Index: llvm/test/Transforms/VectorCombine/AArch64/load-insert-store.ll
===================================================================
--- /dev/null
+++ llvm/test/Transforms/VectorCombine/AArch64/load-insert-store.ll
@@ -0,0 +1,124 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
+; RUN: opt -mtriple=aarch64 -passes=vector-combine -S %s | FileCheck %s
+
+;; Whether these tests can be optimized depends on AArch64's TTI.
+
+define void @load_insert_x2_store(ptr %A, i16 %B, i16 %C) {
+; CHECK-LABEL: define void @load_insert_x2_store
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[C]], i32 6
+; CHECK-NEXT:    store <16 x i16> [[T2]], ptr [[A]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %C, i32 6
+  store <16 x i16> %t2, ptr %A
+  ret void
+}
+
+define <16 x i16> @load_insert_x2_store_multiple_uses(ptr %A, i16 %B, i16 %C) {
+; CHECK-LABEL: define <16 x i16> @load_insert_x2_store_multiple_uses
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[C]], i32 6
+; CHECK-NEXT:    store <16 x i16> [[T2]], ptr [[A]], align 32
+; CHECK-NEXT:    ret <16 x i16> [[T2]]
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %C, i32 6
+  store <16 x i16> %t2, ptr %A
+  ret <16 x i16> %t2
+}
+
+define void @load_insert_x3_store(ptr %A, i16 %B, i16 %C, i16 %D) {
+; CHECK-LABEL: define void @load_insert_x3_store
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]], i16 [[D:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[C]], i32 6
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <16 x i16> [[T2]], i16 [[D]], i32 8
+; CHECK-NEXT:    store <16 x i16> [[T3]], ptr [[A]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %C, i32 6
+  %t3 = insertelement <16 x i16> %t2, i16 %D, i32 8
+  store <16 x i16> %t3, ptr %A
+  ret void
+}
+
+define <8 x float> @load_insert_x3_store_multiple_uses(ptr %A, float %B, float %C, float %D) {
+; CHECK-LABEL: define <8 x float> @load_insert_x3_store_multiple_uses
+; CHECK-SAME: (ptr [[A:%.*]], float [[B:%.*]], float [[C:%.*]], float [[D:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <8 x float>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <8 x float> [[T0]], float [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <8 x float> [[T1]], float [[C]], i32 6
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <8 x float> [[T2]], float [[D]], i32 8
+; CHECK-NEXT:    store <8 x float> [[T3]], ptr [[A]], align 32
+; CHECK-NEXT:    ret <8 x float> [[T3]]
+;
+entry:
+  %t0 = load <8 x float>, ptr %A
+  %t1 = insertelement <8 x float> %t0, float %B, i32 4
+  %t2 = insertelement <8 x float> %t1, float %C, i32 6
+  %t3 = insertelement <8 x float> %t2, float %D, i32 8
+  store <8 x float> %t3, ptr %A
+  ret <8 x float> %t3
+}
+
+define void @load_insert_x4_store(ptr %A, i16 %B, i16 %C) {
+; CHECK-LABEL: define void @load_insert_x4_store
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[B]], i32 5
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <16 x i16> [[T2]], i16 [[C]], i32 6
+; CHECK-NEXT:    [[T4:%.*]] = insertelement <16 x i16> [[T3]], i16 [[C]], i32 7
+; CHECK-NEXT:    store <16 x i16> [[T4]], ptr [[A]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %B, i32 5
+  %t3 = insertelement <16 x i16> %t2, i16 %C, i32 6
+  %t4 = insertelement <16 x i16> %t3, i16 %C, i32 7
+  store <16 x i16> %t4, ptr %A
+  ret void
+}
+
+define <8 x float> @load_insert_x4_store_multiple_uses(ptr %A, float %B, float %C) {
+; CHECK-LABEL: define <8 x float> @load_insert_x4_store_multiple_uses
+; CHECK-SAME: (ptr [[A:%.*]], float [[B:%.*]], float [[C:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <8 x float>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <8 x float> [[T0]], float [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <8 x float> [[T1]], float [[B]], i32 5
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <8 x float> [[T2]], float [[C]], i32 6
+; CHECK-NEXT:    [[T4:%.*]] = insertelement <8 x float> [[T3]], float [[C]], i32 7
+; CHECK-NEXT:    store <8 x float> [[T4]], ptr [[A]], align 32
+; CHECK-NEXT:    ret <8 x float> [[T4]]
+;
+entry:
+  %t0 = load <8 x float>, ptr %A
+  %t1 = insertelement <8 x float> %t0, float %B, i32 4
+  %t2 = insertelement <8 x float> %t1, float %B, i32 5
+  %t3 = insertelement <8 x float> %t2, float %C, i32 6
+  %t4 = insertelement <8 x float> %t3, float %C, i32 7
+  store <8 x float> %t4, ptr %A
+  ret <8 x float> %t4
+}
Index: llvm/test/Transforms/VectorCombine/RISCV/load-insert-store.ll
===================================================================
--- /dev/null
+++ llvm/test/Transforms/VectorCombine/RISCV/load-insert-store.ll
@@ -0,0 +1,124 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
+; RUN: opt -mtriple=riscv64 -mattr=+v -passes=vector-combine -S %s | FileCheck %s
+
+;; Whether these tests can be optimized depends on RISCV's TTI.
