[ARM] More MVE load/store tests for offsets around the negative limit. NFC

llvm-svn: 370726

3 files changed, 1264 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/Thumb2/mve-ldst-offset.ll b/llvm/test/CodeGen/Thumb2/mve-ldst-offset.ll
index 5a0a6051867..8e4f5172778 100644
--- a/llvm/test/CodeGen/Thumb2/mve-ldst-offset.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-ldst-offset.ll
@@ -196,6 +196,40 @@ entry:
   ret i8* %x
 }
 
+define i8* @ldrhu32_m254(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu32_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #254
+; CHECK-NEXT:    vldrh.u32 q0, [r2]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -254
+  %0 = bitcast i8* %z to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = zext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %x
+}
+
+define i8* @ldrhu32_m256(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu32_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #256
+; CHECK-NEXT:    vldrh.u32 q0, [r2]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -256
+  %0 = bitcast i8* %z to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = zext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %x
+}
+
 
 define i8* @ldrhs32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrhs32_4:
@@ -280,6 +314,40 @@ entry:
   ret i8* %x
 }
 
+define i8* @ldrhs32_m254(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhs32_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #254
+; CHECK-NEXT:    vldrh.s32 q0, [r2]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -254
+  %0 = bitcast i8* %z to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = sext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %x
+}
+
+define i8* @ldrhs32_m256(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhs32_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #256
+; CHECK-NEXT:    vldrh.s32 q0, [r2]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -256
+  %0 = bitcast i8* %z to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = sext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %x
+}
+
 
 define i8* @ldrhu16_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrhu16_4:
@@ -359,6 +427,38 @@ entry:
   ret i8* %x
 }
 
+define i8* @ldrhu16_m254(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu16_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #254
+; CHECK-NEXT:    vldrh.u16 q0, [r2]
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -254
+  %0 = bitcast i8* %z to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %x
+}
+
+define i8* @ldrhu16_m256(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu16_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #256
+; CHECK-NEXT:    vldrh.u16 q0, [r2]
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -256
+  %0 = bitcast i8* %z to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %x
+}
+
 
 define i8* @ldrbu32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbu32_4:
@@ -426,6 +526,40 @@ entry:
   ret i8* %x
 }
 
+define i8* @ldrbu32_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu32_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #127
+; CHECK-NEXT:    vldrb.u32 q0, [r2]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = zext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %x
+}
+
+define i8* @ldrbu32_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu32_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #128
+; CHECK-NEXT:    vldrb.u32 q0, [r2]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = zext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %x
+}
+
 
 define i8* @ldrbs32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbs32_4:
@@ -493,6 +627,40 @@ entry:
   ret i8* %x
 }
 
+define i8* @ldrbs32_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs32_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #127
+; CHECK-NEXT:    vldrb.s32 q0, [r2]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = sext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %x
+}
+
+define i8* @ldrbs32_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs32_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #128
+; CHECK-NEXT:    vldrb.s32 q0, [r2]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = sext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %x
+}
+
 
 define i8* @ldrbu16_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbu16_4:
@@ -560,6 +728,40 @@ entry:
   ret i8* %x
 }
 
+define i8* @ldrbu16_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu16_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #127
+; CHECK-NEXT:    vldrb.u16 q0, [r2]
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = zext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %x
+}
+
+define i8* @ldrbu16_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu16_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #128
+; CHECK-NEXT:    vldrb.u16 q0, [r2]
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = zext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %x
+}
+
 
 define i8* @ldrbs16_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbs16_4:
@@ -627,6 +829,40 @@ entry:
   ret i8* %x
 }
 
+define i8* @ldrbs16_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs16_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #127
+; CHECK-NEXT:    vldrb.s16 q0, [r2]
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = sext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %x
+}
+
+define i8* @ldrbs16_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs16_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #128
+; CHECK-NEXT:    vldrb.s16 q0, [r2]
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = sext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %x
+}
+
 
 define i8* @ldrbu8_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbu8_4:
@@ -690,6 +926,39 @@ entry:
   ret i8* %x
 }
 
