[ARM] Fix loads and stores for predicate vectors

[dmgreen]

Summary

These predicate vectors can usually be loaded and stored with a single instruction, a VSTR_P0. However this instruction will store the entire P0 predicate, 16 bits, zeroextended to 32bits. Each lane of the the v4i1/v8i1 representing 4/2 bits.

As far as I understand, when llvm says "store this v4i1", it really does need to store 4 bits (or 8, that being the size of a byte, with this bottom 4 as the interesting bits). For example a bitcast from a v8i1 to a i8 is defined as a store followed by a load, which is how the code is expanded.

So this instead lowers the v4i1/v8i1 load/store through some shuffles to get the bits into the correct positions. This, as you might imagine, is not as efficient as a single instruction. But I believe it is needed for correctness. v16i1 equally should not load/store 32bits, only storing the 16bits of data. Stack loads/stores are still using the VSTR_P0 (as can be seen by the test not changing). This is fine as they are self-consistent, it is only "externally observable loads/stores" (from our point of view) that need to be corrected.

The test changes here are in pred-bitcast (which is no longer incorrect), pred-ldst (which is obviously a lot larger, but I don't believe will be generated a lot), and masked ld/st. The masked ld/st test we should be able to optimise better with a few folds, and we should not be generating masked ld/st only to expand them like this.

[SjoerdMeijer]

Hi Dave, I agree with your analysis. The codegen for this looks horrible, but it is what it is.

One question though. Loading and storing data.....do we need to worry about LE and BE here?

llvm/lib/Target/ARM/ARMISelLowering.cpp

8745 ↗

(On Diff #218385)

nit, typo: itto

[samparker]

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

23 ↗

(On Diff #218385)

I'm missing something here... from my understanding:

• Select 16 bytes, taken from q2 (0xff) and q1(0x0), building a vector predicate mask in q1.
• Then we take the bottom 4 bytes from q1, the mask, putting each into a 32-bit lane of q2.
• Then we compare the 32-bit lanes of q2 against zero.
• Then we select bytes from q0 (%a) and q1 (zero).

It's the second point that I don't understand... why do we only access the lower lanes of q1?

[dmgreen]

One question though. Loading and story and story data.....do we need to worry about LE and BE here?

I believe that this is OK. At least for these two we are only storing a single byte, so it shouldn't be an issue.

It does bring up the issue of whether the v16i1 is correct. It looks like the instruction is loading/storing 32bits, so might not be really right to use if we should only be storing 16bits, even in LE. And for BE would be putting the bits in the wrong place. Let me try and adjust that one too.

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

23 ↗

(On Diff #218385)

This is converting an i4, as in the bottom 4 bits of r0, into a v4i1, as in the 16 bits of P0. It needs to be sort of "shuffled" or a "signext_inreg'd" to get those bits into the correct places.

So the "step 0" in you list would be copy the bits from r0 to p0 using the msr, and the 4 bits of interest are in bottom 4 bits of p0. The rest is the awkward sign extend.

The alternative is to do this in integer instructions. That may be better, depending on the circumstances. I don't think (hope) this will come up a lot though. We just need to not get it wrong.

[samparker]

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

23 ↗

(On Diff #218385)

Ah, yes, hadn't considered the i4 properly. cheers.

[dmgreen]

Now with v16i1, which doesn't need the extracts/buildvector, just going through the vmrs into a vstrh store.

This has the nice effect for bitcasts of eliding the load and store, giving us the bitcast lowering to vmsr for free (unfortunately the stack is still realigned. That should be fixable with a better preferred vector alignment though).

[SjoerdMeijer]

I am happy with this if Sam is happy too.

[samparker]

Yeah, LGTM.

llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp

for (auto VT : pTypes) {
     setOperationAction(ISD::BUILD_VECTOR, VT, Custom);
     setOperationAction(ISD::VECTOR_SHUFFLE, VT, Custom);
     setOperationAction(ISD::EXTRACT_SUBVECTOR, VT, Custom);
     setOperationAction(ISD::CONCAT_VECTORS, VT, Custom);
     setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Custom);
     setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Custom);
     setOperationAction(ISD::SETCC, VT, Custom);
     setOperationAction(ISD::SCALAR_TO_VECTOR, VT, Expand);
+    setOperationAction(ISD::LOAD, VT, Custom);
+    setOperationAction(ISD::STORE, VT, Custom);
   }
 }
 
 ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM,
                                      const ARMSubtarget &STI)
     : TargetLowering(TM), Subtarget(&STI) {
   RegInfo = Subtarget->getRegisterInfo();
   Itins = Subtarget->getInstrItineraryData();
void ARMTargetLowering::ExpandDIV_Windows(
   SDValue Upper = DAG.getNode(ISD::SRL, dl, MVT::i64, Result,
                               DAG.getConstant(32, dl, TLI.getPointerTy(DL)));
   Upper = DAG.getNode(ISD::TRUNCATE, dl, MVT::i32, Upper);
 
   Results.push_back(Lower);
   Results.push_back(Upper);
 }
 
+static SDValue LowerPredicateLoad(SDValue Op, SelectionDAG &DAG) {
+  LoadSDNode *LD = cast<LoadSDNode>(Op.getNode());
+  EVT MemVT = LD->getMemoryVT();
+  assert((MemVT == MVT::v4i1 || MemVT == MVT::v8i1 || MemVT == MVT::v16i1) &&
+         "Expected a predicate type!");
+  assert(MemVT == Op.getValueType());
+  assert(LD->getExtensionType() == ISD::NON_EXTLOAD &&
+         "Expected a non-extending load");
+  assert(LD->isUnindexed() && "Expected a unindexed load");
+
+  // The basic MVE VLDR on a v4i1/v8i1 actually loads the entire 16bit
+  // predicate, with the "v4i1" bits spread out over the 16 bits loaded. We
+  // need to make sure that 8/4 bits are actually loaded into the correct
+  // place, which means loading the value and then shuffling the values into
+  // the bottom bits of the predicate.
+  // Equally, VLDR for an v16i1 will actually load 32bits (so will be incorrect
+  // for BE).
+
+  SDLoc dl(Op);
+  SDValue Load = DAG.getExtLoad(
+      ISD::EXTLOAD, dl, MVT::i32, LD->getChain(), LD->getBasePtr(),
+      EVT::getIntegerVT(*DAG.getContext(), MemVT.getSizeInBits()),
+      LD->getMemOperand());
+  SDValue Pred = DAG.getNode(ARMISD::PREDICATE_CAST, dl, MVT::v16i1, Load);
+  if (MemVT != MVT::v16i1)
+    Pred = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MemVT, Pred,
+                       DAG.getConstant(0, dl, MVT::i32));
+  return DAG.getMergeValues({Pred, Load.getValue(1)}, dl);
+}
+
+static SDValue LowerPredicateStore(SDValue Op, SelectionDAG &DAG) {
+  StoreSDNode *ST = cast<StoreSDNode>(Op.getNode());
+  EVT MemVT = ST->getMemoryVT();
+  assert((MemVT == MVT::v4i1 || MemVT == MVT::v8i1 || MemVT == MVT::v16i1) &&
+         "Expected a predicate type!");
+  assert(MemVT == ST->getValue().getValueType());
+  assert(!ST->isTruncatingStore() && "Expected a non-extending store");
+  assert(ST->isUnindexed() && "Expected a unindexed store");
+
+  // Only store the v4i1 or v8i1 worth of bits, via a buildvector with top bits
+  // unset and a scalar store.
+  SDLoc dl(Op);
+  SDValue Build = ST->getValue();
+  if (MemVT != MVT::v16i1) {
+    SmallVector<SDValue, 16> Ops;
+    for (unsigned I = 0; I < MemVT.getVectorNumElements(); I++)
+      Ops.push_back(DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::i32, Build,
+                                DAG.getConstant(I, dl, MVT::i32)));
+    for (unsigned I = MemVT.getVectorNumElements(); I < 16; I++)
+      Ops.push_back(DAG.getUNDEF(MVT::i32));
+    Build = DAG.getNode(ISD::BUILD_VECTOR, dl, MVT::v16i1, Ops);
+  }
+  SDValue GRP = DAG.getNode(ARMISD::PREDICATE_CAST, dl, MVT::i32, Build);
+  return DAG.getTruncStore(
+      ST->getChain(), dl, GRP, ST->getBasePtr(),
+      EVT::getIntegerVT(*DAG.getContext(), MemVT.getSizeInBits()),
+      ST->getMemOperand());
+}
+
 static SDValue LowerAtomicLoadStore(SDValue Op, SelectionDAG &DAG) {
   if (isStrongerThanMonotonic(cast<AtomicSDNode>(Op)->getOrdering()))
     // Acquire/Release load/store is not legal for targets without a dmb or
     // equivalent available.
     return SDValue();
 
   // Monotonic load/store is legal for all targets.
   return Op;
SDValue ARMTargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) const {
   case ISD::ADDCARRY:
   case ISD::SUBCARRY:      return LowerADDSUBCARRY(Op, DAG);
   case ISD::SADDO:
   case ISD::SSUBO:
     return LowerSignedALUO(Op, DAG);
   case ISD::UADDO:
   case ISD::USUBO:
     return LowerUnsignedALUO(Op, DAG);
+  case ISD::LOAD:
+    return LowerPredicateLoad(Op, DAG);
+  case ISD::STORE:
+    return LowerPredicateStore(Op, DAG);
   case ISD::ATOMIC_LOAD:
   case ISD::ATOMIC_STORE:  return LowerAtomicLoadStore(Op, DAG);
   case ISD::FSINCOS:       return LowerFSINCOS(Op, DAG);
   case ISD::SDIVREM:
   case ISD::UDIVREM:       return LowerDivRem(Op, DAG);
   case ISD::DYNAMIC_STACKALLOC:
     if (Subtarget->isTargetWindows())
       return LowerDYNAMIC_STACKALLOC(Op, DAG);

llvm/trunk/lib/Target/ARM/ARMInstrMVE.td

let Predicates = [HasMVEInt, IsBE] in {
   def : MVE_vector_offset_store_typed<v8f16, MVE_VSTRHU16_pre, aligned16_pre_store, 1>;
   def : MVE_vector_offset_store_typed<v8f16, MVE_VSTRHU16_post, aligned16_post_store, 1>;
   def : MVE_vector_offset_store_typed<v4i32, MVE_VSTRWU32_pre, aligned32_pre_store, 2>;
   def : MVE_vector_offset_store_typed<v4i32, MVE_VSTRWU32_post, aligned32_post_store, 2>;
   def : MVE_vector_offset_store_typed<v4f32, MVE_VSTRWU32_pre, aligned32_pre_store, 2>;
   def : MVE_vector_offset_store_typed<v4f32, MVE_VSTRWU32_post, aligned32_post_store, 2>;
 }
 
-let Predicates = [HasMVEInt] in {
-  // Predicate loads
-  def  : Pat<(v16i1 (load t2addrmode_imm7<2>:$addr)),
-             (v16i1 (VLDR_P0_off t2addrmode_imm7<2>:$addr))>;
-  def  : Pat<(v8i1 (load t2addrmode_imm7<2>:$addr)),
-             (v8i1 (VLDR_P0_off t2addrmode_imm7<2>:$addr))>;
-  def  : Pat<(v4i1 (load t2addrmode_imm7<2>:$addr)),
-             (v4i1 (VLDR_P0_off t2addrmode_imm7<2>:$addr))>;
-
-  // Predicate stores
-  def  : Pat<(store (v4i1 VCCR:$val), t2addrmode_imm7<2>:$addr),
-             (VSTR_P0_off VCCR:$val, t2addrmode_imm7<2>:$addr)>;
-  def  : Pat<(store (v8i1 VCCR:$val), t2addrmode_imm7<2>:$addr),
-             (VSTR_P0_off VCCR:$val, t2addrmode_imm7<2>:$addr)>;
-  def  : Pat<(store (v16i1 VCCR:$val), t2addrmode_imm7<2>:$addr),
-             (VSTR_P0_off VCCR:$val, t2addrmode_imm7<2>:$addr)>;
-}
-
 
 // Widening/Narrowing Loads/Stores
 
 let MinAlignment = 2 in {
   def truncstorevi16_align2 : PatFrag<(ops node:$val, node:$ptr),
                                       (truncstorevi16 node:$val, node:$ptr)>;
   def post_truncstvi16_align2 : PatFrag<(ops node:$val, node:$base, node:$offset),
                                         (post_truncstvi16 node:$val, node:$base, node:$offset)>;

llvm/trunk/test/CodeGen/Thumb2/mve-masked-ldst.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve,+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-LE
 ; RUN: llc -mtriple=thumbebv8.1m.main-arm-none-eabi -mattr=+mve,+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-BE
 
 define void @foo_v4i32_v4i32(<4 x i32> *%dest, <4 x i32> *%mask, <4 x i32> *%src) {
 ; CHECK-LABEL: foo_v4i32_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #8
 ; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #4
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s32 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #4]
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    and r1, r3, #15
 ; CHECK-NEXT:    lsls r3, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrne r3, [r2]
 ; CHECK-NEXT:    vmovne.32 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrmi r3, [r2, #4]
 ; CHECK-NEXT:    vmovmi.32 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrmi r3, [r2, #8]
 ; CHECK-NEXT:    vmovmi.32 q0[2], r3
 ; CHECK-NEXT:    lsls r1, r1, #28
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrmi r1, [r2, #12]
 ; CHECK-NEXT:    vmovmi.32 q0[3], r1
-; CHECK-NEXT:    mov r1, sp
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    movs r1, #0
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #2, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    bfi r1, r2, #3, #1
+; CHECK-NEXT:    and r1, r1, #15
 ; CHECK-NEXT:    lsls r2, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne r2, s0
 ; CHECK-NEXT:    strne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi r2, s1
 ; CHECK-NEXT:    strmi r2, [r0, #4]