+
+define void @load_insert_x2_store(ptr %A, i16 %B, i16 %C) {
+; CHECK-LABEL: define void @load_insert_x2_store
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[C]], i32 6
+; CHECK-NEXT:    store <16 x i16> [[T2]], ptr [[A]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %C, i32 6
+  store <16 x i16> %t2, ptr %A
+  ret void
+}
+
+define <16 x i16> @load_insert_x2_store_multiple_uses(ptr %A, i16 %B, i16 %C) {
+; CHECK-LABEL: define <16 x i16> @load_insert_x2_store_multiple_uses
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[C]], i32 6
+; CHECK-NEXT:    store <16 x i16> [[T2]], ptr [[A]], align 32
+; CHECK-NEXT:    ret <16 x i16> [[T2]]
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %C, i32 6
+  store <16 x i16> %t2, ptr %A
+  ret <16 x i16> %t2
+}
+
+define void @load_insert_x3_store(ptr %A, i16 %B, i16 %C, i16 %D) {
+; CHECK-LABEL: define void @load_insert_x3_store
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]], i16 [[D:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[C]], i32 6
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <16 x i16> [[T2]], i16 [[D]], i32 8
+; CHECK-NEXT:    store <16 x i16> [[T3]], ptr [[A]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %C, i32 6
+  %t3 = insertelement <16 x i16> %t2, i16 %D, i32 8
+  store <16 x i16> %t3, ptr %A
+  ret void
+}
+
+define <8 x float> @load_insert_x3_store_multiple_uses(ptr %A, float %B, float %C, float %D) {
+; CHECK-LABEL: define <8 x float> @load_insert_x3_store_multiple_uses
+; CHECK-SAME: (ptr [[A:%.*]], float [[B:%.*]], float [[C:%.*]], float [[D:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <8 x float>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <8 x float> [[T0]], float [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <8 x float> [[T1]], float [[C]], i32 6
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <8 x float> [[T2]], float [[D]], i32 8
+; CHECK-NEXT:    store <8 x float> [[T3]], ptr [[A]], align 32
+; CHECK-NEXT:    ret <8 x float> [[T3]]
+;
+entry:
+  %t0 = load <8 x float>, ptr %A
+  %t1 = insertelement <8 x float> %t0, float %B, i32 4
+  %t2 = insertelement <8 x float> %t1, float %C, i32 6
+  %t3 = insertelement <8 x float> %t2, float %D, i32 8
+  store <8 x float> %t3, ptr %A
+  ret <8 x float> %t3
+}
+
+define void @load_insert_x4_store(ptr %A, i16 %B, i16 %C) {
+; CHECK-LABEL: define void @load_insert_x4_store
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[B]], i32 5
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <16 x i16> [[T2]], i16 [[C]], i32 6
+; CHECK-NEXT:    [[T4:%.*]] = insertelement <16 x i16> [[T3]], i16 [[C]], i32 7
+; CHECK-NEXT:    store <16 x i16> [[T4]], ptr [[A]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %B, i32 5
+  %t3 = insertelement <16 x i16> %t2, i16 %C, i32 6
+  %t4 = insertelement <16 x i16> %t3, i16 %C, i32 7
+  store <16 x i16> %t4, ptr %A
+  ret void
+}
+
+define <8 x float> @load_insert_x4_store_multiple_uses(ptr %A, float %B, float %C) {
+; CHECK-LABEL: define <8 x float> @load_insert_x4_store_multiple_uses
+; CHECK-SAME: (ptr [[A:%.*]], float [[B:%.*]], float [[C:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <8 x float>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <8 x float> [[T0]], float [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <8 x float> [[T1]], float [[B]], i32 5