+define i8* @ldrbu8_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu8_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #127
+; CHECK-NEXT:    vldrb.u8 q0, [r2]
+; CHECK-NEXT:    vstrb.8 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %x
+}
+
+define i8* @ldrbu8_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu8_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r2, r0, #128
+; CHECK-NEXT:    vldrb.u8 q0, [r2]
+; CHECK-NEXT:    vstrb.8 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %x
+}
+
+
 define i8* @ldrwf32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrwf32_4:
 ; CHECK:       @ %bb.0: @ %entry
@@ -1045,6 +1314,38 @@ entry:
   ret i8* %y
 }
 
+define i8* @strh32_m254(i8* %y, i8* %x) {
+; CHECK-LABEL: strh32_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u32 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #254
+; CHECK-NEXT:    vstrh.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -254
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = bitcast i8* %z to <4 x i16>*
+  store <4 x i16> %1, <4 x i16>* %2, align 2
+  ret i8* %y
+}
+
+define i8* @strh32_m256(i8* %y, i8* %x) {
+; CHECK-LABEL: strh32_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u32 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #256
+; CHECK-NEXT:    vstrh.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -256
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = bitcast i8* %z to <4 x i16>*
+  store <4 x i16> %1, <4 x i16>* %2, align 2
+  ret i8* %y
+}
+
 
 define i8* @strh16_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strh16_4:
@@ -1124,6 +1425,38 @@ entry:
   ret i8* %y
 }
 
+define i8* @strh16_m254(i8* %y, i8* %x) {
+; CHECK-LABEL: strh16_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u16 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #254
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -254
+  %0 = bitcast i8* %x to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %z to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %y
+}
+
+define i8* @strh16_m256(i8* %y, i8* %x) {
+; CHECK-LABEL: strh16_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u16 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #256
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -256
+  %0 = bitcast i8* %x to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %z to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %y
+}
+
 
 define i8* @strb32_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strb32_4:
@@ -1187,6 +1520,38 @@ entry:
   ret i8* %y
 }
 
+define i8* @strb32_m127(i8* %y, i8* %x) {
+; CHECK-LABEL: strb32_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u32 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #127
+; CHECK-NEXT:    vstrb.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -127
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <4 x i8>*
+  store <4 x i8> %1, <4 x i8>* %2, align 1
+  ret i8* %y
+}
+
+define i8* @strb32_m128(i8* %y, i8* %x) {
+; CHECK-LABEL: strb32_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u32 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #128
+; CHECK-NEXT:    vstrb.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -128
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <4 x i8>*
+  store <4 x i8> %1, <4 x i8>* %2, align 1
+  ret i8* %y
+}
+
 
 define i8* @strb16_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strb16_4:
@@ -1250,6 +1615,38 @@ entry:
   ret i8* %y
 }
 
+define i8* @strb16_m127(i8* %y, i8* %x) {
+; CHECK-LABEL: strb16_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u16 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #127
+; CHECK-NEXT:    vstrb.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -127
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <8 x i8>*
+  store <8 x i8> %1, <8 x i8>* %2, align 1
+  ret i8* %y
+}
+
+define i8* @strb16_m128(i8* %y, i8* %x) {
+; CHECK-LABEL: strb16_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u16 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #128
+; CHECK-NEXT:    vstrb.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -128
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <8 x i8>*
+  store <8 x i8> %1, <8 x i8>* %2, align 1
+  ret i8* %y
+}
+
 
 define i8* @strb8_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strb8_4:
@@ -1313,6 +1710,39 @@ entry:
   ret i8* %y
 }
 
+define i8* @strb8_m127(i8* %y, i8* %x) {
+; CHECK-LABEL: strb8_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u8 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #127
+; CHECK-NEXT:    vstrb.8 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -127
+  %0 = bitcast i8* %x to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %y
+}
+
+define i8* @strb8_m128(i8* %y, i8* %x) {
+; CHECK-LABEL: strb8_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u8 q0, [r1]
+; CHECK-NEXT:    sub.w r1, r0, #128
+; CHECK-NEXT:    vstrb.8 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -128
+  %0 = bitcast i8* %x to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %y
+}
+
+
 define i8* @strf32_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strf32_4:
 ; CHECK:       @ %bb.0: @ %entry
diff --git a/llvm/test/CodeGen/Thumb2/mve-ldst-postinc.ll b/llvm/test/CodeGen/Thumb2/mve-ldst-postinc.ll
index 61afa727c62..d84724a0b52 100644
--- a/llvm/test/CodeGen/Thumb2/mve-ldst-postinc.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-ldst-postinc.ll
@@ -199,6 +199,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrhu32_m254(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu32_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u32 q0, [r0], #-254
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -254
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = zext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
+define i8* @ldrhu32_m256(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu32_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u32 q0, [r0]
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -256
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = zext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
 
 define i8* @ldrhs32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrhs32_4:
@@ -282,6 +315,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrhs32_m254(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhs32_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.s32 q0, [r0], #-254
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -254
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = sext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
+define i8* @ldrhs32_m256(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhs32_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.s32 q0, [r0]
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -256
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = sext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
 
 define i8* @ldrhu16_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrhu16_4:
@@ -366,6 +432,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrhu16_m254(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu16_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u16 q0, [r0], #-254
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -254
+  %0 = bitcast i8* %x to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %z
+}
+
+define i8* @ldrhu16_m256(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu16_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u16 q0, [r0]
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -256
+  %0 = bitcast i8* %x to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %z
+}
+
 
 define i8* @ldrbu32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbu32_4:
@@ -432,6 +529,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbu32_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu32_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u32 q0, [r0], #-127
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = zext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
+define i8* @ldrbu32_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu32_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u32 q0, [r0]
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = zext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
 
 define i8* @ldrbs32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbs32_4:
@@ -498,6 +628,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbs32_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs32_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.s32 q0, [r0], #-127
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = sext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
+define i8* @ldrbs32_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs32_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.s32 q0, [r0]
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = sext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
 
 define i8* @ldrbu16_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbu16_4:
@@ -564,6 +727,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbu16_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu16_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u16 q0, [r0], #-127
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = zext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %z
+}
+
+define i8* @ldrbu16_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu16_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u16 q0, [r0]
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = zext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %z
+}
+
 
 define i8* @ldrbs16_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbs16_4:
@@ -630,6 +826,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbs16_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs16_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.s16 q0, [r0], #-127
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = sext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %z
+}
+
+define i8* @ldrbs16_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs16_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.s16 q0, [r0]
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = sext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %z
+}
+
 
 define i8* @ldrbu8_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbu8_4:
@@ -692,6 +921,38 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbu8_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu8_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u8 q0, [r0], #-127
+; CHECK-NEXT:    vstrb.8 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %x to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %z
+}
+
+define i8* @ldrbu8_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu8_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u8 q0, [r0]
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vstrb.8 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %x to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %z
+}
+
+
 define i8* @ldrwf32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrwf32_4:
 ; CHECK:       @ %bb.0: @ %entry
@@ -1067,6 +1328,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @strh32_m254(i8* %y, i8* %x) {
+; CHECK-LABEL: strh32_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u32 q0, [r1]
+; CHECK-NEXT:    vstrh.32 q0, [r0], #-254
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -254
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <4 x i16>*
+  store <4 x i16> %1, <4 x i16>* %2, align 2
+  ret i8* %z
+}
+
+define i8* @strh32_m256(i8* %y, i8* %x) {
+; CHECK-LABEL: strh32_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u32 q0, [r1]
+; CHECK-NEXT:    vstrh.32 q0, [r0]
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -256
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <4 x i16>*
+  store <4 x i16> %1, <4 x i16>* %2, align 2
+  ret i8* %z
+}
+
 
 define i8* @strh16_4(i8* %y, i8* %x) {
 ; CHECK-LE-LABEL: strh16_4:
@@ -1163,6 +1455,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @strh16_m254(i8* %y, i8* %x) {
+; CHECK-LABEL: strh16_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u16 q0, [r1]
+; CHECK-NEXT:    vstrh.16 q0, [r0], #-254
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -254
+  %0 = bitcast i8* %x to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %z
+}
+
+define i8* @strh16_m256(i8* %y, i8* %x) {
+; CHECK-LABEL: strh16_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u16 q0, [r1]
+; CHECK-NEXT:    vstrh.16 q0, [r0]
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -256
+  %0 = bitcast i8* %x to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %z
+}
+
 
 define i8* @strb32_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strb32_4:
@@ -1225,6 +1548,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @strb32_m127(i8* %y, i8* %x) {
+; CHECK-LABEL: strb32_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u32 q0, [r1]
+; CHECK-NEXT:    vstrb.32 q0, [r0], #-127
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -127
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <4 x i8>*
+  store <4 x i8> %1, <4 x i8>* %2, align 1
+  ret i8* %z
+}
+
+define i8* @strb32_m128(i8* %y, i8* %x) {
+; CHECK-LABEL: strb32_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u32 q0, [r1]
+; CHECK-NEXT:    vstrb.32 q0, [r0]
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -128
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <4 x i8>*
+  store <4 x i8> %1, <4 x i8>* %2, align 1
+  ret i8* %z
+}
+
 
 define i8* @strb16_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strb16_4:
@@ -1287,6 +1641,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @strb16_m127(i8* %y, i8* %x) {
+; CHECK-LABEL: strb16_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u16 q0, [r1]
+; CHECK-NEXT:    vstrb.16 q0, [r0], #-127
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -127
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <8 x i8>*
+  store <8 x i8> %1, <8 x i8>* %2, align 1
+  ret i8* %z
+}
+
+define i8* @strb16_m128(i8* %y, i8* %x) {
+; CHECK-LABEL: strb16_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u16 q0, [r1]
+; CHECK-NEXT:    vstrb.16 q0, [r0]
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -128
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <8 x i8>*
+  store <8 x i8> %1, <8 x i8>* %2, align 1
+  ret i8* %z
+}
+
 
 define i8* @strb8_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strb8_4:
@@ -1349,6 +1734,38 @@ entry:
   ret i8* %z
 }
 
+define i8* @strb8_m127(i8* %y, i8* %x) {
+; CHECK-LABEL: strb8_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u8 q0, [r1]
+; CHECK-NEXT:    vstrb.8 q0, [r0], #-127
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -127
+  %0 = bitcast i8* %x to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %z
+}
+
+define i8* @strb8_m128(i8* %y, i8* %x) {
+; CHECK-LABEL: strb8_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u8 q0, [r1]
+; CHECK-NEXT:    vstrb.8 q0, [r0]
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -128
+  %0 = bitcast i8* %x to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %z
+}
+
+
 define i8* @strf32_4(i8* %y, i8* %x) {
 ; CHECK-LE-LABEL: strf32_4:
 ; CHECK-LE:       @ %bb.0: @ %entry
diff --git a/llvm/test/CodeGen/Thumb2/mve-ldst-preinc.ll b/llvm/test/CodeGen/Thumb2/mve-ldst-preinc.ll
index ca1731a23d3..bf732c628ee 100644
--- a/llvm/test/CodeGen/Thumb2/mve-ldst-preinc.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-ldst-preinc.ll
@@ -199,6 +199,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrhu32_m254(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu32_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u32 q0, [r0, #-254]!
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -254
+  %0 = bitcast i8* %z to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = zext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
+define i8* @ldrhu32_m256(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu32_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    vldrh.u32 q0, [r0]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -256
+  %0 = bitcast i8* %z to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = zext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
 
 define i8* @ldrhs32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrhs32_4:
@@ -282,6 +315,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrhs32_m254(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhs32_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.s32 q0, [r0, #-254]!
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -254
+  %0 = bitcast i8* %z to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = sext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
+define i8* @ldrhs32_m256(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhs32_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    vldrh.s32 q0, [r0]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -256
+  %0 = bitcast i8* %z to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = sext <4 x i16> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
 
 define i8* @ldrhu16_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrhu16_4:
@@ -366,6 +432,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrhu16_m254(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu16_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u16 q0, [r0, #-254]!
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -254
+  %0 = bitcast i8* %z to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %z
+}
+
+define i8* @ldrhu16_m256(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrhu16_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    vldrh.u16 q0, [r0]
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -256
+  %0 = bitcast i8* %z to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %z
+}
+
 
 define i8* @ldrbu32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbu32_4:
@@ -432,6 +529,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbu32_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu32_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u32 q0, [r0, #-127]!
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = zext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
+define i8* @ldrbu32_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu32_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vldrb.u32 q0, [r0]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = zext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
 
 define i8* @ldrbs32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbs32_4:
@@ -498,6 +628,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbs32_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs32_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.s32 q0, [r0, #-127]!
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = sext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
+define i8* @ldrbs32_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs32_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vldrb.s32 q0, [r0]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = sext <4 x i8> %1 to <4 x i32>
+  %3 = bitcast i8* %y to <4 x i32>*
+  store <4 x i32> %2, <4 x i32>* %3, align 4
+  ret i8* %z
+}
+
 
 define i8* @ldrbu16_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbu16_4:
@@ -564,6 +727,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbu16_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu16_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u16 q0, [r0, #-127]!
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = zext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %z
+}
+
+define i8* @ldrbu16_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu16_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vldrb.u16 q0, [r0]
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = zext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %z
+}
+
 
 define i8* @ldrbs16_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbs16_4:
@@ -630,6 +826,39 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbs16_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs16_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.s16 q0, [r0, #-127]!
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = sext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %z
+}
+
+define i8* @ldrbs16_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbs16_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vldrb.s16 q0, [r0]
+; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = sext <8 x i8> %1 to <8 x i16>
+  %3 = bitcast i8* %y to <8 x i16>*
+  store <8 x i16> %2, <8 x i16>* %3, align 2
+  ret i8* %z
+}
+
 
 define i8* @ldrbu8_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrbu8_4:
@@ -692,6 +921,38 @@ entry:
   ret i8* %z
 }
 
+define i8* @ldrbu8_m127(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu8_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u8 q0, [r0, #-127]!
+; CHECK-NEXT:    vstrb.8 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -127
+  %0 = bitcast i8* %z to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %z
+}
+
+define i8* @ldrbu8_m128(i8* %x, i8* %y) {
+; CHECK-LABEL: ldrbu8_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vldrb.u8 q0, [r0]
+; CHECK-NEXT:    vstrb.8 q0, [r1]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %x, i32 -128
+  %0 = bitcast i8* %z to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %y to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %z
+}
+
+
 define i8* @ldrwf32_4(i8* %x, i8* %y) {
 ; CHECK-LABEL: ldrwf32_4:
 ; CHECK:       @ %bb.0: @ %entry
@@ -1067,6 +1328,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @strh32_m254(i8* %y, i8* %x) {
+; CHECK-LABEL: strh32_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u32 q0, [r1]
+; CHECK-NEXT:    vstrh.32 q0, [r0, #-254]!
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -254
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = bitcast i8* %z to <4 x i16>*
+  store <4 x i16> %1, <4 x i16>* %2, align 2
+  ret i8* %z
+}
+
+define i8* @strh32_m256(i8* %y, i8* %x) {
+; CHECK-LABEL: strh32_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    vldrh.u32 q0, [r1]
+; CHECK-NEXT:    vstrh.32 q0, [r0]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -256
+  %0 = bitcast i8* %x to <4 x i16>*
+  %1 = load <4 x i16>, <4 x i16>* %0, align 2
+  %2 = bitcast i8* %z to <4 x i16>*
+  store <4 x i16> %1, <4 x i16>* %2, align 2
+  ret i8* %z
+}
+
 
 define i8* @strh16_4(i8* %y, i8* %x) {
 ; CHECK-LE-LABEL: strh16_4:
@@ -1163,6 +1455,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @strh16_m254(i8* %y, i8* %x) {
+; CHECK-LABEL: strh16_m254:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrh.u16 q0, [r1]
+; CHECK-NEXT:    vstrh.16 q0, [r0, #-254]!
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -254
+  %0 = bitcast i8* %x to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %z to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %z
+}
+
+define i8* @strh16_m256(i8* %y, i8* %x) {
+; CHECK-LABEL: strh16_m256:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    sub.w r0, r0, #256
+; CHECK-NEXT:    vldrh.u16 q0, [r1]
+; CHECK-NEXT:    vstrh.16 q0, [r0]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -256
+  %0 = bitcast i8* %x to <8 x i16>*
+  %1 = load <8 x i16>, <8 x i16>* %0, align 2
+  %2 = bitcast i8* %z to <8 x i16>*
+  store <8 x i16> %1, <8 x i16>* %2, align 2
+  ret i8* %z
+}
+
 
 define i8* @strb32_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strb32_4:
@@ -1225,6 +1548,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @strb32_m127(i8* %y, i8* %x) {
+; CHECK-LABEL: strb32_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u32 q0, [r1]
+; CHECK-NEXT:    vstrb.32 q0, [r0, #-127]!
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -127
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <4 x i8>*
+  store <4 x i8> %1, <4 x i8>* %2, align 1
+  ret i8* %z
+}
+
+define i8* @strb32_m128(i8* %y, i8* %x) {
+; CHECK-LABEL: strb32_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vldrb.u32 q0, [r1]
+; CHECK-NEXT:    vstrb.32 q0, [r0]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -128
+  %0 = bitcast i8* %x to <4 x i8>*
+  %1 = load <4 x i8>, <4 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <4 x i8>*
+  store <4 x i8> %1, <4 x i8>* %2, align 1
+  ret i8* %z
+}
+
 
 define i8* @strb16_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strb16_4:
@@ -1287,6 +1641,37 @@ entry:
   ret i8* %z
 }
 
+define i8* @strb16_m127(i8* %y, i8* %x) {
+; CHECK-LABEL: strb16_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u16 q0, [r1]
+; CHECK-NEXT:    vstrb.16 q0, [r0, #-127]!
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -127
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <8 x i8>*
+  store <8 x i8> %1, <8 x i8>* %2, align 1
+  ret i8* %z
+}
+
+define i8* @strb16_m128(i8* %y, i8* %x) {
+; CHECK-LABEL: strb16_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vldrb.u16 q0, [r1]
+; CHECK-NEXT:    vstrb.16 q0, [r0]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -128
+  %0 = bitcast i8* %x to <8 x i8>*
+  %1 = load <8 x i8>, <8 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <8 x i8>*
+  store <8 x i8> %1, <8 x i8>* %2, align 1
+  ret i8* %z
+}
+
 
 define i8* @strb8_4(i8* %y, i8* %x) {
 ; CHECK-LABEL: strb8_4:
@@ -1349,6 +1734,38 @@ entry:
   ret i8* %z
 }
 
+define i8* @strb8_m127(i8* %y, i8* %x) {
+; CHECK-LABEL: strb8_m127:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vldrb.u8 q0, [r1]
+; CHECK-NEXT:    vstrb.8 q0, [r0, #-127]!
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -127
+  %0 = bitcast i8* %x to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %z
+}
+
+define i8* @strb8_m128(i8* %y, i8* %x) {
+; CHECK-LABEL: strb8_m128:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    subs r0, #128
+; CHECK-NEXT:    vldrb.u8 q0, [r1]
+; CHECK-NEXT:    vstrb.8 q0, [r0]
+; CHECK-NEXT:    bx lr
+entry:
+  %z = getelementptr inbounds i8, i8* %y, i32 -128
+  %0 = bitcast i8* %x to <16 x i8>*
+  %1 = load <16 x i8>, <16 x i8>* %0, align 1
+  %2 = bitcast i8* %z to <16 x i8>*
+  store <16 x i8> %1, <16 x i8>* %2, align 1
+  ret i8* %z
+}
+
+
 define i8* @strf32_4(i8* %y, i8* %x) {
 ; CHECK-LE-LABEL: strf32_4:
 ; CHECK-LE:       @ %bb.0: @ %entry