, i32 8, i32 40, i32 7, i32 6], align 4
 
-define i32 @ctz6(ptr nocapture readonly %b) {
+define i32 @ctz6(ptr nocapture readonly %b) !prof !0 {
 ; CHECK-LABEL: @ctz6(
+; CHECK: !prof [[PROF_0:![0-9]+]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr [[B:%.*]], align 8
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i64 @llvm.cttz.i64(i64 [[TMP0]], i1 true)
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP0]], 0
-; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 0, i64 [[TMP1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 0, i64 [[TMP1]], !prof [[PROF_1:![0-9]+]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
 ; CHECK-NEXT:    ret i32 [[TMP4]]
 ;
@@ -40,3 +41,7 @@ entry:
   %1 = load i32, ptr %arrayidx, align 4
   ret i32 %1
 }
+
+!0 = !{!"function_entry_count", i64 1000}
+; CHECK: [[PROF_0]] = !{!"function_entry_count", i64 1000}
+; CHECK: [[PROF_1]] = !{!"branch_weights", i32 1, i32 1048575}

llvm/test/Transforms/AggressiveInstCombine/lower-table-based-cttz-non-argument-value.ll

@@ -20,13 +20,14 @@
 @.str = private constant [3 x i8] c"%u\00", align 1
 @test.table = internal constant [32 x i8] c"\00\01\1C\02\1D\0E\18\03\1E\16\14\0F\19\11\04\08\1F\1B\0D\17\15\13\10\07\1A\0C\12\06\0B\05\0A\09", align 1
 
-define i32 @test() {
+define i32 @test() !prof !0 {
 ; CHECK-LABEL: @test(
+; CHECK: !prof [[PROF_0:![0-9]+]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr @x, align 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.cttz.i32(i32 [[TMP0]], i1 true)
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[TMP0]], 0
-; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 0, i32 [[TMP1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 0, i32 [[TMP1]], !prof [[PROF_1:![0-9]+]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = trunc i32 [[TMP3]] to i8
 ; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP4]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV]]
@@ -43,3 +44,7 @@ entry:
   %conv = zext i8 %1 to i32
   ret i32 %conv
 }
+
+!0 = !{!"function_entry_count", i64 1000}
+; CHECK: [[PROF_0]] = !{!"function_entry_count", i64 1000}
+; CHECK: [[PROF_1]] = !{!"branch_weights", i32 1, i32 1048575}

llvm/test/Transforms/AggressiveInstCombine/lower-table-based-cttz-zero-element.ll

@@ -3,12 +3,13 @@
 
 @ctz1.table = internal constant [32 x i8] c"\00\01\1C\02\1D\0E\18\03\1E\16\14\0F\19\11\04\08\1F\1B\0D\17\15\13\10\07\1A\0C\12\06\0B\05\0A\09", align 1
 
-define i32 @ctz1(i32 %x) {
+define i32 @ctz1(i32 %x) !prof !0 {
 ; CHECK-LABEL: @ctz1(
+; CHECK: !prof [[PROF_0:![0-9]+]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.cttz.i32(i32 [[X:%.*]], i1 true)
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[X]], 0
-; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[TMP0]], !prof [[PROF_1:![0-9]+]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = trunc i32 [[TMP2]] to i8
 ; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP3]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV]]
@@ -24,3 +25,7 @@ entry:
   %conv = zext i8 %0 to i32
   ret i32 %conv
 }
+
+!0 = !{!"function_entry_count", i64 1000}
+; CHECK: [[PROF_0]] = !{!"function_entry_count", i64 1000}
+; CHECK: [[PROF_1]] = !{!"branch_weights", i32 1, i32 1048575}