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <8 x float> [[T2]], float [[C]], i32 6
+; CHECK-NEXT:    [[T4:%.*]] = insertelement <8 x float> [[T3]], float [[C]], i32 7
+; CHECK-NEXT:    store <8 x float> [[T4]], ptr [[A]], align 32
+; CHECK-NEXT:    ret <8 x float> [[T4]]
+;
+entry:
+  %t0 = load <8 x float>, ptr %A
+  %t1 = insertelement <8 x float> %t0, float %B, i32 4
+  %t2 = insertelement <8 x float> %t1, float %B, i32 5
+  %t3 = insertelement <8 x float> %t2, float %C, i32 6
+  %t4 = insertelement <8 x float> %t3, float %C, i32 7
+  store <8 x float> %t4, ptr %A
+  ret <8 x float> %t4
+}
Index: llvm/test/Transforms/VectorCombine/X86/load-insert-store.ll
===================================================================
--- /dev/null
+++ llvm/test/Transforms/VectorCombine/X86/load-insert-store.ll
@@ -0,0 +1,124 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
+; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64 -mattr=AVX2 -S %s | FileCheck %s
+
+;; Whether these tests can be optimized depends on x86_64's TTI.
+
+define void @load_insert_x2_store(ptr %A, i16 %B, i16 %C) {
+; CHECK-LABEL: define void @load_insert_x2_store
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[C]], i32 6
+; CHECK-NEXT:    store <16 x i16> [[T2]], ptr [[A]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %C, i32 6
+  store <16 x i16> %t2, ptr %A
+  ret void
+}
+
+define <16 x i16> @load_insert_x2_store_multiple_uses(ptr %A, i16 %B, i16 %C) {
+; CHECK-LABEL: define <16 x i16> @load_insert_x2_store_multiple_uses
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[C]], i32 6
+; CHECK-NEXT:    store <16 x i16> [[T2]], ptr [[A]], align 32
+; CHECK-NEXT:    ret <16 x i16> [[T2]]
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %C, i32 6
+  store <16 x i16> %t2, ptr %A
+  ret <16 x i16> %t2
+}
+
+define void @load_insert_x3_store(ptr %A, i16 %B, i16 %C, i16 %D) {
+; CHECK-LABEL: define void @load_insert_x3_store
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]], i16 [[D:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[C]], i32 6
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <16 x i16> [[T2]], i16 [[D]], i32 8
+; CHECK-NEXT:    store <16 x i16> [[T3]], ptr [[A]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %C, i32 6
+  %t3 = insertelement <16 x i16> %t2, i16 %D, i32 8
+  store <16 x i16> %t3, ptr %A
+  ret void
+}
+
+define <8 x float> @load_insert_x3_store_multiple_uses(ptr %A, float %B, float %C, float %D) {
+; CHECK-LABEL: define <8 x float> @load_insert_x3_store_multiple_uses
+; CHECK-SAME: (ptr [[A:%.*]], float [[B:%.*]], float [[C:%.*]], float [[D:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <8 x float>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <8 x float> [[T0]], float [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <8 x float> [[T1]], float [[C]], i32 6
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <8 x float> [[T2]], float [[D]], i32 8
+; CHECK-NEXT:    store <8 x float> [[T3]], ptr [[A]], align 32
+; CHECK-NEXT:    ret <8 x float> [[T3]]
+;
+entry:
+  %t0 = load <8 x float>, ptr %A
+  %t1 = insertelement <8 x float> %t0, float %B, i32 4
+  %t2 = insertelement <8 x float> %t1, float %C, i32 6
