[llvm] e6ab86a - [Matrix] Fix IsSupported check in lowerDotProduct.
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Thu Apr 13 11:00:44 PDT 2023
Author: Florian Hahn
Date: 2023-04-13T19:00:30+01:00
New Revision: e6ab86a88796667ede83bf839a8e1bdb103d4a80
URL: https://github.com/llvm/llvm-project/commit/e6ab86a88796667ede83bf839a8e1bdb103d4a80
DIFF: https://github.com/llvm/llvm-project/commit/e6ab86a88796667ede83bf839a8e1bdb103d4a80.diff
LOG: [Matrix] Fix IsSupported check in lowerDotProduct.
The check incorrectly checks the RHS while LHS is transformed later.
Update to check LHS, which fixes a crash in the newly added test cases.
Added:
Modified:
llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
llvm/test/Transforms/LowerMatrixIntrinsics/dot-product-int.ll
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp b/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
index 82e717fedb704..0c5d7eed16234 100644
--- a/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
+++ b/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
@@ -1344,12 +1344,13 @@ class LowerMatrixIntrinsics {
auto IsSupportedArg = [](Value *Op, unsigned N) {
if (!isa<Instruction>(Op))
return true;
- return match(Op, m_OneUse(m_CombineOr(
- m_Load(m_Value()),
- m_Intrinsic<Intrinsic::matrix_column_major_load>(
- m_Value(), m_SpecificInt(N)))));
+ return match(
+ Op, m_OneUse(m_CombineOr(
+ m_Load(m_Value()),
+ m_Intrinsic<Intrinsic::matrix_column_major_load>(
+ m_Value(), m_SpecificInt(N)))));
};
- if (!IsSupportedArg(RHS, RShape.NumColumns))
+ if (!IsSupportedArg(LHS, RShape.NumColumns))
return;
// We compare the costs of a vector.reduce.add to sequential add.
diff --git a/llvm/test/Transforms/LowerMatrixIntrinsics/dot-product-int.ll b/llvm/test/Transforms/LowerMatrixIntrinsics/dot-product-int.ll
index dd5c4cf3f4b4f..5a00e54972ba1 100644
--- a/llvm/test/Transforms/LowerMatrixIntrinsics/dot-product-int.ll
+++ b/llvm/test/Transforms/LowerMatrixIntrinsics/dot-product-int.ll
@@ -75,17 +75,57 @@ define <1 x i32> @add_feeding_dotproduct_i32_v8_1(<8 x i32> %a, <8 x i32> %b, <8
; CHECK-NEXT: [[TMP5:%.*]] = add <1 x i32> [[SPLIT5]], [[SPLIT13]]
; CHECK-NEXT: [[TMP6:%.*]] = add <1 x i32> [[SPLIT6]], [[SPLIT14]]
; CHECK-NEXT: [[TMP7:%.*]] = add <1 x i32> [[SPLIT7]], [[SPLIT15]]
-; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <1 x i32> [[TMP0]], <1 x i32> [[TMP1]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <1 x i32> [[TMP2]], <1 x i32> [[TMP3]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <1 x i32> [[TMP4]], <1 x i32> [[TMP5]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <1 x i32> [[TMP6]], <1 x i32> [[TMP7]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <2 x i32> [[TMP8]], <2 x i32> [[TMP9]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT: [[TMP14:%.*]] = shufflevector <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT: [[TMP15:%.*]] = mul <8 x i32> [[TMP14]], [[C:%.*]]
-; CHECK-NEXT: [[TMP16:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP15]])
-; CHECK-NEXT: [[TMP17:%.*]] = insertelement <1 x i32> poison, i32 [[TMP16]], i64 0
-; CHECK-NEXT: ret <1 x i32> [[TMP17]]
+; CHECK-NEXT: [[SPLIT16:%.*]] = shufflevector <8 x i32> [[C:%.*]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <1 x i32> [[TMP0]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP8:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i32> poison, i32 [[TMP8]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP9:%.*]] = mul <1 x i32> [[BLOCK]], [[SPLAT_SPLAT]]
+; CHECK-NEXT: [[BLOCK17:%.*]] = shufflevector <1 x i32> [[TMP1]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP10:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 1
+; CHECK-NEXT: [[SPLAT_SPLATINSERT18:%.*]] = insertelement <1 x i32> poison, i32 [[TMP10]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT19:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT18]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP11:%.*]] = mul <1 x i32> [[BLOCK17]], [[SPLAT_SPLAT19]]
+; CHECK-NEXT: [[TMP12:%.*]] = add <1 x i32> [[TMP9]], [[TMP11]]
+; CHECK-NEXT: [[BLOCK20:%.*]] = shufflevector <1 x i32> [[TMP2]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP13:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 2
+; CHECK-NEXT: [[SPLAT_SPLATINSERT21:%.*]] = insertelement <1 x i32> poison, i32 [[TMP13]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT22:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT21]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP14:%.*]] = mul <1 x i32> [[BLOCK20]], [[SPLAT_SPLAT22]]
+; CHECK-NEXT: [[TMP15:%.*]] = add <1 x i32> [[TMP12]], [[TMP14]]
+; CHECK-NEXT: [[BLOCK23:%.*]] = shufflevector <1 x i32> [[TMP3]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP16:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 3
+; CHECK-NEXT: [[SPLAT_SPLATINSERT24:%.*]] = insertelement <1 x i32> poison, i32 [[TMP16]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT25:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT24]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP17:%.*]] = mul <1 x i32> [[BLOCK23]], [[SPLAT_SPLAT25]]
+; CHECK-NEXT: [[TMP18:%.*]] = add <1 x i32> [[TMP15]], [[TMP17]]
+; CHECK-NEXT: [[BLOCK26:%.*]] = shufflevector <1 x i32> [[TMP4]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP19:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 4
+; CHECK-NEXT: [[SPLAT_SPLATINSERT27:%.*]] = insertelement <1 x i32> poison, i32 [[TMP19]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT28:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT27]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP20:%.*]] = mul <1 x i32> [[BLOCK26]], [[SPLAT_SPLAT28]]
+; CHECK-NEXT: [[TMP21:%.*]] = add <1 x i32> [[TMP18]], [[TMP20]]
+; CHECK-NEXT: [[BLOCK29:%.*]] = shufflevector <1 x i32> [[TMP5]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP22:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 5
+; CHECK-NEXT: [[SPLAT_SPLATINSERT30:%.*]] = insertelement <1 x i32> poison, i32 [[TMP22]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT31:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT30]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP23:%.*]] = mul <1 x i32> [[BLOCK29]], [[SPLAT_SPLAT31]]
+; CHECK-NEXT: [[TMP24:%.*]] = add <1 x i32> [[TMP21]], [[TMP23]]
+; CHECK-NEXT: [[BLOCK32:%.*]] = shufflevector <1 x i32> [[TMP6]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP25:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 6
+; CHECK-NEXT: [[SPLAT_SPLATINSERT33:%.*]] = insertelement <1 x i32> poison, i32 [[TMP25]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT34:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT33]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP26:%.*]] = mul <1 x i32> [[BLOCK32]], [[SPLAT_SPLAT34]]
+; CHECK-NEXT: [[TMP27:%.*]] = add <1 x i32> [[TMP24]], [[TMP26]]
+; CHECK-NEXT: [[BLOCK35:%.*]] = shufflevector <1 x i32> [[TMP7]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP28:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 7
+; CHECK-NEXT: [[SPLAT_SPLATINSERT36:%.*]] = insertelement <1 x i32> poison, i32 [[TMP28]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT37:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT36]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP29:%.*]] = mul <1 x i32> [[BLOCK35]], [[SPLAT_SPLAT37]]
+; CHECK-NEXT: [[TMP30:%.*]] = add <1 x i32> [[TMP27]], [[TMP29]]
+; CHECK-NEXT: [[TMP31:%.*]] = shufflevector <1 x i32> [[TMP30]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP32:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP31]], <1 x i32> <i32 1>
+; CHECK-NEXT: ret <1 x i32> [[TMP32]]
;
entry:
%add = add <8 x i32> %a, %b
@@ -99,64 +139,10 @@ define <1 x i32> @add_feeding_dotproduct_i32_v8_2(<8 x i32> %a, <8 x i32> %b, <8
; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <8 x i32> [[A:%.*]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT: [[SPLIT1:%.*]] = shufflevector <8 x i32> [[B:%.*]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT: [[TMP0:%.*]] = add <8 x i32> [[SPLIT]], [[SPLIT1]]
-; CHECK-NEXT: [[SPLIT2:%.*]] = shufflevector <8 x i32> [[C:%.*]], <8 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[SPLIT3:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 1>
-; CHECK-NEXT: [[SPLIT4:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 2>
-; CHECK-NEXT: [[SPLIT5:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 3>
-; CHECK-NEXT: [[SPLIT6:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 4>
-; CHECK-NEXT: [[SPLIT7:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 5>
-; CHECK-NEXT: [[SPLIT8:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 6>
-; CHECK-NEXT: [[SPLIT9:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 7>
-; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <1 x i32> [[SPLIT2]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP1:%.*]] = extractelement <8 x i32> [[TMP0]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i32> poison, i32 [[TMP1]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP2:%.*]] = mul <1 x i32> [[BLOCK]], [[SPLAT_SPLAT]]
-; CHECK-NEXT: [[BLOCK10:%.*]] = shufflevector <1 x i32> [[SPLIT3]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP3:%.*]] = extractelement <8 x i32> [[TMP0]], i64 1
-; CHECK-NEXT: [[SPLAT_SPLATINSERT11:%.*]] = insertelement <1 x i32> poison, i32 [[TMP3]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT12:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT11]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP4:%.*]] = mul <1 x i32> [[BLOCK10]], [[SPLAT_SPLAT12]]
-; CHECK-NEXT: [[TMP5:%.*]] = add <1 x i32> [[TMP2]], [[TMP4]]
-; CHECK-NEXT: [[BLOCK13:%.*]] = shufflevector <1 x i32> [[SPLIT4]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP6:%.*]] = extractelement <8 x i32> [[TMP0]], i64 2
-; CHECK-NEXT: [[SPLAT_SPLATINSERT14:%.*]] = insertelement <1 x i32> poison, i32 [[TMP6]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT15:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT14]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP7:%.*]] = mul <1 x i32> [[BLOCK13]], [[SPLAT_SPLAT15]]
-; CHECK-NEXT: [[TMP8:%.*]] = add <1 x i32> [[TMP5]], [[TMP7]]
-; CHECK-NEXT: [[BLOCK16:%.*]] = shufflevector <1 x i32> [[SPLIT5]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP9:%.*]] = extractelement <8 x i32> [[TMP0]], i64 3
-; CHECK-NEXT: [[SPLAT_SPLATINSERT17:%.*]] = insertelement <1 x i32> poison, i32 [[TMP9]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT18:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT17]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP10:%.*]] = mul <1 x i32> [[BLOCK16]], [[SPLAT_SPLAT18]]
-; CHECK-NEXT: [[TMP11:%.*]] = add <1 x i32> [[TMP8]], [[TMP10]]
-; CHECK-NEXT: [[BLOCK19:%.*]] = shufflevector <1 x i32> [[SPLIT6]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP12:%.*]] = extractelement <8 x i32> [[TMP0]], i64 4
-; CHECK-NEXT: [[SPLAT_SPLATINSERT20:%.*]] = insertelement <1 x i32> poison, i32 [[TMP12]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT21:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT20]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP13:%.*]] = mul <1 x i32> [[BLOCK19]], [[SPLAT_SPLAT21]]
-; CHECK-NEXT: [[TMP14:%.*]] = add <1 x i32> [[TMP11]], [[TMP13]]
-; CHECK-NEXT: [[BLOCK22:%.*]] = shufflevector <1 x i32> [[SPLIT7]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP15:%.*]] = extractelement <8 x i32> [[TMP0]], i64 5
-; CHECK-NEXT: [[SPLAT_SPLATINSERT23:%.*]] = insertelement <1 x i32> poison, i32 [[TMP15]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT24:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT23]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP16:%.*]] = mul <1 x i32> [[BLOCK22]], [[SPLAT_SPLAT24]]
-; CHECK-NEXT: [[TMP17:%.*]] = add <1 x i32> [[TMP14]], [[TMP16]]
-; CHECK-NEXT: [[BLOCK25:%.*]] = shufflevector <1 x i32> [[SPLIT8]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP18:%.*]] = extractelement <8 x i32> [[TMP0]], i64 6
-; CHECK-NEXT: [[SPLAT_SPLATINSERT26:%.*]] = insertelement <1 x i32> poison, i32 [[TMP18]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT27:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT26]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP19:%.*]] = mul <1 x i32> [[BLOCK25]], [[SPLAT_SPLAT27]]
-; CHECK-NEXT: [[TMP20:%.*]] = add <1 x i32> [[TMP17]], [[TMP19]]
-; CHECK-NEXT: [[BLOCK28:%.*]] = shufflevector <1 x i32> [[SPLIT9]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP21:%.*]] = extractelement <8 x i32> [[TMP0]], i64 7
-; CHECK-NEXT: [[SPLAT_SPLATINSERT29:%.*]] = insertelement <1 x i32> poison, i32 [[TMP21]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT30:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT29]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP22:%.*]] = mul <1 x i32> [[BLOCK28]], [[SPLAT_SPLAT30]]
-; CHECK-NEXT: [[TMP23:%.*]] = add <1 x i32> [[TMP20]], [[TMP22]]
-; CHECK-NEXT: [[TMP24:%.*]] = shufflevector <1 x i32> [[TMP23]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP25:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP24]], <1 x i32> <i32 1>
-; CHECK-NEXT: ret <1 x i32> [[TMP25]]
+; CHECK-NEXT: [[TMP1:%.*]] = mul <8 x i32> [[C:%.*]], [[TMP0]]
+; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP1]])
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <1 x i32> poison, i32 [[TMP2]], i64 0
+; CHECK-NEXT: ret <1 x i32> [[TMP3]]
;
entry:
