[SLP]Fix mask building for alternate node cost estimation

[alexey-bataev]

Need to to use same functionality in cost model, as for the codegen, to
correctly build the shuffle mask and estimate the cost.

[llvmbot]

@llvm/pr-subscribers-llvm-transforms

Author: Alexey Bataev (alexey-bataev)

Changes

Need to to use same functionality in cost model, as for the codegen, to
correctly build the shuffle mask and estimate the cost.

---

Full diff: https://github.com/llvm/llvm-project/pull/102966.diff

3 Files Affected:

• (modified) llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp (+4-3)
• (modified) llvm/test/Transforms/PhaseOrdering/X86/vector-reductions-logical.ll (+22-19)
• (modified) llvm/test/Transforms/SLPVectorizer/X86/reduction-logical.ll (+10-22)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index c05e4b822d2c4e..f30a76af832790 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -10135,13 +10135,14 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
       }
       SmallVector<int> Mask;
       E->buildAltOpShuffleMask(
-          [E](Instruction *I) {
+          [&](Instruction *I) {
             assert(E->isOpcodeOrAlt(I) && "Unexpected main/alternate opcode");
-            return I->getOpcode() == E->getAltOpcode();
+            return isAlternateInstruction(I, E->getMainOp(), E->getAltOp(),
+                                          *TLI);
           },
           Mask);
       VecCost += ::getShuffleCost(TTIRef, TargetTransformInfo::SK_PermuteTwoSrc,
-                                  FinalVecTy, Mask);
+                                  FinalVecTy, Mask, CostKind);
       // Patterns like [fadd,fsub] can be combined into a single instruction
       // in x86. Reordering them into [fsub,fadd] blocks this pattern. So we
       // need to take into account their order when looking for the most used
diff --git a/llvm/test/Transforms/PhaseOrdering/X86/vector-reductions-logical.ll b/llvm/test/Transforms/PhaseOrdering/X86/vector-reductions-logical.ll
index 9c3e1fbec663ba..c97d8da58be441 100644
--- a/llvm/test/Transforms/PhaseOrdering/X86/vector-reductions-logical.ll
+++ b/llvm/test/Transforms/PhaseOrdering/X86/vector-reductions-logical.ll
@@ -86,16 +86,17 @@ return:
 define float @test_merge_anyof_v4sf(<4 x float> %t) {
 ; CHECK-LABEL: @test_merge_anyof_v4sf(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[T_FR:%.*]] = freeze <4 x float> [[T:%.*]]
-; CHECK-NEXT:    [[TMP0:%.*]] = fcmp olt <4 x float> [[T_FR]], zeroinitializer
-; CHECK-NEXT:    [[TMP1:%.*]] = fcmp ogt <4 x float> [[T_FR]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
-; CHECK-NEXT:    [[TMP2:%.*]] = or <4 x i1> [[TMP1]], [[TMP0]]
-; CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i1> [[TMP2]] to i4
-; CHECK-NEXT:    [[OP_RDX_NOT:%.*]] = icmp eq i4 [[TMP3]], 0
-; CHECK-NEXT:    [[SHIFT:%.*]] = shufflevector <4 x float> [[T_FR]], <4 x float> poison, <4 x i32> <i32 1, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP4:%.*]] = fadd <4 x float> [[T_FR]], [[SHIFT]]
-; CHECK-NEXT:    [[ADD:%.*]] = extractelement <4 x float> [[TMP4]], i64 0
-; CHECK-NEXT:    [[RETVAL_0:%.*]] = select i1 [[OP_RDX_NOT]], float [[ADD]], float 0.000000e+00
+; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x float> [[T:%.*]], <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = fcmp ogt <8 x float> [[TMP0]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>
+; CHECK-NEXT:    [[TMP2:%.*]] = fcmp olt <8 x float> [[TMP0]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i1> [[TMP1]], <8 x i1> [[TMP2]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEXT:    [[TMP4:%.*]] = freeze <8 x i1> [[TMP3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i1> [[TMP4]] to i8
+; CHECK-NEXT:    [[DOTNOT:%.*]] = icmp eq i8 [[TMP5]], 0
+; CHECK-NEXT:    [[SHIFT:%.*]] = shufflevector <4 x float> [[T]], <4 x float> poison, <4 x i32> <i32 1, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP6:%.*]] = fadd <4 x float> [[SHIFT]], [[T]]
+; CHECK-NEXT:    [[ADD:%.*]] = extractelement <4 x float> [[TMP6]], i64 0
+; CHECK-NEXT:    [[RETVAL_0:%.*]] = select i1 [[DOTNOT]], float [[ADD]], float 0.000000e+00
 ; CHECK-NEXT:    ret float [[RETVAL_0]]
 ;
 entry:
@@ -400,16 +401,18 @@ return:
 define float @test_merge_anyof_v4si(<4 x i32> %t) {
 ; CHECK-LABEL: @test_merge_anyof_v4si(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[T_FR:%.*]] = freeze <4 x i32> [[T:%.*]]
-; CHECK-NEXT:    [[TMP0:%.*]] = add <4 x i32> [[T_FR]], <i32 -256, i32 -256, i32 -256, i32 -256>
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult <4 x i32> [[TMP0]], <i32 -255, i32 -255, i32 -255, i32 -255>
-; CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x i1> [[TMP1]] to i4
-; CHECK-NEXT:    [[OP_RDX_NOT:%.*]] = icmp eq i4 [[TMP2]], 0
-; CHECK-NEXT:    [[SHIFT:%.*]] = shufflevector <4 x i32> [[T_FR]], <4 x i32> poison, <4 x i32> <i32 1, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <4 x i32> [[T_FR]], [[SHIFT]]
-; CHECK-NEXT:    [[ADD:%.*]] = extractelement <4 x i32> [[TMP3]], i64 0
+; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i32> [[T:%.*]], <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp sgt <8 x i32> [[TMP0]], <i32 255, i32 255, i32 255, i32 255, i32 1, i32 1, i32 1, i32 1>
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp slt <8 x i32> [[TMP0]], <i32 255, i32 255, i32 255, i32 255, i32 1, i32 1, i32 1, i32 1>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i1> [[TMP1]], <8 x i1> [[TMP2]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEXT:    [[TMP4:%.*]] = freeze <8 x i1> [[TMP3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i1> [[TMP4]] to i8
+; CHECK-NEXT:    [[DOTNOT:%.*]] = icmp eq i8 [[TMP5]], 0
+; CHECK-NEXT:    [[SHIFT:%.*]] = shufflevector <4 x i32> [[T]], <4 x i32> poison, <4 x i32> <i32 1, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <4 x i32> [[SHIFT]], [[T]]
+; CHECK-NEXT:    [[ADD:%.*]] = extractelement <4 x i32> [[TMP6]], i64 0
 ; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[ADD]] to float
-; CHECK-NEXT:    [[RETVAL_0:%.*]] = select i1 [[OP_RDX_NOT]], float [[CONV]], float 0.000000e+00
+; CHECK-NEXT:    [[RETVAL_0:%.*]] = select i1 [[DOTNOT]], float [[CONV]], float 0.000000e+00
 ; CHECK-NEXT:    ret float [[RETVAL_0]]
 ;
 entry:
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/reduction-logical.ll b/llvm/test/Transforms/SLPVectorizer/X86/reduction-logical.ll
index c25e07cfd451fc..865d8178667167 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/reduction-logical.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/reduction-logical.ll
@@ -388,28 +388,16 @@ define i1 @logical_and_icmp_clamp_partial(<4 x i32> %x) {
 }
 
 define i1 @logical_and_icmp_clamp_pred_diff(<4 x i32> %x) {
-; SSE-LABEL: @logical_and_icmp_clamp_pred_diff(
-; SSE-NEXT:    [[TMP1:%.*]] = icmp slt <4 x i32> [[X:%.*]], <i32 42, i32 42, i32 42, i32 42>
-; SSE-NEXT:    [[TMP2:%.*]] = icmp ult <4 x i32> [[X]], <i32 42, i32 42, i32 42, i32 42>
-; SSE-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i1> [[TMP1]], <4 x i1> [[TMP2]], <4 x i32> <i32 0, i32 1, i32 2, i32 7>
-; SSE-NEXT:    [[TMP4:%.*]] = icmp sgt <4 x i32> [[X]], <i32 17, i32 17, i32 17, i32 17>
-; SSE-NEXT:    [[TMP5:%.*]] = freeze <4 x i1> [[TMP4]]
-; SSE-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.and.v4i1(<4 x i1> [[TMP5]])
-; SSE-NEXT:    [[TMP7:%.*]] = freeze <4 x i1> [[TMP3]]
-; SSE-NEXT:    [[TMP8:%.*]] = call i1 @llvm.vector.reduce.and.v4i1(<4 x i1> [[TMP7]])
-; SSE-NEXT:    [[OP_RDX:%.*]] = select i1 [[TMP6]], i1 [[TMP8]], i1 false
-; SSE-NEXT:    ret i1 [[OP_RDX]]
-;
-; AVX-LABEL: @logical_and_icmp_clamp_pred_diff(
-; AVX-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X:%.*]], <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
-; AVX-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[TMP1]], <8 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 42, i32 42, i32 42, i32 poison>, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 12, i32 13, i32 14, i32 3>
-; AVX-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[TMP1]], <8 x i32> <i32 17, i32 17, i32 17, i32 17, i32 poison, i32 poison, i32 poison, i32 42>, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 0, i32 1, i32 2, i32 15>
-; AVX-NEXT:    [[TMP4:%.*]] = icmp sgt <8 x i32> [[TMP2]], [[TMP3]]
-; AVX-NEXT:    [[TMP5:%.*]] = icmp ult <8 x i32> [[TMP2]], [[TMP3]]
-; AVX-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i1> [[TMP4]], <8 x i1> [[TMP5]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 15>
-; AVX-NEXT:    [[TMP7:%.*]] = freeze <8 x i1> [[TMP6]]
-; AVX-NEXT:    [[TMP8:%.*]] = call i1 @llvm.vector.reduce.and.v8i1(<8 x i1> [[TMP7]])
-; AVX-NEXT:    ret i1 [[TMP8]]
+; CHECK-LABEL: @logical_and_icmp_clamp_pred_diff(
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X:%.*]], <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[TMP1]], <8 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 42, i32 42, i32 42, i32 poison>, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 12, i32 13, i32 14, i32 3>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[TMP1]], <8 x i32> <i32 17, i32 17, i32 17, i32 17, i32 poison, i32 poison, i32 poison, i32 42>, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 0, i32 1, i32 2, i32 15>
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp sgt <8 x i32> [[TMP2]], [[TMP3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp ult <8 x i32> [[TMP2]], [[TMP3]]
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i1> [[TMP4]], <8 x i1> [[TMP5]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 15>
+; CHECK-NEXT:    [[TMP7:%.*]] = freeze <8 x i1> [[TMP6]]
+; CHECK-NEXT:    [[TMP8:%.*]] = call i1 @llvm.vector.reduce.and.v8i1(<8 x i1> [[TMP7]])
+; CHECK-NEXT:    ret i1 [[TMP8]]
 ;
   %x0 = extractelement <4 x i32> %x, i32 0
   %x1 = extractelement <4 x i32> %x, i32 1

[RKSimon]

LGTM