[ARM] Generate VDUP(Const) from constant buildvectors

[dmgreen]

Summary

If we cannot otherwise use a VMOVimm/VMOVFPimm/VMVNimm, fall back to producing a VDUP(const) as opposed to a constant pool load. This will at least be smaller codesize and can allow the VDUP to be folded into other instructions.

[simon_tatham]

Looks like an obviously good thing, and I only have one nitpick.

llvm/lib/Target/ARM/ARMISelLowering.cpp

7648

You've used VECTOR_REG_CAST where other branches of this code have BITCAST.

As far as I can see, either one will work provided the constant is constructed right (e.g. if you wanted to make a v16i8 containing 1,2,3,4,1,2,3,4,... then you might have to vdup 0x01020304 or 0x04030201 depending which cast you wanted to use afterwards). But I don't see any big-endian test to demonstrate it picking the right one. Unless I've missed one, could you add it?

[dmgreen]

Added two new test cases, mov_int8_1234 that does like you said i8 <1,2,3,4,1,2,3,4,..> and mov_int32_16908546 which is 0x1020102 VDUP'd as a i16.

[simon_tatham]

llvm/test/CodeGen/Thumb2/mve-vmovimm.ll

37–39

I think this output is right, but it confused me completely for a while and I had to try it in emulation to convince myself!

In the middle of a larger function, I think that if you wanted to make this 1,2,3,4,1,2,3,4,... vector and then immediately apply another v16i8 operation to it, you would vdup the same 32-bit constant 0x04030201 regardless of endianness, because the logical 'lane 0' of the vector always occupies the low-order bits.

And the reason why the output is different between LE and BE in this context is that the vdup is immediately followed by a function return, which in BE requires an extra vrev due to the vector register PCS. And that function-return vrev has been folded into the constant, which is why it's the other way round here.

So, I think this is the right output, but it might benefit from a comment in case the next reader gets as confused as I did!

llvm/lib/Target/ARM/ARMISelLowering.cpp

if ((ST->hasNEON() && SplatBitSize <= 64) ||
       // Use vmov.f32 to materialize other v2f32 and v4f32 splats.
       if ((VT == MVT::v2f32 || VT == MVT::v4f32) && SplatBitSize == 32) {
         int ImmVal = ARM_AM::getFP32Imm(SplatBits);
         if (ImmVal != -1) {
           SDValue Val = DAG.getTargetConstant(ImmVal, dl, MVT::i32);
           return DAG.getNode(ARMISD::VMOVFPIMM, dl, VT, Val);
         }
       }
+
+      // If we are under MVE, generate a VDUP(constant), bitcast to the original
+      // type.
+      if (ST->hasMVEIntegerOps() &&
+          (SplatBitSize == 8 || SplatBitSize == 16 || SplatBitSize == 32)) {
+        EVT DupVT = SplatBitSize == 32   ? MVT::v4i32
+                    : SplatBitSize == 16 ? MVT::v8i16
+                                         : MVT::v16i8;
+        SDValue Const = DAG.getConstant(SplatBits.getZExtValue(), dl, MVT::i32);
+        SDValue VDup = DAG.getNode(ARMISD::VDUP, dl, DupVT, Const);
+        return DAG.getNode(ARMISD::VECTOR_REG_CAST, dl, VT, VDup);
+      }
     }
   }
 
   // Scan through the operands to see if only one value is used.
   //
   // As an optimisation, even if more than one value is used it may be more
   // profitable to splat with one value then change some lanes.
   //

llvm/test/CodeGen/Thumb2/LowOverheadLoops/predicated-liveout-unknown-lanes.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp -O3 -tail-predication=force-enabled-no-reductions %s -o - | FileCheck %s
 
 define arm_aapcs_vfpcc <4 x float> @arm_max_no_idx_f32_mve(float* %pSrc, i32 %blockSize, float* nocapture %pResult) {
 ; CHECK-LABEL: arm_max_no_idx_f32_mve:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    subs r2, r1, #4
-; CHECK-NEXT:    adr r3, .LCPI0_0
-; CHECK-NEXT:    vldrw.u32 q0, [r3]
+; CHECK-NEXT:    movw r3, #0
+; CHECK-NEXT:    movt r3, #65408
+; CHECK-NEXT:    vdup.32 q0, r3
 ; CHECK-NEXT:    dlstp.32 lr, r1
 ; CHECK-NEXT:  .LBB0_1: @ %do.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    vldrw.u32 q1, [r0], #16
 ; CHECK-NEXT:    vmaxnm.f32 q0, q1, q0
 ; CHECK-NEXT:    letp lr, .LBB0_1
 ; CHECK-NEXT:  @ %bb.2: @ %do.end
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI0_0:
-; CHECK-NEXT:    .long 0xff800000 @ float -Inf
-; CHECK-NEXT:    .long 0xff800000 @ float -Inf
-; CHECK-NEXT:    .long 0xff800000 @ float -Inf
-; CHECK-NEXT:    .long 0xff800000 @ float -Inf
 entry:
   br label %do.body
 
 do.body:                                          ; preds = %do.body, %entry
   %blockSize.addr.0 = phi i32 [ %blockSize, %entry ], [ %sub, %do.body ]
   %curExtremValVec.0 = phi <4 x float> [ <float 0xFFF0000000000000, float 0xFFF0000000000000, float 0xFFF0000000000000, float 0xFFF0000000000000>, %entry ], [ %3, %do.body ]
   %pSrc.addr.0 = phi float* [ %pSrc, %entry ], [ %add.ptr, %do.body ]
   %0 = tail call <4 x i1> @llvm.arm.mve.vctp32(i32 %blockSize.addr.0)

llvm/test/CodeGen/Thumb2/mve-fp16convertloops.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp -verify-machineinstrs %s -o - | FileCheck %s
 
 define void @to_4(float* nocapture readonly %x, half* noalias nocapture %y) {
 ; CHECK-LABEL: to_4:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #256
-; CHECK-NEXT:    adr r2, .LCPI0_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB0_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrw.u32 q1, [r0], #16
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
-; CHECK-NEXT:    vstrh.32 q1, [r1], #8
+; CHECK-NEXT:    vldrw.u32 q0, [r0], #16
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
+; CHECK-NEXT:    vstrh.32 q0, [r1], #8
 ; CHECK-NEXT:    le lr, .LBB0_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI0_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds float, float* %x, i32 %index
   %1 = bitcast float* %0 to <4 x float>*
   %wide.load = load <4 x float>, <4 x float>* %1, align 4

 }
 
 define void @to_8(float* nocapture readonly %x, half* noalias nocapture %y) {
 ; CHECK-LABEL: to_8:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #128
-; CHECK-NEXT:    adr r2, .LCPI1_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB1_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
-; CHECK-NEXT:    vstrh.32 q1, [r1, #8]
-; CHECK-NEXT:    vldrw.u32 q1, [r0], #32
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
-; CHECK-NEXT:    vstrh.32 q1, [r1], #16
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
+; CHECK-NEXT:    vstrh.32 q0, [r1, #8]
+; CHECK-NEXT:    vldrw.u32 q0, [r0], #32
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
+; CHECK-NEXT:    vstrh.32 q0, [r1], #16
 ; CHECK-NEXT:    le lr, .LBB1_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI1_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds float, float* %x, i32 %index
   %1 = bitcast float* %0 to <8 x float>*
   %wide.load = load <8 x float>, <8 x float>* %1, align 4

