diff --git a/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp b/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
--- a/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
+++ b/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
@@ -400,9 +400,68 @@
     }
   }
 
+  /// Propagate the shape to operands of instructions with shape information.
+  void propagateShapeBackward() {
+    SmallVector<Value *, 8> WorkList;
+    // Worklist contains instruction for which we already know the shape.
+    for (auto &V : ShapeMap)
+      WorkList.push_back(V.first);
+
+    // Pop an element with known shape.  Traverse the operands, if their shape
+    // derives from the result shape and is unknown, add it and add them to the
+    // worklist.
+    LLVM_DEBUG(dbgs() << "Backward-propagate shapes:\n");
+    while (!WorkList.empty()) {
+      Value *V = WorkList.back();
+      WorkList.pop_back();
+
+      if (!isa<Instruction>(V))
+        continue;
+
+      Value *MatrixA;
+      Value *MatrixB;
+      Value *M;
+      Value *N;
+      Value *K;
+      if (match(V, m_Intrinsic<Intrinsic::matrix_multiply>(
+                       m_Value(MatrixA), m_Value(MatrixB), m_Value(M),
+                       m_Value(N), m_Value(K)))) {
+        if (setShapeInfo(MatrixA, {M, N}))
+          WorkList.push_back(MatrixA);
+
+        if (setShapeInfo(MatrixB, {N, K}))
+          WorkList.push_back(MatrixB);
+
+      } else if (match(V, m_Intrinsic<Intrinsic::matrix_transpose>(
+                              m_Value(MatrixA), m_Value(M), m_Value(N)))) {
+        // Flip dimensions.
+        if (setShapeInfo(MatrixA, {M, N}))
+          WorkList.push_back(MatrixA);
+      } else if (match(V, m_Intrinsic<Intrinsic::matrix_columnwise_store>(
+                              m_Value(MatrixA), m_Value(), m_Value(),
+                              m_Value(M), m_Value(N)))) {
+        if (setShapeInfo(MatrixA, {M, N})) {
+          WorkList.push_back(MatrixA);
+        }
+      } else if (isa<StoreInst>(V)) {
+        // Nothing to do.  We forward-propagated to this so we would just
+        // backward propagate to an instruction with an already known shape.
+      } else if (isUniformShape(V)) {
+        // Propagate to all operands.
+        ShapeInfo Shape = ShapeMap[V];
+        for (Use &U : cast<Instruction>(V)->operands()) {
+          if (setShapeInfo(U.get(), Shape))
+            WorkList.push_back(U.get());
+        }
+      }
+    }
+  }
+
   bool Visit() {
-    if (EnableShapePropagation)
+    if (EnableShapePropagation) {
       propagateShapeForward();
+      propagateShapeBackward();
+    }
 
     ReversePostOrderTraversal<Function *> RPOT(&Func);
     bool Changed = false;
diff --git a/llvm/test/Transforms/LowerMatrixIntrinsics/propagate-backward.ll b/llvm/test/Transforms/LowerMatrixIntrinsics/propagate-backward.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/Transforms/LowerMatrixIntrinsics/propagate-backward.ll
@@ -0,0 +1,46 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -lower-matrix-intrinsics -S < %s | FileCheck  %s
+; RUN: opt -passes='lower-matrix-intrinsics' -S < %s | FileCheck %s
+
+
+define <8 x double> @fadd_transpose(<8 x double> %a, <8 x double> %b) {
+; CHECK-LABEL: @fadd_transpose(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x double> [[A:%.*]], <8 x double> undef, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <2 x i32> <i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <8 x double> [[A]], <8 x double> undef, <2 x i32> <i32 6, i32 7>
+; CHECK-NEXT:    [[SPLIT4:%.*]] = shufflevector <8 x double> [[B:%.*]], <8 x double> undef, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[SPLIT5:%.*]] = shufflevector <8 x double> [[B]], <8 x double> undef, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[SPLIT6:%.*]] = shufflevector <8 x double> [[B]], <8 x double> undef, <2 x i32> <i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT7:%.*]] = shufflevector <8 x double> [[B]], <8 x double> undef, <2 x i32> <i32 6, i32 7>
+; CHECK-NEXT:    [[TMP0:%.*]] = fadd <2 x double> [[SPLIT]], [[SPLIT4]]
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <2 x double> [[SPLIT1]], [[SPLIT5]]
+; CHECK-NEXT:    [[TMP2:%.*]] = fadd <2 x double> [[SPLIT2]], [[SPLIT6]]
+; CHECK-NEXT:    [[TMP3:%.*]] = fadd <2 x double> [[SPLIT3]], [[SPLIT7]]
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x double> [[TMP0]], i64 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x double> undef, double [[TMP4]], i64 0
+; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x double> [[TMP1]], i64 0
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x double> [[TMP5]], double [[TMP6]], i64 1
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x double> [[TMP2]], i64 0
+; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x double> [[TMP7]], double [[TMP8]], i64 2
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x double> [[TMP3]], i64 0
+; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <4 x double> [[TMP9]], double [[TMP10]], i64 3
+; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <2 x double> [[TMP0]], i64 1
+; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x double> undef, double [[TMP12]], i64 0
+; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x double> [[TMP1]], i64 1
+; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x double> [[TMP13]], double [[TMP14]], i64 1
+; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <2 x double> [[TMP2]], i64 1
+; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x double> [[TMP15]], double [[TMP16]], i64 2
+; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <2 x double> [[TMP3]], i64 1
+; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP18]], i64 3
+; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <4 x double> [[TMP11]], <4 x double> [[TMP19]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    ret <8 x double> [[TMP20]]
+;
+entry:
+  %add = fadd <8 x double> %a, %b
+  %c  = call <8 x double> @llvm.matrix.transpose(<8 x double> %add, i32 2, i32 4)
+  ret <8 x double> %c
+}
+
+declare <8 x double> @llvm.matrix.transpose(<8 x double>, i32, i32)
diff --git a/llvm/test/Transforms/LowerMatrixIntrinsics/propagate-backwards-unsupported.ll b/llvm/test/Transforms/LowerMatrixIntrinsics/propagate-backwards-unsupported.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/Transforms/LowerMatrixIntrinsics/propagate-backwards-unsupported.ll
@@ -0,0 +1,480 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -lower-matrix-intrinsics -S < %s | FileCheck %s
+; RUN: opt -passes='lower-matrix-intrinsics' -S < %s | FileCheck %s
+
+define <9 x double> @unsupported_phi(i1 %cond, <9 x double> %A, <9 x double> %B, <9 x double> %C) {
+; CHECK-LABEL: @unsupported_phi(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 [[COND:%.*]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <9 x double> [[A:%.*]], <9 x double> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT:    [[SPLIT4:%.*]] = shufflevector <9 x double> [[A]], <9 x double> undef, <3 x i32> <i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT5:%.*]] = shufflevector <9 x double> [[A]], <9 x double> undef, <3 x i32> <i32 6, i32 7, i32 8>
+; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <3 x double> [[SPLIT3]], i64 0
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <3 x double> undef, double [[TMP0]], i64 0
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <3 x double> [[SPLIT4]], i64 0
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <3 x double> [[TMP1]], double [[TMP2]], i64 1
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <3 x double> [[SPLIT5]], i64 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <3 x double> [[TMP3]], double [[TMP4]], i64 2
+; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <3 x double> [[SPLIT3]], i64 1
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <3 x double> undef, double [[TMP6]], i64 0
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <3 x double> [[SPLIT4]], i64 1
+; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <3 x double> [[TMP7]], double [[TMP8]], i64 1
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <3 x double> [[SPLIT5]], i64 1
+; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <3 x double> [[TMP9]], double [[TMP10]], i64 2
+; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <3 x double> [[SPLIT3]], i64 2
+; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <3 x double> undef, double [[TMP12]], i64 0
+; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <3 x double> [[SPLIT4]], i64 2
+; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <3 x double> [[TMP13]], double [[TMP14]], i64 1
+; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <3 x double> [[SPLIT5]], i64 2
+; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <3 x double> [[TMP15]], double [[TMP16]], i64 2
+; CHECK-NEXT:    [[TMP18:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> [[TMP11]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[TMP19:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> undef, <6 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <6 x double> [[TMP18]], <6 x double> [[TMP19]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
+; CHECK-NEXT:    br label [[IF_END:%.*]]
+; CHECK:       if.else:
+; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <9 x double> [[B:%.*]], <9 x double> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <9 x double> [[B]], <9 x double> undef, <3 x i32> <i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <9 x double> [[B]], <9 x double> undef, <3 x i32> <i32 6, i32 7, i32 8>
+; CHECK-NEXT:    [[TMP21:%.*]] = extractelement <3 x double> [[SPLIT]], i64 0
+; CHECK-NEXT:    [[TMP22:%.*]] = insertelement <3 x double> undef, double [[TMP21]], i64 0
+; CHECK-NEXT:    [[TMP23:%.*]] = extractelement <3 x double> [[SPLIT1]], i64 0
+; CHECK-NEXT:    [[TMP24:%.*]] = insertelement <3 x double> [[TMP22]], double [[TMP23]], i64 1
+; CHECK-NEXT:    [[TMP25:%.*]] = extractelement <3 x double> [[SPLIT2]], i64 0
+; CHECK-NEXT:    [[TMP26:%.*]] = insertelement <3 x double> [[TMP24]], double [[TMP25]], i64 2
+; CHECK-NEXT:    [[TMP27:%.*]] = extractelement <3 x double> [[SPLIT]], i64 1
+; CHECK-NEXT:    [[TMP28:%.*]] = insertelement <3 x double> undef, double [[TMP27]], i64 0
+; CHECK-NEXT:    [[TMP29:%.*]] = extractelement <3 x double> [[SPLIT1]], i64 1
+; CHECK-NEXT:    [[TMP30:%.*]] = insertelement <3 x double> [[TMP28]], double [[TMP29]], i64 1
+; CHECK-NEXT:    [[TMP31:%.*]] = extractelement <3 x double> [[SPLIT2]], i64 1
+; CHECK-NEXT:    [[TMP32:%.*]] = insertelement <3 x double> [[TMP30]], double [[TMP31]], i64 2
+; CHECK-NEXT:    [[TMP33:%.*]] = extractelement <3 x double> [[SPLIT]], i64 2
+; CHECK-NEXT:    [[TMP34:%.*]] = insertelement <3 x double> undef, double [[TMP33]], i64 0
+; CHECK-NEXT:    [[TMP35:%.*]] = extractelement <3 x double> [[SPLIT1]], i64 2
+; CHECK-NEXT:    [[TMP36:%.*]] = insertelement <3 x double> [[TMP34]], double [[TMP35]], i64 1
+; CHECK-NEXT:    [[TMP37:%.*]] = extractelement <3 x double> [[SPLIT2]], i64 2
+; CHECK-NEXT:    [[TMP38:%.*]] = insertelement <3 x double> [[TMP36]], double [[TMP37]], i64 2
+; CHECK-NEXT:    [[TMP39:%.*]] = shufflevector <3 x double> [[TMP26]], <3 x double> [[TMP32]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[TMP40:%.*]] = shufflevector <3 x double> [[TMP38]], <3 x double> undef, <6 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP41:%.*]] = shufflevector <6 x double> [[TMP39]], <6 x double> [[TMP40]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
+; CHECK-NEXT:    br label [[IF_END]]
+; CHECK:       if.end:
+; CHECK-NEXT:    [[MERGE:%.*]] = phi <9 x double> [ [[TMP20]], [[IF_THEN]] ], [ [[TMP41]], [[IF_ELSE]] ]
+; CHECK-NEXT:    [[SPLIT6:%.*]] = shufflevector <9 x double> [[C:%.*]], <9 x double> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT:    [[SPLIT7:%.*]] = shufflevector <9 x double> [[C]], <9 x double> undef, <3 x i32> <i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT8:%.*]] = shufflevector <9 x double> [[C]], <9 x double> undef, <3 x i32> <i32 6, i32 7, i32 8>
+; CHECK-NEXT:    [[SPLIT9:%.*]] = shufflevector <9 x double> [[MERGE]], <9 x double> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT:    [[SPLIT10:%.*]] = shufflevector <9 x double> [[MERGE]], <9 x double> undef, <3 x i32> <i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT11:%.*]] = shufflevector <9 x double> [[MERGE]], <9 x double> undef, <3 x i32> <i32 6, i32 7, i32 8>