+  %t3 = insertelement <8 x float> %t2, float %D, i32 8
+  store <8 x float> %t3, ptr %A
+  ret <8 x float> %t3
+}
+
+define void @load_insert_x4_store(ptr %A, i16 %B, i16 %C) {
+; CHECK-LABEL: define void @load_insert_x4_store
+; CHECK-SAME: (ptr [[A:%.*]], i16 [[B:%.*]], i16 [[C:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <16 x i16>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <16 x i16> [[T0]], i16 [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <16 x i16> [[T1]], i16 [[B]], i32 5
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <16 x i16> [[T2]], i16 [[C]], i32 6
+; CHECK-NEXT:    [[T4:%.*]] = insertelement <16 x i16> [[T3]], i16 [[C]], i32 7
+; CHECK-NEXT:    store <16 x i16> [[T4]], ptr [[A]], align 32
+; CHECK-NEXT:    ret void
+;
+entry:
+  %t0 = load <16 x i16>, ptr %A
+  %t1 = insertelement <16 x i16> %t0, i16 %B, i32 4
+  %t2 = insertelement <16 x i16> %t1, i16 %B, i32 5
+  %t3 = insertelement <16 x i16> %t2, i16 %C, i32 6
+  %t4 = insertelement <16 x i16> %t3, i16 %C, i32 7
+  store <16 x i16> %t4, ptr %A
+  ret void
+}
+
+define <8 x float> @load_insert_x4_store_multiple_uses(ptr %A, float %B, float %C) {
+; CHECK-LABEL: define <8 x float> @load_insert_x4_store_multiple_uses
+; CHECK-SAME: (ptr [[A:%.*]], float [[B:%.*]], float [[C:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[T0:%.*]] = load <8 x float>, ptr [[A]], align 32
+; CHECK-NEXT:    [[T1:%.*]] = insertelement <8 x float> [[T0]], float [[B]], i32 4
+; CHECK-NEXT:    [[T2:%.*]] = insertelement <8 x float> [[T1]], float [[B]], i32 5
+; CHECK-NEXT:    [[T3:%.*]] = insertelement <8 x float> [[T2]], float [[C]], i32 6
+; CHECK-NEXT:    [[T4:%.*]] = insertelement <8 x float> [[T3]], float [[C]], i32 7
+; CHECK-NEXT:    store <8 x float> [[T4]], ptr [[A]], align 32
+; CHECK-NEXT:    ret <8 x float> [[T4]]
+;
+entry:
+  %t0 = load <8 x float>, ptr %A
+  %t1 = insertelement <8 x float> %t0, float %B, i32 4
+  %t2 = insertelement <8 x float> %t1, float %B, i32 5
+  %t3 = insertelement <8 x float> %t2, float %C, i32 6
+  %t4 = insertelement <8 x float> %t3, float %C, i32 7
+  store <8 x float> %t4, ptr %A
+  ret <8 x float> %t4
+}
Index: llvm/test/Transforms/VectorCombine/load-insert-store.ll
===================================================================
--- llvm/test/Transforms/VectorCombine/load-insert-store.ll
+++ llvm/test/Transforms/VectorCombine/load-insert-store.ll
@@ -16,6 +16,23 @@
   ret void
 }
 
+; To verify the case that there are multiple uses of the vector.
+define <16 x i8> @insert_store_multiple_uses(ptr %q, i8 zeroext %s) {
+; CHECK-LABEL: @insert_store_multiple_uses(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = load <16 x i8>, ptr [[Q:%.*]], align 16
+; CHECK-NEXT:    [[VECINS:%.*]] = insertelement <16 x i8> [[TMP0]], i8 [[S:%.*]], i32 3
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds <16 x i8>, ptr [[Q]], i32 0, i32 3
+; CHECK-NEXT:    store i8 [[S]], ptr [[TMP1]], align 1
+; CHECK-NEXT:    ret <16 x i8> [[VECINS]]
+;
+entry:
+  %0 = load <16 x i8>, ptr %q
+  %vecins = insertelement <16 x i8> %0, i8 %s, i32 3
+  store <16 x i8> %vecins, ptr %q, align 16
+  ret <16 x i8> %vecins
+}
+
 define void @insert_store_i16_align1(ptr %q, i16 zeroext %s) {
 ; CHECK-LABEL: @insert_store_i16_align1(
 ; CHECK-NEXT:  entry: