[llvm] [SLP]Support Mul as the main copyable instruction (PR #207622)
via llvm-commits
llvm-commits at lists.llvm.org
Sun Jul 5 16:01:39 PDT 2026
llvmorg-github-actions[bot] wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-llvm-transforms
Author: Alexey Bataev (alexey-bataev)
<details>
<summary>Changes</summary>
Added support for the missed Mul as a main copyable instruction opcode
---
Patch is 53.76 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/207622.diff
11 Files Affected:
- (modified) llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp (+7-6)
- (modified) llvm/test/Transforms/SLPVectorizer/AArch64/externally-used-copyables.ll (+93-95)
- (modified) llvm/test/Transforms/SLPVectorizer/X86/buildvector-postpone-for-dependency.ll (+1-1)
- (modified) llvm/test/Transforms/SLPVectorizer/X86/copyable-operands-reordering.ll (+9-12)
- (modified) llvm/test/Transforms/SLPVectorizer/X86/gathered-node-with-in-order-parent.ll (+5-8)
- (modified) llvm/test/Transforms/SLPVectorizer/X86/non-schedulable-before-main.ll (+3-3)
- (modified) llvm/test/Transforms/SLPVectorizer/X86/reduction-with-removed-extracts.ll (+7-5)
- (modified) llvm/test/Transforms/SLPVectorizer/X86/vec_list_bias-inseltpoison.ll (+5-7)
- (modified) llvm/test/Transforms/SLPVectorizer/X86/vec_list_bias.ll (+5-7)
- (modified) llvm/test/Transforms/SLPVectorizer/X86/vec_list_bias_external_insert_shuffled.ll (+5-7)
- (modified) llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll (+3-12)
``````````diff
diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 92befc3c2ff9b..ea97a18403dc8 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -11893,12 +11893,12 @@ class InstructionsCompatibilityAnalysis {
/// elements.
static bool isSupportedOpcode(const unsigned Opcode) {
return Opcode == Instruction::Add || Opcode == Instruction::Sub ||
- Opcode == Instruction::LShr || Opcode == Instruction::Shl ||
- Opcode == Instruction::SDiv || Opcode == Instruction::UDiv ||
- Opcode == Instruction::And || Opcode == Instruction::Or ||
- Opcode == Instruction::Xor || Opcode == Instruction::FAdd ||
- Opcode == Instruction::FSub || Opcode == Instruction::FMul ||
- Opcode == Instruction::FDiv;
+ Opcode == Instruction::Mul || Opcode == Instruction::LShr ||
+ Opcode == Instruction::Shl || Opcode == Instruction::SDiv ||
+ Opcode == Instruction::UDiv || Opcode == Instruction::And ||
+ Opcode == Instruction::Or || Opcode == Instruction::Xor ||
+ Opcode == Instruction::FAdd || Opcode == Instruction::FSub ||
+ Opcode == Instruction::FMul || Opcode == Instruction::FDiv;
}
/// Identifies the best candidate value, which represents main opcode
@@ -12520,6 +12520,7 @@ class InstructionsCompatibilityAnalysis {
switch (MainOpcode) {
case Instruction::Add:
case Instruction::Sub:
+ case Instruction::Mul:
case Instruction::LShr:
case Instruction::Shl:
case Instruction::SDiv:
diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/externally-used-copyables.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/externally-used-copyables.ll
index b32d606b7d46a..2c9eea2571f07 100644
--- a/llvm/test/Transforms/SLPVectorizer/AArch64/externally-used-copyables.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/externally-used-copyables.ll
@@ -5,94 +5,84 @@ define void @test(i64 %0, i64 %1, i64 %2, i64 %3, i64 %.sroa.3341.0.copyload, i6
; CHECK-LABEL: define void @test(
; CHECK-SAME: i64 [[TMP0:%.*]], i64 [[TMP1:%.*]], i64 [[TMP2:%.*]], i64 [[TMP3:%.*]], i64 [[DOTSROA_3341_0_COPYLOAD:%.*]], i64 [[DOTSROA_3308_0_COPYLOAD:%.*]], i64 [[DOTNEG1:%.*]], i64 [[INDVAR3788:%.*]], i64 [[TMP4:%.*]], i64 [[TMP5:%.*]], i64 [[TMP6:%.*]], i64 [[TMP7:%.*]], i64 [[TMP8:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT: [[_LR_PH_PREHEADER:.*:]]
-; CHECK-NEXT: [[TMP10:%.*]] = mul i64 [[TMP0]], 40
+; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x i64> poison, i64 [[TMP1]], i32 0
+; CHECK-NEXT: [[TMP10:%.*]] = insertelement <2 x i64> [[TMP9]], i64 [[TMP0]], i32 1
+; CHECK-NEXT: [[TMP13:%.*]] = mul <2 x i64> [[TMP10]], <i64 1, i64 24>
+; CHECK-NEXT: [[TMP17:%.*]] = mul <2 x i64> [[TMP10]], <i64 1, i64 40>
; CHECK-NEXT: [[TMP11:%.*]] = mul i64 [[TMP0]], 48
+; CHECK-NEXT: [[TMP14:%.*]] = insertelement <2 x i64> <i64 poison, i64 1>, i64 [[TMP0]], i32 0
+; CHECK-NEXT: [[TMP15:%.*]] = shufflevector <2 x i64> [[TMP14]], <2 x i64> <i64 -1, i64 poison>, <2 x i32> <i32 2, i32 0>
+; CHECK-NEXT: [[TMP16:%.*]] = sub <2 x i64> [[TMP14]], [[TMP15]]
+; CHECK-NEXT: [[TMP20:%.*]] = shl i64 [[TMP0]], 11
; CHECK-NEXT: [[TMP35:%.*]] = insertelement <2 x i64> poison, i64 [[TMP0]], i32 0
; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <2 x i64> [[TMP35]], <2 x i64> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP13:%.*]] = add <2 x i64> [[TMP12]], <i64 0, i64 1>
-; CHECK-NEXT: [[TMP14:%.*]] = sub i64 1, [[TMP0]]
-; CHECK-NEXT: [[TMP20:%.*]] = shl i64 [[TMP0]], 11
+; CHECK-NEXT: [[TMP25:%.*]] = shl <2 x i64> [[TMP12]], <i64 11, i64 0>
; CHECK-NEXT: [[TMP21:%.*]] = sub i64 1, [[TMP20]]
-; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP20]], 8
+; CHECK-NEXT: [[TMP22:%.*]] = add <2 x i64> [[TMP25]], <i64 8, i64 1>
+; CHECK-NEXT: [[TMP23:%.*]] = or <2 x i64> [[TMP25]], <i64 8, i64 1>
+; CHECK-NEXT: [[TMP33:%.*]] = shufflevector <2 x i64> [[TMP22]], <2 x i64> [[TMP23]], <2 x i32> <i32 0, i32 3>
; CHECK-NEXT: [[TMP18:%.*]] = shl i64 [[TMP0]], 1
; CHECK-NEXT: [[TMP19:%.*]] = or i64 [[TMP18]], [[TMP0]]
-; CHECK-NEXT: [[TMP15:%.*]] = insertelement <2 x i64> [[TMP35]], i64 [[TMP19]], i32 1
-; CHECK-NEXT: [[TMP16:%.*]] = or <2 x i64> [[TMP15]], splat (i64 1)
-; CHECK-NEXT: [[TMP23:%.*]] = shufflevector <2 x i64> [[TMP12]], <2 x i64> <i64 24, i64 poison>, <2 x i32> <i32 2, i32 0>
-; CHECK-NEXT: [[TMP41:%.*]] = mul <2 x i64> [[TMP12]], [[TMP23]]
-; CHECK-NEXT: [[TMP44:%.*]] = insertelement <32 x i64> poison, i64 [[TMP1]], i32 5
-; CHECK-NEXT: [[TMP25:%.*]] = insertelement <16 x i64> poison, i64 [[TMP1]], i32 2
-; CHECK-NEXT: [[TMP31:%.*]] = insertelement <16 x i64> [[TMP25]], i64 [[TMP10]], i32 3
-; CHECK-NEXT: [[TMP27:%.*]] = shufflevector <2 x i64> [[TMP12]], <2 x i64> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP34:%.*]] = or i64 [[TMP19]], 1
+; CHECK-NEXT: [[TMP55:%.*]] = mul i64 [[TMP0]], [[TMP0]]
+; CHECK-NEXT: [[TMP58:%.*]] = insertelement <8 x i64> poison, i64 [[TMP0]], i32 0
+; CHECK-NEXT: [[TMP30:%.*]] = shufflevector <8 x i64> [[TMP58]], <8 x i64> poison, <8 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP31:%.*]] = shufflevector <8 x i64> [[TMP30]], <8 x i64> <i64 poison, i64 1, i64 1, i64 1, i64 poison, i64 poison, i64 poison, i64 poison>, <6 x i32> <i32 0, i32 9, i32 10, i32 11, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP32:%.*]] = shufflevector <8 x i64> [[TMP30]], <8 x i64> <i64 0, i64 0, i64 0, i64 0, i64 poison, i64 1, i64 1, i64 1>, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 0, i32 13, i32 14, i32 15>
+; CHECK-NEXT: [[TMP27:%.*]] = shufflevector <8 x i64> [[TMP30]], <8 x i64> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT: [[TMP28:%.*]] = insertelement <3 x i64> [[TMP27]], i64 [[TMP1]], i32 1
; CHECK-NEXT: [[TMP29:%.*]] = insertelement <3 x i64> [[TMP28]], i64 [[DOTNEG1]], i32 2
-; CHECK-NEXT: [[TMP30:%.*]] = shufflevector <3 x i64> [[TMP29]], <3 x i64> poison, <8 x i32> <i32 0, i32 0, i32 1, i32 1, i32 1, i32 2, i32 0, i32 0>
-; CHECK-NEXT: [[TMP85:%.*]] = shufflevector <2 x i64> [[TMP15]], <2 x i64> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP32:%.*]] = insertelement <4 x i64> poison, i64 [[TMP1]], i32 1
-; CHECK-NEXT: [[TMP33:%.*]] = insertelement <2 x i64> [[TMP15]], i64 [[TMP2]], i32 1
-; CHECK-NEXT: [[TMP34:%.*]] = shufflevector <2 x i64> [[TMP33]], <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 0>
-; CHECK-NEXT: [[TMP40:%.*]] = insertelement <2 x i64> poison, i64 [[TMP21]], i32 0
-; CHECK-NEXT: [[TMP43:%.*]] = insertelement <2 x i64> [[TMP40]], i64 [[TMP17]], i32 1
-; CHECK-NEXT: [[TMP37:%.*]] = insertelement <2 x i64> <i64 poison, i64 1>, i64 [[TMP0]], i32 0
-; CHECK-NEXT: [[TMP38:%.*]] = shufflevector <2 x i64> [[TMP41]], <2 x i64> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP74:%.*]] = shufflevector <32 x i64> [[TMP44]], <32 x i64> [[TMP38]], <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 5, i32 poison, i32 poison, i32 poison, i32 poison, i32 32, i32 33, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP36:%.*]] = shufflevector <3 x i64> [[TMP29]], <3 x i64> poison, <8 x i32> <i32 0, i32 1, i32 1, i32 1, i32 2, i32 0, i32 0, i32 0>
+; CHECK-NEXT: [[TMP37:%.*]] = insertelement <4 x i64> poison, i64 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP38:%.*]] = insertelement <4 x i64> [[TMP37]], i64 [[TMP8]], i32 1
+; CHECK-NEXT: [[TMP39:%.*]] = insertelement <4 x i64> [[TMP38]], i64 [[DOTSROA_3308_0_COPYLOAD]], i32 2
+; CHECK-NEXT: [[TMP40:%.*]] = insertelement <4 x i64> [[TMP39]], i64 [[TMP0]], i32 3
+; CHECK-NEXT: [[TMP41:%.*]] = shufflevector <2 x i64> [[TMP17]], <2 x i64> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP74:%.*]] = insertelement <4 x i64> poison, i64 [[TMP0]], i32 1
+; CHECK-NEXT: [[TMP43:%.*]] = mul i64 [[TMP0]], [[TMP0]]
+; CHECK-NEXT: [[TMP44:%.*]] = shufflevector <2 x i64> [[TMP13]], <2 x i64> poison, <32 x i32> <i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP45:%.*]] = shufflevector <32 x i64> [[TMP44]], <32 x i64> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 5, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP46:%.*]] = shufflevector <2 x i64> [[TMP33]], <2 x i64> poison, <6 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP47:%.*]] = shufflevector <6 x i64> [[TMP31]], <6 x i64> [[TMP46]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 6, i32 7>
+; CHECK-NEXT: [[TMP48:%.*]] = shufflevector <2 x i64> [[TMP16]], <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; CHECK-NEXT: br label %[[DOTLR_PH1977_US:.*]]
; CHECK: [[_LR_PH1977_US:.*:]]
; CHECK-NEXT: [[INDVAR37888:%.*]] = phi i64 [ 0, [[DOTLR_PH_PREHEADER:%.*]] ], [ 1, %[[DOTLR_PH1977_US]] ]
-; CHECK-NEXT: [[TMP36:%.*]] = mul <2 x i64> [[TMP16]], [[TMP85]]
-; CHECK-NEXT: [[TMP73:%.*]] = mul <2 x i64> [[TMP43]], [[TMP12]]
+; CHECK-NEXT: [[TMP49:%.*]] = shufflevector <6 x i64> [[TMP47]], <6 x i64> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 5, i32 5>
+; CHECK-NEXT: [[TMP50:%.*]] = mul <8 x i64> [[TMP30]], [[TMP49]]
+; CHECK-NEXT: [[TMP51:%.*]] = or <8 x i64> [[TMP30]], [[TMP49]]
+; CHECK-NEXT: [[TMP52:%.*]] = shufflevector <8 x i64> [[TMP50]], <8 x i64> [[TMP51]], <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT: [[TMP53:%.*]] = mul i64 [[TMP34]], [[TMP0]]
+; CHECK-NEXT: [[TMP54:%.*]] = mul i64 [[TMP21]], [[TMP0]]
; CHECK-NEXT: [[TMP26:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-NEXT: [[DIFF_CHECK3783:%.*]] = icmp ult i64 [[TMP11]], [[TMP26]]
-; CHECK-NEXT: [[TMP39:%.*]] = mul i64 [[TMP0]], [[TMP0]]
-; CHECK-NEXT: [[TMP90:%.*]] = or <2 x i64> [[TMP12]], splat (i64 1)
+; CHECK-NEXT: [[DIFF_CHECK3790:%.*]] = icmp ult i64 [[INDVAR37888]], [[TMP0]]
; CHECK-NEXT: [[DIFF_CHECK3805:%.*]] = icmp ugt i64 [[TMP26]], 1
-; CHECK-NEXT: [[TMP45:%.*]] = shufflevector <2 x i64> [[TMP36]], <2 x i64> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP46:%.*]] = add <2 x i64> [[TMP45]], splat (i64 1)
-; CHECK-NEXT: [[TMP47:%.*]] = insertelement <8 x i64> poison, i64 [[INDVAR37888]], i32 0
-; CHECK-NEXT: [[TMP91:%.*]] = insertelement <8 x i64> [[TMP47]], i64 [[TMP39]], i32 1
-; CHECK-NEXT: [[TMP92:%.*]] = shufflevector <2 x i64> [[TMP46]], <2 x i64> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP93:%.*]] = shufflevector <2 x i64> [[TMP90]], <2 x i64> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP24:%.*]] = extractelement <2 x i64> [[TMP36]], i32 0
+; CHECK-NEXT: [[TMP56:%.*]] = add <8 x i64> [[TMP52]], [[TMP32]]
+; CHECK-NEXT: [[TMP57:%.*]] = icmp ult <8 x i64> [[TMP56]], [[TMP36]]
+; CHECK-NEXT: [[TMP24:%.*]] = extractelement <8 x i64> [[TMP52]], i32 5
; CHECK-NEXT: [[TMP42:%.*]] = add i64 [[TMP24]], 1
; CHECK-NEXT: [[DIFF_CHECK3842:%.*]] = icmp ult i64 [[TMP42]], [[TMP0]]
-; CHECK-NEXT: [[TMP94:%.*]] = add <2 x i64> [[TMP36]], splat (i64 1)
-; CHECK-NEXT: [[TMP95:%.*]] = shufflevector <2 x i64> [[TMP36]], <2 x i64> poison, <4 x i32> <i32 0, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP55:%.*]] = shufflevector <2 x i64> [[TMP36]], <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP56:%.*]] = shufflevector <4 x i64> [[TMP32]], <4 x i64> [[TMP55]], <4 x i32> <i32 poison, i32 1, i32 poison, i32 4>
-; CHECK-NEXT: [[TMP57:%.*]] = shufflevector <2 x i64> [[TMP94]], <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP96:%.*]] = shufflevector <4 x i64> [[TMP57]], <4 x i64> [[TMP56]], <4 x i32> <i32 0, i32 5, i32 1, i32 7>
-; CHECK-NEXT: [[TMP97:%.*]] = icmp ult <4 x i64> [[TMP96]], [[TMP34]]
+; CHECK-NEXT: [[DIFF_CHECK3817:%.*]] = icmp ult i64 [[TMP1]], [[TMP2]]
+; CHECK-NEXT: [[TMP76:%.*]] = add i64 [[TMP53]], 1
+; CHECK-NEXT: [[DIFF_CHECK3820:%.*]] = icmp ult i64 [[TMP76]], [[TMP0]]
; CHECK-NEXT: [[TMP98:%.*]] = shl <2 x i64> [[TMP12]], splat (i64 1)
-; CHECK-NEXT: [[TMP48:%.*]] = mul <2 x i64> [[TMP12]], [[TMP13]]
-; CHECK-NEXT: [[TMP99:%.*]] = shufflevector <2 x i64> [[TMP73]], <2 x i64> poison, <2 x i32> <i32 poison, i32 0>
-; CHECK-NEXT: [[TMP100:%.*]] = insertelement <2 x i64> [[TMP99]], i64 [[TMP14]], i32 0
-; CHECK-NEXT: [[TMP101:%.*]] = mul <2 x i64> [[TMP100]], [[TMP37]]
-; CHECK-NEXT: [[TMP102:%.*]] = add <2 x i64> [[TMP100]], [[TMP37]]
-; CHECK-NEXT: [[TMP49:%.*]] = shufflevector <2 x i64> [[TMP101]], <2 x i64> [[TMP102]], <2 x i32> <i32 0, i32 3>
-; CHECK-NEXT: [[TMP50:%.*]] = insertelement <32 x i64> [[TMP74]], i64 [[INDVAR37888]], i32 0
-; CHECK-NEXT: [[TMP51:%.*]] = shufflevector <2 x i64> [[TMP48]], <2 x i64> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP52:%.*]] = shufflevector <32 x i64> [[TMP50]], <32 x i64> [[TMP51]], <32 x i32> <i32 0, i32 32, i32 33, i32 poison, i32 poison, i32 5, i32 poison, i32 poison, i32 poison, i32 poison, i32 10, i32 11, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP53:%.*]] = shufflevector <2 x i64> [[TMP49]], <2 x i64> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP54:%.*]] = shufflevector <32 x i64> [[TMP52]], <32 x i64> [[TMP53]], <32 x i32> <i32 0, i32 1, i32 2, i32 32, i32 33, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; CHECK-NEXT: [[TMP103:%.*]] = shufflevector <2 x i64> [[TMP36]], <2 x i64> poison, <32 x i32> <i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT: [[TMP58:%.*]] = shufflevector <2 x i64> [[TMP36]], <2 x i64> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP77:%.*]] = insertelement <4 x i64> [[TMP74]], i64 [[INDVAR37888]], i32 0
+; CHECK-NEXT: [[TMP78:%.*]] = shufflevector <4 x i64> [[TMP77]], <4 x i64> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 1>
+; CHECK-NEXT: [[TMP80:%.*]] = shufflevector <4 x i64> [[TMP77]], <4 x i64> <i64 1, i64 poison, i64 poison, i64 poison>, <4 x i32> <i32 4, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP81:%.*]] = shufflevector <4 x i64> [[TMP80]], <4 x i64> [[TMP48]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT: [[TMP82:%.*]] = mul <4 x i64> [[TMP78]], [[TMP81]]
+; CHECK-NEXT: [[TMP84:%.*]] = add i64 [[TMP54]], 1
+; CHECK-NEXT: [[TMP85:%.*]] = insertelement <32 x i64> [[TMP45]], i64 [[TMP55]], i32 11
+; CHECK-NEXT: [[TMP86:%.*]] = shufflevector <4 x i64> [[TMP82]], <4 x i64> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP89:%.*]] = shufflevector <32 x i64> [[TMP85]], <32 x i64> [[TMP86]], <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 poison, i32 5, i32 poison, i32 poison, i32 poison, i32 poison, i32 10, i32 11, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP90:%.*]] = insertelement <32 x i64> [[TMP89]], i64 [[TMP84]], i32 4
+; CHECK-NEXT: [[TMP100:%.*]] = shufflevector <8 x i64> [[TMP52]], <8 x i64> poison, <32 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 4, i32 5, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP73:%.*]] = shufflevector <8 x i64> [[TMP52]], <8 x i64> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT: [[TMP87:%.*]] = shufflevector <32 x i64> [[TMP90]], <32 x i64> [[TMP73]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 36, i32 37, i32 poison, i32 poison, i32 10, i32 11, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; CHECK-NEXT: [[TMP125:%.*]] = shufflevector <2 x i64> [[TMP98]], <2 x i64> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison,...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/207622
More information about the llvm-commits
mailing list