 }
 
 define void @to_16(float* nocapture readonly %x, half* noalias nocapture %y) {
 ; CHECK-LABEL: to_16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #64
-; CHECK-NEXT:    adr r2, .LCPI2_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB2_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
-; CHECK-NEXT:    vstrh.32 q1, [r1, #24]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
-; CHECK-NEXT:    vstrh.32 q1, [r1, #16]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
-; CHECK-NEXT:    vstrh.32 q1, [r1, #8]
-; CHECK-NEXT:    vldrw.u32 q1, [r0], #64
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
-; CHECK-NEXT:    vstrh.32 q1, [r1], #32
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
+; CHECK-NEXT:    vstrh.32 q0, [r1, #24]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
+; CHECK-NEXT:    vstrh.32 q0, [r1, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
+; CHECK-NEXT:    vstrh.32 q0, [r1, #8]
+; CHECK-NEXT:    vldrw.u32 q0, [r0], #64
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
+; CHECK-NEXT:    vstrh.32 q0, [r1], #32
 ; CHECK-NEXT:    le lr, .LBB2_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI2_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds float, float* %x, i32 %index
   %1 = bitcast float* %0 to <16 x float>*
   %wide.load = load <16 x float>, <16 x float>* %1, align 4

 }
 
 define void @from_4(half* nocapture readonly %x, float* noalias nocapture %y) {
 ; CHECK-LABEL: from_4:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #256
-; CHECK-NEXT:    adr r2, .LCPI3_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB3_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrh.u32 q1, [r0], #8
-; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vstrb.8 q1, [r1], #16
+; CHECK-NEXT:    vldrh.u32 q0, [r0], #8
+; CHECK-NEXT:    vcvtb.f32.f16 q0, q0
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vstrb.8 q0, [r1], #16
 ; CHECK-NEXT:    le lr, .LBB3_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI3_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds half, half* %x, i32 %index
   %1 = bitcast half* %0 to <4 x half>*
   %wide.load = load <4 x half>, <4 x half>* %1, align 2

 }
 
 define void @from_8(half* nocapture readonly %x, float* noalias nocapture %y) {
 ; CHECK-LABEL: from_8:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #128
-; CHECK-NEXT:    adr r2, .LCPI4_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB4_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrh.u32 q1, [r0, #8]
+; CHECK-NEXT:    vldrh.u32 q0, [r0], #16
+; CHECK-NEXT:    vldrh.u32 q1, [r0, #-8]
+; CHECK-NEXT:    vcvtb.f32.f16 q0, q0
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
 ; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
+; CHECK-NEXT:    vmul.f32 q1, q1, r2
 ; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vldrh.u32 q1, [r0], #16
-; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vstrw.32 q1, [r1], #32
+; CHECK-NEXT:    vstrw.32 q0, [r1], #32
 ; CHECK-NEXT:    le lr, .LBB4_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI4_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds half, half* %x, i32 %index
   %1 = bitcast half* %0 to <8 x half>*
   %wide.load = load <8 x half>, <8 x half>* %1, align 2

 }
 
 define void @from_16(half* nocapture readonly %x, float* noalias nocapture %y) {
 ; CHECK-LABEL: from_16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #64
-; CHECK-NEXT:    adr r2, .LCPI5_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB5_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrh.u32 q1, [r0, #24]
-; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vstrw.32 q1, [r1, #48]
-; CHECK-NEXT:    vldrh.u32 q1, [r0, #16]
-; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vstrw.32 q1, [r1, #32]
-; CHECK-NEXT:    vldrh.u32 q1, [r0, #8]
+; CHECK-NEXT:    vldrh.u32 q0, [r0], #32
+; CHECK-NEXT:    vldrh.u32 q1, [r0, #-24]
+; CHECK-NEXT:    vldrh.u32 q2, [r0, #-16]
+; CHECK-NEXT:    vldrh.u32 q3, [r0, #-8]
+; CHECK-NEXT:    vcvtb.f32.f16 q0, q0
 ; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
+; CHECK-NEXT:    vcvtb.f32.f16 q2, q2
+; CHECK-NEXT:    vcvtb.f32.f16 q3, q3
+; CHECK-NEXT:    vmul.f32 q2, q2, r2
+; CHECK-NEXT:    vmul.f32 q3, q3, r2
+; CHECK-NEXT:    vmul.f32 q1, q1, r2
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vstrw.32 q3, [r1, #48]
+; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
 ; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vldrh.u32 q1, [r0], #32
-; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vstrw.32 q1, [r1], #64
+; CHECK-NEXT:    vstrw.32 q0, [r1], #64
 ; CHECK-NEXT:    le lr, .LBB5_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI5_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds half, half* %x, i32 %index
   %1 = bitcast half* %0 to <16 x half>*
   %wide.load = load <16 x half>, <16 x half>* %1, align 2

 }
 
 define void @both_4(half* nocapture readonly %x, half* noalias nocapture %y) {
 ; CHECK-LABEL: both_4:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #256
-; CHECK-NEXT:    adr r2, .LCPI6_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB6_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrh.u32 q1, [r0], #8
-; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
-; CHECK-NEXT:    vstrh.32 q1, [r1], #8
+; CHECK-NEXT:    vldrh.u32 q0, [r0], #8
+; CHECK-NEXT:    vcvtb.f32.f16 q0, q0
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
+; CHECK-NEXT:    vstrh.32 q0, [r1], #8
 ; CHECK-NEXT:    le lr, .LBB6_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI6_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds half, half* %x, i32 %index
   %1 = bitcast half* %0 to <4 x half>*
   %wide.load = load <4 x half>, <4 x half>* %1, align 2

 }
 
 define void @both_8(half* nocapture readonly %x, half* noalias nocapture %y) {
 ; CHECK-LABEL: both_8:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #128
-; CHECK-NEXT:    adr r2, .LCPI7_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB7_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrh.u16 q1, [r0], #16
-; CHECK-NEXT:    vcvtb.f32.f16 q2, q1
-; CHECK-NEXT:    vcvtt.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q2, q2, q0
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q2, q2
-; CHECK-NEXT:    vcvtt.f16.f32 q2, q1
-; CHECK-NEXT:    vstrb.8 q2, [r1], #16
+; CHECK-NEXT:    vldrh.u16 q0, [r0], #16
+; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
+; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
+; CHECK-NEXT:    vmul.f32 q1, q1, r2
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
+; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
+; CHECK-NEXT:    vstrb.8 q1, [r1], #16
 ; CHECK-NEXT:    le lr, .LBB7_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI7_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds half, half* %x, i32 %index
   %1 = bitcast half* %0 to <8 x half>*
   %wide.load = load <8 x half>, <8 x half>* %1, align 2

