[SLP]Do not include the cost of and -1, <v> and emit just <v> after MinBitWidth.

After minbitwidth analysis, and <v>, (power_of_2 - 1 const) can be
transformed into just an <v>, (all_ones const), which can be ignored at
the cost estimation and at the codegen. x264 benchmark has this pattern.

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/90739

3 files changed, 26 insertions, 4 deletions

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index c33d90d531bf..038bceb39cb8 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -9484,6 +9484,16 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
                                          Op1Info, Op2Info, Operands, VI);
     };
     auto GetVectorCost = [=](InstructionCost CommonCost) {
+      if (ShuffleOrOp == Instruction::And && It != MinBWs.end()) {
+        for (unsigned I : seq<unsigned>(0, E->getNumOperands())) {
+          ArrayRef<Value *> Ops = E->getOperand(I);
+          if (all_of(Ops, [&](Value *Op) {
+                auto *CI = dyn_cast<ConstantInt>(Op);
+                return CI && CI->getValue().countr_one() >= It->second.first;
+              }))
+            return CommonCost;
+        }
+      }
       unsigned OpIdx = isa<UnaryOperator>(VL0) ? 0 : 1;
       TTI::OperandValueInfo Op1Info = getOperandInfo(E->getOperand(0));
       TTI::OperandValueInfo Op2Info = getOperandInfo(E->getOperand(OpIdx));
@@ -12969,6 +12979,20 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
         LLVM_DEBUG(dbgs() << "SLP: Diamond merged for " << *VL0 << ".\n");
         return E->VectorizedValue;
       }
+      if (ShuffleOrOp == Instruction::And && It != MinBWs.end()) {
+        for (unsigned I : seq<unsigned>(0, E->getNumOperands())) {
+          ArrayRef<Value *> Ops = E->getOperand(I);
+          if (all_of(Ops, [&](Value *Op) {
+                auto *CI = dyn_cast<ConstantInt>(Op);
+                return CI && CI->getValue().countr_one() >= It->second.first;
+              })) {
+            V = FinalShuffle(I == 0 ? RHS : LHS, E, VecTy);
+            E->VectorizedValue = V;
+            ++NumVectorInstructions;
+            return V;
+          }
+        }
+      }
       if (LHS->getType() != VecTy || RHS->getType() != VecTy) {
         assert((It != MinBWs.end() ||
                 getOperandEntry(E, 0)->State == TreeEntry::NeedToGather ||
diff --git a/llvm/test/Transforms/SLPVectorizer/RISCV/minbw-with-and-and-scalar-trunc.ll b/llvm/test/Transforms/SLPVectorizer/RISCV/minbw-with-and-and-scalar-trunc.ll
index fc977585614b..d6dc3bcc3354 100644
--- a/llvm/test/Transforms/SLPVectorizer/RISCV/minbw-with-and-and-scalar-trunc.ll
+++ b/llvm/test/Transforms/SLPVectorizer/RISCV/minbw-with-and-and-scalar-trunc.ll
@@ -12,8 +12,7 @@ define i16 @test() {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = call <4 x i64> @llvm.experimental.vp.strided.load.v4i64.p0.i64(ptr align 8 @c, i64 24, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, i32 4)
 ; CHECK-NEXT:    [[TMP1:%.*]] = trunc <4 x i64> [[TMP0]] to <4 x i16>
-; CHECK-NEXT:    [[TMP2:%.*]] = and <4 x i16> [[TMP1]], <i16 -1, i16 -1, i16 -1, i16 -1>
-; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i16> [[TMP2]], <i16 -1, i16 -1, i16 -1, i16 -1>
+; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i16> [[TMP1]], <i16 -1, i16 -1, i16 -1, i16 -1>
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i16 @llvm.vector.reduce.umax.v4i16(<4 x i16> [[TMP3]])
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext i16 [[TMP4]] to i32
 ; CHECK-NEXT:    [[T:%.*]] = trunc i32 [[TMP5]] to i16
diff --git a/llvm/test/Transforms/SLPVectorizer/RISCV/trunc-to-large-than-bw.ll b/llvm/test/Transforms/SLPVectorizer/RISCV/trunc-to-large-than-bw.ll
index 04d275742832..0c0c723e6699 100644
--- a/llvm/test/Transforms/SLPVectorizer/RISCV/trunc-to-large-than-bw.ll
+++ b/llvm/test/Transforms/SLPVectorizer/RISCV/trunc-to-large-than-bw.ll
@@ -9,8 +9,7 @@ define i32 @test() {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = call <4 x i64> @llvm.experimental.vp.strided.load.v4i64.p0.i64(ptr align 8 @c, i64 24, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, i32 4)
 ; CHECK-NEXT:    [[TMP1:%.*]] = trunc <4 x i64> [[TMP0]] to <4 x i16>
-; CHECK-NEXT:    [[TMP2:%.*]] = and <4 x i16> [[TMP1]], <i16 -1, i16 -1, i16 -1, i16 -1>
-; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i16> [[TMP2]], <i16 -1, i16 -1, i16 -1, i16 -1>
+; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i16> [[TMP1]], <i16 -1, i16 -1, i16 -1, i16 -1>
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i16 @llvm.vector.reduce.umax.v4i16(<4 x i16> [[TMP3]])
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext i16 [[TMP4]] to i32
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.umax.i32(i32 [[TMP5]], i32 1)