+; CHECK-NEXT:    [[BLOCK:%.*]] = shufflevector <3 x double> [[SPLIT6]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP42:%.*]] = extractelement <3 x double> [[SPLIT9]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x double> undef, double [[TMP42]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP43:%.*]] = fmul <1 x double> [[BLOCK]], [[SPLAT_SPLAT]]
+; CHECK-NEXT:    [[BLOCK12:%.*]] = shufflevector <3 x double> [[SPLIT7]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP44:%.*]] = extractelement <3 x double> [[SPLIT9]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT13:%.*]] = insertelement <1 x double> undef, double [[TMP44]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT14:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT13]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP45:%.*]] = fmul <1 x double> [[BLOCK12]], [[SPLAT_SPLAT14]]
+; CHECK-NEXT:    [[TMP46:%.*]] = fadd <1 x double> [[TMP43]], [[TMP45]]
+; CHECK-NEXT:    [[BLOCK15:%.*]] = shufflevector <3 x double> [[SPLIT8]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP47:%.*]] = extractelement <3 x double> [[SPLIT9]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT16:%.*]] = insertelement <1 x double> undef, double [[TMP47]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT17:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT16]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP48:%.*]] = fmul <1 x double> [[BLOCK15]], [[SPLAT_SPLAT17]]
+; CHECK-NEXT:    [[TMP49:%.*]] = fadd <1 x double> [[TMP46]], [[TMP48]]
+; CHECK-NEXT:    [[TMP50:%.*]] = shufflevector <1 x double> [[TMP49]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP51:%.*]] = shufflevector <3 x double> undef, <3 x double> [[TMP50]], <3 x i32> <i32 3, i32 1, i32 2>
+; CHECK-NEXT:    [[BLOCK18:%.*]] = shufflevector <3 x double> [[SPLIT6]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP52:%.*]] = extractelement <3 x double> [[SPLIT9]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT19:%.*]] = insertelement <1 x double> undef, double [[TMP52]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT20:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT19]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP53:%.*]] = fmul <1 x double> [[BLOCK18]], [[SPLAT_SPLAT20]]
+; CHECK-NEXT:    [[BLOCK21:%.*]] = shufflevector <3 x double> [[SPLIT7]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP54:%.*]] = extractelement <3 x double> [[SPLIT9]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT22:%.*]] = insertelement <1 x double> undef, double [[TMP54]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT23:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT22]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP55:%.*]] = fmul <1 x double> [[BLOCK21]], [[SPLAT_SPLAT23]]
+; CHECK-NEXT:    [[TMP56:%.*]] = fadd <1 x double> [[TMP53]], [[TMP55]]
+; CHECK-NEXT:    [[BLOCK24:%.*]] = shufflevector <3 x double> [[SPLIT8]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP57:%.*]] = extractelement <3 x double> [[SPLIT9]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT25:%.*]] = insertelement <1 x double> undef, double [[TMP57]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT26:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT25]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP58:%.*]] = fmul <1 x double> [[BLOCK24]], [[SPLAT_SPLAT26]]
+; CHECK-NEXT:    [[TMP59:%.*]] = fadd <1 x double> [[TMP56]], [[TMP58]]
+; CHECK-NEXT:    [[TMP60:%.*]] = shufflevector <1 x double> [[TMP59]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP61:%.*]] = shufflevector <3 x double> [[TMP51]], <3 x double> [[TMP60]], <3 x i32> <i32 0, i32 3, i32 2>
+; CHECK-NEXT:    [[BLOCK27:%.*]] = shufflevector <3 x double> [[SPLIT6]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP62:%.*]] = extractelement <3 x double> [[SPLIT9]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT28:%.*]] = insertelement <1 x double> undef, double [[TMP62]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT29:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT28]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP63:%.*]] = fmul <1 x double> [[BLOCK27]], [[SPLAT_SPLAT29]]
+; CHECK-NEXT:    [[BLOCK30:%.*]] = shufflevector <3 x double> [[SPLIT7]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP64:%.*]] = extractelement <3 x double> [[SPLIT9]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT31:%.*]] = insertelement <1 x double> undef, double [[TMP64]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT32:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT31]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP65:%.*]] = fmul <1 x double> [[BLOCK30]], [[SPLAT_SPLAT32]]
+; CHECK-NEXT:    [[TMP66:%.*]] = fadd <1 x double> [[TMP63]], [[TMP65]]
+; CHECK-NEXT:    [[BLOCK33:%.*]] = shufflevector <3 x double> [[SPLIT8]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP67:%.*]] = extractelement <3 x double> [[SPLIT9]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT34:%.*]] = insertelement <1 x double> undef, double [[TMP67]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT35:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT34]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP68:%.*]] = fmul <1 x double> [[BLOCK33]], [[SPLAT_SPLAT35]]
+; CHECK-NEXT:    [[TMP69:%.*]] = fadd <1 x double> [[TMP66]], [[TMP68]]
+; CHECK-NEXT:    [[TMP70:%.*]] = shufflevector <1 x double> [[TMP69]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP71:%.*]] = shufflevector <3 x double> [[TMP61]], <3 x double> [[TMP70]], <3 x i32> <i32 0, i32 1, i32 3>
+; CHECK-NEXT:    [[BLOCK36:%.*]] = shufflevector <3 x double> [[SPLIT6]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP72:%.*]] = extractelement <3 x double> [[SPLIT10]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT37:%.*]] = insertelement <1 x double> undef, double [[TMP72]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT38:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT37]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP73:%.*]] = fmul <1 x double> [[BLOCK36]], [[SPLAT_SPLAT38]]
+; CHECK-NEXT:    [[BLOCK39:%.*]] = shufflevector <3 x double> [[SPLIT7]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP74:%.*]] = extractelement <3 x double> [[SPLIT10]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT40:%.*]] = insertelement <1 x double> undef, double [[TMP74]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT41:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT40]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP75:%.*]] = fmul <1 x double> [[BLOCK39]], [[SPLAT_SPLAT41]]
+; CHECK-NEXT:    [[TMP76:%.*]] = fadd <1 x double> [[TMP73]], [[TMP75]]
+; CHECK-NEXT:    [[BLOCK42:%.*]] = shufflevector <3 x double> [[SPLIT8]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP77:%.*]] = extractelement <3 x double> [[SPLIT10]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT43:%.*]] = insertelement <1 x double> undef, double [[TMP77]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT44:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT43]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP78:%.*]] = fmul <1 x double> [[BLOCK42]], [[SPLAT_SPLAT44]]
+; CHECK-NEXT:    [[TMP79:%.*]] = fadd <1 x double> [[TMP76]], [[TMP78]]
+; CHECK-NEXT:    [[TMP80:%.*]] = shufflevector <1 x double> [[TMP79]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP81:%.*]] = shufflevector <3 x double> undef, <3 x double> [[TMP80]], <3 x i32> <i32 3, i32 1, i32 2>
+; CHECK-NEXT:    [[BLOCK45:%.*]] = shufflevector <3 x double> [[SPLIT6]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP82:%.*]] = extractelement <3 x double> [[SPLIT10]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT46:%.*]] = insertelement <1 x double> undef, double [[TMP82]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT47:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT46]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP83:%.*]] = fmul <1 x double> [[BLOCK45]], [[SPLAT_SPLAT47]]
+; CHECK-NEXT:    [[BLOCK48:%.*]] = shufflevector <3 x double> [[SPLIT7]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP84:%.*]] = extractelement <3 x double> [[SPLIT10]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT49:%.*]] = insertelement <1 x double> undef, double [[TMP84]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT50:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT49]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP85:%.*]] = fmul <1 x double> [[BLOCK48]], [[SPLAT_SPLAT50]]
+; CHECK-NEXT:    [[TMP86:%.*]] = fadd <1 x double> [[TMP83]], [[TMP85]]
+; CHECK-NEXT:    [[BLOCK51:%.*]] = shufflevector <3 x double> [[SPLIT8]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP87:%.*]] = extractelement <3 x double> [[SPLIT10]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT52:%.*]] = insertelement <1 x double> undef, double [[TMP87]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT53:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT52]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP88:%.*]] = fmul <1 x double> [[BLOCK51]], [[SPLAT_SPLAT53]]
+; CHECK-NEXT:    [[TMP89:%.*]] = fadd <1 x double> [[TMP86]], [[TMP88]]
+; CHECK-NEXT:    [[TMP90:%.*]] = shufflevector <1 x double> [[TMP89]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP91:%.*]] = shufflevector <3 x double> [[TMP81]], <3 x double> [[TMP90]], <3 x i32> <i32 0, i32 3, i32 2>
+; CHECK-NEXT:    [[BLOCK54:%.*]] = shufflevector <3 x double> [[SPLIT6]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP92:%.*]] = extractelement <3 x double> [[SPLIT10]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT55:%.*]] = insertelement <1 x double> undef, double [[TMP92]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT56:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT55]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP93:%.*]] = fmul <1 x double> [[BLOCK54]], [[SPLAT_SPLAT56]]
+; CHECK-NEXT:    [[BLOCK57:%.*]] = shufflevector <3 x double> [[SPLIT7]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP94:%.*]] = extractelement <3 x double> [[SPLIT10]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT58:%.*]] = insertelement <1 x double> undef, double [[TMP94]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT59:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT58]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP95:%.*]] = fmul <1 x double> [[BLOCK57]], [[SPLAT_SPLAT59]]
+; CHECK-NEXT:    [[TMP96:%.*]] = fadd <1 x double> [[TMP93]], [[TMP95]]
+; CHECK-NEXT:    [[BLOCK60:%.*]] = shufflevector <3 x double> [[SPLIT8]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP97:%.*]] = extractelement <3 x double> [[SPLIT10]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT61:%.*]] = insertelement <1 x double> undef, double [[TMP97]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT62:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT61]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP98:%.*]] = fmul <1 x double> [[BLOCK60]], [[SPLAT_SPLAT62]]
+; CHECK-NEXT:    [[TMP99:%.*]] = fadd <1 x double> [[TMP96]], [[TMP98]]
+; CHECK-NEXT:    [[TMP100:%.*]] = shufflevector <1 x double> [[TMP99]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP101:%.*]] = shufflevector <3 x double> [[TMP91]], <3 x double> [[TMP100]], <3 x i32> <i32 0, i32 1, i32 3>
+; CHECK-NEXT:    [[BLOCK63:%.*]] = shufflevector <3 x double> [[SPLIT6]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP102:%.*]] = extractelement <3 x double> [[SPLIT11]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT64:%.*]] = insertelement <1 x double> undef, double [[TMP102]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT65:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT64]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP103:%.*]] = fmul <1 x double> [[BLOCK63]], [[SPLAT_SPLAT65]]
+; CHECK-NEXT:    [[BLOCK66:%.*]] = shufflevector <3 x double> [[SPLIT7]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP104:%.*]] = extractelement <3 x double> [[SPLIT11]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT67:%.*]] = insertelement <1 x double> undef, double [[TMP104]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT68:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT67]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP105:%.*]] = fmul <1 x double> [[BLOCK66]], [[SPLAT_SPLAT68]]
+; CHECK-NEXT:    [[TMP106:%.*]] = fadd <1 x double> [[TMP103]], [[TMP105]]
+; CHECK-NEXT:    [[BLOCK69:%.*]] = shufflevector <3 x double> [[SPLIT8]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP107:%.*]] = extractelement <3 x double> [[SPLIT11]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT70:%.*]] = insertelement <1 x double> undef, double [[TMP107]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT71:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT70]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP108:%.*]] = fmul <1 x double> [[BLOCK69]], [[SPLAT_SPLAT71]]
+; CHECK-NEXT:    [[TMP109:%.*]] = fadd <1 x double> [[TMP106]], [[TMP108]]
+; CHECK-NEXT:    [[TMP110:%.*]] = shufflevector <1 x double> [[TMP109]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP111:%.*]] = shufflevector <3 x double> undef, <3 x double> [[TMP110]], <3 x i32> <i32 3, i32 1, i32 2>
+; CHECK-NEXT:    [[BLOCK72:%.*]] = shufflevector <3 x double> [[SPLIT6]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP112:%.*]] = extractelement <3 x double> [[SPLIT11]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT73:%.*]] = insertelement <1 x double> undef, double [[TMP112]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT74:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT73]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP113:%.*]] = fmul <1 x double> [[BLOCK72]], [[SPLAT_SPLAT74]]
+; CHECK-NEXT:    [[BLOCK75:%.*]] = shufflevector <3 x double> [[SPLIT7]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP114:%.*]] = extractelement <3 x double> [[SPLIT11]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT76:%.*]] = insertelement <1 x double> undef, double [[TMP114]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT77:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT76]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP115:%.*]] = fmul <1 x double> [[BLOCK75]], [[SPLAT_SPLAT77]]
+; CHECK-NEXT:    [[TMP116:%.*]] = fadd <1 x double> [[TMP113]], [[TMP115]]
+; CHECK-NEXT:    [[BLOCK78:%.*]] = shufflevector <3 x double> [[SPLIT8]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP117:%.*]] = extractelement <3 x double> [[SPLIT11]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT79:%.*]] = insertelement <1 x double> undef, double [[TMP117]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT80:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT79]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP118:%.*]] = fmul <1 x double> [[BLOCK78]], [[SPLAT_SPLAT80]]
+; CHECK-NEXT:    [[TMP119:%.*]] = fadd <1 x double> [[TMP116]], [[TMP118]]
+; CHECK-NEXT:    [[TMP120:%.*]] = shufflevector <1 x double> [[TMP119]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP121:%.*]] = shufflevector <3 x double> [[TMP111]], <3 x double> [[TMP120]], <3 x i32> <i32 0, i32 3, i32 2>
+; CHECK-NEXT:    [[BLOCK81:%.*]] = shufflevector <3 x double> [[SPLIT6]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP122:%.*]] = extractelement <3 x double> [[SPLIT11]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT82:%.*]] = insertelement <1 x double> undef, double [[TMP122]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT83:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT82]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP123:%.*]] = fmul <1 x double> [[BLOCK81]], [[SPLAT_SPLAT83]]
+; CHECK-NEXT:    [[BLOCK84:%.*]] = shufflevector <3 x double> [[SPLIT7]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP124:%.*]] = extractelement <3 x double> [[SPLIT11]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT85:%.*]] = insertelement <1 x double> undef, double [[TMP124]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT86:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT85]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP125:%.*]] = fmul <1 x double> [[BLOCK84]], [[SPLAT_SPLAT86]]
+; CHECK-NEXT:    [[TMP126:%.*]] = fadd <1 x double> [[TMP123]], [[TMP125]]
+; CHECK-NEXT:    [[BLOCK87:%.*]] = shufflevector <3 x double> [[SPLIT8]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP127:%.*]] = extractelement <3 x double> [[SPLIT11]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT88:%.*]] = insertelement <1 x double> undef, double [[TMP127]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT89:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT88]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP128:%.*]] = fmul <1 x double> [[BLOCK87]], [[SPLAT_SPLAT89]]
+; CHECK-NEXT:    [[TMP129:%.*]] = fadd <1 x double> [[TMP126]], [[TMP128]]
+; CHECK-NEXT:    [[TMP130:%.*]] = shufflevector <1 x double> [[TMP129]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP131:%.*]] = shufflevector <3 x double> [[TMP121]], <3 x double> [[TMP130]], <3 x i32> <i32 0, i32 1, i32 3>
+; CHECK-NEXT:    [[TMP132:%.*]] = shufflevector <3 x double> [[TMP71]], <3 x double> [[TMP101]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[TMP133:%.*]] = shufflevector <3 x double> [[TMP131]], <3 x double> undef, <6 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP134:%.*]] = shufflevector <6 x double> [[TMP132]], <6 x double> [[TMP133]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
+; CHECK-NEXT:    ret <9 x double> [[TMP134]]
+;
+entry:
+  br i1 %cond, label %if.then, label %if.else
+
+if.then:                    ; preds = %entry
+  %A.trans = tail call <9 x double> @llvm.matrix.transpose.v9f64(<9 x double> %A, i32 3, i32 3)
+  br label %if.end
+
+if.else:                                       ; preds = %entry
+  %B.trans = tail call <9 x double> @llvm.matrix.transpose.v9f64(<9 x double> %B, i32 3, i32 3)
+  br label %if.end
+
+if.end:                                        ; preds = %if.then, %if.else
+  %merge = phi <9 x double> [ %A.trans, %if.then], [ %B.trans, %if.else ]
+  %res = tail call <9 x double> @llvm.matrix.multiply.v9f64.v9f64.v9f64(<9 x double> %C, <9 x double> %merge, i32 3, i32 3, i32 3)
+  ret <9 x double> %res
+}
+
+; Make sure we use a flattened vector when calling @foo and the use its flat vector result properly.