%add = add <8 x i32> %a, %b
@@ -191,17 +177,57 @@ define <1 x i32> @sub_feeding_dotproduct_i32_v8_1(<8 x i32> %a, <8 x i32> %b, <8
; CHECK-NEXT: [[TMP5:%.*]] = sub <1 x i32> [[SPLIT5]], [[SPLIT13]]
; CHECK-NEXT: [[TMP6:%.*]] = sub <1 x i32> [[SPLIT6]], [[SPLIT14]]
; CHECK-NEXT: [[TMP7:%.*]] = sub <1 x i32> [[SPLIT7]], [[SPLIT15]]
-; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <1 x i32> [[TMP0]], <1 x i32> [[TMP1]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <1 x i32> [[TMP2]], <1 x i32> [[TMP3]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <1 x i32> [[TMP4]], <1 x i32> [[TMP5]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <1 x i32> [[TMP6]], <1 x i32> [[TMP7]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <2 x i32> [[TMP8]], <2 x i32> [[TMP9]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT: [[TMP14:%.*]] = shufflevector <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT: [[TMP15:%.*]] = mul <8 x i32> [[TMP14]], [[C:%.*]]
-; CHECK-NEXT: [[TMP16:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP15]])
-; CHECK-NEXT: [[TMP17:%.*]] = insertelement <1 x i32> poison, i32 [[TMP16]], i64 0
-; CHECK-NEXT: ret <1 x i32> [[TMP17]]
+; CHECK-NEXT: [[SPLIT16:%.*]] = shufflevector <8 x i32> [[C:%.*]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <1 x i32> [[TMP0]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP8:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i32> poison, i32 [[TMP8]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP9:%.*]] = mul <1 x i32> [[BLOCK]], [[SPLAT_SPLAT]]
+; CHECK-NEXT: [[BLOCK17:%.*]] = shufflevector <1 x i32> [[TMP1]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP10:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 1
+; CHECK-NEXT: [[SPLAT_SPLATINSERT18:%.*]] = insertelement <1 x i32> poison, i32 [[TMP10]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT19:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT18]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP11:%.*]] = mul <1 x i32> [[BLOCK17]], [[SPLAT_SPLAT19]]
+; CHECK-NEXT: [[TMP12:%.*]] = add <1 x i32> [[TMP9]], [[TMP11]]
+; CHECK-NEXT: [[BLOCK20:%.*]] = shufflevector <1 x i32> [[TMP2]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP13:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 2
+; CHECK-NEXT: [[SPLAT_SPLATINSERT21:%.*]] = insertelement <1 x i32> poison, i32 [[TMP13]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT22:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT21]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP14:%.*]] = mul <1 x i32> [[BLOCK20]], [[SPLAT_SPLAT22]]
+; CHECK-NEXT: [[TMP15:%.*]] = add <1 x i32> [[TMP12]], [[TMP14]]
+; CHECK-NEXT: [[BLOCK23:%.*]] = shufflevector <1 x i32> [[TMP3]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP16:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 3
+; CHECK-NEXT: [[SPLAT_SPLATINSERT24:%.*]] = insertelement <1 x i32> poison, i32 [[TMP16]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT25:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT24]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP17:%.*]] = mul <1 x i32> [[BLOCK23]], [[SPLAT_SPLAT25]]
+; CHECK-NEXT: [[TMP18:%.*]] = add <1 x i32> [[TMP15]], [[TMP17]]
+; CHECK-NEXT: [[BLOCK26:%.*]] = shufflevector <1 x i32> [[TMP4]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP19:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 4
+; CHECK-NEXT: [[SPLAT_SPLATINSERT27:%.*]] = insertelement <1 x i32> poison, i32 [[TMP19]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT28:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT27]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP20:%.*]] = mul <1 x i32> [[BLOCK26]], [[SPLAT_SPLAT28]]
+; CHECK-NEXT: [[TMP21:%.*]] = add <1 x i32> [[TMP18]], [[TMP20]]
+; CHECK-NEXT: [[BLOCK29:%.*]] = shufflevector <1 x i32> [[TMP5]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP22:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 5
+; CHECK-NEXT: [[SPLAT_SPLATINSERT30:%.*]] = insertelement <1 x i32> poison, i32 [[TMP22]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT31:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT30]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP23:%.*]] = mul <1 x i32> [[BLOCK29]], [[SPLAT_SPLAT31]]
+; CHECK-NEXT: [[TMP24:%.*]] = add <1 x i32> [[TMP21]], [[TMP23]]
+; CHECK-NEXT: [[BLOCK32:%.*]] = shufflevector <1 x i32> [[TMP6]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP25:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 6
+; CHECK-NEXT: [[SPLAT_SPLATINSERT33:%.*]] = insertelement <1 x i32> poison, i32 [[TMP25]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT34:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT33]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP26:%.*]] = mul <1 x i32> [[BLOCK32]], [[SPLAT_SPLAT34]]
+; CHECK-NEXT: [[TMP27:%.*]] = add <1 x i32> [[TMP24]], [[TMP26]]
+; CHECK-NEXT: [[BLOCK35:%.*]] = shufflevector <1 x i32> [[TMP7]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP28:%.*]] = extractelement <8 x i32> [[SPLIT16]], i64 7
+; CHECK-NEXT: [[SPLAT_SPLATINSERT36:%.*]] = insertelement <1 x i32> poison, i32 [[TMP28]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT37:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT36]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP29:%.*]] = mul <1 x i32> [[BLOCK35]], [[SPLAT_SPLAT37]]
+; CHECK-NEXT: [[TMP30:%.*]] = add <1 x i32> [[TMP27]], [[TMP29]]
+; CHECK-NEXT: [[TMP31:%.*]] = shufflevector <1 x i32> [[TMP30]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP32:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP31]], <1 x i32> <i32 1>
+; CHECK-NEXT: ret <1 x i32> [[TMP32]]
;
entry:
%sub = sub <8 x i32> %a, %b
@@ -215,64 +241,10 @@ define <1 x i32> @sub_feeding_dotproduct_i32_v8_2(<8 x i32> %a, <8 x i32> %b, <8
; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <8 x i32> [[A:%.*]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT: [[SPLIT1:%.*]] = shufflevector <8 x i32> [[B:%.*]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT: [[TMP0:%.*]] = sub <8 x i32> [[SPLIT]], [[SPLIT1]]
-; CHECK-NEXT: [[SPLIT2:%.*]] = shufflevector <8 x i32> [[C:%.*]], <8 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[SPLIT3:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 1>
-; CHECK-NEXT: [[SPLIT4:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 2>
-; CHECK-NEXT: [[SPLIT5:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 3>
-; CHECK-NEXT: [[SPLIT6:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 4>
-; CHECK-NEXT: [[SPLIT7:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 5>
-; CHECK-NEXT: [[SPLIT8:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 6>
-; CHECK-NEXT: [[SPLIT9:%.*]] = shufflevector <8 x i32> [[C]], <8 x i32> poison, <1 x i32> <i32 7>
-; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <1 x i32> [[SPLIT2]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP1:%.*]] = extractelement <8 x i32> [[TMP0]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i32> poison, i32 [[TMP1]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP2:%.*]] = mul <1 x i32> [[BLOCK]], [[SPLAT_SPLAT]]
-; CHECK-NEXT: [[BLOCK10:%.*]] = shufflevector <1 x i32> [[SPLIT3]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP3:%.*]] = extractelement <8 x i32> [[TMP0]], i64 1
-; CHECK-NEXT: [[SPLAT_SPLATINSERT11:%.*]] = insertelement <1 x i32> poison, i32 [[TMP3]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT12:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT11]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP4:%.*]] = mul <1 x i32> [[BLOCK10]], [[SPLAT_SPLAT12]]
-; CHECK-NEXT: [[TMP5:%.*]] = add <1 x i32> [[TMP2]], [[TMP4]]
-; CHECK-NEXT: [[BLOCK13:%.*]] = shufflevector <1 x i32> [[SPLIT4]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP6:%.*]] = extractelement <8 x i32> [[TMP0]], i64 2
-; CHECK-NEXT: [[SPLAT_SPLATINSERT14:%.*]] = insertelement <1 x i32> poison, i32 [[TMP6]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT15:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT14]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP7:%.*]] = mul <1 x i32> [[BLOCK13]], [[SPLAT_SPLAT15]]
-; CHECK-NEXT: [[TMP8:%.*]] = add <1 x i32> [[TMP5]], [[TMP7]]
-; CHECK-NEXT: [[BLOCK16:%.*]] = shufflevector <1 x i32> [[SPLIT5]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP9:%.*]] = extractelement <8 x i32> [[TMP0]], i64 3
-; CHECK-NEXT: [[SPLAT_SPLATINSERT17:%.*]] = insertelement <1 x i32> poison, i32 [[TMP9]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT18:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT17]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP10:%.*]] = mul <1 x i32> [[BLOCK16]], [[SPLAT_SPLAT18]]
-; CHECK-NEXT: [[TMP11:%.*]] = add <1 x i32> [[TMP8]], [[TMP10]]
-; CHECK-NEXT: [[BLOCK19:%.*]] = shufflevector <1 x i32> [[SPLIT6]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP12:%.*]] = extractelement <8 x i32> [[TMP0]], i64 4
-; CHECK-NEXT: [[SPLAT_SPLATINSERT20:%.*]] = insertelement <1 x i32> poison, i32 [[TMP12]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT21:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT20]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP13:%.*]] = mul <1 x i32> [[BLOCK19]], [[SPLAT_SPLAT21]]
-; CHECK-NEXT: [[TMP14:%.*]] = add <1 x i32> [[TMP11]], [[TMP13]]
-; CHECK-NEXT: [[BLOCK22:%.*]] = shufflevector <1 x i32> [[SPLIT7]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP15:%.*]] = extractelement <8 x i32> [[TMP0]], i64 5
-; CHECK-NEXT: [[SPLAT_SPLATINSERT23:%.*]] = insertelement <1 x i32> poison, i32 [[TMP15]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT24:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT23]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP16:%.*]] = mul <1 x i32> [[BLOCK22]], [[SPLAT_SPLAT24]]
-; CHECK-NEXT: [[TMP17:%.*]] = add <1 x i32> [[TMP14]], [[TMP16]]
-; CHECK-NEXT: [[BLOCK25:%.*]] = shufflevector <1 x i32> [[SPLIT8]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP18:%.*]] = extractelement <8 x i32> [[TMP0]], i64 6
-; CHECK-NEXT: [[SPLAT_SPLATINSERT26:%.*]] = insertelement <1 x i32> poison, i32 [[TMP18]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT27:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT26]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP19:%.*]] = mul <1 x i32> [[BLOCK25]], [[SPLAT_SPLAT27]]
-; CHECK-NEXT: [[TMP20:%.*]] = add <1 x i32> [[TMP17]], [[TMP19]]
-; CHECK-NEXT: [[BLOCK28:%.*]] = shufflevector <1 x i32> [[SPLIT9]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP21:%.*]] = extractelement <8 x i32> [[TMP0]], i64 7
-; CHECK-NEXT: [[SPLAT_SPLATINSERT29:%.*]] = insertelement <1 x i32> poison, i32 [[TMP21]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT30:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT29]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP22:%.*]] = mul <1 x i32> [[BLOCK28]], [[SPLAT_SPLAT30]]
-; CHECK-NEXT: [[TMP23:%.*]] = add <1 x i32> [[TMP20]], [[TMP22]]
-; CHECK-NEXT: [[TMP24:%.*]] = shufflevector <1 x i32> [[TMP23]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP25:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP24]], <1 x i32> <i32 1>
-; CHECK-NEXT: ret <1 x i32> [[TMP25]]
+; CHECK-NEXT: [[TMP1:%.*]] = mul <8 x i32> [[C:%.*]], [[TMP0]]
+; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP1]])
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <1 x i32> poison, i32 [[TMP2]], i64 0
+; CHECK-NEXT: ret <1 x i32> [[TMP3]]
;
entry:
%sub = sub <8 x i32> %a, %b
@@ -323,17 +295,57 @@ define <1 x i32> @add_chain_feeding_dotproduct_i32_v8_1(<8 x i32> %a, <8 x i32>
; CHECK-NEXT: [[TMP13:%.*]] = add <1 x i32> [[TMP5]], [[SPLIT21]]
; CHECK-NEXT: [[TMP14:%.*]] = add <1 x i32> [[TMP6]], [[SPLIT22]]
; CHECK-NEXT: [[TMP15:%.*]] = add <1 x i32> [[TMP7]], [[SPLIT23]]
-; CHECK-NEXT: [[TMP16:%.*]] = shufflevector <1 x i32> [[TMP8]], <1 x i32> [[TMP9]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP17:%.*]] = shufflevector <1 x i32> [[TMP10]], <1 x i32> [[TMP11]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP18:%.*]] = shufflevector <1 x i32> [[TMP12]], <1 x i32> [[TMP13]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP19:%.*]] = shufflevector <1 x i32> [[TMP14]], <1 x i32> [[TMP15]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <2 x i32> [[TMP16]], <2 x i32> [[TMP17]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT: [[TMP21:%.*]] = shufflevector <2 x i32> [[TMP18]], <2 x i32> [[TMP19]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT: [[TMP22:%.*]] = shufflevector <4 x i32> [[TMP20]], <4 x i32> [[TMP21]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT: [[TMP23:%.*]] = mul <8 x i32> [[TMP22]], [[D:%.*]]
-; CHECK-NEXT: [[TMP24:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP23]])
-; CHECK-NEXT: [[TMP25:%.*]] = insertelement <1 x i32> poison, i32 [[TMP24]], i64 0
-; CHECK-NEXT: ret <1 x i32> [[TMP25]]
+; CHECK-NEXT: [[SPLIT24:%.*]] = shufflevector <8 x i32> [[D:%.*]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <1 x i32> [[TMP8]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP16:%.*]] = extractelement <8 x i32> [[SPLIT24]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i32> poison, i32 [[TMP16]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP17:%.*]] = mul <1 x i32> [[BLOCK]], [[SPLAT_SPLAT]]
+; CHECK-NEXT: [[BLOCK25:%.*]] = shufflevector <1 x i32> [[TMP9]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP18:%.*]] = extractelement <8 x i32> [[SPLIT24]], i64 1
+; CHECK-NEXT: [[SPLAT_SPLATINSERT26:%.*]] = insertelement <1 x i32> poison, i32 [[TMP18]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT27:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT26]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP19:%.*]] = mul <1 x i32> [[BLOCK25]], [[SPLAT_SPLAT27]]
+; CHECK-NEXT: [[TMP20:%.*]] = add <1 x i32> [[TMP17]], [[TMP19]]
+; CHECK-NEXT: [[BLOCK28:%.*]] = shufflevector <1 x i32> [[TMP10]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP21:%.*]] = extractelement <8 x i32> [[SPLIT24]], i64 2
+; CHECK-NEXT: [[SPLAT_SPLATINSERT29:%.*]] = insertelement <1 x i32> poison, i32 [[TMP21]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT30:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT29]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP22:%.*]] = mul <1 x i32> [[BLOCK28]], [[SPLAT_SPLAT30]]
+; CHECK-NEXT: [[TMP23:%.*]] = add <1 x i32> [[TMP20]], [[TMP22]]
+; CHECK-NEXT: [[BLOCK31:%.*]] = shufflevector <1 x i32> [[TMP11]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP24:%.*]] = extractelement <8 x i32> [[SPLIT24]], i64 3
+; CHECK-NEXT: [[SPLAT_SPLATINSERT32:%.*]] = insertelement <1 x i32> poison, i32 [[TMP24]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT33:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT32]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP25:%.*]] = mul <1 x i32> [[BLOCK31]], [[SPLAT_SPLAT33]]
+; CHECK-NEXT: [[TMP26:%.*]] = add <1 x i32> [[TMP23]], [[TMP25]]
+; CHECK-NEXT: [[BLOCK34:%.*]] = shufflevector <1 x i32> [[TMP12]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP27:%.*]] = extractelement <8 x i32> [[SPLIT24]], i64 4
+; CHECK-NEXT: [[SPLAT_SPLATINSERT35:%.*]] = insertelement <1 x i32> poison, i32 [[TMP27]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT36:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT35]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP28:%.*]] = mul <1 x i32> [[BLOCK34]], [[SPLAT_SPLAT36]]
+; CHECK-NEXT: [[TMP29:%.*]] = add <1 x i32> [[TMP26]], [[TMP28]]
+; CHECK-NEXT: [[BLOCK37:%.*]] = shufflevector <1 x i32> [[TMP13]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP30:%.*]] = extractelement <8 x i32> [[SPLIT24]], i64 5
+; CHECK-NEXT: [[SPLAT_SPLATINSERT38:%.*]] = insertelement <1 x i32> poison, i32 [[TMP30]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT39:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT38]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP31:%.*]] = mul <1 x i32> [[BLOCK37]], [[SPLAT_SPLAT39]]
+; CHECK-NEXT: [[TMP32:%.*]] = add <1 x i32> [[TMP29]], [[TMP31]]
+; CHECK-NEXT: [[BLOCK40:%.*]] = shufflevector <1 x i32> [[TMP14]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP33:%.*]] = extractelement <8 x i32> [[SPLIT24]], i64 6
+; CHECK-NEXT: [[SPLAT_SPLATINSERT41:%.*]] = insertelement <1 x i32> poison, i32 [[TMP33]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT42:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT41]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP34:%.*]] = mul <1 x i32> [[BLOCK40]], [[SPLAT_SPLAT42]]
+; CHECK-NEXT: [[TMP35:%.*]] = add <1 x i32> [[TMP32]], [[TMP34]]
+; CHECK-NEXT: [[BLOCK43:%.*]] = shufflevector <1 x i32> [[TMP15]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP36:%.*]] = extractelement <8 x i32> [[SPLIT24]], i64 7
+; CHECK-NEXT: [[SPLAT_SPLATINSERT44:%.*]] = insertelement <1 x i32> poison, i32 [[TMP36]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT45:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT44]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP37:%.*]] = mul <1 x i32> [[BLOCK43]], [[SPLAT_SPLAT45]]
+; CHECK-NEXT: [[TMP38:%.*]] = add <1 x i32> [[TMP35]], [[TMP37]]
+; CHECK-NEXT: [[TMP39:%.*]] = shufflevector <1 x i32> [[TMP38]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP40:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP39]], <1 x i32> <i32 1>
+; CHECK-NEXT: ret <1 x i32> [[TMP40]]
;
entry:
%add.1 = add <8 x i32> %a, %b
@@ -350,64 +362,10 @@ define <1 x i32> @add_chain_feeding_dotproduct_i32_v8_2(<8 x i32> %a, <8 x i32>
; CHECK-NEXT: [[TMP0:%.*]] = add <8 x i32> [[SPLIT]], [[SPLIT1]]
; CHECK-NEXT: [[SPLIT2:%.*]] = shufflevector <8 x i32> [[C:%.*]], <8 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT: [[TMP1:%.*]] = add <8 x i32> [[TMP0]], [[SPLIT2]]
-; CHECK-NEXT: [[SPLIT3:%.*]] = shufflevector <8 x i32> [[D:%.*]], <8 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[SPLIT4:%.*]] = shufflevector <8 x i32> [[D]], <8 x i32> poison, <1 x i32> <i32 1>
-; CHECK-NEXT: [[SPLIT5:%.*]] = shufflevector <8 x i32> [[D]], <8 x i32> poison, <1 x i32> <i32 2>
-; CHECK-NEXT: [[SPLIT6:%.*]] = shufflevector <8 x i32> [[D]], <8 x i32> poison, <1 x i32> <i32 3>
-; CHECK-NEXT: [[SPLIT7:%.*]] = shufflevector <8 x i32> [[D]], <8 x i32> poison, <1 x i32> <i32 4>
-; CHECK-NEXT: [[SPLIT8:%.*]] = shufflevector <8 x i32> [[D]], <8 x i32> poison, <1 x i32> <i32 5>
-; CHECK-NEXT: [[SPLIT9:%.*]] = shufflevector <8 x i32> [[D]], <8 x i32> poison, <1 x i32> <i32 6>
-; CHECK-NEXT: [[SPLIT10:%.*]] = shufflevector <8 x i32> [[D]], <8 x i32> poison, <1 x i32> <i32 7>
-; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <1 x i32> [[SPLIT3]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP2:%.*]] = extractelement <8 x i32> [[TMP1]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i32> poison, i32 [[TMP2]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP3:%.*]] = mul <1 x i32> [[BLOCK]], [[SPLAT_SPLAT]]
-; CHECK-NEXT: [[BLOCK11:%.*]] = shufflevector <1 x i32> [[SPLIT4]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP4:%.*]] = extractelement <8 x i32> [[TMP1]], i64 1
-; CHECK-NEXT: [[SPLAT_SPLATINSERT12:%.*]] = insertelement <1 x i32> poison, i32 [[TMP4]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT13:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT12]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP5:%.*]] = mul <1 x i32> [[BLOCK11]], [[SPLAT_SPLAT13]]
-; CHECK-NEXT: [[TMP6:%.*]] = add <1 x i32> [[TMP3]], [[TMP5]]
-; CHECK-NEXT: [[BLOCK14:%.*]] = shufflevector <1 x i32> [[SPLIT5]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP7:%.*]] = extractelement <8 x i32> [[TMP1]], i64 2
-; CHECK-NEXT: [[SPLAT_SPLATINSERT15:%.*]] = insertelement <1 x i32> poison, i32 [[TMP7]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT16:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT15]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP8:%.*]] = mul <1 x i32> [[BLOCK14]], [[SPLAT_SPLAT16]]
-; CHECK-NEXT: [[TMP9:%.*]] = add <1 x i32> [[TMP6]], [[TMP8]]
-; CHECK-NEXT: [[BLOCK17:%.*]] = shufflevector <1 x i32> [[SPLIT6]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP10:%.*]] = extractelement <8 x i32> [[TMP1]], i64 3
-; CHECK-NEXT: [[SPLAT_SPLATINSERT18:%.*]] = insertelement <1 x i32> poison, i32 [[TMP10]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT19:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT18]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP11:%.*]] = mul <1 x i32> [[BLOCK17]], [[SPLAT_SPLAT19]]
-; CHECK-NEXT: [[TMP12:%.*]] = add <1 x i32> [[TMP9]], [[TMP11]]
-; CHECK-NEXT: [[BLOCK20:%.*]] = shufflevector <1 x i32> [[SPLIT7]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP13:%.*]] = extractelement <8 x i32> [[TMP1]], i64 4
-; CHECK-NEXT: [[SPLAT_SPLATINSERT21:%.*]] = insertelement <1 x i32> poison, i32 [[TMP13]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT22:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT21]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP14:%.*]] = mul <1 x i32> [[BLOCK20]], [[SPLAT_SPLAT22]]
-; CHECK-NEXT: [[TMP15:%.*]] = add <1 x i32> [[TMP12]], [[TMP14]]
-; CHECK-NEXT: [[BLOCK23:%.*]] = shufflevector <1 x i32> [[SPLIT8]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP16:%.*]] = extractelement <8 x i32> [[TMP1]], i64 5
-; CHECK-NEXT: [[SPLAT_SPLATINSERT24:%.*]] = insertelement <1 x i32> poison, i32 [[TMP16]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT25:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT24]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP17:%.*]] = mul <1 x i32> [[BLOCK23]], [[SPLAT_SPLAT25]]
-; CHECK-NEXT: [[TMP18:%.*]] = add <1 x i32> [[TMP15]], [[TMP17]]
-; CHECK-NEXT: [[BLOCK26:%.*]] = shufflevector <1 x i32> [[SPLIT9]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP19:%.*]] = extractelement <8 x i32> [[TMP1]], i64 6
-; CHECK-NEXT: [[SPLAT_SPLATINSERT27:%.*]] = insertelement <1 x i32> poison, i32 [[TMP19]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT28:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT27]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP20:%.*]] = mul <1 x i32> [[BLOCK26]], [[SPLAT_SPLAT28]]
-; CHECK-NEXT: [[TMP21:%.*]] = add <1 x i32> [[TMP18]], [[TMP20]]
-; CHECK-NEXT: [[BLOCK29:%.*]] = shufflevector <1 x i32> [[SPLIT10]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP22:%.*]] = extractelement <8 x i32> [[TMP1]], i64 7
-; CHECK-NEXT: [[SPLAT_SPLATINSERT30:%.*]] = insertelement <1 x i32> poison, i32 [[TMP22]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT31:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT30]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP23:%.*]] = mul <1 x i32> [[BLOCK29]], [[SPLAT_SPLAT31]]
-; CHECK-NEXT: [[TMP24:%.*]] = add <1 x i32> [[TMP21]], [[TMP23]]
-; CHECK-NEXT: [[TMP25:%.*]] = shufflevector <1 x i32> [[TMP24]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP26:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP25]], <1 x i32> <i32 1>
-; CHECK-NEXT: ret <1 x i32> [[TMP26]]
+; CHECK-NEXT: [[TMP2:%.*]] = mul <8 x i32> [[D:%.*]], [[TMP1]]
+; CHECK-NEXT: [[TMP3:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP2]])
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <1 x i32> poison, i32 [[TMP3]], i64 0
+; CHECK-NEXT: ret <1 x i32> [[TMP4]]
;
entry:
%add.1 = add <8 x i32> %a, %b
@@ -434,72 +392,12 @@ declare <1 x i64> @llvm.matrix.multiply.v1i64.v8i64.v8i64(<8 x i64>, <8 x i64>,
define <1 x i32> @intrinsic_column_major_load_dot_product_i32_v8(ptr %lhs_address, ptr %rhs_address) {
; CHECK-LABEL: @intrinsic_column_major_load_dot_product_i32_v8(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[COL_LOAD:%.*]] = load <1 x i32>, ptr [[LHS_ADDRESS:%.*]], align 4
-; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr i32, ptr [[LHS_ADDRESS]], i64 1
-; CHECK-NEXT: [[COL_LOAD1:%.*]] = load <1 x i32>, ptr [[VEC_GEP]], align 4
-; CHECK-NEXT: [[VEC_GEP2:%.*]] = getelementptr i32, ptr [[LHS_ADDRESS]], i64 2
-; CHECK-NEXT: [[COL_LOAD3:%.*]] = load <1 x i32>, ptr [[VEC_GEP2]], align 4
-; CHECK-NEXT: [[VEC_GEP4:%.*]] = getelementptr i32, ptr [[LHS_ADDRESS]], i64 3
-; CHECK-NEXT: [[COL_LOAD5:%.*]] = load <1 x i32>, ptr [[VEC_GEP4]], align 4
-; CHECK-NEXT: [[VEC_GEP6:%.*]] = getelementptr i32, ptr [[LHS_ADDRESS]], i64 4
-; CHECK-NEXT: [[COL_LOAD7:%.*]] = load <1 x i32>, ptr [[VEC_GEP6]], align 4
-; CHECK-NEXT: [[VEC_GEP8:%.*]] = getelementptr i32, ptr [[LHS_ADDRESS]], i64 5
-; CHECK-NEXT: [[COL_LOAD9:%.*]] = load <1 x i32>, ptr [[VEC_GEP8]], align 4
-; CHECK-NEXT: [[VEC_GEP10:%.*]] = getelementptr i32, ptr [[LHS_ADDRESS]], i64 6
-; CHECK-NEXT: [[COL_LOAD11:%.*]] = load <1 x i32>, ptr [[VEC_GEP10]], align 4
-; CHECK-NEXT: [[VEC_GEP12:%.*]] = getelementptr i32, ptr [[LHS_ADDRESS]], i64 7
-; CHECK-NEXT: [[COL_LOAD13:%.*]] = load <1 x i32>, ptr [[VEC_GEP12]], align 4
-; CHECK-NEXT: [[COL_LOAD14:%.*]] = load <8 x i32>, ptr [[RHS_ADDRESS:%.*]], align 4
-; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <1 x i32> [[COL_LOAD]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP0:%.*]] = extractelement <8 x i32> [[COL_LOAD14]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i32> poison, i32 [[TMP0]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP1:%.*]] = mul <1 x i32> [[BLOCK]], [[SPLAT_SPLAT]]
-; CHECK-NEXT: [[BLOCK15:%.*]] = shufflevector <1 x i32> [[COL_LOAD1]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP2:%.*]] = extractelement <8 x i32> [[COL_LOAD14]], i64 1
-; CHECK-NEXT: [[SPLAT_SPLATINSERT16:%.*]] = insertelement <1 x i32> poison, i32 [[TMP2]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT17:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT16]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP3:%.*]] = mul <1 x i32> [[BLOCK15]], [[SPLAT_SPLAT17]]
-; CHECK-NEXT: [[TMP4:%.*]] = add <1 x i32> [[TMP1]], [[TMP3]]
-; CHECK-NEXT: [[BLOCK18:%.*]] = shufflevector <1 x i32> [[COL_LOAD3]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP5:%.*]] = extractelement <8 x i32> [[COL_LOAD14]], i64 2
-; CHECK-NEXT: [[SPLAT_SPLATINSERT19:%.*]] = insertelement <1 x i32> poison, i32 [[TMP5]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT20:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT19]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP6:%.*]] = mul <1 x i32> [[BLOCK18]], [[SPLAT_SPLAT20]]
-; CHECK-NEXT: [[TMP7:%.*]] = add <1 x i32> [[TMP4]], [[TMP6]]
-; CHECK-NEXT: [[BLOCK21:%.*]] = shufflevector <1 x i32> [[COL_LOAD5]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP8:%.*]] = extractelement <8 x i32> [[COL_LOAD14]], i64 3
-; CHECK-NEXT: [[SPLAT_SPLATINSERT22:%.*]] = insertelement <1 x i32> poison, i32 [[TMP8]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT23:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT22]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP9:%.*]] = mul <1 x i32> [[BLOCK21]], [[SPLAT_SPLAT23]]