 }
 
 define void @both_16(half* nocapture readonly %x, half* noalias nocapture %y) {
 ; CHECK-LABEL: both_16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #64
-; CHECK-NEXT:    adr r2, .LCPI8_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB8_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrh.u16 q1, [r0, #16]
-; CHECK-NEXT:    vcvtb.f32.f16 q2, q1
-; CHECK-NEXT:    vcvtt.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q2, q2, q0
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q2, q2
-; CHECK-NEXT:    vcvtt.f16.f32 q2, q1
-; CHECK-NEXT:    vldrh.u16 q1, [r0], #32
-; CHECK-NEXT:    vstrh.16 q2, [r1, #16]
-; CHECK-NEXT:    vcvtb.f32.f16 q2, q1
-; CHECK-NEXT:    vcvtt.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q2, q2, q0
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q2, q2
-; CHECK-NEXT:    vcvtt.f16.f32 q2, q1
-; CHECK-NEXT:    vstrh.16 q2, [r1], #32
+; CHECK-NEXT:    vldrh.u16 q0, [r0, #16]
+; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
+; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
+; CHECK-NEXT:    vmul.f32 q1, q1, r2
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
+; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
+; CHECK-NEXT:    vldrh.u16 q0, [r0], #32
+; CHECK-NEXT:    vstrh.16 q1, [r1, #16]
+; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
+; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
+; CHECK-NEXT:    vmul.f32 q1, q1, r2
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
+; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
+; CHECK-NEXT:    vstrh.16 q1, [r1], #32
 ; CHECK-NEXT:    le lr, .LBB8_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI8_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds half, half* %x, i32 %index
   %1 = bitcast half* %0 to <16 x half>*
   %wide.load = load <16 x half>, <16 x half>* %1, align 2

 }
 
 define void @both_8_I(half* nocapture readonly %x, half* noalias nocapture %y) {
 ; CHECK-LABEL: both_8_I:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #128
-; CHECK-NEXT:    adr r2, .LCPI9_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB9_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrh.u16 q1, [r0], #16
-; CHECK-NEXT:    vcvtb.f32.f16 q2, q1
-; CHECK-NEXT:    vcvtt.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q2, q2, q0
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q2, q2
-; CHECK-NEXT:    vcvtt.f16.f32 q2, q1
-; CHECK-NEXT:    vstrb.8 q2, [r1], #16
+; CHECK-NEXT:    vldrh.u16 q0, [r0], #16
+; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
+; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
+; CHECK-NEXT:    vmul.f32 q1, q1, r2
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
+; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
+; CHECK-NEXT:    vstrb.8 q1, [r1], #16
 ; CHECK-NEXT:    le lr, .LBB9_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI9_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds half, half* %x, i32 %index
   %1 = bitcast half* %0 to <8 x half>*
   %wide.load = load <8 x half>, <8 x half>* %1, align 2

 }
 
 define void @both_16_I(half* nocapture readonly %x, half* noalias nocapture %y) {
 ; CHECK-LABEL: both_16_I:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    mov.w lr, #128
-; CHECK-NEXT:    adr r2, .LCPI10_0
-; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    movw r2, #26214
+; CHECK-NEXT:    movt r2, #16390
 ; CHECK-NEXT:  .LBB10_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrh.u16 q1, [r0]
-; CHECK-NEXT:    vcvtb.f32.f16 q2, q1
-; CHECK-NEXT:    vcvtt.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q2, q2, q0
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q2, q2
-; CHECK-NEXT:    vcvtt.f16.f32 q2, q1
-; CHECK-NEXT:    vldrh.u16 q1, [r0, #16]!
-; CHECK-NEXT:    vstrh.16 q2, [r1]
-; CHECK-NEXT:    vcvtb.f32.f16 q2, q1
-; CHECK-NEXT:    vcvtt.f32.f16 q1, q1
-; CHECK-NEXT:    vmul.f32 q2, q2, q0
-; CHECK-NEXT:    vmul.f32 q1, q1, q0
-; CHECK-NEXT:    vcvtb.f16.f32 q2, q2
-; CHECK-NEXT:    vcvtt.f16.f32 q2, q1
-; CHECK-NEXT:    vstrb.8 q2, [r1, #16]!
+; CHECK-NEXT:    vldrh.u16 q0, [r0]
+; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
+; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
+; CHECK-NEXT:    vmul.f32 q1, q1, r2
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
+; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
+; CHECK-NEXT:    vldrh.u16 q0, [r0, #16]!
+; CHECK-NEXT:    vstrh.16 q1, [r1]
+; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
+; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
+; CHECK-NEXT:    vmul.f32 q1, q1, r2
+; CHECK-NEXT:    vmul.f32 q0, q0, r2
+; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
+; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
+; CHECK-NEXT:    vstrb.8 q1, [r1, #16]!
 ; CHECK-NEXT:    le lr, .LBB10_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    pop {r7, pc}
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.3:
-; CHECK-NEXT:  .LCPI10_0:
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
-; CHECK-NEXT:    .long 0x40066666 @ float 2.0999999
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds half, half* %x, i32 %index
   %1 = bitcast half* %0 to <16 x half>*
   %wide.load = load <16 x half>, <16 x half>* %1, align 2

llvm/test/CodeGen/Thumb2/mve-gather-scatter-tailpred.ll

middle.block:                                     ; preds = %vector.body
   br label %end
 end:                                 ; preds = %middle.block
   ret void
 }
 
 define void @justoffsets(i8* noalias nocapture readonly %r, i8* noalias nocapture %w, i32 %N) {
 ; CHECK-LABEL: justoffsets:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, lr}
-; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, lr}
+; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, lr}
+; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, lr}
+; CHECK-NEXT:    .pad #4
+; CHECK-NEXT:    sub sp, #4
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #216
-; CHECK-NEXT:    sub sp, #216
+; CHECK-NEXT:    .pad #32
+; CHECK-NEXT:    sub sp, #32
 ; CHECK-NEXT:    cmp r2, #0
-; CHECK-NEXT:    beq.w .LBB3_3
+; CHECK-NEXT:    beq .LBB3_3
 ; CHECK-NEXT:  @ %bb.1: @ %vector.ph
+; CHECK-NEXT:    adr r5, .LCPI3_2
+; CHECK-NEXT:    vldrw.u32 q1, [r5]
+; CHECK-NEXT:    adr r4, .LCPI3_1
+; CHECK-NEXT:    movw r5, #50417
+; CHECK-NEXT:    adr r3, .LCPI3_0
+; CHECK-NEXT:    movw r7, #32769
+; CHECK-NEXT:    vldrw.u32 q2, [r4]
+; CHECK-NEXT:    movw r4, #47888
+; CHECK-NEXT:    vldrw.u32 q3, [r3]
+; CHECK-NEXT:    vstrw.32 q1, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vmov.i32 q1, #0x7fff
 ; CHECK-NEXT:    vmov.i32 q0, #0x8000
-; CHECK-NEXT:    adr r7, .LCPI3_5
-; CHECK-NEXT:    vstrw.32 q0, [sp, #160] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r7]
-; CHECK-NEXT:    adr r6, .LCPI3_4
-; CHECK-NEXT:    adr r5, .LCPI3_3
-; CHECK-NEXT:    vstrw.32 q0, [sp, #176] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r6]
-; CHECK-NEXT:    adr.w r8, .LCPI3_2
-; CHECK-NEXT:    vstrw.32 q0, [sp, #144] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r5]
-; CHECK-NEXT:    adr.w lr, .LCPI3_1
-; CHECK-NEXT:    vstrw.32 q0, [sp, #128] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r8]
-; CHECK-NEXT:    vstrw.32 q0, [sp, #112] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [lr]
-; CHECK-NEXT:    adr.w r12, .LCPI3_0
-; CHECK-NEXT:    vstrw.32 q0, [sp, #96] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r12]
-; CHECK-NEXT:    adr r7, .LCPI3_7
-; CHECK-NEXT:    adr r5, .LCPI3_10
-; CHECK-NEXT:    vstrw.32 q0, [sp, #80] @ 16-byte Spill
-; CHECK-NEXT:    vmov.i32 q0, #0x7fff
-; CHECK-NEXT:    vstrw.32 q0, [sp, #64] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r7]
-; CHECK-NEXT:    adr r6, .LCPI3_9
-; CHECK-NEXT:    adr r4, .LCPI3_6
-; CHECK-NEXT:    vstrw.32 q0, [sp, #48] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r5]
-; CHECK-NEXT:    adr r7, .LCPI3_8
-; CHECK-NEXT:    vldrw.u32 q1, [r4]
-; CHECK-NEXT:    vstrw.32 q0, [sp, #32] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r6]
-; CHECK-NEXT:    vstrw.32 q1, [sp, #192] @ 16-byte Spill
-; CHECK-NEXT:    vstrw.32 q0, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r7]
-; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
+; CHECK-NEXT:    movw r12, #7471
+; CHECK-NEXT:    movw r9, #19595
+; CHECK-NEXT:    movw r8, #38470
+; CHECK-NEXT:    movt r4, #65535
+; CHECK-NEXT:    movt r5, #65535
+; CHECK-NEXT:    movw r6, #19485
+; CHECK-NEXT:    movt r7, #65535
+; CHECK-NEXT:    movw r3, #13282
+; CHECK-NEXT:    vstrw.32 q1, [sp] @ 16-byte Spill
 ; CHECK-NEXT:    dlstp.32 lr, r2
 ; CHECK-NEXT:  .LBB3_2: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #192] @ 16-byte Reload
-; CHECK-NEXT:    vldrb.u32 q4, [r0, q0]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #176] @ 16-byte Reload
-; CHECK-NEXT:    vldrb.u32 q7, [r0, q0]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #144] @ 16-byte Reload
-; CHECK-NEXT:    vldrw.u32 q5, [sp, #112] @ 16-byte Reload
-; CHECK-NEXT:    vmul.i32 q6, q7, q0
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #128] @ 16-byte Reload
-; CHECK-NEXT:    vldrb.u32 q1, [r0, q5]
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #80] @ 16-byte Reload
-; CHECK-NEXT:    vmul.i32 q3, q4, q0
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #96] @ 16-byte Reload
-; CHECK-NEXT:    vadd.i32 q3, q3, q6
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #16] @ 16-byte Reload
+; CHECK-NEXT:    vldrb.u32 q7, [r0, q1]
+; CHECK-NEXT:    vldrb.u32 q5, [r0, q2]
+; CHECK-NEXT:    vmul.i32 q4, q5, r8
+; CHECK-NEXT:    vmla.u32 q4, q7, r9
+; CHECK-NEXT:    vldrb.u32 q6, [r0, q3]
+; CHECK-NEXT:    vmla.u32 q4, q6, r12
 ; CHECK-NEXT:    adds r0, #12
-; CHECK-NEXT:    vmul.i32 q6, q1, q0
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #160] @ 16-byte Reload
-; CHECK-NEXT:    vadd.i32 q3, q3, q6
-; CHECK-NEXT:    vadd.i32 q3, q3, q0
-; CHECK-NEXT:    vshr.u32 q6, q3, #16
-; CHECK-NEXT:    vmul.i32 q3, q7, q2
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    vmul.i32 q2, q4, q2
-; CHECK-NEXT:    vadd.i32 q2, q2, q3
-; CHECK-NEXT:    vldrw.u32 q3, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vmul.i32 q3, q1, q3
-; CHECK-NEXT:    vadd.i32 q2, q2, q3
-; CHECK-NEXT:    vldrw.u32 q3, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    vadd.i32 q2, q2, q0
-; CHECK-NEXT:    vmul.i32 q3, q7, q3
-; CHECK-NEXT:    vldrw.u32 q7, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vshr.u32 q2, q2, #16
-; CHECK-NEXT:    vmul.i32 q4, q4, q7
-; CHECK-NEXT:    vadd.i32 q3, q4, q3
-; CHECK-NEXT:    vldrw.u32 q4, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmul.i32 q1, q1, q4
-; CHECK-NEXT:    vadd.i32 q1, q3, q1
+; CHECK-NEXT:    vadd.i32 q4, q4, q0
+; CHECK-NEXT:    vshr.u32 q4, q4, #16
+; CHECK-NEXT:    vstrb.32 q4, [r1, q1]
+; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
+; CHECK-NEXT:    vmul.i32 q4, q7, q1
+; CHECK-NEXT:    vmul.i32 q1, q5, r7
+; CHECK-NEXT:    vmla.u32 q1, q7, r3
+; CHECK-NEXT:    vmla.u32 q4, q5, r5
+; CHECK-NEXT:    vmla.u32 q1, q6, r6
+; CHECK-NEXT:    vmla.u32 q4, q6, r4
 ; CHECK-NEXT:    vadd.i32 q1, q1, q0
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #192] @ 16-byte Reload
+; CHECK-NEXT:    vadd.i32 q4, q4, q0
 ; CHECK-NEXT:    vshr.u32 q1, q1, #16
-; CHECK-NEXT:    vstrb.32 q1, [r1, q0]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #176] @ 16-byte Reload
-; CHECK-NEXT:    vstrb.32 q2, [r1, q0]
-; CHECK-NEXT:    vstrb.32 q6, [r1, q5]
+; CHECK-NEXT:    vshr.u32 q4, q4, #16
+; CHECK-NEXT:    vstrb.32 q4, [r1, q2]
+; CHECK-NEXT:    vstrb.32 q1, [r1, q3]
 ; CHECK-NEXT:    adds r1, #12
 ; CHECK-NEXT:    letp lr, .LBB3_2
 ; CHECK-NEXT:  .LBB3_3: @ %for.cond.cleanup
-; CHECK-NEXT:    add sp, #216
+; CHECK-NEXT:    add sp, #32
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, pc}
+; CHECK-NEXT:    add sp, #4
+; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, pc}
 ; CHECK-NEXT:    .p2align 4
 ; CHECK-NEXT:  @ %bb.4:
 ; CHECK-NEXT:  .LCPI3_0:
-; CHECK-NEXT:    .long 4294952177 @ 0xffffc4f1
-; CHECK-NEXT:    .long 4294952177 @ 0xffffc4f1
-; CHECK-NEXT:    .long 4294952177 @ 0xffffc4f1
-; CHECK-NEXT:    .long 4294952177 @ 0xffffc4f1
-; CHECK-NEXT:  .LCPI3_1:
-; CHECK-NEXT:    .long 19485 @ 0x4c1d
-; CHECK-NEXT:    .long 19485 @ 0x4c1d
-; CHECK-NEXT:    .long 19485 @ 0x4c1d
-; CHECK-NEXT:    .long 19485 @ 0x4c1d
-; CHECK-NEXT:  .LCPI3_2:
 ; CHECK-NEXT:    .long 2 @ 0x2
 ; CHECK-NEXT:    .long 5 @ 0x5
 ; CHECK-NEXT:    .long 8 @ 0x8
 ; CHECK-NEXT:    .long 11 @ 0xb
-; CHECK-NEXT:  .LCPI3_3:
-; CHECK-NEXT:    .long 13282 @ 0x33e2
-; CHECK-NEXT:    .long 13282 @ 0x33e2
-; CHECK-NEXT:    .long 13282 @ 0x33e2
-; CHECK-NEXT:    .long 13282 @ 0x33e2
-; CHECK-NEXT:  .LCPI3_4:
-; CHECK-NEXT:    .long 4294934529 @ 0xffff8001
-; CHECK-NEXT:    .long 4294934529 @ 0xffff8001
-; CHECK-NEXT:    .long 4294934529 @ 0xffff8001
-; CHECK-NEXT:    .long 4294934529 @ 0xffff8001
-; CHECK-NEXT:  .LCPI3_5:
+; CHECK-NEXT:  .LCPI3_1:
 ; CHECK-NEXT:    .long 1 @ 0x1
 ; CHECK-NEXT:    .long 4 @ 0x4
 ; CHECK-NEXT:    .long 7 @ 0x7
 ; CHECK-NEXT:    .long 10 @ 0xa
-; CHECK-NEXT:  .LCPI3_6:
+; CHECK-NEXT:  .LCPI3_2:
 ; CHECK-NEXT:    .long 0 @ 0x0
 ; CHECK-NEXT:    .long 3 @ 0x3
 ; CHECK-NEXT:    .long 6 @ 0x6
 ; CHECK-NEXT:    .long 9 @ 0x9
-; CHECK-NEXT:  .LCPI3_7:
-; CHECK-NEXT:    .long 4294949648 @ 0xffffbb10
-; CHECK-NEXT:    .long 4294949648 @ 0xffffbb10
-; CHECK-NEXT:    .long 4294949648 @ 0xffffbb10
-; CHECK-NEXT:    .long 4294949648 @ 0xffffbb10
-; CHECK-NEXT:  .LCPI3_8:
-; CHECK-NEXT:    .long 7471 @ 0x1d2f
-; CHECK-NEXT:    .long 7471 @ 0x1d2f
-; CHECK-NEXT:    .long 7471 @ 0x1d2f
-; CHECK-NEXT:    .long 7471 @ 0x1d2f
-; CHECK-NEXT:  .LCPI3_9:
-; CHECK-NEXT:    .long 19595 @ 0x4c8b
-; CHECK-NEXT:    .long 19595 @ 0x4c8b
-; CHECK-NEXT:    .long 19595 @ 0x4c8b
-; CHECK-NEXT:    .long 19595 @ 0x4c8b
-; CHECK-NEXT:  .LCPI3_10:
-; CHECK-NEXT:    .long 38470 @ 0x9646
-; CHECK-NEXT:    .long 38470 @ 0x9646
-; CHECK-NEXT:    .long 38470 @ 0x9646
-; CHECK-NEXT:    .long 38470 @ 0x9646
 entry:
   %cmp47.not = icmp eq i32 %N, 0
   br i1 %cmp47.not, label %for.cond.cleanup, label %vector.ph
 
 vector.ph:                                        ; preds = %vector.memcheck
   %n.rnd.up = add i32 %N, 3
   %n.vec = and i32 %n.rnd.up, -4
   br label %vector.body

llvm/test/CodeGen/Thumb2/mve-pred-vctpvpsel.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp %s -verify-machineinstrs -o - | FileCheck %s
 
 define void @arm_min_helium_f32(float* %pSrc, i32 %blockSize, float* nocapture %pResult, i32* nocapture %pIndex) {
 ; CHECK-LABEL: arm_min_helium_f32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r6, r7, lr}
-; CHECK-NEXT:    push {r4, r6, r7, lr}
+; CHECK-NEXT:    .save {r4, lr}
+; CHECK-NEXT:    push {r4, lr}
 ; CHECK-NEXT:    .vsave {d8, d9}
 ; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    movs r6, #0
-; CHECK-NEXT:    vidup.u32 q2, r6, #1
-; CHECK-NEXT:    adr r4, .LCPI0_0
-; CHECK-NEXT:    vldrw.u32 q1, [r4]
+; CHECK-NEXT:    movs r4, #0
+; CHECK-NEXT:    vidup.u32 q2, r4, #1
+; CHECK-NEXT:    movw r4, #54437
+; CHECK-NEXT:    movt r4, #21352
+; CHECK-NEXT:    vdup.32 q1, r4
 ; CHECK-NEXT:    vmov.i32 q0, #0x0
 ; CHECK-NEXT:    vmov.i32 q3, #0x4
 ; CHECK-NEXT:    dlstp.32 lr, r1
 ; CHECK-NEXT:  .LBB0_1: @ %do.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    vldrw.u32 q4, [r0], #16
 ; CHECK-NEXT:    vptt.f32 ge, q1, q4
 ; CHECK-NEXT:    vmovt q1, q4
 ; CHECK-NEXT:    vmovt q0, q2
 ; CHECK-NEXT:    vadd.i32 q2, q2, q3
 ; CHECK-NEXT:    letp lr, .LBB0_1
 ; CHECK-NEXT:  @ %bb.2: @ %do.end
-; CHECK-NEXT:    vldr s8, .LCPI0_1
+; CHECK-NEXT:    vldr s8, .LCPI0_0
 ; CHECK-NEXT:    vdup.32 q3, r1
 ; CHECK-NEXT:    vmov r0, s8
 ; CHECK-NEXT:    vminnmv.f32 r0, q1
 ; CHECK-NEXT:    vcmp.f32 le, q1, r0
 ; CHECK-NEXT:    vmov s8, r0
 ; CHECK-NEXT:    vpsel q0, q0, q3
 ; CHECK-NEXT:    vminv.u32 r1, q0
 ; CHECK-NEXT:    str r1, [r3]
 ; CHECK-NEXT:    vstr s8, [r2]
 ; CHECK-NEXT:    vpop {d8, d9}
-; CHECK-NEXT:    pop {r4, r6, r7, pc}
-; CHECK-NEXT:    .p2align 4
+; CHECK-NEXT:    pop {r4, pc}
+; CHECK-NEXT:    .p2align 2
 ; CHECK-NEXT:  @ %bb.3:
 ; CHECK-NEXT:  .LCPI0_0:
 ; CHECK-NEXT:    .long 0x5368d4a5 @ float 9.99999995E+11
-; CHECK-NEXT:    .long 0x5368d4a5 @ float 9.99999995E+11
-; CHECK-NEXT:    .long 0x5368d4a5 @ float 9.99999995E+11
-; CHECK-NEXT:    .long 0x5368d4a5 @ float 9.99999995E+11
-; CHECK-NEXT:  .LCPI0_1:
-; CHECK-NEXT:    .long 0x5368d4a5 @ float 9.99999995E+11
 entry:
   %0 = tail call { <4 x i32>, i32 } @llvm.arm.mve.vidup.v4i32(i32 0, i32 1)
   %1 = extractvalue { <4 x i32>, i32 } %0, 0
   br label %do.body
 
 do.body:                                          ; preds = %do.body, %entry
   %curExtremValVec.0 = phi <4 x float> [ <float 0x426D1A94A0000000, float 0x426D1A94A0000000, float 0x426D1A94A0000000, float 0x426D1A94A0000000>, %entry ], [ %8, %do.body ]
   %indexVec.0 = phi <4 x i32> [ %1, %entry ], [ %11, %do.body ]

llvm/test/CodeGen/Thumb2/mve-shifts.ll

entry:
   %s = shufflevector <2 x i64> %i, <2 x i64> undef, <2 x i32> zeroinitializer
   %0 = ashr <2 x i64> %src1, %s
   ret <2 x i64> %0
 }
 
 define arm_aapcs_vfpcc <16 x i8> @shl_qiv_int8_t(<16 x i8> %src1) {
 ; CHECK-LABEL: shl_qiv_int8_t:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    adr r0, .LCPI36_0
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
+; CHECK-NEXT:    movw r0, #513
+; CHECK-NEXT:    movt r0, #1027
+; CHECK-NEXT:    vdup.32 q1, r0
 ; CHECK-NEXT:    vshl.u8 q0, q0, q1
 ; CHECK-NEXT:    bx lr
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.1:
-; CHECK-NEXT:  .LCPI36_0:
-; CHECK-NEXT:    .byte 1 @ 0x1
-; CHECK-NEXT:    .byte 2 @ 0x2
-; CHECK-NEXT:    .byte 3 @ 0x3
-; CHECK-NEXT:    .byte 4 @ 0x4
-; CHECK-NEXT:    .byte 1 @ 0x1
-; CHECK-NEXT:    .byte 2 @ 0x2
-; CHECK-NEXT:    .byte 3 @ 0x3
-; CHECK-NEXT:    .byte 4 @ 0x4
-; CHECK-NEXT:    .byte 1 @ 0x1
-; CHECK-NEXT:    .byte 2 @ 0x2
-; CHECK-NEXT:    .byte 3 @ 0x3
-; CHECK-NEXT:    .byte 4 @ 0x4
-; CHECK-NEXT:    .byte 1 @ 0x1
-; CHECK-NEXT:    .byte 2 @ 0x2
-; CHECK-NEXT:    .byte 3 @ 0x3
-; CHECK-NEXT:    .byte 4 @ 0x4
 entry:
   %0 = shl <16 x i8> %src1, <i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4>
   ret <16 x i8> %0
 }
 
 define arm_aapcs_vfpcc <8 x i16> @shl_qiv_int16_t(<8 x i16> %src1) {
 ; CHECK-LABEL: shl_qiv_int16_t:
 ; CHECK:       @ %bb.0: @ %entry

llvm/test/CodeGen/Thumb2/mve-vmovimm.ll

 ; CHECK-LABEL: mov_int8_m1:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i8 q0, #0xff
 ; CHECK-NEXT:    bx lr
 entry:
   ret <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
 }
 
+; This has 0x01020304 or 0x04030201 vdup.32'd to q reg depending on endianness.
+; The big endian is different as there is an implicit vrev64.8 out of the
+; function, which gets constant folded away.
 define arm_aapcs_vfpcc <16 x i8> @mov_int8_1234() {
 ; CHECKLE-LABEL: mov_int8_1234:
 ; CHECKLE:       @ %bb.0: @ %entry
-; CHECKLE-NEXT:    adr r0, .LCPI2_0
-; CHECKLE-NEXT:    vldrw.u32 q0, [r0]
+; CHECKLE-NEXT:    movw r0, #513
+; CHECKLE-NEXT:    movt r0, #1027
+; CHECKLE-NEXT:    vdup.32 q0, r0
 ; CHECKLE-NEXT:    bx lr
-; CHECKLE-NEXT:    .p2align 4
-; CHECKLE-NEXT:  @ %bb.1:
-; CHECKLE-NEXT:  .LCPI2_0:
-; CHECKLE-NEXT:    .long 67305985 @ double 2.4380727978175888E-289
-; CHECKLE-NEXT:    .long 67305985
-; CHECKLE-NEXT:    .long 67305985 @ double 2.4380727978175888E-289
-; CHECKLE-NEXT:    .long 67305985
 ;
 ; CHECKBE-LABEL: mov_int8_1234:
 ; CHECKBE:       @ %bb.0: @ %entry
-; CHECKBE-NEXT:    adr r0, .LCPI2_0
-; CHECKBE-NEXT:    vldrb.u8 q1, [r0]
-; CHECKBE-NEXT:    vrev64.8 q0, q1
+; CHECKBE-NEXT:    movw r0, #772
+; CHECKBE-NEXT:    movt r0, #258
+; CHECKBE-NEXT:    vdup.32 q0, r0
 ; CHECKBE-NEXT:    bx lr
-; CHECKBE-NEXT:    .p2align 4
-; CHECKBE-NEXT:  @ %bb.1:
-; CHECKBE-NEXT:  .LCPI2_0:
-; CHECKBE-NEXT:    .long 16909060 @ double 8.2078802900595913E-304
-; CHECKBE-NEXT:    .long 16909060
-; CHECKBE-NEXT:    .long 16909060 @ double 8.2078802900595913E-304
-; CHECKBE-NEXT:    .long 16909060
 entry:
   ret <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4>
 }
 
 define arm_aapcs_vfpcc <8 x i16> @mov_int16_1() {
 ; CHECK-LABEL: mov_int16_1:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i16 q0, #0x1

 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i8 q0, #0x1
 ; CHECK-NEXT:    bx lr
 entry:
   ret <8 x i16> <i16 257, i16 257, i16 257, i16 257, i16 257, i16 257, i16 257, i16 257>
 }
 
 define arm_aapcs_vfpcc <8 x i16> @mov_int16_258() {
-; CHECKLE-LABEL: mov_int16_258:
-; CHECKLE:       @ %bb.0: @ %entry
-; CHECKLE-NEXT:    adr r0, .LCPI7_0
-; CHECKLE-NEXT:    vldrw.u32 q0, [r0]
-; CHECKLE-NEXT:    bx lr
-; CHECKLE-NEXT:    .p2align 4
-; CHECKLE-NEXT:  @ %bb.1:
-; CHECKLE-NEXT:  .LCPI7_0:
-; CHECKLE-NEXT:    .long 16908546 @ double 8.204306265173532E-304
-; CHECKLE-NEXT:    .long 16908546
-; CHECKLE-NEXT:    .long 16908546 @ double 8.204306265173532E-304
-; CHECKLE-NEXT:    .long 16908546
-;
-; CHECKBE-LABEL: mov_int16_258:
-; CHECKBE:       @ %bb.0: @ %entry
-; CHECKBE-NEXT:    adr r0, .LCPI7_0
-; CHECKBE-NEXT:    vldrb.u8 q1, [r0]
-; CHECKBE-NEXT:    vrev64.8 q0, q1
-; CHECKBE-NEXT:    bx lr
-; CHECKBE-NEXT:    .p2align 4
-; CHECKBE-NEXT:  @ %bb.1:
-; CHECKBE-NEXT:  .LCPI7_0:
-; CHECKBE-NEXT:    .long 16908546 @ double 8.204306265173532E-304
-; CHECKBE-NEXT:    .long 16908546
-; CHECKBE-NEXT:    .long 16908546 @ double 8.204306265173532E-304
-; CHECKBE-NEXT:    .long 16908546
+; CHECK-LABEL: mov_int16_258:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    mov.w r0, #258
+; CHECK-NEXT:    vdup.16 q0, r0
+; CHECK-NEXT:    bx lr
 entry:
   ret <8 x i16> <i16 258, i16 258, i16 258, i16 258, i16 258, i16 258, i16 258, i16 258>
 }
 
 define arm_aapcs_vfpcc <4 x i32> @mov_int32_1() {
 ; CHECK-LABEL: mov_int32_1:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i32 q0, #0x1

 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i32 q0, #0x1000000
 ; CHECK-NEXT:    bx lr
 entry:
   ret <4 x i32> <i32 16777216, i32 16777216, i32 16777216, i32 16777216>
 }
 
 define arm_aapcs_vfpcc <4 x i32> @mov_int32_16777217() {
-; CHECKLE-LABEL: mov_int32_16777217:
-; CHECKLE:       @ %bb.0: @ %entry
-; CHECKLE-NEXT:    adr r0, .LCPI12_0
-; CHECKLE-NEXT:    vldrw.u32 q0, [r0]
-; CHECKLE-NEXT:    bx lr
-; CHECKLE-NEXT:    .p2align 4
-; CHECKLE-NEXT:  @ %bb.1:
-; CHECKLE-NEXT:  .LCPI12_0:
-; CHECKLE-NEXT:    .long 16777217 @ double 7.2911290000737531E-304
-; CHECKLE-NEXT:    .long 16777217
-; CHECKLE-NEXT:    .long 16777217 @ double 7.2911290000737531E-304
-; CHECKLE-NEXT:    .long 16777217
-;
-; CHECKBE-LABEL: mov_int32_16777217:
-; CHECKBE:       @ %bb.0: @ %entry
-; CHECKBE-NEXT:    adr r0, .LCPI12_0
-; CHECKBE-NEXT:    vldrb.u8 q1, [r0]
-; CHECKBE-NEXT:    vrev64.8 q0, q1
-; CHECKBE-NEXT:    bx lr
-; CHECKBE-NEXT:    .p2align 4
-; CHECKBE-NEXT:  @ %bb.1:
-; CHECKBE-NEXT:  .LCPI12_0:
-; CHECKBE-NEXT:    .long 16777217 @ double 7.2911290000737531E-304
-; CHECKBE-NEXT:    .long 16777217
-; CHECKBE-NEXT:    .long 16777217 @ double 7.2911290000737531E-304
-; CHECKBE-NEXT:    .long 16777217
+; CHECK-LABEL: mov_int32_16777217:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    movs r0, #1
+; CHECK-NEXT:    movt r0, #256
+; CHECK-NEXT:    vdup.32 q0, r0
+; CHECK-NEXT:    bx lr
 entry:
   ret <4 x i32> <i32 16777217, i32 16777217, i32 16777217, i32 16777217>
 }
 
 define arm_aapcs_vfpcc <4 x i32> @mov_int32_17919() {
 ; CHECK-LABEL: mov_int32_17919:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i32 q0, #0x45ff

 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmvn.i32 q0, #0xffff
 ; CHECK-NEXT:    bx lr
 entry:
   ret <4 x i32> <i32 4294901760, i32 4294901760, i32 4294901760, i32 4294901760>
 }
 
 define arm_aapcs_vfpcc <4 x i32> @mov_int32_4278190335() {
-; CHECKLE-LABEL: mov_int32_4278190335:
-; CHECKLE:       @ %bb.0: @ %entry
-; CHECKLE-NEXT:    adr r0, .LCPI17_0
-; CHECKLE-NEXT:    vldrw.u32 q0, [r0]
-; CHECKLE-NEXT:    bx lr
-; CHECKLE-NEXT:    .p2align 4
-; CHECKLE-NEXT:  @ %bb.1:
-; CHECKLE-NEXT:  .LCPI17_0:
-; CHECKLE-NEXT:    .long 4278190335 @ double -5.4874634341155774E+303
-; CHECKLE-NEXT:    .long 4278190335
-; CHECKLE-NEXT:    .long 4278190335 @ double -5.4874634341155774E+303
-; CHECKLE-NEXT:    .long 4278190335
-;
-; CHECKBE-LABEL: mov_int32_4278190335:
-; CHECKBE:       @ %bb.0: @ %entry
-; CHECKBE-NEXT:    adr r0, .LCPI17_0
-; CHECKBE-NEXT:    vldrb.u8 q1, [r0]
-; CHECKBE-NEXT:    vrev64.8 q0, q1
-; CHECKBE-NEXT:    bx lr
-; CHECKBE-NEXT:    .p2align 4
-; CHECKBE-NEXT:  @ %bb.1:
-; CHECKBE-NEXT:  .LCPI17_0:
-; CHECKBE-NEXT:    .long 4278190335 @ double -5.4874634341155774E+303
-; CHECKBE-NEXT:    .long 4278190335
-; CHECKBE-NEXT:    .long 4278190335 @ double -5.4874634341155774E+303
-; CHECKBE-NEXT:    .long 4278190335
+; CHECK-LABEL: mov_int32_4278190335:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    movs r0, #255
+; CHECK-NEXT:    movt r0, #65280
+; CHECK-NEXT:    vdup.32 q0, r0
+; CHECK-NEXT:    bx lr
 entry:
   ret <4 x i32> <i32 4278190335, i32 4278190335, i32 4278190335, i32 4278190335>
 }
 
 define arm_aapcs_vfpcc <4 x i32> @mov_int32_4278255615() {
 ; CHECK-LABEL: mov_int32_4278255615:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmvn.i32 q0, #0xff0000
 ; CHECK-NEXT:    bx lr
 entry:
   ret <4 x i32> <i32 4278255615, i32 4278255615, i32 4278255615, i32 4278255615>
 }
 
 define arm_aapcs_vfpcc <4 x i32> @mov_int32_16908546() {
-; CHECKLE-LABEL: mov_int32_16908546:
-; CHECKLE:       @ %bb.0: @ %entry
-; CHECKLE-NEXT:    adr r0, .LCPI19_0
-; CHECKLE-NEXT:    vldrw.u32 q0, [r0]
-; CHECKLE-NEXT:    bx lr
-; CHECKLE-NEXT:    .p2align 4
-; CHECKLE-NEXT:  @ %bb.1:
-; CHECKLE-NEXT:  .LCPI19_0:
-; CHECKLE-NEXT:    .long 16908546 @ double 8.204306265173532E-304
-; CHECKLE-NEXT:    .long 16908546
-; CHECKLE-NEXT:    .long 16908546 @ double 8.204306265173532E-304
-; CHECKLE-NEXT:    .long 16908546
-;
-; CHECKBE-LABEL: mov_int32_16908546:
-; CHECKBE:       @ %bb.0: @ %entry
-; CHECKBE-NEXT:    adr r0, .LCPI19_0
-; CHECKBE-NEXT:    vldrb.u8 q1, [r0]
-; CHECKBE-NEXT:    vrev64.8 q0, q1
-; CHECKBE-NEXT:    bx lr
-; CHECKBE-NEXT:    .p2align 4
-; CHECKBE-NEXT:  @ %bb.1:
-; CHECKBE-NEXT:  .LCPI19_0:
-; CHECKBE-NEXT:    .long 16908546 @ double 8.204306265173532E-304
-; CHECKBE-NEXT:    .long 16908546
-; CHECKBE-NEXT:    .long 16908546 @ double 8.204306265173532E-304
-; CHECKBE-NEXT:    .long 16908546
+; CHECK-LABEL: mov_int32_16908546:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    mov.w r0, #258
+; CHECK-NEXT:    vdup.16 q0, r0
+; CHECK-NEXT:    bx lr
 entry:
   ret <4 x i32> <i32 16908546, i32 16908546, i32 16908546, i32 16908546>
 }
 
 define arm_aapcs_vfpcc <2 x i64> @mov_int64_1() {
 ; CHECKLE-LABEL: mov_int64_1:
 ; CHECKLE:       @ %bb.0: @ %entry
 ; CHECKLE-NEXT:    adr r0, .LCPI20_0

 ; CHECKBE:       @ %bb.0: @ %entry
 ; CHECKBE-NEXT:    vmov.i16 q0, #0xff00
 ; CHECKBE-NEXT:    bx lr
 entry:
   ret <16 x i8> <i8 -1, i8 0, i8 -1, i8 0, i8 -1, i8 0, i8 -1, i8 0, i8 -1, i8 0, i8 -1, i8 0, i8 -1, i8 0, i8 -1, i8 0>
 }
 
 define arm_aapcs_vfpcc <4 x float> @mov_float_1() {
-; CHECKLE-LABEL: mov_float_1:
-; CHECKLE:       @ %bb.0: @ %entry
-; CHECKLE-NEXT:    adr r0, .LCPI28_0
-; CHECKLE-NEXT:    vldrw.u32 q0, [r0]
-; CHECKLE-NEXT:    bx lr
-; CHECKLE-NEXT:    .p2align 4
-; CHECKLE-NEXT:  @ %bb.1:
-; CHECKLE-NEXT:  .LCPI28_0:
-; CHECKLE-NEXT:    .long 1065353216 @ double 0.007812501848093234
-; CHECKLE-NEXT:    .long 1065353216
-; CHECKLE-NEXT:    .long 1065353216 @ double 0.007812501848093234
-; CHECKLE-NEXT:    .long 1065353216
-;
-; CHECKBE-LABEL: mov_float_1:
-; CHECKBE:       @ %bb.0: @ %entry
-; CHECKBE-NEXT:    adr r0, .LCPI28_0
-; CHECKBE-NEXT:    vldrb.u8 q1, [r0]
-; CHECKBE-NEXT:    vrev64.8 q0, q1
-; CHECKBE-NEXT:    bx lr
-; CHECKBE-NEXT:    .p2align 4
-; CHECKBE-NEXT:  @ %bb.1:
-; CHECKBE-NEXT:  .LCPI28_0:
-; CHECKBE-NEXT:    .long 1065353216 @ double 0.007812501848093234
-; CHECKBE-NEXT:    .long 1065353216
-; CHECKBE-NEXT:    .long 1065353216 @ double 0.007812501848093234
-; CHECKBE-NEXT:    .long 1065353216
+; CHECK-LABEL: mov_float_1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    mov.w r0, #1065353216
+; CHECK-NEXT:    vdup.32 q0, r0
+; CHECK-NEXT:    bx lr
 entry:
   ret <4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
 }
 
 define arm_aapcs_vfpcc <4 x float> @mov_float_m3() {
-; CHECKLE-LABEL: mov_float_m3:
-; CHECKLE:       @ %bb.0: @ %entry
-; CHECKLE-NEXT:    adr r0, .LCPI29_0
-; CHECKLE-NEXT:    vldrw.u32 q0, [r0]
-; CHECKLE-NEXT:    bx lr
-; CHECKLE-NEXT:    .p2align 4
-; CHECKLE-NEXT:  @ %bb.1:
-; CHECKLE-NEXT:  .LCPI29_0:
-; CHECKLE-NEXT:    .long 3225419776 @ double -32.000022917985916
-; CHECKLE-NEXT:    .long 3225419776
-; CHECKLE-NEXT:    .long 3225419776 @ double -32.000022917985916
-; CHECKLE-NEXT:    .long 3225419776
-;
-; CHECKBE-LABEL: mov_float_m3:
-; CHECKBE:       @ %bb.0: @ %entry
-; CHECKBE-NEXT:    adr r0, .LCPI29_0
-; CHECKBE-NEXT:    vldrb.u8 q1, [r0]
-; CHECKBE-NEXT:    vrev64.8 q0, q1
-; CHECKBE-NEXT:    bx lr
-; CHECKBE-NEXT:    .p2align 4
-; CHECKBE-NEXT:  @ %bb.1:
-; CHECKBE-NEXT:  .LCPI29_0:
-; CHECKBE-NEXT:    .long 3225419776 @ double -32.000022917985916
-; CHECKBE-NEXT:    .long 3225419776
-; CHECKBE-NEXT:    .long 3225419776 @ double -32.000022917985916
-; CHECKBE-NEXT:    .long 3225419776
+; CHECK-LABEL: mov_float_m3:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    movs r0, #0
+; CHECK-NEXT:    movt r0, #49216
+; CHECK-NEXT:    vdup.32 q0, r0
+; CHECK-NEXT:    bx lr
 entry:
   ret <4 x float> <float -3.000000e+00, float -3.000000e+00, float -3.000000e+00, float -3.000000e+00>
 }
 
 define arm_aapcs_vfpcc <8 x half> @mov_float16_1() {
 ; CHECK-LABEL: mov_float16_1:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i16 q0, #0x3c00

llvm/test/CodeGen/Thumb2/mve-vmvnimm.ll

 ; CHECK-NEXT:    bx lr
 entry:
   ret <4 x i32> <i32 4294902528, i32 4294902528, i32 4294902528, i32 4294902528>
 }
 
 define arm_aapcs_vfpcc <4 x i32> @mov_int32_4278386688() {
 ; CHECK-LABEL: mov_int32_4278386688:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    adr r0, .LCPI7_0
-; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    movs r0, #0
+; CHECK-NEXT:    movt r0, #65283
+; CHECK-NEXT:    vdup.32 q0, r0
 ; CHECK-NEXT:    bx lr
-; CHECK-NEXT:    .p2align 4
-; CHECK-NEXT:  @ %bb.1:
-; CHECK-NEXT:  .LCPI7_0:
-; CHECK-NEXT:    .long 4278386688 @ double -6.5147775434702224E+303
-; CHECK-NEXT:    .long 4278386688
-; CHECK-NEXT:    .long 4278386688 @ double -6.5147775434702224E+303
-; CHECK-NEXT:    .long 4278386688
 entry:
   ret <4 x i32> <i32 4278386688, i32 4278386688, i32 4278386688, i32 4278386688>
 }