+define <9 x double> @unsupported_call(i1 %cond, <9 x double> %A, <9 x double> %B) {
+; CHECK-LABEL: @unsupported_call(
+; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <9 x double> [[A:%.*]], <9 x double> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <9 x double> [[A]], <9 x double> undef, <3 x i32> <i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <9 x double> [[A]], <9 x double> undef, <3 x i32> <i32 6, i32 7, i32 8>
+; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <3 x double> [[SPLIT]], i64 0
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <3 x double> undef, double [[TMP1]], i64 0
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <3 x double> [[SPLIT1]], i64 0
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <3 x double> [[TMP2]], double [[TMP3]], i64 1
+; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <3 x double> [[SPLIT2]], i64 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <3 x double> [[TMP4]], double [[TMP5]], i64 2
+; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <3 x double> [[SPLIT]], i64 1
+; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <3 x double> undef, double [[TMP7]], i64 0
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <3 x double> [[SPLIT1]], i64 1
+; CHECK-NEXT:    [[TMP10:%.*]] = insertelement <3 x double> [[TMP8]], double [[TMP9]], i64 1
+; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <3 x double> [[SPLIT2]], i64 1
+; CHECK-NEXT:    [[TMP12:%.*]] = insertelement <3 x double> [[TMP10]], double [[TMP11]], i64 2
+; CHECK-NEXT:    [[TMP13:%.*]] = extractelement <3 x double> [[SPLIT]], i64 2
+; CHECK-NEXT:    [[TMP14:%.*]] = insertelement <3 x double> undef, double [[TMP13]], i64 0
+; CHECK-NEXT:    [[TMP15:%.*]] = extractelement <3 x double> [[SPLIT1]], i64 2
+; CHECK-NEXT:    [[TMP16:%.*]] = insertelement <3 x double> [[TMP14]], double [[TMP15]], i64 1
+; CHECK-NEXT:    [[TMP17:%.*]] = extractelement <3 x double> [[SPLIT2]], i64 2
+; CHECK-NEXT:    [[TMP18:%.*]] = insertelement <3 x double> [[TMP16]], double [[TMP17]], i64 2
+; CHECK-NEXT:    [[TMP19:%.*]] = shufflevector <3 x double> [[TMP6]], <3 x double> [[TMP12]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <3 x double> [[TMP18]], <3 x double> undef, <6 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP21:%.*]] = shufflevector <6 x double> [[TMP19]], <6 x double> [[TMP20]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
+; CHECK-NEXT:    [[A_FOO:%.*]] = call <9 x double> @foo(<9 x double> [[TMP21]])
+; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <9 x double> [[B:%.*]], <9 x double> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT:    [[SPLIT4:%.*]] = shufflevector <9 x double> [[B]], <9 x double> undef, <3 x i32> <i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT5:%.*]] = shufflevector <9 x double> [[B]], <9 x double> undef, <3 x i32> <i32 6, i32 7, i32 8>
+; CHECK-NEXT:    [[SPLIT6:%.*]] = shufflevector <9 x double> [[A_FOO]], <9 x double> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT:    [[SPLIT7:%.*]] = shufflevector <9 x double> [[A_FOO]], <9 x double> undef, <3 x i32> <i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[SPLIT8:%.*]] = shufflevector <9 x double> [[A_FOO]], <9 x double> undef, <3 x i32> <i32 6, i32 7, i32 8>
+; CHECK-NEXT:    [[BLOCK:%.*]] = shufflevector <3 x double> [[SPLIT3]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP22:%.*]] = extractelement <3 x double> [[SPLIT6]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x double> undef, double [[TMP22]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP23:%.*]] = fmul <1 x double> [[BLOCK]], [[SPLAT_SPLAT]]
+; CHECK-NEXT:    [[BLOCK9:%.*]] = shufflevector <3 x double> [[SPLIT4]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP24:%.*]] = extractelement <3 x double> [[SPLIT6]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT10:%.*]] = insertelement <1 x double> undef, double [[TMP24]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT11:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT10]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP25:%.*]] = fmul <1 x double> [[BLOCK9]], [[SPLAT_SPLAT11]]
+; CHECK-NEXT:    [[TMP26:%.*]] = fadd <1 x double> [[TMP23]], [[TMP25]]
+; CHECK-NEXT:    [[BLOCK12:%.*]] = shufflevector <3 x double> [[SPLIT5]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP27:%.*]] = extractelement <3 x double> [[SPLIT6]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT13:%.*]] = insertelement <1 x double> undef, double [[TMP27]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT14:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT13]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP28:%.*]] = fmul <1 x double> [[BLOCK12]], [[SPLAT_SPLAT14]]
+; CHECK-NEXT:    [[TMP29:%.*]] = fadd <1 x double> [[TMP26]], [[TMP28]]
+; CHECK-NEXT:    [[TMP30:%.*]] = shufflevector <1 x double> [[TMP29]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP31:%.*]] = shufflevector <3 x double> undef, <3 x double> [[TMP30]], <3 x i32> <i32 3, i32 1, i32 2>
+; CHECK-NEXT:    [[BLOCK15:%.*]] = shufflevector <3 x double> [[SPLIT3]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP32:%.*]] = extractelement <3 x double> [[SPLIT6]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT16:%.*]] = insertelement <1 x double> undef, double [[TMP32]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT17:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT16]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP33:%.*]] = fmul <1 x double> [[BLOCK15]], [[SPLAT_SPLAT17]]
+; CHECK-NEXT:    [[BLOCK18:%.*]] = shufflevector <3 x double> [[SPLIT4]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP34:%.*]] = extractelement <3 x double> [[SPLIT6]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT19:%.*]] = insertelement <1 x double> undef, double [[TMP34]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT20:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT19]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP35:%.*]] = fmul <1 x double> [[BLOCK18]], [[SPLAT_SPLAT20]]
+; CHECK-NEXT:    [[TMP36:%.*]] = fadd <1 x double> [[TMP33]], [[TMP35]]
+; CHECK-NEXT:    [[BLOCK21:%.*]] = shufflevector <3 x double> [[SPLIT5]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP37:%.*]] = extractelement <3 x double> [[SPLIT6]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT22:%.*]] = insertelement <1 x double> undef, double [[TMP37]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT23:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT22]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP38:%.*]] = fmul <1 x double> [[BLOCK21]], [[SPLAT_SPLAT23]]
+; CHECK-NEXT:    [[TMP39:%.*]] = fadd <1 x double> [[TMP36]], [[TMP38]]
+; CHECK-NEXT:    [[TMP40:%.*]] = shufflevector <1 x double> [[TMP39]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP41:%.*]] = shufflevector <3 x double> [[TMP31]], <3 x double> [[TMP40]], <3 x i32> <i32 0, i32 3, i32 2>
+; CHECK-NEXT:    [[BLOCK24:%.*]] = shufflevector <3 x double> [[SPLIT3]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP42:%.*]] = extractelement <3 x double> [[SPLIT6]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT25:%.*]] = insertelement <1 x double> undef, double [[TMP42]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT26:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT25]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP43:%.*]] = fmul <1 x double> [[BLOCK24]], [[SPLAT_SPLAT26]]
+; CHECK-NEXT:    [[BLOCK27:%.*]] = shufflevector <3 x double> [[SPLIT4]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP44:%.*]] = extractelement <3 x double> [[SPLIT6]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT28:%.*]] = insertelement <1 x double> undef, double [[TMP44]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT29:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT28]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP45:%.*]] = fmul <1 x double> [[BLOCK27]], [[SPLAT_SPLAT29]]
+; CHECK-NEXT:    [[TMP46:%.*]] = fadd <1 x double> [[TMP43]], [[TMP45]]
+; CHECK-NEXT:    [[BLOCK30:%.*]] = shufflevector <3 x double> [[SPLIT5]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP47:%.*]] = extractelement <3 x double> [[SPLIT6]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT31:%.*]] = insertelement <1 x double> undef, double [[TMP47]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT32:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT31]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP48:%.*]] = fmul <1 x double> [[BLOCK30]], [[SPLAT_SPLAT32]]
+; CHECK-NEXT:    [[TMP49:%.*]] = fadd <1 x double> [[TMP46]], [[TMP48]]
+; CHECK-NEXT:    [[TMP50:%.*]] = shufflevector <1 x double> [[TMP49]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP51:%.*]] = shufflevector <3 x double> [[TMP41]], <3 x double> [[TMP50]], <3 x i32> <i32 0, i32 1, i32 3>
+; CHECK-NEXT:    [[BLOCK33:%.*]] = shufflevector <3 x double> [[SPLIT3]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP52:%.*]] = extractelement <3 x double> [[SPLIT7]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT34:%.*]] = insertelement <1 x double> undef, double [[TMP52]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT35:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT34]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP53:%.*]] = fmul <1 x double> [[BLOCK33]], [[SPLAT_SPLAT35]]
+; CHECK-NEXT:    [[BLOCK36:%.*]] = shufflevector <3 x double> [[SPLIT4]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP54:%.*]] = extractelement <3 x double> [[SPLIT7]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT37:%.*]] = insertelement <1 x double> undef, double [[TMP54]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT38:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT37]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP55:%.*]] = fmul <1 x double> [[BLOCK36]], [[SPLAT_SPLAT38]]
+; CHECK-NEXT:    [[TMP56:%.*]] = fadd <1 x double> [[TMP53]], [[TMP55]]
+; CHECK-NEXT:    [[BLOCK39:%.*]] = shufflevector <3 x double> [[SPLIT5]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP57:%.*]] = extractelement <3 x double> [[SPLIT7]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT40:%.*]] = insertelement <1 x double> undef, double [[TMP57]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT41:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT40]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP58:%.*]] = fmul <1 x double> [[BLOCK39]], [[SPLAT_SPLAT41]]
+; CHECK-NEXT:    [[TMP59:%.*]] = fadd <1 x double> [[TMP56]], [[TMP58]]
+; CHECK-NEXT:    [[TMP60:%.*]] = shufflevector <1 x double> [[TMP59]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP61:%.*]] = shufflevector <3 x double> undef, <3 x double> [[TMP60]], <3 x i32> <i32 3, i32 1, i32 2>
+; CHECK-NEXT:    [[BLOCK42:%.*]] = shufflevector <3 x double> [[SPLIT3]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP62:%.*]] = extractelement <3 x double> [[SPLIT7]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT43:%.*]] = insertelement <1 x double> undef, double [[TMP62]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT44:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT43]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP63:%.*]] = fmul <1 x double> [[BLOCK42]], [[SPLAT_SPLAT44]]
+; CHECK-NEXT:    [[BLOCK45:%.*]] = shufflevector <3 x double> [[SPLIT4]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP64:%.*]] = extractelement <3 x double> [[SPLIT7]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT46:%.*]] = insertelement <1 x double> undef, double [[TMP64]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT47:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT46]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP65:%.*]] = fmul <1 x double> [[BLOCK45]], [[SPLAT_SPLAT47]]
+; CHECK-NEXT:    [[TMP66:%.*]] = fadd <1 x double> [[TMP63]], [[TMP65]]
+; CHECK-NEXT:    [[BLOCK48:%.*]] = shufflevector <3 x double> [[SPLIT5]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP67:%.*]] = extractelement <3 x double> [[SPLIT7]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT49:%.*]] = insertelement <1 x double> undef, double [[TMP67]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT50:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT49]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP68:%.*]] = fmul <1 x double> [[BLOCK48]], [[SPLAT_SPLAT50]]
+; CHECK-NEXT:    [[TMP69:%.*]] = fadd <1 x double> [[TMP66]], [[TMP68]]
+; CHECK-NEXT:    [[TMP70:%.*]] = shufflevector <1 x double> [[TMP69]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP71:%.*]] = shufflevector <3 x double> [[TMP61]], <3 x double> [[TMP70]], <3 x i32> <i32 0, i32 3, i32 2>
+; CHECK-NEXT:    [[BLOCK51:%.*]] = shufflevector <3 x double> [[SPLIT3]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP72:%.*]] = extractelement <3 x double> [[SPLIT7]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT52:%.*]] = insertelement <1 x double> undef, double [[TMP72]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT53:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT52]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP73:%.*]] = fmul <1 x double> [[BLOCK51]], [[SPLAT_SPLAT53]]
+; CHECK-NEXT:    [[BLOCK54:%.*]] = shufflevector <3 x double> [[SPLIT4]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP74:%.*]] = extractelement <3 x double> [[SPLIT7]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT55:%.*]] = insertelement <1 x double> undef, double [[TMP74]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT56:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT55]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP75:%.*]] = fmul <1 x double> [[BLOCK54]], [[SPLAT_SPLAT56]]
+; CHECK-NEXT:    [[TMP76:%.*]] = fadd <1 x double> [[TMP73]], [[TMP75]]
+; CHECK-NEXT:    [[BLOCK57:%.*]] = shufflevector <3 x double> [[SPLIT5]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP77:%.*]] = extractelement <3 x double> [[SPLIT7]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT58:%.*]] = insertelement <1 x double> undef, double [[TMP77]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT59:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT58]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP78:%.*]] = fmul <1 x double> [[BLOCK57]], [[SPLAT_SPLAT59]]
+; CHECK-NEXT:    [[TMP79:%.*]] = fadd <1 x double> [[TMP76]], [[TMP78]]
+; CHECK-NEXT:    [[TMP80:%.*]] = shufflevector <1 x double> [[TMP79]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP81:%.*]] = shufflevector <3 x double> [[TMP71]], <3 x double> [[TMP80]], <3 x i32> <i32 0, i32 1, i32 3>
+; CHECK-NEXT:    [[BLOCK60:%.*]] = shufflevector <3 x double> [[SPLIT3]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP82:%.*]] = extractelement <3 x double> [[SPLIT8]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT61:%.*]] = insertelement <1 x double> undef, double [[TMP82]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT62:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT61]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP83:%.*]] = fmul <1 x double> [[BLOCK60]], [[SPLAT_SPLAT62]]
+; CHECK-NEXT:    [[BLOCK63:%.*]] = shufflevector <3 x double> [[SPLIT4]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP84:%.*]] = extractelement <3 x double> [[SPLIT8]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT64:%.*]] = insertelement <1 x double> undef, double [[TMP84]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT65:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT64]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP85:%.*]] = fmul <1 x double> [[BLOCK63]], [[SPLAT_SPLAT65]]
+; CHECK-NEXT:    [[TMP86:%.*]] = fadd <1 x double> [[TMP83]], [[TMP85]]
+; CHECK-NEXT:    [[BLOCK66:%.*]] = shufflevector <3 x double> [[SPLIT5]], <3 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP87:%.*]] = extractelement <3 x double> [[SPLIT8]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT67:%.*]] = insertelement <1 x double> undef, double [[TMP87]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT68:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT67]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP88:%.*]] = fmul <1 x double> [[BLOCK66]], [[SPLAT_SPLAT68]]
+; CHECK-NEXT:    [[TMP89:%.*]] = fadd <1 x double> [[TMP86]], [[TMP88]]
+; CHECK-NEXT:    [[TMP90:%.*]] = shufflevector <1 x double> [[TMP89]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP91:%.*]] = shufflevector <3 x double> undef, <3 x double> [[TMP90]], <3 x i32> <i32 3, i32 1, i32 2>
+; CHECK-NEXT:    [[BLOCK69:%.*]] = shufflevector <3 x double> [[SPLIT3]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP92:%.*]] = extractelement <3 x double> [[SPLIT8]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT70:%.*]] = insertelement <1 x double> undef, double [[TMP92]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT71:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT70]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP93:%.*]] = fmul <1 x double> [[BLOCK69]], [[SPLAT_SPLAT71]]
+; CHECK-NEXT:    [[BLOCK72:%.*]] = shufflevector <3 x double> [[SPLIT4]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP94:%.*]] = extractelement <3 x double> [[SPLIT8]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT73:%.*]] = insertelement <1 x double> undef, double [[TMP94]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT74:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT73]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP95:%.*]] = fmul <1 x double> [[BLOCK72]], [[SPLAT_SPLAT74]]
+; CHECK-NEXT:    [[TMP96:%.*]] = fadd <1 x double> [[TMP93]], [[TMP95]]
+; CHECK-NEXT:    [[BLOCK75:%.*]] = shufflevector <3 x double> [[SPLIT5]], <3 x double> undef, <1 x i32> <i32 1>
+; CHECK-NEXT:    [[TMP97:%.*]] = extractelement <3 x double> [[SPLIT8]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT76:%.*]] = insertelement <1 x double> undef, double [[TMP97]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT77:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT76]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP98:%.*]] = fmul <1 x double> [[BLOCK75]], [[SPLAT_SPLAT77]]
+; CHECK-NEXT:    [[TMP99:%.*]] = fadd <1 x double> [[TMP96]], [[TMP98]]
+; CHECK-NEXT:    [[TMP100:%.*]] = shufflevector <1 x double> [[TMP99]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP101:%.*]] = shufflevector <3 x double> [[TMP91]], <3 x double> [[TMP100]], <3 x i32> <i32 0, i32 3, i32 2>
+; CHECK-NEXT:    [[BLOCK78:%.*]] = shufflevector <3 x double> [[SPLIT3]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP102:%.*]] = extractelement <3 x double> [[SPLIT8]], i64 0
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT79:%.*]] = insertelement <1 x double> undef, double [[TMP102]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT80:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT79]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP103:%.*]] = fmul <1 x double> [[BLOCK78]], [[SPLAT_SPLAT80]]
+; CHECK-NEXT:    [[BLOCK81:%.*]] = shufflevector <3 x double> [[SPLIT4]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP104:%.*]] = extractelement <3 x double> [[SPLIT8]], i64 1
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT82:%.*]] = insertelement <1 x double> undef, double [[TMP104]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT83:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT82]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP105:%.*]] = fmul <1 x double> [[BLOCK81]], [[SPLAT_SPLAT83]]
+; CHECK-NEXT:    [[TMP106:%.*]] = fadd <1 x double> [[TMP103]], [[TMP105]]
+; CHECK-NEXT:    [[BLOCK84:%.*]] = shufflevector <3 x double> [[SPLIT5]], <3 x double> undef, <1 x i32> <i32 2>
+; CHECK-NEXT:    [[TMP107:%.*]] = extractelement <3 x double> [[SPLIT8]], i64 2
+; CHECK-NEXT:    [[SPLAT_SPLATINSERT85:%.*]] = insertelement <1 x double> undef, double [[TMP107]], i32 0
+; CHECK-NEXT:    [[SPLAT_SPLAT86:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT85]], <1 x double> undef, <1 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP108:%.*]] = fmul <1 x double> [[BLOCK84]], [[SPLAT_SPLAT86]]
+; CHECK-NEXT:    [[TMP109:%.*]] = fadd <1 x double> [[TMP106]], [[TMP108]]
+; CHECK-NEXT:    [[TMP110:%.*]] = shufflevector <1 x double> [[TMP109]], <1 x double> undef, <3 x i32> <i32 0, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP111:%.*]] = shufflevector <3 x double> [[TMP101]], <3 x double> [[TMP110]], <3 x i32> <i32 0, i32 1, i32 3>
+; CHECK-NEXT:    [[TMP112:%.*]] = shufflevector <3 x double> [[TMP51]], <3 x double> [[TMP81]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
+; CHECK-NEXT:    [[TMP113:%.*]] = shufflevector <3 x double> [[TMP111]], <3 x double> undef, <6 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[TMP114:%.*]] = shufflevector <6 x double> [[TMP112]], <6 x double> [[TMP113]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
+; CHECK-NEXT:    ret <9 x double> [[TMP114]]
+;
+  %A.trans = tail call <9 x double> @llvm.matrix.transpose.v9f64(<9 x double> %A, i32 3, i32 3)
+  %A.foo = call <9 x double> @foo(<9 x double> %A.trans)
+  %res = tail call <9 x double> @llvm.matrix.multiply.v9f64.v9f64.v9f64(<9 x double> %B, <9 x double> %A.foo, i32 3, i32 3, i32 3)
+  ret <9 x double> %res
+}
+
+declare <9 x double> @llvm.matrix.multiply.v9f64.v9f64.v9f64(<9 x double>, <9 x double>, i32 immarg, i32 immarg, i32 immarg)
+declare <9 x double> @llvm.matrix.transpose.v9f64(<9 x double>, i32 immarg, i32 immarg)
+declare <9 x double> @foo(<9 x double>)