-; CHECK-NEXT: [[TMP10:%.*]] = add <1 x i32> [[TMP7]], [[TMP9]]
-; CHECK-NEXT: [[BLOCK24:%.*]] = shufflevector <1 x i32> [[COL_LOAD7]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP11:%.*]] = extractelement <8 x i32> [[COL_LOAD14]], i64 4
-; CHECK-NEXT: [[SPLAT_SPLATINSERT25:%.*]] = insertelement <1 x i32> poison, i32 [[TMP11]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT26:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT25]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP12:%.*]] = mul <1 x i32> [[BLOCK24]], [[SPLAT_SPLAT26]]
-; CHECK-NEXT: [[TMP13:%.*]] = add <1 x i32> [[TMP10]], [[TMP12]]
-; CHECK-NEXT: [[BLOCK27:%.*]] = shufflevector <1 x i32> [[COL_LOAD9]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP14:%.*]] = extractelement <8 x i32> [[COL_LOAD14]], i64 5
-; CHECK-NEXT: [[SPLAT_SPLATINSERT28:%.*]] = insertelement <1 x i32> poison, i32 [[TMP14]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT29:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT28]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP15:%.*]] = mul <1 x i32> [[BLOCK27]], [[SPLAT_SPLAT29]]
-; CHECK-NEXT: [[TMP16:%.*]] = add <1 x i32> [[TMP13]], [[TMP15]]
-; CHECK-NEXT: [[BLOCK30:%.*]] = shufflevector <1 x i32> [[COL_LOAD11]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP17:%.*]] = extractelement <8 x i32> [[COL_LOAD14]], i64 6
-; CHECK-NEXT: [[SPLAT_SPLATINSERT31:%.*]] = insertelement <1 x i32> poison, i32 [[TMP17]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT32:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT31]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP18:%.*]] = mul <1 x i32> [[BLOCK30]], [[SPLAT_SPLAT32]]
-; CHECK-NEXT: [[TMP19:%.*]] = add <1 x i32> [[TMP16]], [[TMP18]]
-; CHECK-NEXT: [[BLOCK33:%.*]] = shufflevector <1 x i32> [[COL_LOAD13]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP20:%.*]] = extractelement <8 x i32> [[COL_LOAD14]], i64 7
-; CHECK-NEXT: [[SPLAT_SPLATINSERT34:%.*]] = insertelement <1 x i32> poison, i32 [[TMP20]], i64 0
-; CHECK-NEXT: [[SPLAT_SPLAT35:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT34]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP21:%.*]] = mul <1 x i32> [[BLOCK33]], [[SPLAT_SPLAT35]]
-; CHECK-NEXT: [[TMP22:%.*]] = add <1 x i32> [[TMP19]], [[TMP21]]
-; CHECK-NEXT: [[TMP23:%.*]] = shufflevector <1 x i32> [[TMP22]], <1 x i32> poison, <1 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP24:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP23]], <1 x i32> <i32 1>
-; CHECK-NEXT: ret <1 x i32> [[TMP24]]
+; CHECK-NEXT: [[COL_LOAD:%.*]] = load <8 x i32>, ptr [[RHS_ADDRESS:%.*]], align 4
+; CHECK-NEXT: [[TMP0:%.*]] = load <8 x i32>, ptr [[LHS_ADDRESS:%.*]], align 32
+; CHECK-NEXT: [[TMP1:%.*]] = mul <8 x i32> [[TMP0]], [[COL_LOAD]]
+; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP1]])
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <1 x i32> poison, i32 [[TMP2]], i64 0
+; CHECK-NEXT: ret <1 x i32> [[TMP3]]
;
entry:
%lhs = tail call <8 x i32> @llvm.matrix.column.major.load.v8i32.i64(ptr nonnull align 4 %lhs_address, i64 1, i1 false, i32 1, i32 8)
@@ -646,9 +544,23 @@ define void @transposed_multiply_feeding_dot_produc_v4i32() {
; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <2 x i32> [[TMP7]], <2 x i32> poison, <2 x i32> <i32 0, i32 1>
; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <2 x i32> undef, <2 x i32> [[TMP8]], <2 x i32> <i32 2, i32 3>
; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <2 x i32> [[TMP4]], <2 x i32> [[TMP9]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT: [[TMP11:%.*]] = mul <4 x i32> [[TMP10]], zeroinitializer
-; CHECK-NEXT: [[TMP12:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP11]])
-; CHECK-NEXT: [[TMP13:%.*]] = insertelement <1 x i32> poison, i32 [[TMP12]], i64 0
+; CHECK-NEXT: [[SPLIT5:%.*]] = shufflevector <4 x i32> [[TMP10]], <4 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[SPLIT6:%.*]] = shufflevector <4 x i32> [[TMP10]], <4 x i32> poison, <1 x i32> <i32 1>
+; CHECK-NEXT: [[SPLIT7:%.*]] = shufflevector <4 x i32> [[TMP10]], <4 x i32> poison, <1 x i32> <i32 2>
+; CHECK-NEXT: [[SPLIT8:%.*]] = shufflevector <4 x i32> [[TMP10]], <4 x i32> poison, <1 x i32> <i32 3>
+; CHECK-NEXT: [[BLOCK9:%.*]] = shufflevector <1 x i32> [[SPLIT5]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP11:%.*]] = mul <1 x i32> [[BLOCK9]], zeroinitializer
+; CHECK-NEXT: [[BLOCK10:%.*]] = shufflevector <1 x i32> [[SPLIT6]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP12:%.*]] = mul <1 x i32> [[BLOCK10]], zeroinitializer
+; CHECK-NEXT: [[TMP13:%.*]] = add <1 x i32> [[TMP11]], [[TMP12]]
+; CHECK-NEXT: [[BLOCK11:%.*]] = shufflevector <1 x i32> [[SPLIT7]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP14:%.*]] = mul <1 x i32> [[BLOCK11]], zeroinitializer
+; CHECK-NEXT: [[TMP15:%.*]] = add <1 x i32> [[TMP13]], [[TMP14]]
+; CHECK-NEXT: [[BLOCK12:%.*]] = shufflevector <1 x i32> [[SPLIT8]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP16:%.*]] = mul <1 x i32> [[BLOCK12]], zeroinitializer
+; CHECK-NEXT: [[TMP17:%.*]] = add <1 x i32> [[TMP15]], [[TMP16]]
+; CHECK-NEXT: [[TMP18:%.*]] = shufflevector <1 x i32> [[TMP17]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP19:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP18]], <1 x i32> <i32 1>
; CHECK-NEXT: ret void
;
entry:
@@ -662,3 +574,110 @@ entry:
declare <4 x i32> @llvm.matrix.transpose.v4i32(<4 x i32>, i32 immarg, i32 immarg)
declare <4 x i32> @llvm.matrix.multiply.v4i32.v4i32.v4i32(<4 x i32>, <4 x i32>, i32 immarg, i32 immarg, i32 immarg)
+
+define <1 x i32> @test_load_multiuse(ptr %src, <4 x i32> %b) {
+; CHECK-LABEL: @test_load_multiuse(
+; CHECK-NEXT: [[COL_LOAD:%.*]] = load <1 x i32>, ptr [[SRC:%.*]], align 16
+; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr i32, ptr [[SRC]], i64 1
+; CHECK-NEXT: [[COL_LOAD1:%.*]] = load <1 x i32>, ptr [[VEC_GEP]], align 4
+; CHECK-NEXT: [[VEC_GEP2:%.*]] = getelementptr i32, ptr [[SRC]], i64 2
+; CHECK-NEXT: [[COL_LOAD3:%.*]] = load <1 x i32>, ptr [[VEC_GEP2]], align 8
+; CHECK-NEXT: [[VEC_GEP4:%.*]] = getelementptr i32, ptr [[SRC]], i64 3
+; CHECK-NEXT: [[COL_LOAD5:%.*]] = load <1 x i32>, ptr [[VEC_GEP4]], align 4
+; CHECK-NEXT: [[TMP1:%.*]] = extractelement <1 x i32> [[COL_LOAD]], i64 0
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i64 0
+; CHECK-NEXT: [[TMP3:%.*]] = extractelement <1 x i32> [[COL_LOAD1]], i64 0
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[TMP3]], i64 1
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <1 x i32> [[COL_LOAD3]], i64 0
+; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x i32> [[TMP4]], i32 [[TMP5]], i64 2
+; CHECK-NEXT: [[TMP7:%.*]] = extractelement <1 x i32> [[COL_LOAD5]], i64 0
+; CHECK-NEXT: [[TMP8:%.*]] = insertelement <4 x i32> [[TMP6]], i32 [[TMP7]], i64 3
+; CHECK-NEXT: call void @use.v4i32(<4 x i32> [[TMP8]])
+; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <1 x i32> [[COL_LOAD]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i32> [[SPLIT]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i32> poison, i32 [[TMP9]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP10:%.*]] = mul <1 x i32> [[BLOCK]], [[SPLAT_SPLAT]]
+; CHECK-NEXT: [[BLOCK6:%.*]] = shufflevector <1 x i32> [[COL_LOAD1]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP11:%.*]] = extractelement <4 x i32> [[SPLIT]], i64 1
+; CHECK-NEXT: [[SPLAT_SPLATINSERT7:%.*]] = insertelement <1 x i32> poison, i32 [[TMP11]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT8:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT7]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP12:%.*]] = mul <1 x i32> [[BLOCK6]], [[SPLAT_SPLAT8]]
+; CHECK-NEXT: [[TMP13:%.*]] = add <1 x i32> [[TMP10]], [[TMP12]]
+; CHECK-NEXT: [[BLOCK9:%.*]] = shufflevector <1 x i32> [[COL_LOAD3]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i32> [[SPLIT]], i64 2
+; CHECK-NEXT: [[SPLAT_SPLATINSERT10:%.*]] = insertelement <1 x i32> poison, i32 [[TMP14]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT11:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT10]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP15:%.*]] = mul <1 x i32> [[BLOCK9]], [[SPLAT_SPLAT11]]
+; CHECK-NEXT: [[TMP16:%.*]] = add <1 x i32> [[TMP13]], [[TMP15]]
+; CHECK-NEXT: [[BLOCK12:%.*]] = shufflevector <1 x i32> [[COL_LOAD5]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP17:%.*]] = extractelement <4 x i32> [[SPLIT]], i64 3
+; CHECK-NEXT: [[SPLAT_SPLATINSERT13:%.*]] = insertelement <1 x i32> poison, i32 [[TMP17]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT14:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT13]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP18:%.*]] = mul <1 x i32> [[BLOCK12]], [[SPLAT_SPLAT14]]
+; CHECK-NEXT: [[TMP19:%.*]] = add <1 x i32> [[TMP16]], [[TMP18]]
+; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <1 x i32> [[TMP19]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP21:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP20]], <1 x i32> <i32 1>
+; CHECK-NEXT: ret <1 x i32> [[TMP21]]
+;
+ %l = load <4 x i32>, ptr %src
+ %t = call <4 x i32> @llvm.matrix.transpose.v4i32(<4 x i32> %l, i32 1, i32 4)
+ call void @use.v4i32(<4 x i32> %t)
+ %res = call <1 x i32> @llvm.matrix.multiply.v1i32.v4i32.v4i32(<4 x i32> %l, <4 x i32> %b, i32 1, i32 4, i32 1)
+ ret <1 x i32> %res
+}
+define <1 x i32> @test_builtin_column_major_load_multiuse(ptr %src, <4 x i32> %b) {
+; CHECK-LABEL: @test_builtin_column_major_load_multiuse(
+; CHECK-NEXT: [[COL_LOAD:%.*]] = load <1 x i32>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr i32, ptr [[SRC]], i64 1
+; CHECK-NEXT: [[COL_LOAD1:%.*]] = load <1 x i32>, ptr [[VEC_GEP]], align 4
+; CHECK-NEXT: [[VEC_GEP2:%.*]] = getelementptr i32, ptr [[SRC]], i64 2
+; CHECK-NEXT: [[COL_LOAD3:%.*]] = load <1 x i32>, ptr [[VEC_GEP2]], align 4
+; CHECK-NEXT: [[VEC_GEP4:%.*]] = getelementptr i32, ptr [[SRC]], i64 3
+; CHECK-NEXT: [[COL_LOAD5:%.*]] = load <1 x i32>, ptr [[VEC_GEP4]], align 4
+; CHECK-NEXT: [[TMP1:%.*]] = extractelement <1 x i32> [[COL_LOAD]], i64 0
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i64 0
+; CHECK-NEXT: [[TMP3:%.*]] = extractelement <1 x i32> [[COL_LOAD1]], i64 0
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[TMP3]], i64 1
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <1 x i32> [[COL_LOAD3]], i64 0
+; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x i32> [[TMP4]], i32 [[TMP5]], i64 2
+; CHECK-NEXT: [[TMP7:%.*]] = extractelement <1 x i32> [[COL_LOAD5]], i64 0
+; CHECK-NEXT: [[TMP8:%.*]] = insertelement <4 x i32> [[TMP6]], i32 [[TMP7]], i64 3
+; CHECK-NEXT: call void @use.v4i32(<4 x i32> [[TMP8]])
+; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <4 x i32> [[B:%.*]], <4 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <1 x i32> [[COL_LOAD]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i32> [[SPLIT]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i32> poison, i32 [[TMP9]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP10:%.*]] = mul <1 x i32> [[BLOCK]], [[SPLAT_SPLAT]]
+; CHECK-NEXT: [[BLOCK6:%.*]] = shufflevector <1 x i32> [[COL_LOAD1]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP11:%.*]] = extractelement <4 x i32> [[SPLIT]], i64 1
+; CHECK-NEXT: [[SPLAT_SPLATINSERT7:%.*]] = insertelement <1 x i32> poison, i32 [[TMP11]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT8:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT7]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP12:%.*]] = mul <1 x i32> [[BLOCK6]], [[SPLAT_SPLAT8]]
+; CHECK-NEXT: [[TMP13:%.*]] = add <1 x i32> [[TMP10]], [[TMP12]]
+; CHECK-NEXT: [[BLOCK9:%.*]] = shufflevector <1 x i32> [[COL_LOAD3]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i32> [[SPLIT]], i64 2
+; CHECK-NEXT: [[SPLAT_SPLATINSERT10:%.*]] = insertelement <1 x i32> poison, i32 [[TMP14]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT11:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT10]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP15:%.*]] = mul <1 x i32> [[BLOCK9]], [[SPLAT_SPLAT11]]
+; CHECK-NEXT: [[TMP16:%.*]] = add <1 x i32> [[TMP13]], [[TMP15]]
+; CHECK-NEXT: [[BLOCK12:%.*]] = shufflevector <1 x i32> [[COL_LOAD5]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP17:%.*]] = extractelement <4 x i32> [[SPLIT]], i64 3
+; CHECK-NEXT: [[SPLAT_SPLATINSERT13:%.*]] = insertelement <1 x i32> poison, i32 [[TMP17]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLAT14:%.*]] = shufflevector <1 x i32> [[SPLAT_SPLATINSERT13]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP18:%.*]] = mul <1 x i32> [[BLOCK12]], [[SPLAT_SPLAT14]]
+; CHECK-NEXT: [[TMP19:%.*]] = add <1 x i32> [[TMP16]], [[TMP18]]
+; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <1 x i32> [[TMP19]], <1 x i32> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP21:%.*]] = shufflevector <1 x i32> undef, <1 x i32> [[TMP20]], <1 x i32> <i32 1>
+; CHECK-NEXT: ret <1 x i32> [[TMP21]]
+;
+ %l = call <4 x i32> @llvm.matrix.column.major.load.v4i32.i64(ptr %src, i64 1, i1 false, i32 1, i32 4)
+ %t = call <4 x i32> @llvm.matrix.transpose.v4i32(<4 x i32> %l, i32 1, i32 4)
+ call void @use.v4i32(<4 x i32> %t)
+ %res = call <1 x i32> @llvm.matrix.multiply.v1i32.v4i32.v4i32(<4 x i32> %l, <4 x i32> %b, i32 1, i32 4, i32 1)
+ ret <1 x i32> %res
+}
+
+declare void @use.v4i32(<4 x i32>)
More information about the llvm-commits
mailing list