 }
 
 define void @foo_sext_v4i32_v4i8(<4 x i32> *%dest, <4 x i32> *%mask, <4 x i8> *%src) {
 ; CHECK-LABEL: foo_sext_v4i32_v4i8:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #8
 ; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #4
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s32 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #4]
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    and r1, r3, #15
 ; CHECK-NEXT:    lsls r3, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrbne r3, [r2]
 ; CHECK-NEXT:    vmovne.32 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #1]
 ; CHECK-NEXT:    vmovmi.32 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #2]
 ; CHECK-NEXT:    vmovmi.32 q0[2], r3
 ; CHECK-NEXT:    lsls r1, r1, #28
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r1, [r2, #3]
 ; CHECK-NEXT:    vmovmi.32 q0[3], r1
-; CHECK-NEXT:    mov r1, sp
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    vmovlb.s8 q0, q0
-; CHECK-NEXT:    vstr p0, [r1]
 ; CHECK-NEXT:    vmovlb.s16 q0, q0
-; CHECK-NEXT:    ldrb.w r1, [sp]
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #2, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    bfi r1, r2, #3, #1
+; CHECK-NEXT:    and r1, r1, #15
 ; CHECK-NEXT:    lsls r2, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne r2, s0
 ; CHECK-NEXT:    strne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi r2, s1
 ; CHECK-NEXT:    strmi r2, [r0, #4]

 }
 
 define void @foo_sext_v4i32_v4i16(<4 x i32> *%dest, <4 x i32> *%mask, <4 x i16> *%src) {
 ; CHECK-LABEL: foo_sext_v4i32_v4i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #8
 ; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #4
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s32 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #4]
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    and r1, r3, #15
 ; CHECK-NEXT:    lsls r3, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrhne r3, [r2]
 ; CHECK-NEXT:    vmovne.32 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r3, [r2, #2]
 ; CHECK-NEXT:    vmovmi.32 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r3, [r2, #4]
 ; CHECK-NEXT:    vmovmi.32 q0[2], r3
 ; CHECK-NEXT:    lsls r1, r1, #28
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r1, [r2, #6]
 ; CHECK-NEXT:    vmovmi.32 q0[3], r1
-; CHECK-NEXT:    mov r1, sp
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    vmovlb.s16 q0, q0
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #2, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    bfi r1, r2, #3, #1
+; CHECK-NEXT:    and r1, r1, #15
 ; CHECK-NEXT:    lsls r2, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne r2, s0
 ; CHECK-NEXT:    strne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi r2, s1
 ; CHECK-NEXT:    strmi r2, [r0, #4]

 }
 
 define void @foo_zext_v4i32_v4i8(<4 x i32> *%dest, <4 x i32> *%mask, <4 x i8> *%src) {
 ; CHECK-LABEL: foo_zext_v4i32_v4i8:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #8
 ; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #4
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vmov.i32 q1, #0xff
 ; CHECK-NEXT:    vcmp.s32 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #4]
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    and r1, r3, #15
 ; CHECK-NEXT:    lsls r3, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrbne r3, [r2]
 ; CHECK-NEXT:    vmovne.32 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #1]
 ; CHECK-NEXT:    vmovmi.32 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #2]
 ; CHECK-NEXT:    vmovmi.32 q0[2], r3
 ; CHECK-NEXT:    lsls r1, r1, #28
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r1, [r2, #3]
 ; CHECK-NEXT:    vmovmi.32 q0[3], r1
-; CHECK-NEXT:    mov r1, sp
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #2, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    bfi r1, r2, #3, #1
+; CHECK-NEXT:    and r1, r1, #15
 ; CHECK-NEXT:    lsls r2, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne r2, s0
 ; CHECK-NEXT:    strne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi r2, s1
 ; CHECK-NEXT:    strmi r2, [r0, #4]

 }
 
 define void @foo_zext_v4i32_v4i16(<4 x i32> *%dest, <4 x i32> *%mask, <4 x i16> *%src) {
 ; CHECK-LABEL: foo_zext_v4i32_v4i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #8
 ; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #4
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s32 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #4]
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    and r1, r3, #15
 ; CHECK-NEXT:    lsls r3, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrhne r3, [r2]
 ; CHECK-NEXT:    vmovne.32 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r3, [r2, #2]
 ; CHECK-NEXT:    vmovmi.32 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r3, [r2, #4]
 ; CHECK-NEXT:    vmovmi.32 q0[2], r3
 ; CHECK-NEXT:    lsls r1, r1, #28
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r1, [r2, #6]
 ; CHECK-NEXT:    vmovmi.32 q0[3], r1
-; CHECK-NEXT:    mov r1, sp
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    vmovlb.u16 q0, q0
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #2, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    bfi r1, r2, #3, #1
+; CHECK-NEXT:    and r1, r1, #15
 ; CHECK-NEXT:    lsls r2, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne r2, s0
 ; CHECK-NEXT:    strne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi r2, s1
 ; CHECK-NEXT:    strmi r2, [r0, #4]

 }
 
 define void @foo_v8i16_v8i16(<8 x i16> *%dest, <8 x i16> *%mask, <8 x i16> *%src) {
 ; CHECK-LABEL: foo_v8i16_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #16
 ; CHECK-NEXT:    sub sp, #16
 ; CHECK-NEXT:    vldrh.u16 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #8
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s16 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #8]
-; CHECK-NEXT:    lsls r3, r1, #31
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #2, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #4, #1
+; CHECK-NEXT:    ubfx r1, r12, #10, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #5, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #6, #1
+; CHECK-NEXT:    ubfx r1, r12, #14, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r3
+; CHECK-NEXT:    lsls r3, r3, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrhne r3, [r2]
 ; CHECK-NEXT:    vmovne.16 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r3, [r2, #2]
 ; CHECK-NEXT:    vmovmi.16 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29

 ; CHECK-NEXT:    lsls r3, r1, #25
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r3, [r2, #12]
 ; CHECK-NEXT:    vmovmi.16 q0[6], r3
 ; CHECK-NEXT:    lsls r1, r1, #24
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r1, [r2, #14]
 ; CHECK-NEXT:    vmovmi.16 q0[7], r1
-; CHECK-NEXT:    mov r1, sp
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
-; CHECK-NEXT:    lsls r2, r1, #31
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    and r3, r1, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #2, #1
+; CHECK-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #3, #1
+; CHECK-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #4, #1
+; CHECK-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #5, #1
+; CHECK-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r2, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r2
+; CHECK-NEXT:    lsls r2, r2, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne.u16 r2, q0[0]
 ; CHECK-NEXT:    strhne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi.u16 r2, q0[1]
 ; CHECK-NEXT:    strhmi r2, [r0, #2]
 ; CHECK-NEXT:    lsls r2, r1, #29

 }
 
 define void @foo_sext_v8i16_v8i8(<8 x i16> *%dest, <8 x i16> *%mask, <8 x i8> *%src) {
 ; CHECK-LABEL: foo_sext_v8i16_v8i8:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #16
 ; CHECK-NEXT:    sub sp, #16
 ; CHECK-NEXT:    vldrh.u16 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #8
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s16 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #8]
-; CHECK-NEXT:    lsls r3, r1, #31
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #2, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #4, #1
+; CHECK-NEXT:    ubfx r1, r12, #10, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #5, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #6, #1
+; CHECK-NEXT:    ubfx r1, r12, #14, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r3
+; CHECK-NEXT:    lsls r3, r3, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrbne r3, [r2]
 ; CHECK-NEXT:    vmovne.16 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #1]
 ; CHECK-NEXT:    vmovmi.16 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29

 ; CHECK-NEXT:    lsls r3, r1, #25
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #6]
 ; CHECK-NEXT:    vmovmi.16 q0[6], r3
 ; CHECK-NEXT:    lsls r1, r1, #24
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r1, [r2, #7]
 ; CHECK-NEXT:    vmovmi.16 q0[7], r1
-; CHECK-NEXT:    mov r1, sp
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vmrs r1, p0
 ; CHECK-NEXT:    vmovlb.s8 q0, q0
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
-; CHECK-NEXT:    lsls r2, r1, #31
+; CHECK-NEXT:    and r3, r1, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #2, #1
+; CHECK-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #3, #1
+; CHECK-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #4, #1
+; CHECK-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #5, #1
+; CHECK-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r2, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r2
+; CHECK-NEXT:    lsls r2, r2, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne.u16 r2, q0[0]
 ; CHECK-NEXT:    strhne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi.u16 r2, q0[1]
 ; CHECK-NEXT:    strhmi r2, [r0, #2]
 ; CHECK-NEXT:    lsls r2, r1, #29

 }
 
 define void @foo_zext_v8i16_v8i8(<8 x i16> *%dest, <8 x i16> *%mask, <8 x i8> *%src) {
 ; CHECK-LABEL: foo_zext_v8i16_v8i8:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #16
 ; CHECK-NEXT:    sub sp, #16
 ; CHECK-NEXT:    vldrh.u16 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #8
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s16 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #8]
-; CHECK-NEXT:    lsls r3, r1, #31
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #2, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #4, #1
+; CHECK-NEXT:    ubfx r1, r12, #10, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #5, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #6, #1
+; CHECK-NEXT:    ubfx r1, r12, #14, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r3
+; CHECK-NEXT:    lsls r3, r3, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrbne r3, [r2]
 ; CHECK-NEXT:    vmovne.16 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #1]
 ; CHECK-NEXT:    vmovmi.16 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29

 ; CHECK-NEXT:    lsls r3, r1, #25
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #6]
 ; CHECK-NEXT:    vmovmi.16 q0[6], r3
 ; CHECK-NEXT:    lsls r1, r1, #24
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r1, [r2, #7]
 ; CHECK-NEXT:    vmovmi.16 q0[7], r1
-; CHECK-NEXT:    mov r1, sp
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vmrs r1, p0
 ; CHECK-NEXT:    vmovlb.u8 q0, q0
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
-; CHECK-NEXT:    lsls r2, r1, #31
+; CHECK-NEXT:    and r3, r1, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #2, #1
+; CHECK-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #3, #1
+; CHECK-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #4, #1
+; CHECK-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #5, #1
+; CHECK-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r2, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r2
+; CHECK-NEXT:    lsls r2, r2, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne.u16 r2, q0[0]
 ; CHECK-NEXT:    strhne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi.u16 r2, q0[1]
 ; CHECK-NEXT:    strhmi r2, [r0, #2]
 ; CHECK-NEXT:    lsls r2, r1, #29

 ; CHECK-NEXT:    .setfp r7, sp, #8
 ; CHECK-NEXT:    add r7, sp, #8
 ; CHECK-NEXT:    .pad #32
 ; CHECK-NEXT:    sub sp, #32
 ; CHECK-NEXT:    mov r4, sp
 ; CHECK-NEXT:    bfc r4, #0, #4
 ; CHECK-NEXT:    mov sp, r4
 ; CHECK-NEXT:    vldrb.u8 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #16
 ; CHECK-NEXT:    sub.w r4, r7, #8
 ; CHECK-NEXT:    vcmp.s8 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrh.w r1, [sp, #16]
-; CHECK-NEXT:    lsls r3, r1, #31
+; CHECK-NEXT:    vmrs r3, p0
+; CHECK-NEXT:    uxth r1, r3
+; CHECK-NEXT:    lsls r3, r3, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrbne r3, [r2]
 ; CHECK-NEXT:    vmovne.8 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #1]
 ; CHECK-NEXT:    vmovmi.8 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29

 ; CHECK-NEXT:    lsls r3, r1, #17
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r3, [r2, #14]
 ; CHECK-NEXT:    vmovmi.8 q0[14], r3
 ; CHECK-NEXT:    lsls r1, r1, #16
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrbmi r1, [r2, #15]
 ; CHECK-NEXT:    vmovmi.8 q0[15], r1
-; CHECK-NEXT:    mov r1, sp
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrh.w r1, [sp]
-; CHECK-NEXT:    lsls r2, r1, #31
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    uxth r1, r2
+; CHECK-NEXT:    lsls r2, r2, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne.u8 r2, q0[0]
 ; CHECK-NEXT:    strbne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi.u8 r2, q0[1]
 ; CHECK-NEXT:    strbmi r2, [r0, #1]
 ; CHECK-NEXT:    lsls r2, r1, #29

 }
 
 define void @foo_trunc_v8i8_v8i16(<8 x i8> *%dest, <8 x i16> *%mask, <8 x i16> *%src) {
 ; CHECK-LABEL: foo_trunc_v8i8_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #16
 ; CHECK-NEXT:    sub sp, #16
 ; CHECK-NEXT:    vldrh.u16 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #8
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s16 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #8]
-; CHECK-NEXT:    lsls r3, r1, #31
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #2, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #4, #1
+; CHECK-NEXT:    ubfx r1, r12, #10, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #5, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #6, #1
+; CHECK-NEXT:    ubfx r1, r12, #14, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r3
+; CHECK-NEXT:    lsls r3, r3, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrhne r3, [r2]
 ; CHECK-NEXT:    vmovne.16 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r3, [r2, #2]
 ; CHECK-NEXT:    vmovmi.16 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29

 ; CHECK-NEXT:    lsls r3, r1, #25
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r3, [r2, #12]
 ; CHECK-NEXT:    vmovmi.16 q0[6], r3
 ; CHECK-NEXT:    lsls r1, r1, #24
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrhmi r1, [r2, #14]
 ; CHECK-NEXT:    vmovmi.16 q0[7], r1
-; CHECK-NEXT:    mov r1, sp
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
-; CHECK-NEXT:    lsls r2, r1, #31
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    and r3, r1, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #2, #1
+; CHECK-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #3, #1
+; CHECK-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #4, #1
+; CHECK-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #5, #1
+; CHECK-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r2, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r2
+; CHECK-NEXT:    lsls r2, r2, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne.u16 r2, q0[0]
 ; CHECK-NEXT:    strbne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi.u16 r2, q0[1]
 ; CHECK-NEXT:    strbmi r2, [r0, #1]
 ; CHECK-NEXT:    lsls r2, r1, #29

 }
 
 define void @foo_trunc_v4i8_v4i32(<4 x i8> *%dest, <4 x i32> *%mask, <4 x i32> *%src) {
 ; CHECK-LABEL: foo_trunc_v4i8_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #8
 ; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #4
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s32 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #4]
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    and r1, r3, #15
 ; CHECK-NEXT:    lsls r3, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrne r3, [r2]
 ; CHECK-NEXT:    vmovne.32 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrmi r3, [r2, #4]
 ; CHECK-NEXT:    vmovmi.32 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrmi r3, [r2, #8]
 ; CHECK-NEXT:    vmovmi.32 q0[2], r3
 ; CHECK-NEXT:    lsls r1, r1, #28
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrmi r1, [r2, #12]
 ; CHECK-NEXT:    vmovmi.32 q0[3], r1
-; CHECK-NEXT:    mov r1, sp
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    movs r1, #0
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #2, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    bfi r1, r2, #3, #1
+; CHECK-NEXT:    and r1, r1, #15
 ; CHECK-NEXT:    lsls r2, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne r2, s0
 ; CHECK-NEXT:    strbne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi r2, s1
 ; CHECK-NEXT:    strbmi r2, [r0, #1]

 }
 
 define void @foo_trunc_v4i16_v4i32(<4 x i16> *%dest, <4 x i32> *%mask, <4 x i32> *%src) {
 ; CHECK-LABEL: foo_trunc_v4i16_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #8
 ; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #4
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s32 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #4]
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    and r1, r3, #15
 ; CHECK-NEXT:    lsls r3, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    ldrne r3, [r2]
 ; CHECK-NEXT:    vmovne.32 q0[0], r3
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrmi r3, [r2, #4]
 ; CHECK-NEXT:    vmovmi.32 q0[1], r3
 ; CHECK-NEXT:    lsls r3, r1, #29
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrmi r3, [r2, #8]
 ; CHECK-NEXT:    vmovmi.32 q0[2], r3
 ; CHECK-NEXT:    lsls r1, r1, #28
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    ldrmi r1, [r2, #12]
 ; CHECK-NEXT:    vmovmi.32 q0[3], r1
-; CHECK-NEXT:    mov r1, sp
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    movs r1, #0
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #2, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    bfi r1, r2, #3, #1
+; CHECK-NEXT:    and r1, r1, #15
 ; CHECK-NEXT:    lsls r2, r1, #31
 ; CHECK-NEXT:    itt ne
 ; CHECK-NEXT:    vmovne r2, s0
 ; CHECK-NEXT:    strhne r2, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    itt mi
 ; CHECK-NEXT:    vmovmi r2, s1
 ; CHECK-NEXT:    strhmi r2, [r0, #2]

 }
 
 define void @foo_v4f32_v4f32(<4 x float> *%dest, <4 x i32> *%mask, <4 x float> *%src) {
 ; CHECK-LABEL: foo_v4f32_v4f32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #8
 ; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #4
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s32 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #4]
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    and r1, r3, #15
 ; CHECK-NEXT:    lsls r3, r1, #31
 ; CHECK-NEXT:    it ne
 ; CHECK-NEXT:    vldrne s0, [r2]
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    it mi
 ; CHECK-NEXT:    vldrmi s1, [r2, #4]
 ; CHECK-NEXT:    lsls r3, r1, #29
 ; CHECK-NEXT:    it mi
 ; CHECK-NEXT:    vldrmi s2, [r2, #8]
 ; CHECK-NEXT:    lsls r1, r1, #28
 ; CHECK-NEXT:    it mi
 ; CHECK-NEXT:    vldrmi s3, [r2, #12]
-; CHECK-NEXT:    mov r1, sp
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    movs r1, #0
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r1, r3, #2, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    bfi r1, r2, #3, #1
+; CHECK-NEXT:    and r1, r1, #15
 ; CHECK-NEXT:    lsls r2, r1, #31
 ; CHECK-NEXT:    it ne
 ; CHECK-NEXT:    vstrne s0, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    it mi
 ; CHECK-NEXT:    vstrmi s1, [r0, #4]
 ; CHECK-NEXT:    lsls r2, r1, #29
 ; CHECK-NEXT:    it mi

 }
 
 define void @foo_v8f16_v8f16(<8 x half> *%dest, <8 x i16> *%mask, <8 x half> *%src) {
 ; CHECK-LABEL: foo_v8f16_v8f16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .pad #16
 ; CHECK-NEXT:    sub sp, #16
 ; CHECK-NEXT:    vldrh.u16 q0, [r1]
-; CHECK-NEXT:    add r3, sp, #8
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vcmp.s16 gt, q0, zr
 ; CHECK-NEXT:    @ implicit-def: $q0
-; CHECK-NEXT:    vstr p0, [r3]
-; CHECK-NEXT:    ldrb.w r1, [sp, #8]
-; CHECK-NEXT:    lsls r3, r1, #31
+; CHECK-NEXT:    vmrs r12, p0
+; CHECK-NEXT:    and r1, r12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #0, #1
+; CHECK-NEXT:    ubfx r1, r12, #2, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #1, #1
+; CHECK-NEXT:    ubfx r1, r12, #4, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #2, #1
+; CHECK-NEXT:    ubfx r1, r12, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #3, #1
+; CHECK-NEXT:    ubfx r1, r12, #8, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #4, #1
+; CHECK-NEXT:    ubfx r1, r12, #10, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #5, #1
+; CHECK-NEXT:    ubfx r1, r12, #12, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #6, #1
+; CHECK-NEXT:    ubfx r1, r12, #14, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r3, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r3
+; CHECK-NEXT:    lsls r3, r3, #31
 ; CHECK-NEXT:    bne .LBB13_18
 ; CHECK-NEXT:  @ %bb.1: @ %else
 ; CHECK-NEXT:    lsls r3, r1, #30
 ; CHECK-NEXT:    bmi .LBB13_19
 ; CHECK-NEXT:  .LBB13_2: @ %else2
 ; CHECK-NEXT:    lsls r3, r1, #29
 ; CHECK-NEXT:    bmi .LBB13_20
 ; CHECK-NEXT:  .LBB13_3: @ %else5

 ; CHECK-NEXT:  .LBB13_7: @ %else17
 ; CHECK-NEXT:    lsls r1, r1, #24
 ; CHECK-NEXT:    bpl .LBB13_9
 ; CHECK-NEXT:  .LBB13_8: @ %cond.load19
 ; CHECK-NEXT:    vldr.16 s4, [r2, #14]
 ; CHECK-NEXT:    vmov r1, s4
 ; CHECK-NEXT:    vmov.16 q0[7], r1
 ; CHECK-NEXT:  .LBB13_9: @ %else20
-; CHECK-NEXT:    mov r1, sp
-; CHECK-NEXT:    vstr p0, [r1]
-; CHECK-NEXT:    ldrb.w r1, [sp]
-; CHECK-NEXT:    lsls r2, r1, #31
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    and r3, r1, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #0, #1
+; CHECK-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #1, #1
+; CHECK-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #2, #1
+; CHECK-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #3, #1
+; CHECK-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #4, #1
+; CHECK-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #5, #1
+; CHECK-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    bfi r2, r3, #6, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    bfi r2, r1, #7, #1
+; CHECK-NEXT:    uxtb r1, r2
+; CHECK-NEXT:    lsls r2, r2, #31
 ; CHECK-NEXT:    bne .LBB13_25
 ; CHECK-NEXT:  @ %bb.10: @ %else23
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    bmi .LBB13_26
 ; CHECK-NEXT:  .LBB13_11: @ %else25
 ; CHECK-NEXT:    lsls r2, r1, #29
 ; CHECK-NEXT:    bmi .LBB13_27
 ; CHECK-NEXT:  .LBB13_12: @ %else27

 ; CHECK-NEXT:    vmov.16 q0[4], r3
 ; CHECK-NEXT:    lsls r3, r1, #26
 ; CHECK-NEXT:    bpl .LBB13_6
 ; CHECK-NEXT:  .LBB13_23: @ %cond.load13
 ; CHECK-NEXT:    vldr.16 s4, [r2, #10]
 ; CHECK-NEXT:    vmov r3, s4
 ; CHECK-NEXT:    vmov.16 q0[5], r3
 ; CHECK-NEXT:    lsls r3, r1, #25
-; CHECK-NEXT:    bpl .LBB13_7
+; CHECK-NEXT:    bpl.w .LBB13_7
 ; CHECK-NEXT:  .LBB13_24: @ %cond.load16
 ; CHECK-NEXT:    vldr.16 s4, [r2, #12]
 ; CHECK-NEXT:    vmov r3, s4
 ; CHECK-NEXT:    vmov.16 q0[6], r3
 ; CHECK-NEXT:    lsls r1, r1, #24
-; CHECK-NEXT:    bmi .LBB13_8
+; CHECK-NEXT:    bmi.w .LBB13_8
 ; CHECK-NEXT:    b .LBB13_9
 ; CHECK-NEXT:  .LBB13_25: @ %cond.store
 ; CHECK-NEXT:    vstr.16 s0, [r0]
 ; CHECK-NEXT:    lsls r2, r1, #30
 ; CHECK-NEXT:    bpl .LBB13_11
 ; CHECK-NEXT:  .LBB13_26: @ %cond.store24
 ; CHECK-NEXT:    vmovx.f16 s4, s0
 ; CHECK-NEXT:    vstr.16 s4, [r0, #2]

llvm/trunk/test/CodeGen/Thumb2/mve-masked-load.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve,+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-LE
 ; RUN: llc -mtriple=thumbebv8.1m.main-arm-none-eabi -mattr=+mve,+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-BE
 
 define arm_aapcs_vfpcc <4 x i32> @masked_v4i32_align4_zero(<4 x i32> *%dest, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32_align4_zero:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    mov.w r12, #0
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #3, #1
+; CHECK-LE-NEXT:    and r1, r2, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    beq .LBB0_2
 ; CHECK-LE-NEXT:  @ %bb.1: @ %cond.load
-; CHECK-LE-NEXT:    movs r2, #0
-; CHECK-LE-NEXT:    ldr r3, [r0]
-; CHECK-LE-NEXT:    vdup.32 q0, r2
-; CHECK-LE-NEXT:    vmov.32 q0[0], r3
+; CHECK-LE-NEXT:    ldr r2, [r0]
+; CHECK-LE-NEXT:    vdup.32 q0, r12
+; CHECK-LE-NEXT:    vmov.32 q0[0], r2
 ; CHECK-LE-NEXT:    b .LBB0_3
 ; CHECK-LE-NEXT:  .LBB0_2:
 ; CHECK-LE-NEXT:    vmov.i32 q0, #0x0
 ; CHECK-LE-NEXT:  .LBB0_3: @ %else
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #8]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[2], r2
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r0, [r0, #12]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[3], r0
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32_align4_zero:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    mov.w r12, #0
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #3, #1
+; CHECK-BE-NEXT:    and r1, r2, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    beq .LBB0_2
 ; CHECK-BE-NEXT:  @ %bb.1: @ %cond.load
-; CHECK-BE-NEXT:    movs r2, #0
-; CHECK-BE-NEXT:    ldr r3, [r0]
-; CHECK-BE-NEXT:    vdup.32 q1, r2
-; CHECK-BE-NEXT:    vmov.32 q1[0], r3
+; CHECK-BE-NEXT:    ldr r2, [r0]
+; CHECK-BE-NEXT:    vdup.32 q1, r12
+; CHECK-BE-NEXT:    vmov.32 q1[0], r2
 ; CHECK-BE-NEXT:    b .LBB0_3
 ; CHECK-BE-NEXT:  .LBB0_2:
 ; CHECK-BE-NEXT:    vmov.i32 q1, #0x0
 ; CHECK-BE-NEXT:  .LBB0_3: @ %else
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-BE-NEXT:    vmovmi.32 q1[1], r2

 }
 
 define arm_aapcs_vfpcc <4 x i32> @masked_v4i32_align4_undef(<4 x i32> *%dest, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32_align4_undef:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    mov r1, sp
-; CHECK-LE-NEXT:    vstr p0, [r1]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.32 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #8]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[2], r2
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r0, [r0, #12]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[3], r0
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32_align4_undef:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.32 q1[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-BE-NEXT:    vmovmi.32 q1[1], r2

 }
 
 define arm_aapcs_vfpcc <4 x i32> @masked_v4i32_align1_undef(<4 x i32> *%dest, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32_align1_undef:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    mov r1, sp
-; CHECK-LE-NEXT:    vstr p0, [r1]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.32 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #8]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[2], r2
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r0, [r0, #12]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[3], r0
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32_align1_undef:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.32 q1[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-BE-NEXT:    vmovmi.32 q1[1], r2
entry:
   ret <4 x i32> %l
 }
 
 define arm_aapcs_vfpcc <4 x i32> @masked_v4i32_align4_other(<4 x i32> *%dest, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32_align4_other:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.32 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #8]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[2], r2
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r0, [r0, #12]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[3], r0
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32_align4_other:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.32 q1[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-BE-NEXT:    vmovmi.32 q1[1], r2

 }
 
 define arm_aapcs_vfpcc i8* @masked_v4i32_preinc(i8* %x, i8* %y, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32_preinc:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vstr p0, [r2]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r12, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
 ; CHECK-LE-NEXT:    adds r0, #4
-; CHECK-LE-NEXT:    ldrb.w r2, [sp]
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    and r2, r2, #15
 ; CHECK-LE-NEXT:    lsls r3, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrne r3, [r0]
 ; CHECK-LE-NEXT:    vmovne.32 q0[0], r3
 ; CHECK-LE-NEXT:    lsls r3, r2, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r3, [r0, #4]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[1], r3

 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32_preinc:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
 ; CHECK-BE-NEXT:    adds r0, #4
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    and r2, r2, #15
 ; CHECK-BE-NEXT:    lsls r3, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrne r3, [r0]
 ; CHECK-BE-NEXT:    vmovne.32 q0[0], r3
 ; CHECK-BE-NEXT:    lsls r3, r2, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrmi r3, [r0, #4]
 ; CHECK-BE-NEXT:    vmovmi.32 q0[1], r3

 }
 
 define arm_aapcs_vfpcc i8* @masked_v4i32_postinc(i8* %x, i8* %y, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32_postinc:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vstr p0, [r2]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r12, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
 ; CHECK-LE-NEXT:    add.w r12, r0, #4
-; CHECK-LE-NEXT:    ldrb.w r3, [sp]
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    and r3, r2, #15
 ; CHECK-LE-NEXT:    lsls r2, r3, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.32 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r3, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-LE-NEXT:    vmovmi.32 q0[1], r2

 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32_postinc:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
 ; CHECK-BE-NEXT:    add.w r12, r0, #4
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r3, [sp]
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    and r3, r2, #15
 ; CHECK-BE-NEXT:    lsls r2, r3, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.32 q0[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r3, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-BE-NEXT:    vmovmi.32 q0[1], r2

 
 
 
 define arm_aapcs_vfpcc <8 x i16> @masked_v8i16_align4_zero(<8 x i16> *%dest, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16_align4_zero:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    mov.w r12, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r2, r3, #0
+; CHECK-LE-NEXT:    movs r3, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r3, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r3
+; CHECK-LE-NEXT:    lsls r2, r3, #31
 ; CHECK-LE-NEXT:    beq .LBB6_2
 ; CHECK-LE-NEXT:  @ %bb.1: @ %cond.load
-; CHECK-LE-NEXT:    movs r2, #0
-; CHECK-LE-NEXT:    ldrh r3, [r0]
-; CHECK-LE-NEXT:    vdup.16 q0, r2
-; CHECK-LE-NEXT:    vmov.16 q0[0], r3
+; CHECK-LE-NEXT:    ldrh r2, [r0]
+; CHECK-LE-NEXT:    vdup.16 q0, r12
+; CHECK-LE-NEXT:    vmov.16 q0[0], r2
 ; CHECK-LE-NEXT:    b .LBB6_3
 ; CHECK-LE-NEXT:  .LBB6_2:
 ; CHECK-LE-NEXT:    vmov.i32 q0, #0x0
 ; CHECK-LE-NEXT:  .LBB6_3: @ %else
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrhmi r2, [r0, #2]
 ; CHECK-LE-NEXT:    vmovmi.16 q0[1], r2

 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8i16_align4_zero:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    mov.w r12, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r2, r3, #0
+; CHECK-BE-NEXT:    movs r3, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r3, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r3
+; CHECK-BE-NEXT:    lsls r2, r3, #31
 ; CHECK-BE-NEXT:    beq .LBB6_2
 ; CHECK-BE-NEXT:  @ %bb.1: @ %cond.load
-; CHECK-BE-NEXT:    movs r2, #0
-; CHECK-BE-NEXT:    ldrh r3, [r0]
-; CHECK-BE-NEXT:    vdup.16 q1, r2
-; CHECK-BE-NEXT:    vmov.16 q1[0], r3
+; CHECK-BE-NEXT:    ldrh r2, [r0]
+; CHECK-BE-NEXT:    vdup.16 q1, r12
+; CHECK-BE-NEXT:    vmov.16 q1[0], r2
 ; CHECK-BE-NEXT:    b .LBB6_3
 ; CHECK-BE-NEXT:  .LBB6_2:
 ; CHECK-BE-NEXT:    vmov.i32 q0, #0x0
 ; CHECK-BE-NEXT:    vrev32.16 q1, q0
 ; CHECK-BE-NEXT:  .LBB6_3: @ %else
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrhmi r2, [r0, #2]

 }
 
 define arm_aapcs_vfpcc <8 x i16> @masked_v8i16_align4_undef(<8 x i16> *%dest, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16_align4_undef:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    mov r1, sp
-; CHECK-LE-NEXT:    vstr p0, [r1]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrhne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.16 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrhmi r2, [r0, #2]
 ; CHECK-LE-NEXT:    vmovmi.16 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29

 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8i16_align4_undef:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrhne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.16 q1[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrhmi r2, [r0, #2]
 ; CHECK-BE-NEXT:    vmovmi.16 q1[1], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29

 }
 
 define arm_aapcs_vfpcc <8 x i16> @masked_v8i16_align1_undef(<8 x i16> *%dest, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16_align1_undef:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    mov r1, sp
-; CHECK-LE-NEXT:    vstr p0, [r1]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrhne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.16 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrhmi r2, [r0, #2]
 ; CHECK-LE-NEXT:    vmovmi.16 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29

 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8i16_align1_undef:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrhne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.16 q1[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrhmi r2, [r0, #2]
 ; CHECK-BE-NEXT:    vmovmi.16 q1[1], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
entry:
   ret <8 x i16> %l
 }
 
 define arm_aapcs_vfpcc <8 x i16> @masked_v8i16_align4_other(<8 x i16> *%dest, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16_align4_other:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrhne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.16 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrhmi r2, [r0, #2]
 ; CHECK-LE-NEXT:    vmovmi.16 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29

 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8i16_align4_other:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrhne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.16 q1[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrhmi r2, [r0, #2]
 ; CHECK-BE-NEXT:    vmovmi.16 q1[1], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29

 define i8* @masked_v8i16_preinc(i8* %x, i8* %y, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16_preinc:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-LE-NEXT:    adds r0, #4
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    mov r2, sp
+; CHECK-LE-NEXT:    movs r3, #0
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrb.w r2, [sp]
-; CHECK-LE-NEXT:    lsls r3, r2, #31
+; CHECK-LE-NEXT:    vmrs r12, p0
+; CHECK-LE-NEXT:    and r2, r12, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #6, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #10, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #14, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #7, #1
+; CHECK-LE-NEXT:    uxtb r2, r3
+; CHECK-LE-NEXT:    lsls r3, r3, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrhne r3, [r0]
 ; CHECK-LE-NEXT:    vmovne.16 q0[0], r3
 ; CHECK-LE-NEXT:    lsls r3, r2, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrhmi r3, [r0, #2]
 ; CHECK-LE-NEXT:    vmovmi.16 q0[1], r3
 ; CHECK-LE-NEXT:    lsls r3, r2, #29

 ;
 ; CHECK-BE-LABEL: masked_v8i16_preinc:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-BE-NEXT:    adds r0, #4
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r3, #0
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp]
-; CHECK-BE-NEXT:    lsls r3, r2, #31
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r2, r12, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #6, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #10, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #14, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #7, #1
+; CHECK-BE-NEXT:    uxtb r2, r3
+; CHECK-BE-NEXT:    lsls r3, r3, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrhne r3, [r0]
 ; CHECK-BE-NEXT:    vmovne.16 q0[0], r3
 ; CHECK-BE-NEXT:    lsls r3, r2, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrhmi r3, [r0, #2]
 ; CHECK-BE-NEXT:    vmovmi.16 q0[1], r3
 ; CHECK-BE-NEXT:    lsls r3, r2, #29

 }
 
 define arm_aapcs_vfpcc i8* @masked_v8i16_postinc(i8* %x, i8* %y, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16_postinc:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vstr p0, [r2]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r12, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
 ; CHECK-LE-NEXT:    add.w r12, r0, #4
-; CHECK-LE-NEXT:    ldrb.w r3, [sp]
-; CHECK-LE-NEXT:    lsls r2, r3, #31
+; CHECK-LE-NEXT:    bfi r2, r3, #7, #1
+; CHECK-LE-NEXT:    uxtb r3, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrhne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.16 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r3, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrhmi r2, [r0, #2]
 ; CHECK-LE-NEXT:    vmovmi.16 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r3, #29

 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8i16_postinc:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
 ; CHECK-BE-NEXT:    add.w r12, r0, #4
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r3, [sp]
-; CHECK-BE-NEXT:    lsls r2, r3, #31
+; CHECK-BE-NEXT:    bfi r2, r3, #7, #1
+; CHECK-BE-NEXT:    uxtb r3, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrhne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.16 q0[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r3, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrhmi r2, [r0, #2]
 ; CHECK-BE-NEXT:    vmovmi.16 q0[1], r2
 ; CHECK-BE-NEXT:    lsls r2, r3, #29

 ; CHECK-LE-NEXT:    push {r4, r6, r7, lr}
 ; CHECK-LE-NEXT:    .setfp r7, sp, #8
 ; CHECK-LE-NEXT:    add r7, sp, #8
 ; CHECK-LE-NEXT:    .pad #16
 ; CHECK-LE-NEXT:    sub sp, #16
 ; CHECK-LE-NEXT:    mov r4, sp
 ; CHECK-LE-NEXT:    bfc r4, #0, #4
 ; CHECK-LE-NEXT:    mov sp, r4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s8 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrh.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    uxth r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    beq .LBB12_2
 ; CHECK-LE-NEXT:  @ %bb.1: @ %cond.load
 ; CHECK-LE-NEXT:    movs r2, #0
 ; CHECK-LE-NEXT:    ldrb r3, [r0]
 ; CHECK-LE-NEXT:    vdup.8 q0, r2
 ; CHECK-LE-NEXT:    vmov.8 q0[0], r3
 ; CHECK-LE-NEXT:    b .LBB12_3
 ; CHECK-LE-NEXT:  .LBB12_2:

 ; CHECK-BE-NEXT:    .setfp r7, sp, #8
 ; CHECK-BE-NEXT:    add r7, sp, #8
 ; CHECK-BE-NEXT:    .pad #16
 ; CHECK-BE-NEXT:    sub sp, #16
 ; CHECK-BE-NEXT:    mov r4, sp
 ; CHECK-BE-NEXT:    bfc r4, #0, #4
 ; CHECK-BE-NEXT:    mov sp, r4
 ; CHECK-BE-NEXT:    vrev64.8 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
 ; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrh.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    uxth r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    beq .LBB12_2
 ; CHECK-BE-NEXT:  @ %bb.1: @ %cond.load
 ; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    ldrb r3, [r0]
 ; CHECK-BE-NEXT:    vdup.8 q1, r2
 ; CHECK-BE-NEXT:    vmov.8 q1[0], r3
 ; CHECK-BE-NEXT:    b .LBB12_3
 ; CHECK-BE-NEXT:  .LBB12_2:

 ; CHECK-LE-NEXT:    .setfp r7, sp, #8
 ; CHECK-LE-NEXT:    add r7, sp, #8
 ; CHECK-LE-NEXT:    .pad #16
 ; CHECK-LE-NEXT:    sub sp, #16
 ; CHECK-LE-NEXT:    mov r4, sp
 ; CHECK-LE-NEXT:    bfc r4, #0, #4
 ; CHECK-LE-NEXT:    mov sp, r4
 ; CHECK-LE-NEXT:    vcmp.s8 gt, q0, zr
-; CHECK-LE-NEXT:    mov r1, sp
-; CHECK-LE-NEXT:    vstr p0, [r1]
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
 ; CHECK-LE-NEXT:    sub.w r4, r7, #8
-; CHECK-LE-NEXT:    ldrh.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    uxth r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrbne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.8 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrbmi r2, [r0, #1]
 ; CHECK-LE-NEXT:    vmovmi.8 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29

 ; CHECK-BE-NEXT:    .setfp r7, sp, #8
 ; CHECK-BE-NEXT:    add r7, sp, #8
 ; CHECK-BE-NEXT:    .pad #16
 ; CHECK-BE-NEXT:    sub sp, #16
 ; CHECK-BE-NEXT:    mov r4, sp
 ; CHECK-BE-NEXT:    bfc r4, #0, #4
 ; CHECK-BE-NEXT:    mov sp, r4
 ; CHECK-BE-NEXT:    vrev64.8 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    sub.w r4, r7, #8
 ; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q1
-; CHECK-BE-NEXT:    sub.w r4, r7, #8
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrh.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    uxth r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrbne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.8 q1[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrbmi r2, [r0, #1]
 ; CHECK-BE-NEXT:    vmovmi.8 q1[1], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29

 ; CHECK-LE-NEXT:    push {r4, r6, r7, lr}
 ; CHECK-LE-NEXT:    .setfp r7, sp, #8
 ; CHECK-LE-NEXT:    add r7, sp, #8
 ; CHECK-LE-NEXT:    .pad #16
 ; CHECK-LE-NEXT:    sub sp, #16
 ; CHECK-LE-NEXT:    mov r4, sp
 ; CHECK-LE-NEXT:    bfc r4, #0, #4
 ; CHECK-LE-NEXT:    mov sp, r4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s8 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
 ; CHECK-LE-NEXT:    sub.w r4, r7, #8
-; CHECK-LE-NEXT:    ldrh.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    uxth r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrbne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.8 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrbmi r2, [r0, #1]
 ; CHECK-LE-NEXT:    vmovmi.8 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29

 ; CHECK-BE-NEXT:    .setfp r7, sp, #8
 ; CHECK-BE-NEXT:    add r7, sp, #8
 ; CHECK-BE-NEXT:    .pad #16
 ; CHECK-BE-NEXT:    sub sp, #16
 ; CHECK-BE-NEXT:    mov r4, sp
 ; CHECK-BE-NEXT:    bfc r4, #0, #4
 ; CHECK-BE-NEXT:    mov sp, r4
 ; CHECK-BE-NEXT:    vrev64.8 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
-; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
 ; CHECK-BE-NEXT:    sub.w r4, r7, #8
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrh.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    uxth r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrbne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.8 q1[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrbmi r2, [r0, #1]
 ; CHECK-BE-NEXT:    vmovmi.8 q1[1], r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29

 ; CHECK-LE-NEXT:    .setfp r7, sp, #8
 ; CHECK-LE-NEXT:    add r7, sp, #8
 ; CHECK-LE-NEXT:    .pad #16
 ; CHECK-LE-NEXT:    sub sp, #16
 ; CHECK-LE-NEXT:    mov r4, sp
 ; CHECK-LE-NEXT:    bfc r4, #0, #4
 ; CHECK-LE-NEXT:    mov sp, r4
 ; CHECK-LE-NEXT:    vcmp.s8 gt, q0, zr
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vstr p0, [r2]
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
 ; CHECK-LE-NEXT:    adds r0, #4
-; CHECK-LE-NEXT:    ldrh.w r2, [sp]
+; CHECK-LE-NEXT:    vmrs r3, p0
 ; CHECK-LE-NEXT:    sub.w r4, r7, #8
-; CHECK-LE-NEXT:    lsls r3, r2, #31
+; CHECK-LE-NEXT:    uxth r2, r3
+; CHECK-LE-NEXT:    lsls r3, r3, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrbne r3, [r0]
 ; CHECK-LE-NEXT:    vmovne.8 q0[0], r3
 ; CHECK-LE-NEXT:    lsls r3, r2, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrbmi r3, [r0, #1]
 ; CHECK-LE-NEXT:    vmovmi.8 q0[1], r3
 ; CHECK-LE-NEXT:    lsls r3, r2, #29

 ; CHECK-BE-NEXT:    .setfp r7, sp, #8
 ; CHECK-BE-NEXT:    add r7, sp, #8
 ; CHECK-BE-NEXT:    .pad #16
 ; CHECK-BE-NEXT:    sub sp, #16
 ; CHECK-BE-NEXT:    mov r4, sp
 ; CHECK-BE-NEXT:    bfc r4, #0, #4
 ; CHECK-BE-NEXT:    mov sp, r4
 ; CHECK-BE-NEXT:    vrev64.8 q1, q0
-; CHECK-BE-NEXT:    mov r2, sp
-; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
 ; CHECK-BE-NEXT:    adds r0, #4
-; CHECK-BE-NEXT:    vstr p0, [r2]
+; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
 ; CHECK-BE-NEXT:    sub.w r4, r7, #8
-; CHECK-BE-NEXT:    ldrh.w r2, [sp]
-; CHECK-BE-NEXT:    lsls r3, r2, #31
+; CHECK-BE-NEXT:    vmrs r3, p0
+; CHECK-BE-NEXT:    uxth r2, r3
+; CHECK-BE-NEXT:    lsls r3, r3, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrbne r3, [r0]
 ; CHECK-BE-NEXT:    vmovne.8 q0[0], r3
 ; CHECK-BE-NEXT:    lsls r3, r2, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrbmi r3, [r0, #1]
 ; CHECK-BE-NEXT:    vmovmi.8 q0[1], r3
 ; CHECK-BE-NEXT:    lsls r3, r2, #29

 ; CHECK-LE-NEXT:    .setfp r7, sp, #8
 ; CHECK-LE-NEXT:    add r7, sp, #8
 ; CHECK-LE-NEXT:    .pad #16
 ; CHECK-LE-NEXT:    sub sp, #16
 ; CHECK-LE-NEXT:    mov r4, sp
 ; CHECK-LE-NEXT:    bfc r4, #0, #4
 ; CHECK-LE-NEXT:    mov sp, r4
 ; CHECK-LE-NEXT:    vcmp.s8 gt, q0, zr
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vstr p0, [r2]
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
 ; CHECK-LE-NEXT:    sub.w r4, r7, #8
-; CHECK-LE-NEXT:    ldrh.w r3, [sp]
+; CHECK-LE-NEXT:    vmrs r2, p0
 ; CHECK-LE-NEXT:    add.w r12, r0, #4
-; CHECK-LE-NEXT:    lsls r2, r3, #31
+; CHECK-LE-NEXT:    uxth r3, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrbne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne.8 q0[0], r2
 ; CHECK-LE-NEXT:    lsls r2, r3, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrbmi r2, [r0, #1]
 ; CHECK-LE-NEXT:    vmovmi.8 q0[1], r2
 ; CHECK-LE-NEXT:    lsls r2, r3, #29

 ; CHECK-BE-NEXT:    .setfp r7, sp, #8
 ; CHECK-BE-NEXT:    add r7, sp, #8
 ; CHECK-BE-NEXT:    .pad #16
 ; CHECK-BE-NEXT:    sub sp, #16
 ; CHECK-BE-NEXT:    mov r4, sp
 ; CHECK-BE-NEXT:    bfc r4, #0, #4
 ; CHECK-BE-NEXT:    mov sp, r4
 ; CHECK-BE-NEXT:    vrev64.8 q1, q0
-; CHECK-BE-NEXT:    mov r2, sp
-; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
 ; CHECK-BE-NEXT:    sub.w r4, r7, #8
-; CHECK-BE-NEXT:    vstr p0, [r2]
+; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
 ; CHECK-BE-NEXT:    add.w r12, r0, #4
-; CHECK-BE-NEXT:    ldrh.w r3, [sp]
-; CHECK-BE-NEXT:    lsls r2, r3, #31
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    uxth r3, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrbne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne.8 q0[0], r2
 ; CHECK-BE-NEXT:    lsls r2, r3, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrbmi r2, [r0, #1]
 ; CHECK-BE-NEXT:    vmovmi.8 q0[1], r2
 ; CHECK-BE-NEXT:    lsls r2, r3, #29

 }
 
 
 define arm_aapcs_vfpcc <4 x float> @masked_v4f32_align4_zero(<4 x float> *%dest, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4f32_align4_zero:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    beq .LBB17_2
 ; CHECK-LE-NEXT:  @ %bb.1: @ %cond.load
 ; CHECK-LE-NEXT:    vldr s0, .LCPI17_0
 ; CHECK-LE-NEXT:    vldr s4, [r0]
 ; CHECK-LE-NEXT:    vmov r2, s0
 ; CHECK-LE-NEXT:    vdup.32 q0, r2
 ; CHECK-LE-NEXT:    vmov.f32 s0, s4

 ; CHECK-LE-NEXT:  .LCPI17_0:
 ; CHECK-LE-NEXT:    .long 0 @ float 0
 ;
 ; CHECK-BE-LABEL: masked_v4f32_align4_zero:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    beq .LBB17_2
 ; CHECK-BE-NEXT:  @ %bb.1: @ %cond.load
 ; CHECK-BE-NEXT:    vldr s0, .LCPI17_0
 ; CHECK-BE-NEXT:    vldr s2, [r0]
 ; CHECK-BE-NEXT:    vmov r2, s0
 ; CHECK-BE-NEXT:    vdup.32 q1, r2
 ; CHECK-BE-NEXT:    vmov.f32 s4, s2

 }
 
 define arm_aapcs_vfpcc <4 x float> @masked_v4f32_align4_undef(<4 x float> *%dest, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4f32_align4_undef:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    mov r1, sp
-; CHECK-LE-NEXT:    vstr p0, [r1]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    it ne
 ; CHECK-LE-NEXT:    vldrne s0, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s1, [r0, #4]
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s2, [r0, #8]
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s3, [r0, #12]
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4f32_align4_undef:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    it ne
 ; CHECK-BE-NEXT:    vldrne s4, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    it mi
 ; CHECK-BE-NEXT:    vldrmi s5, [r0, #4]
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    it mi

 }
 
 define arm_aapcs_vfpcc <4 x float> @masked_v4f32_align1_undef(<4 x float> *%dest, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4f32_align1_undef:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    mov r1, sp
-; CHECK-LE-NEXT:    vstr p0, [r1]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    ldrne r2, [r0]
 ; CHECK-LE-NEXT:    vmovne s0, r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-LE-NEXT:    vmovmi s1, r2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r2, [r0, #8]
 ; CHECK-LE-NEXT:    vmovmi s2, r2
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    ldrmi r0, [r0, #12]
 ; CHECK-LE-NEXT:    vmovmi s3, r0
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4f32_align1_undef:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    ldrne r2, [r0]
 ; CHECK-BE-NEXT:    vmovne s4, r2
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    ldrmi r2, [r0, #4]
 ; CHECK-BE-NEXT:    vmovmi s5, r2
entry:
   ret <4 x float> %l
 }
 
 define arm_aapcs_vfpcc <4 x float> @masked_v4f32_align4_other(<4 x float> *%dest, <4 x i32> %a, <4 x float> %b) {
 ; CHECK-LE-LABEL: masked_v4f32_align4_other:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    it ne
 ; CHECK-LE-NEXT:    vldrne s4, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s5, [r0, #4]
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s6, [r0, #8]
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s7, [r0, #12]
 ; CHECK-LE-NEXT:    vmov q0, q1
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4f32_align4_other:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q2, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q2, zr
 ; CHECK-BE-NEXT:    vrev64.32 q2, q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    it ne
 ; CHECK-BE-NEXT:    vldrne s8, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    it mi
 ; CHECK-BE-NEXT:    vldrmi s9, [r0, #4]
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    it mi

 }
 
 define arm_aapcs_vfpcc i8* @masked_v4f32_preinc(i8* %x, i8* %y, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4f32_preinc:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vstr p0, [r2]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r12, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
 ; CHECK-LE-NEXT:    adds r0, #4
-; CHECK-LE-NEXT:    ldrb.w r2, [sp]
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    and r2, r2, #15
 ; CHECK-LE-NEXT:    lsls r3, r2, #31
 ; CHECK-LE-NEXT:    it ne
 ; CHECK-LE-NEXT:    vldrne s0, [r0]
 ; CHECK-LE-NEXT:    lsls r3, r2, #30
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s1, [r0, #4]
 ; CHECK-LE-NEXT:    lsls r3, r2, #29
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s2, [r0, #8]
 ; CHECK-LE-NEXT:    lsls r2, r2, #28
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s3, [r0, #12]
 ; CHECK-LE-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4f32_preinc:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
 ; CHECK-BE-NEXT:    adds r0, #4
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    and r2, r2, #15
 ; CHECK-BE-NEXT:    lsls r3, r2, #31
 ; CHECK-BE-NEXT:    it ne
 ; CHECK-BE-NEXT:    vldrne s0, [r0]
 ; CHECK-BE-NEXT:    lsls r3, r2, #30
 ; CHECK-BE-NEXT:    it mi
 ; CHECK-BE-NEXT:    vldrmi s1, [r0, #4]
 ; CHECK-BE-NEXT:    lsls r3, r2, #29
 ; CHECK-BE-NEXT:    it mi

 }
 
 define arm_aapcs_vfpcc i8* @masked_v4f32_postinc(i8* %x, i8* %y, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4f32_postinc:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vstr p0, [r2]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r12, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
 ; CHECK-LE-NEXT:    add.w r12, r0, #4
-; CHECK-LE-NEXT:    ldrb.w r3, [sp]
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    and r3, r2, #15
 ; CHECK-LE-NEXT:    lsls r2, r3, #31
 ; CHECK-LE-NEXT:    it ne
 ; CHECK-LE-NEXT:    vldrne s0, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r3, #30
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s1, [r0, #4]
 ; CHECK-LE-NEXT:    lsls r2, r3, #29
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s2, [r0, #8]
 ; CHECK-LE-NEXT:    lsls r2, r3, #28
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vldrmi s3, [r0, #12]
 ; CHECK-LE-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-LE-NEXT:    mov r0, r12
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4f32_postinc:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
 ; CHECK-BE-NEXT:    add.w r12, r0, #4
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r3, [sp]
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    and r3, r2, #15
 ; CHECK-BE-NEXT:    lsls r2, r3, #31
 ; CHECK-BE-NEXT:    it ne
 ; CHECK-BE-NEXT:    vldrne s0, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r3, #30
 ; CHECK-BE-NEXT:    it mi
 ; CHECK-BE-NEXT:    vldrmi s1, [r0, #4]
 ; CHECK-BE-NEXT:    lsls r2, r3, #29
 ; CHECK-BE-NEXT:    it mi

 }
 
 
 define arm_aapcs_vfpcc <8 x half> @masked_v8f16_align4_zero(<8 x half> *%dest, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8f16_align4_zero:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    beq .LBB23_2
 ; CHECK-LE-NEXT:  @ %bb.1: @ %cond.load
 ; CHECK-LE-NEXT:    vldr.16 s0, .LCPI23_0
 ; CHECK-LE-NEXT:    vmov r2, s0
 ; CHECK-LE-NEXT:    vldr.16 s0, [r0]
 ; CHECK-LE-NEXT:    vmov r3, s0
 ; CHECK-LE-NEXT:    vdup.16 q0, r2
 ; CHECK-LE-NEXT:    vmov.16 q0[0], r3

 ; CHECK-LE-NEXT:  .LCPI23_0:
 ; CHECK-LE-NEXT:    .short 0 @ half 0
 ;
 ; CHECK-BE-LABEL: masked_v8f16_align4_zero:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    beq .LBB23_2
 ; CHECK-BE-NEXT:  @ %bb.1: @ %cond.load
 ; CHECK-BE-NEXT:    vldr.16 s0, .LCPI23_0
 ; CHECK-BE-NEXT:    vmov r2, s0
 ; CHECK-BE-NEXT:    vldr.16 s0, [r0]
 ; CHECK-BE-NEXT:    vdup.16 q1, r2
 ; CHECK-BE-NEXT:    vmov r3, s0
 ; CHECK-BE-NEXT:    vmov.16 q1[0], r3

 }
 
 define arm_aapcs_vfpcc <8 x half> @masked_v8f16_align4_undef(<8 x half> *%dest, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8f16_align4_undef:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    mov r1, sp
-; CHECK-LE-NEXT:    vstr p0, [r1]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    bne .LBB24_9
 ; CHECK-LE-NEXT:  @ %bb.1: @ %else
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    bmi .LBB24_10
 ; CHECK-LE-NEXT:  .LBB24_2: @ %else2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    bmi .LBB24_11
 ; CHECK-LE-NEXT:  .LBB24_3: @ %else5

 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8f16_align4_undef:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    bne .LBB24_10
 ; CHECK-BE-NEXT:  @ %bb.1: @ %else
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    bmi .LBB24_11
 ; CHECK-BE-NEXT:  .LBB24_2: @ %else2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    bmi .LBB24_12
 ; CHECK-BE-NEXT:  .LBB24_3: @ %else5

 }
 
 define arm_aapcs_vfpcc <8 x half> @masked_v8f16_align1_undef(<8 x half> *%dest, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8f16_align1_undef:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #40
 ; CHECK-LE-NEXT:    sub sp, #40
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    add r1, sp, #32
-; CHECK-LE-NEXT:    vstr p0, [r1]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    ldrb.w r1, [sp, #32]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    bne .LBB25_9
 ; CHECK-LE-NEXT:  @ %bb.1: @ %else
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    bmi .LBB25_10
 ; CHECK-LE-NEXT:  .LBB25_2: @ %else2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    bmi .LBB25_11
 ; CHECK-LE-NEXT:  .LBB25_3: @ %else5

 ; CHECK-LE-NEXT:    add sp, #40
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8f16_align1_undef:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #40
 ; CHECK-BE-NEXT:    sub sp, #40
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    add r1, sp, #32
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp, #32]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    bne .LBB25_10
 ; CHECK-BE-NEXT:  @ %bb.1: @ %else
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    bmi .LBB25_11
 ; CHECK-BE-NEXT:  .LBB25_2: @ %else2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    bmi .LBB25_12
 ; CHECK-BE-NEXT:  .LBB25_3: @ %else5
entry:
   ret <8 x half> %l
 }
 
 define arm_aapcs_vfpcc <8 x half> @masked_v8f16_align4_other(<8 x half> *%dest, <8 x i16> %a, <8 x half> %b) {
 ; CHECK-LE-LABEL: masked_v8f16_align4_other:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    bne .LBB26_10
 ; CHECK-LE-NEXT:  @ %bb.1: @ %else
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    bmi .LBB26_11
 ; CHECK-LE-NEXT:  .LBB26_2: @ %else2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    bmi .LBB26_12
 ; CHECK-LE-NEXT:  .LBB26_3: @ %else5

 ; CHECK-LE-NEXT:    bmi .LBB26_8
 ; CHECK-LE-NEXT:    b .LBB26_9
 ;
 ; CHECK-BE-LABEL: masked_v8f16_align4_other:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q2, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q2, zr
 ; CHECK-BE-NEXT:    vrev64.16 q2, q1
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    bne .LBB26_10
 ; CHECK-BE-NEXT:  @ %bb.1: @ %else
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    bmi .LBB26_11
 ; CHECK-BE-NEXT:  .LBB26_2: @ %else2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    bmi .LBB26_12
 ; CHECK-BE-NEXT:  .LBB26_3: @ %else5

 }
 
 define arm_aapcs_vfpcc i8* @masked_v8f16_preinc(i8* %x, i8* %y, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8f16_preinc:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vstr p0, [r2]
+; CHECK-LE-NEXT:    movs r3, #0
+; CHECK-LE-NEXT:    vmrs r12, p0
 ; CHECK-LE-NEXT:    adds r0, #4
-; CHECK-LE-NEXT:    ldrb.w r2, [sp]
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    lsls r3, r2, #31
+; CHECK-LE-NEXT:    and r2, r12, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #6, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #10, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #14, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #7, #1
+; CHECK-LE-NEXT:    uxtb r2, r3
+; CHECK-LE-NEXT:    lsls r3, r3, #31
 ; CHECK-LE-NEXT:    bne .LBB27_10
 ; CHECK-LE-NEXT:  @ %bb.1: @ %else
 ; CHECK-LE-NEXT:    lsls r3, r2, #30
 ; CHECK-LE-NEXT:    bmi .LBB27_11
 ; CHECK-LE-NEXT:  .LBB27_2: @ %else2
 ; CHECK-LE-NEXT:    lsls r3, r2, #29
 ; CHECK-LE-NEXT:    bmi .LBB27_12
 ; CHECK-LE-NEXT:  .LBB27_3: @ %else5

 ; CHECK-LE-NEXT:    bmi .LBB27_8
 ; CHECK-LE-NEXT:    b .LBB27_9
 ;
 ; CHECK-BE-LABEL: masked_v8f16_preinc:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r3, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
 ; CHECK-BE-NEXT:    adds r0, #4
-; CHECK-BE-NEXT:    vstr p0, [r2]
+; CHECK-BE-NEXT:    vmrs r12, p0
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
-; CHECK-BE-NEXT:    ldrb.w r2, [sp]
-; CHECK-BE-NEXT:    lsls r3, r2, #31
+; CHECK-BE-NEXT:    and r2, r12, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #6, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #10, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #14, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #7, #1
+; CHECK-BE-NEXT:    uxtb r2, r3
+; CHECK-BE-NEXT:    lsls r3, r3, #31
 ; CHECK-BE-NEXT:    bne .LBB27_10
 ; CHECK-BE-NEXT:  @ %bb.1: @ %else
 ; CHECK-BE-NEXT:    lsls r3, r2, #30
 ; CHECK-BE-NEXT:    bmi .LBB27_11
 ; CHECK-BE-NEXT:  .LBB27_2: @ %else2
 ; CHECK-BE-NEXT:    lsls r3, r2, #29
 ; CHECK-BE-NEXT:    bmi .LBB27_12
 ; CHECK-BE-NEXT:  .LBB27_3: @ %else5

 }
 
 define arm_aapcs_vfpcc i8* @masked_v8f16_postinc(i8* %x, i8* %y, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8f16_postinc:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vstr p0, [r2]
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r12, p0
 ; CHECK-LE-NEXT:    @ implicit-def: $q0
-; CHECK-LE-NEXT:    ldrb.w r3, [sp]
-; CHECK-LE-NEXT:    lsls r2, r3, #31
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #7, #1
+; CHECK-LE-NEXT:    uxtb r3, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    bne .LBB28_12
 ; CHECK-LE-NEXT:  @ %bb.1: @ %else
 ; CHECK-LE-NEXT:    lsls r2, r3, #30
 ; CHECK-LE-NEXT:    bmi .LBB28_13
 ; CHECK-LE-NEXT:  .LBB28_2: @ %else2
 ; CHECK-LE-NEXT:    lsls r2, r3, #29
 ; CHECK-LE-NEXT:    bmi .LBB28_14
 ; CHECK-LE-NEXT:  .LBB28_3: @ %else5

 ; CHECK-LE-NEXT:    bmi .LBB28_6
 ; CHECK-LE-NEXT:    b .LBB28_7
 ;
 ; CHECK-BE-LABEL: masked_v8f16_postinc:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
 ; CHECK-BE-NEXT:    @ implicit-def: $q0
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r3, [sp]
-; CHECK-BE-NEXT:    lsls r2, r3, #31
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #7, #1
+; CHECK-BE-NEXT:    uxtb r3, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    bne .LBB28_12
 ; CHECK-BE-NEXT:  @ %bb.1: @ %else
 ; CHECK-BE-NEXT:    lsls r2, r3, #30
 ; CHECK-BE-NEXT:    bmi .LBB28_13
 ; CHECK-BE-NEXT:  .LBB28_2: @ %else2
 ; CHECK-BE-NEXT:    lsls r2, r3, #29
 ; CHECK-BE-NEXT:    bmi .LBB28_14
 ; CHECK-BE-NEXT:  .LBB28_3: @ %else5

llvm/trunk/test/CodeGen/Thumb2/mve-masked-store.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve,+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-LE
 ; RUN: llc -mtriple=thumbebv8.1m.main-arm-none-eabi -mattr=+mve,+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-BE
 
 define arm_aapcs_vfpcc void @masked_v4i32(<4 x i32> *%dest, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    vmovne r2, s0
 ; CHECK-LE-NEXT:    strne r2, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi r2, s1
 ; CHECK-LE-NEXT:    strmi r2, [r0, #4]
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi r2, s2
 ; CHECK-LE-NEXT:    strmi r2, [r0, #8]
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi r1, s3
 ; CHECK-LE-NEXT:    strmi r1, [r0, #12]
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    vmovne r2, s4
 ; CHECK-BE-NEXT:    strne r2, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi r2, s5
 ; CHECK-BE-NEXT:    strmi r2, [r0, #4]
entry:
   ret void
 }
 
 define arm_aapcs_vfpcc void @masked_v4i32_align1(<4 x i32> *%dest, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32_align1:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    vmovne r2, s0
 ; CHECK-LE-NEXT:    strne r2, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi r2, s1
 ; CHECK-LE-NEXT:    strmi r2, [r0, #4]
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi r2, s2
 ; CHECK-LE-NEXT:    strmi r2, [r0, #8]
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi r1, s3
 ; CHECK-LE-NEXT:    strmi r1, [r0, #12]
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32_align1:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    vmovne r2, s4
 ; CHECK-BE-NEXT:    strne r2, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi r2, s5
 ; CHECK-BE-NEXT:    strmi r2, [r0, #4]

 define i8* @masked_v4i32_pre(i8* %y, i8* %x, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32_pre:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-LE-NEXT:    adds r0, #4
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    add r2, sp, #4
+; CHECK-LE-NEXT:    movs r2, #0
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrb.w r2, [sp, #4]
+; CHECK-LE-NEXT:    vmrs r12, p0
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-LE-NEXT:    lsls r1, r2, #31
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    and r1, r2, #15
+; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    itt ne
-; CHECK-LE-NEXT:    vmovne r1, s0
-; CHECK-LE-NEXT:    strne r1, [r0]
-; CHECK-LE-NEXT:    lsls r1, r2, #30
+; CHECK-LE-NEXT:    vmovne r2, s0
+; CHECK-LE-NEXT:    strne r2, [r0]
+; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi r1, s1
-; CHECK-LE-NEXT:    strmi r1, [r0, #4]
-; CHECK-LE-NEXT:    lsls r1, r2, #29
+; CHECK-LE-NEXT:    vmovmi r2, s1
+; CHECK-LE-NEXT:    strmi r2, [r0, #4]
+; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi r1, s2
-; CHECK-LE-NEXT:    strmi r1, [r0, #8]
-; CHECK-LE-NEXT:    lsls r1, r2, #28
+; CHECK-LE-NEXT:    vmovmi r2, s2
+; CHECK-LE-NEXT:    strmi r2, [r0, #8]
+; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi r1, s3
 ; CHECK-LE-NEXT:    strmi r1, [r0, #12]
 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32_pre:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-BE-NEXT:    adds r0, #4
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    add r2, sp, #4
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp, #4]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
 ; CHECK-BE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-BE-NEXT:    lsls r1, r2, #31
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    and r1, r2, #15
+; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    itt ne
-; CHECK-BE-NEXT:    vmovne r1, s0
-; CHECK-BE-NEXT:    strne r1, [r0]
-; CHECK-BE-NEXT:    lsls r1, r2, #30
+; CHECK-BE-NEXT:    vmovne r2, s0
+; CHECK-BE-NEXT:    strne r2, [r0]
+; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi r1, s1
-; CHECK-BE-NEXT:    strmi r1, [r0, #4]
-; CHECK-BE-NEXT:    lsls r1, r2, #29
+; CHECK-BE-NEXT:    vmovmi r2, s1
+; CHECK-BE-NEXT:    strmi r2, [r0, #4]
+; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi r1, s2
-; CHECK-BE-NEXT:    strmi r1, [r0, #8]
-; CHECK-BE-NEXT:    lsls r1, r2, #28
+; CHECK-BE-NEXT:    vmovmi r2, s2
+; CHECK-BE-NEXT:    strmi r2, [r0, #8]
+; CHECK-BE-NEXT:    lsls r1, r1, #28
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi r1, s3
 ; CHECK-BE-NEXT:    strmi r1, [r0, #12]
 ; CHECK-BE-NEXT:    add sp, #8
 ; CHECK-BE-NEXT:    bx lr
 entry:
   %z = getelementptr inbounds i8, i8* %y, i32 4
   %0 = bitcast i8* %x to <4 x i32>*
   %1 = load <4 x i32>, <4 x i32>* %0, align 4
   %2 = bitcast i8* %z to <4 x i32>*
   %c = icmp sgt <4 x i32> %a, zeroinitializer
   call void @llvm.masked.store.v4i32.p0v4i32(<4 x i32> %1, <4 x i32>* %2, i32 4, <4 x i1> %c)
   ret i8* %z
 }
 
 define i8* @masked_v4i32_post(i8* %y, i8* %x, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4i32_post:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    add r2, sp, #4
+; CHECK-LE-NEXT:    movs r2, #0
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrb.w r2, [sp, #4]
+; CHECK-LE-NEXT:    vmrs r12, p0
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    and r2, r2, #15
 ; CHECK-LE-NEXT:    lsls r1, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    vmovne r1, s0
 ; CHECK-LE-NEXT:    strne r1, [r0]
 ; CHECK-LE-NEXT:    lsls r1, r2, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi r1, s1
 ; CHECK-LE-NEXT:    strmi r1, [r0, #4]

 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4i32_post:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    add r2, sp, #4
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp, #4]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
 ; CHECK-BE-NEXT:    vldrw.u32 q0, [r1]
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    and r2, r2, #15
 ; CHECK-BE-NEXT:    lsls r1, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    vmovne r1, s0
 ; CHECK-BE-NEXT:    strne r1, [r0]
 ; CHECK-BE-NEXT:    lsls r1, r2, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi r1, s1
 ; CHECK-BE-NEXT:    strmi r1, [r0, #4]

 }
 
 
 define arm_aapcs_vfpcc void @masked_v8i16(<8 x i16> *%dest, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    vmovne.u16 r2, q0[0]
 ; CHECK-LE-NEXT:    strhne r2, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r2, q0[1]
 ; CHECK-LE-NEXT:    strhmi r2, [r0, #2]
 ; CHECK-LE-NEXT:    lsls r2, r1, #29

 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8i16:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    vmovne.u16 r2, q1[0]
 ; CHECK-BE-NEXT:    strhne r2, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r2, q1[1]
 ; CHECK-BE-NEXT:    strhmi r2, [r0, #2]
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
entry:
   ret void
 }
 
 define arm_aapcs_vfpcc void @masked_v8i16_align1(<8 x i16> *%dest, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16_align1:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    vmovne.u16 r2, q0[0]
 ; CHECK-LE-NEXT:    strhne r2, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r2, q0[1]
 ; CHECK-LE-NEXT:    strhmi r2, [r0, #2]
 ; CHECK-LE-NEXT:    lsls r2, r1, #29

 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8i16_align1:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    vmovne.u16 r2, q1[0]
 ; CHECK-BE-NEXT:    strhne r2, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r2, q1[1]
 ; CHECK-BE-NEXT:    strhmi r2, [r0, #2]
 ; CHECK-BE-NEXT:    lsls r2, r1, #29

 define i8* @masked_v8i16_pre(i8* %y, i8* %x, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16_pre:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-LE-NEXT:    adds r0, #4
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    mov r2, sp
+; CHECK-LE-NEXT:    movs r2, #0
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrb.w r2, [sp]
+; CHECK-LE-NEXT:    vmrs r12, p0
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-LE-NEXT:    lsls r1, r2, #31
+; CHECK-LE-NEXT:    bfi r2, r3, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
-; CHECK-LE-NEXT:    vmovne.u16 r1, q0[0]
-; CHECK-LE-NEXT:    strhne r1, [r0]
-; CHECK-LE-NEXT:    lsls r1, r2, #30
+; CHECK-LE-NEXT:    vmovne.u16 r2, q0[0]
+; CHECK-LE-NEXT:    strhne r2, [r0]
+; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[1]
-; CHECK-LE-NEXT:    strhmi r1, [r0, #2]
-; CHECK-LE-NEXT:    lsls r1, r2, #29
+; CHECK-LE-NEXT:    vmovmi.u16 r2, q0[1]
+; CHECK-LE-NEXT:    strhmi r2, [r0, #2]
+; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[2]
-; CHECK-LE-NEXT:    strhmi r1, [r0, #4]
-; CHECK-LE-NEXT:    lsls r1, r2, #28
+; CHECK-LE-NEXT:    vmovmi.u16 r2, q0[2]
+; CHECK-LE-NEXT:    strhmi r2, [r0, #4]
+; CHECK-LE-NEXT:    lsls r2, r1, #28
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[3]
-; CHECK-LE-NEXT:    strhmi r1, [r0, #6]
-; CHECK-LE-NEXT:    lsls r1, r2, #27
+; CHECK-LE-NEXT:    vmovmi.u16 r2, q0[3]
+; CHECK-LE-NEXT:    strhmi r2, [r0, #6]
+; CHECK-LE-NEXT:    lsls r2, r1, #27
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[4]
-; CHECK-LE-NEXT:    strhmi r1, [r0, #8]
-; CHECK-LE-NEXT:    lsls r1, r2, #26
+; CHECK-LE-NEXT:    vmovmi.u16 r2, q0[4]
+; CHECK-LE-NEXT:    strhmi r2, [r0, #8]
+; CHECK-LE-NEXT:    lsls r2, r1, #26
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[5]
-; CHECK-LE-NEXT:    strhmi r1, [r0, #10]
-; CHECK-LE-NEXT:    lsls r1, r2, #25
+; CHECK-LE-NEXT:    vmovmi.u16 r2, q0[5]
+; CHECK-LE-NEXT:    strhmi r2, [r0, #10]
+; CHECK-LE-NEXT:    lsls r2, r1, #25
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[6]
-; CHECK-LE-NEXT:    strhmi r1, [r0, #12]
-; CHECK-LE-NEXT:    lsls r1, r2, #24
+; CHECK-LE-NEXT:    vmovmi.u16 r2, q0[6]
+; CHECK-LE-NEXT:    strhmi r2, [r0, #12]
+; CHECK-LE-NEXT:    lsls r1, r1, #24
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[7]
 ; CHECK-LE-NEXT:    strhmi r1, [r0, #14]
 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8i16_pre:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-BE-NEXT:    adds r0, #4
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
 ; CHECK-BE-NEXT:    vldrh.u16 q0, [r1]
-; CHECK-BE-NEXT:    lsls r1, r2, #31
+; CHECK-BE-NEXT:    bfi r2, r3, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
-; CHECK-BE-NEXT:    vmovne.u16 r1, q0[0]
-; CHECK-BE-NEXT:    strhne r1, [r0]
-; CHECK-BE-NEXT:    lsls r1, r2, #30
+; CHECK-BE-NEXT:    vmovne.u16 r2, q0[0]
+; CHECK-BE-NEXT:    strhne r2, [r0]
+; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[1]
-; CHECK-BE-NEXT:    strhmi r1, [r0, #2]
-; CHECK-BE-NEXT:    lsls r1, r2, #29
+; CHECK-BE-NEXT:    vmovmi.u16 r2, q0[1]
+; CHECK-BE-NEXT:    strhmi r2, [r0, #2]
+; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[2]
-; CHECK-BE-NEXT:    strhmi r1, [r0, #4]
-; CHECK-BE-NEXT:    lsls r1, r2, #28
+; CHECK-BE-NEXT:    vmovmi.u16 r2, q0[2]
+; CHECK-BE-NEXT:    strhmi r2, [r0, #4]
+; CHECK-BE-NEXT:    lsls r2, r1, #28
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[3]
-; CHECK-BE-NEXT:    strhmi r1, [r0, #6]
-; CHECK-BE-NEXT:    lsls r1, r2, #27
+; CHECK-BE-NEXT:    vmovmi.u16 r2, q0[3]
+; CHECK-BE-NEXT:    strhmi r2, [r0, #6]
+; CHECK-BE-NEXT:    lsls r2, r1, #27
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[4]
-; CHECK-BE-NEXT:    strhmi r1, [r0, #8]
-; CHECK-BE-NEXT:    lsls r1, r2, #26
+; CHECK-BE-NEXT:    vmovmi.u16 r2, q0[4]
+; CHECK-BE-NEXT:    strhmi r2, [r0, #8]
+; CHECK-BE-NEXT:    lsls r2, r1, #26
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[5]
-; CHECK-BE-NEXT:    strhmi r1, [r0, #10]
-; CHECK-BE-NEXT:    lsls r1, r2, #25
+; CHECK-BE-NEXT:    vmovmi.u16 r2, q0[5]
+; CHECK-BE-NEXT:    strhmi r2, [r0, #10]
+; CHECK-BE-NEXT:    lsls r2, r1, #25
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[6]
-; CHECK-BE-NEXT:    strhmi r1, [r0, #12]
-; CHECK-BE-NEXT:    lsls r1, r2, #24
+; CHECK-BE-NEXT:    vmovmi.u16 r2, q0[6]
+; CHECK-BE-NEXT:    strhmi r2, [r0, #12]
+; CHECK-BE-NEXT:    lsls r1, r1, #24
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[7]
 ; CHECK-BE-NEXT:    strhmi r1, [r0, #14]
 ; CHECK-BE-NEXT:    add sp, #8
 ; CHECK-BE-NEXT:    bx lr
 entry:
   %z = getelementptr inbounds i8, i8* %y, i32 4
   %0 = bitcast i8* %x to <8 x i16>*
   %1 = load <8 x i16>, <8 x i16>* %0, align 4
   %2 = bitcast i8* %z to <8 x i16>*
   %c = icmp sgt <8 x i16> %a, zeroinitializer
   call void @llvm.masked.store.v8i16.p0v8i16(<8 x i16> %1, <8 x i16>* %2, i32 2, <8 x i1> %c)
   ret i8* %z
 }
 
 define i8* @masked_v8i16_post(i8* %y, i8* %x, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8i16_post:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    mov r2, sp
+; CHECK-LE-NEXT:    movs r3, #0
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrb.w r2, [sp]
+; CHECK-LE-NEXT:    vmrs r12, p0
+; CHECK-LE-NEXT:    and r2, r12, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #6, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #10, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #14, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-LE-NEXT:    lsls r1, r2, #31
+; CHECK-LE-NEXT:    bfi r3, r2, #7, #1
+; CHECK-LE-NEXT:    lsls r1, r3, #31
+; CHECK-LE-NEXT:    uxtb r2, r3
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    vmovne.u16 r1, q0[0]
 ; CHECK-LE-NEXT:    strhne r1, [r0]
 ; CHECK-LE-NEXT:    lsls r1, r2, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[1]
 ; CHECK-LE-NEXT:    strhmi r1, [r0, #2]
 ; CHECK-LE-NEXT:    lsls r1, r2, #29
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[2]
 ; CHECK-LE-NEXT:    strhmi r1, [r0, #4]
 ; CHECK-LE-NEXT:    lsls r1, r2, #28
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[3]
 ; CHECK-LE-NEXT:    strhmi r1, [r0, #6]
 ; CHECK-LE-NEXT:    lsls r1, r2, #27
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[4]
 ; CHECK-LE-NEXT:    strhmi r1, [r0, #8]
 ; CHECK-LE-NEXT:    lsls r1, r2, #26
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r1, q0[5]
 ; CHECK-LE-NEXT:    strhmi r1, [r0, #10]
-; CHECK-LE-NEXT:    adds r1, r0, #4
 ; CHECK-LE-NEXT:    lsls r3, r2, #25
+; CHECK-LE-NEXT:    add.w r1, r0, #4
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r3, q0[6]
 ; CHECK-LE-NEXT:    strhmi r3, [r0, #12]
 ; CHECK-LE-NEXT:    lsls r2, r2, #24
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u16 r2, q0[7]
 ; CHECK-LE-NEXT:    strhmi r2, [r0, #14]
 ; CHECK-LE-NEXT:    mov r0, r1
 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8i16_post:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r3, #0
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r2, r12, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #6, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #10, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #14, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
 ; CHECK-BE-NEXT:    vldrh.u16 q0, [r1]
-; CHECK-BE-NEXT:    lsls r1, r2, #31
+; CHECK-BE-NEXT:    bfi r3, r2, #7, #1
+; CHECK-BE-NEXT:    lsls r1, r3, #31
+; CHECK-BE-NEXT:    uxtb r2, r3
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    vmovne.u16 r1, q0[0]
 ; CHECK-BE-NEXT:    strhne r1, [r0]
 ; CHECK-BE-NEXT:    lsls r1, r2, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[1]
 ; CHECK-BE-NEXT:    strhmi r1, [r0, #2]
 ; CHECK-BE-NEXT:    lsls r1, r2, #29
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[2]
 ; CHECK-BE-NEXT:    strhmi r1, [r0, #4]
 ; CHECK-BE-NEXT:    lsls r1, r2, #28
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[3]
 ; CHECK-BE-NEXT:    strhmi r1, [r0, #6]
 ; CHECK-BE-NEXT:    lsls r1, r2, #27
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[4]
 ; CHECK-BE-NEXT:    strhmi r1, [r0, #8]
 ; CHECK-BE-NEXT:    lsls r1, r2, #26
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r1, q0[5]
 ; CHECK-BE-NEXT:    strhmi r1, [r0, #10]
-; CHECK-BE-NEXT:    adds r1, r0, #4
 ; CHECK-BE-NEXT:    lsls r3, r2, #25
+; CHECK-BE-NEXT:    add.w r1, r0, #4
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r3, q0[6]
 ; CHECK-BE-NEXT:    strhmi r3, [r0, #12]
 ; CHECK-BE-NEXT:    lsls r2, r2, #24
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u16 r2, q0[7]
 ; CHECK-BE-NEXT:    strhmi r2, [r0, #14]
 ; CHECK-BE-NEXT:    mov r0, r1

 ; CHECK-LE-NEXT:    push {r4, r6, r7, lr}
 ; CHECK-LE-NEXT:    .setfp r7, sp, #8
 ; CHECK-LE-NEXT:    add r7, sp, #8
 ; CHECK-LE-NEXT:    .pad #16
 ; CHECK-LE-NEXT:    sub sp, #16
 ; CHECK-LE-NEXT:    mov r4, sp
 ; CHECK-LE-NEXT:    bfc r4, #0, #4
 ; CHECK-LE-NEXT:    mov sp, r4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.s8 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
 ; CHECK-LE-NEXT:    sub.w r4, r7, #8
-; CHECK-LE-NEXT:    ldrh.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    uxth r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    vmovne.u8 r2, q0[0]
 ; CHECK-LE-NEXT:    strbne r2, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[1]
 ; CHECK-LE-NEXT:    strbmi r2, [r0, #1]
 ; CHECK-LE-NEXT:    lsls r2, r1, #29

 ; CHECK-BE-NEXT:    .setfp r7, sp, #8
 ; CHECK-BE-NEXT:    add r7, sp, #8
 ; CHECK-BE-NEXT:    .pad #16
 ; CHECK-BE-NEXT:    sub sp, #16
 ; CHECK-BE-NEXT:    mov r4, sp
 ; CHECK-BE-NEXT:    bfc r4, #0, #4
 ; CHECK-BE-NEXT:    mov sp, r4
 ; CHECK-BE-NEXT:    vrev64.8 q1, q0
-; CHECK-BE-NEXT:    mov r1, sp
-; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
 ; CHECK-BE-NEXT:    sub.w r4, r7, #8
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrh.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    uxth r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    vmovne.u8 r2, q1[0]
 ; CHECK-BE-NEXT:    strbne r2, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u8 r2, q1[1]
 ; CHECK-BE-NEXT:    strbmi r2, [r0, #1]
 ; CHECK-BE-NEXT:    lsls r2, r1, #29

 ; CHECK-LE-NEXT:    .pad #16
 ; CHECK-LE-NEXT:    sub sp, #16
 ; CHECK-LE-NEXT:    mov r4, sp
 ; CHECK-LE-NEXT:    bfc r4, #0, #4
 ; CHECK-LE-NEXT:    mov sp, r4
 ; CHECK-LE-NEXT:    vldr d1, [r7, #8]
 ; CHECK-LE-NEXT:    adds r0, #4
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    mov r2, sp
-; CHECK-LE-NEXT:    vcmp.s8 gt, q0, zr
 ; CHECK-LE-NEXT:    sub.w r4, r7, #8
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrh.w r2, [sp]
+; CHECK-LE-NEXT:    vcmp.s8 gt, q0, zr
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-LE-NEXT:    lsls r1, r2, #31
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    uxth r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    itt ne
-; CHECK-LE-NEXT:    vmovne.u8 r1, q0[0]
-; CHECK-LE-NEXT:    strbne r1, [r0]
-; CHECK-LE-NEXT:    lsls r1, r2, #30
+; CHECK-LE-NEXT:    vmovne.u8 r2, q0[0]
+; CHECK-LE-NEXT:    strbne r2, [r0]
+; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[1]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #1]
-; CHECK-LE-NEXT:    lsls r1, r2, #29
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[1]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #1]
+; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[2]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #2]
-; CHECK-LE-NEXT:    lsls r1, r2, #28
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[2]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #2]
+; CHECK-LE-NEXT:    lsls r2, r1, #28
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[3]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #3]
-; CHECK-LE-NEXT:    lsls r1, r2, #27
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[3]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #3]
+; CHECK-LE-NEXT:    lsls r2, r1, #27
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[4]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #4]
-; CHECK-LE-NEXT:    lsls r1, r2, #26
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[4]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #4]
+; CHECK-LE-NEXT:    lsls r2, r1, #26
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[5]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #5]
-; CHECK-LE-NEXT:    lsls r1, r2, #25
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[5]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #5]
+; CHECK-LE-NEXT:    lsls r2, r1, #25
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[6]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #6]
-; CHECK-LE-NEXT:    lsls r1, r2, #24
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[6]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #6]
+; CHECK-LE-NEXT:    lsls r2, r1, #24
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[7]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #7]
-; CHECK-LE-NEXT:    lsls r1, r2, #23
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[7]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #7]
+; CHECK-LE-NEXT:    lsls r2, r1, #23
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[8]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #8]
-; CHECK-LE-NEXT:    lsls r1, r2, #22
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[8]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #8]
+; CHECK-LE-NEXT:    lsls r2, r1, #22
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[9]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #9]
-; CHECK-LE-NEXT:    lsls r1, r2, #21
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[9]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #9]
+; CHECK-LE-NEXT:    lsls r2, r1, #21
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[10]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #10]
-; CHECK-LE-NEXT:    lsls r1, r2, #20
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[10]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #10]
+; CHECK-LE-NEXT:    lsls r2, r1, #20
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[11]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #11]
-; CHECK-LE-NEXT:    lsls r1, r2, #19
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[11]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #11]
+; CHECK-LE-NEXT:    lsls r2, r1, #19
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[12]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #12]
-; CHECK-LE-NEXT:    lsls r1, r2, #18
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[12]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #12]
+; CHECK-LE-NEXT:    lsls r2, r1, #18
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[13]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #13]
-; CHECK-LE-NEXT:    lsls r1, r2, #17
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[13]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #13]
+; CHECK-LE-NEXT:    lsls r2, r1, #17
 ; CHECK-LE-NEXT:    itt mi
-; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[14]
-; CHECK-LE-NEXT:    strbmi r1, [r0, #14]
-; CHECK-LE-NEXT:    lsls r1, r2, #16
+; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[14]
+; CHECK-LE-NEXT:    strbmi r2, [r0, #14]
+; CHECK-LE-NEXT:    lsls r1, r1, #16
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[15]
 ; CHECK-LE-NEXT:    strbmi r1, [r0, #15]
 ; CHECK-LE-NEXT:    mov sp, r4
 ; CHECK-LE-NEXT:    pop {r4, r6, r7, pc}
 ;
 ; CHECK-BE-LABEL: masked_v16i8_pre:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .save {r4, r6, r7, lr}
 ; CHECK-BE-NEXT:    push {r4, r6, r7, lr}
 ; CHECK-BE-NEXT:    .setfp r7, sp, #8
 ; CHECK-BE-NEXT:    add r7, sp, #8
 ; CHECK-BE-NEXT:    .pad #16
 ; CHECK-BE-NEXT:    sub sp, #16
 ; CHECK-BE-NEXT:    mov r4, sp
 ; CHECK-BE-NEXT:    bfc r4, #0, #4
 ; CHECK-BE-NEXT:    mov sp, r4
 ; CHECK-BE-NEXT:    vldr d1, [r7, #8]
 ; CHECK-BE-NEXT:    adds r0, #4
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    mov r2, sp
-; CHECK-BE-NEXT:    vrev64.8 q1, q0
 ; CHECK-BE-NEXT:    sub.w r4, r7, #8
-; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrh.w r2, [sp]
+; CHECK-BE-NEXT:    vrev64.8 q1, q0
 ; CHECK-BE-NEXT:    vldrb.u8 q0, [r1]
-; CHECK-BE-NEXT:    lsls r1, r2, #31
+; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    uxth r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    itt ne
-; CHECK-BE-NEXT:    vmovne.u8 r1, q0[0]
-; CHECK-BE-NEXT:    strbne r1, [r0]
-; CHECK-BE-NEXT:    lsls r1, r2, #30
+; CHECK-BE-NEXT:    vmovne.u8 r2, q0[0]
+; CHECK-BE-NEXT:    strbne r2, [r0]
+; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[1]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #1]
-; CHECK-BE-NEXT:    lsls r1, r2, #29
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[1]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #1]
+; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[2]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #2]
-; CHECK-BE-NEXT:    lsls r1, r2, #28
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[2]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #2]
+; CHECK-BE-NEXT:    lsls r2, r1, #28
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[3]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #3]
-; CHECK-BE-NEXT:    lsls r1, r2, #27
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[3]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #3]
+; CHECK-BE-NEXT:    lsls r2, r1, #27
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[4]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #4]
-; CHECK-BE-NEXT:    lsls r1, r2, #26
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[4]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #4]
+; CHECK-BE-NEXT:    lsls r2, r1, #26
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[5]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #5]
-; CHECK-BE-NEXT:    lsls r1, r2, #25
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[5]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #5]
+; CHECK-BE-NEXT:    lsls r2, r1, #25
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[6]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #6]
-; CHECK-BE-NEXT:    lsls r1, r2, #24
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[6]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #6]
+; CHECK-BE-NEXT:    lsls r2, r1, #24
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[7]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #7]
-; CHECK-BE-NEXT:    lsls r1, r2, #23
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[7]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #7]
+; CHECK-BE-NEXT:    lsls r2, r1, #23
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[8]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #8]
-; CHECK-BE-NEXT:    lsls r1, r2, #22
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[8]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #8]
+; CHECK-BE-NEXT:    lsls r2, r1, #22
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[9]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #9]
-; CHECK-BE-NEXT:    lsls r1, r2, #21
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[9]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #9]
+; CHECK-BE-NEXT:    lsls r2, r1, #21
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[10]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #10]
-; CHECK-BE-NEXT:    lsls r1, r2, #20
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[10]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #10]
+; CHECK-BE-NEXT:    lsls r2, r1, #20
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[11]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #11]
-; CHECK-BE-NEXT:    lsls r1, r2, #19
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[11]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #11]
+; CHECK-BE-NEXT:    lsls r2, r1, #19
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[12]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #12]
-; CHECK-BE-NEXT:    lsls r1, r2, #18
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[12]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #12]
+; CHECK-BE-NEXT:    lsls r2, r1, #18
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[13]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #13]
-; CHECK-BE-NEXT:    lsls r1, r2, #17
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[13]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #13]
+; CHECK-BE-NEXT:    lsls r2, r1, #17
 ; CHECK-BE-NEXT:    itt mi
-; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[14]
-; CHECK-BE-NEXT:    strbmi r1, [r0, #14]
-; CHECK-BE-NEXT:    lsls r1, r2, #16
+; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[14]
+; CHECK-BE-NEXT:    strbmi r2, [r0, #14]
+; CHECK-BE-NEXT:    lsls r1, r1, #16
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[15]
 ; CHECK-BE-NEXT:    strbmi r1, [r0, #15]
 ; CHECK-BE-NEXT:    mov sp, r4
 ; CHECK-BE-NEXT:    pop {r4, r6, r7, pc}
 entry:
   %z = getelementptr inbounds i8, i8* %y, i32 4
   %0 = bitcast i8* %x to <16 x i8>*

 ; CHECK-LE-NEXT:    .pad #16
 ; CHECK-LE-NEXT:    sub sp, #16
 ; CHECK-LE-NEXT:    mov r4, sp
 ; CHECK-LE-NEXT:    bfc r4, #0, #4
 ; CHECK-LE-NEXT:    mov sp, r4
 ; CHECK-LE-NEXT:    vldr d1, [r7, #8]
 ; CHECK-LE-NEXT:    sub.w r4, r7, #8
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    mov r2, sp
 ; CHECK-LE-NEXT:    vcmp.s8 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrh.w r2, [sp]
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-LE-NEXT:    lsls r1, r2, #31
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    uxth r2, r1
+; CHECK-LE-NEXT:    lsls r1, r1, #31
 ; CHECK-LE-NEXT:    itt ne
 ; CHECK-LE-NEXT:    vmovne.u8 r1, q0[0]
 ; CHECK-LE-NEXT:    strbne r1, [r0]
 ; CHECK-LE-NEXT:    lsls r1, r2, #30
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[1]
 ; CHECK-LE-NEXT:    strbmi r1, [r0, #1]
 ; CHECK-LE-NEXT:    lsls r1, r2, #29

 ; CHECK-LE-NEXT:    lsls r1, r2, #19
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[12]
 ; CHECK-LE-NEXT:    strbmi r1, [r0, #12]
 ; CHECK-LE-NEXT:    lsls r1, r2, #18
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u8 r1, q0[13]
 ; CHECK-LE-NEXT:    strbmi r1, [r0, #13]
-; CHECK-LE-NEXT:    adds r1, r0, #4
 ; CHECK-LE-NEXT:    lsls r3, r2, #17
+; CHECK-LE-NEXT:    add.w r1, r0, #4
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u8 r3, q0[14]
 ; CHECK-LE-NEXT:    strbmi r3, [r0, #14]
 ; CHECK-LE-NEXT:    lsls r2, r2, #16
 ; CHECK-LE-NEXT:    itt mi
 ; CHECK-LE-NEXT:    vmovmi.u8 r2, q0[15]
 ; CHECK-LE-NEXT:    strbmi r2, [r0, #15]
 ; CHECK-LE-NEXT:    mov r0, r1

 ; CHECK-BE-NEXT:    .pad #16
 ; CHECK-BE-NEXT:    sub sp, #16
 ; CHECK-BE-NEXT:    mov r4, sp
 ; CHECK-BE-NEXT:    bfc r4, #0, #4
 ; CHECK-BE-NEXT:    mov sp, r4
 ; CHECK-BE-NEXT:    vldr d1, [r7, #8]
 ; CHECK-BE-NEXT:    sub.w r4, r7, #8
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    mov r2, sp
 ; CHECK-BE-NEXT:    vrev64.8 q1, q0
-; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrh.w r2, [sp]
 ; CHECK-BE-NEXT:    vldrb.u8 q0, [r1]
-; CHECK-BE-NEXT:    lsls r1, r2, #31
+; CHECK-BE-NEXT:    vcmp.s8 gt, q1, zr
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    uxth r2, r1
+; CHECK-BE-NEXT:    lsls r1, r1, #31
 ; CHECK-BE-NEXT:    itt ne
 ; CHECK-BE-NEXT:    vmovne.u8 r1, q0[0]
 ; CHECK-BE-NEXT:    strbne r1, [r0]
 ; CHECK-BE-NEXT:    lsls r1, r2, #30
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[1]
 ; CHECK-BE-NEXT:    strbmi r1, [r0, #1]
 ; CHECK-BE-NEXT:    lsls r1, r2, #29

 ; CHECK-BE-NEXT:    lsls r1, r2, #19
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[12]
 ; CHECK-BE-NEXT:    strbmi r1, [r0, #12]
 ; CHECK-BE-NEXT:    lsls r1, r2, #18
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u8 r1, q0[13]
 ; CHECK-BE-NEXT:    strbmi r1, [r0, #13]
-; CHECK-BE-NEXT:    adds r1, r0, #4
 ; CHECK-BE-NEXT:    lsls r3, r2, #17
+; CHECK-BE-NEXT:    add.w r1, r0, #4
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u8 r3, q0[14]
 ; CHECK-BE-NEXT:    strbmi r3, [r0, #14]
 ; CHECK-BE-NEXT:    lsls r2, r2, #16
 ; CHECK-BE-NEXT:    itt mi
 ; CHECK-BE-NEXT:    vmovmi.u8 r2, q0[15]
 ; CHECK-BE-NEXT:    strbmi r2, [r0, #15]
 ; CHECK-BE-NEXT:    mov r0, r1

 }
 
 
 define arm_aapcs_vfpcc void @masked_v4f32(<4 x float> *%dest, <4 x float> %a, <4 x i32> %b) {
 ; CHECK-LE-LABEL: masked_v4f32:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #4
 ; CHECK-LE-NEXT:    sub sp, #4
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.i32 ne, q1, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    it ne
 ; CHECK-LE-NEXT:    vstrne s0, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vstrmi s1, [r0, #4]
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vstrmi s2, [r0, #8]
 ; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vstrmi s3, [r0, #12]
 ; CHECK-LE-NEXT:    add sp, #4
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4f32:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #4
 ; CHECK-BE-NEXT:    sub sp, #4
 ; CHECK-BE-NEXT:    vrev64.32 q2, q1
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.i32 ne, q2, zr
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    it ne
 ; CHECK-BE-NEXT:    vstrne s4, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    it mi
 ; CHECK-BE-NEXT:    vstrmi s5, [r0, #4]
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    it mi
entry:
   ret void
 }
 
 define arm_aapcs_vfpcc void @masked_v4f32_align1(<4 x float> *%dest, <4 x float> %a, <4 x i32> %b) {
 ; CHECK-LE-LABEL: masked_v4f32_align1:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #20
 ; CHECK-LE-NEXT:    sub sp, #20
-; CHECK-LE-NEXT:    add r1, sp, #16
 ; CHECK-LE-NEXT:    vcmp.i32 ne, q1, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp, #16]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
 ; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    ittt ne
 ; CHECK-LE-NEXT:    vstrne s0, [sp, #12]
 ; CHECK-LE-NEXT:    ldrne r2, [sp, #12]
 ; CHECK-LE-NEXT:    strne r2, [r0]
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    ittt mi
 ; CHECK-LE-NEXT:    vstrmi s1, [sp, #8]

 ; CHECK-LE-NEXT:    add sp, #20
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4f32_align1:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #20
 ; CHECK-BE-NEXT:    sub sp, #20
 ; CHECK-BE-NEXT:    vrev64.32 q2, q1
-; CHECK-BE-NEXT:    add r1, sp, #16
+; CHECK-BE-NEXT:    movs r1, #0
 ; CHECK-BE-NEXT:    vcmp.i32 ne, q2, zr
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp, #16]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    and r3, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-BE-NEXT:    and r1, r1, #15
 ; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    ittt ne
 ; CHECK-BE-NEXT:    vstrne s4, [sp, #12]
 ; CHECK-BE-NEXT:    ldrne r2, [sp, #12]
 ; CHECK-BE-NEXT:    strne r2, [r0]
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    ittt mi
 ; CHECK-BE-NEXT:    vstrmi s5, [sp, #8]

 define i8* @masked_v4f32_pre(i8* %y, i8* %x, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4f32_pre:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-LE-NEXT:    adds r0, #4
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    add r2, sp, #4
+; CHECK-LE-NEXT:    movs r2, #0
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrb.w r2, [sp, #4]
+; CHECK-LE-NEXT:    vmrs r12, p0
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-LE-NEXT:    lsls r1, r2, #31
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    and r1, r2, #15
+; CHECK-LE-NEXT:    lsls r2, r1, #31
 ; CHECK-LE-NEXT:    it ne
 ; CHECK-LE-NEXT:    vstrne s0, [r0]
-; CHECK-LE-NEXT:    lsls r1, r2, #30
+; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vstrmi s1, [r0, #4]
-; CHECK-LE-NEXT:    lsls r1, r2, #29
+; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vstrmi s2, [r0, #8]
-; CHECK-LE-NEXT:    lsls r1, r2, #28
+; CHECK-LE-NEXT:    lsls r1, r1, #28
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vstrmi s3, [r0, #12]
 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4f32_pre:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-BE-NEXT:    adds r0, #4
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    add r2, sp, #4
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp, #4]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
 ; CHECK-BE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-BE-NEXT:    lsls r1, r2, #31
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    and r1, r2, #15
+; CHECK-BE-NEXT:    lsls r2, r1, #31
 ; CHECK-BE-NEXT:    it ne
 ; CHECK-BE-NEXT:    vstrne s0, [r0]
-; CHECK-BE-NEXT:    lsls r1, r2, #30
+; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    it mi
 ; CHECK-BE-NEXT:    vstrmi s1, [r0, #4]
-; CHECK-BE-NEXT:    lsls r1, r2, #29
+; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    it mi
 ; CHECK-BE-NEXT:    vstrmi s2, [r0, #8]
-; CHECK-BE-NEXT:    lsls r1, r2, #28
+; CHECK-BE-NEXT:    lsls r1, r1, #28
 ; CHECK-BE-NEXT:    it mi
 ; CHECK-BE-NEXT:    vstrmi s3, [r0, #12]
 ; CHECK-BE-NEXT:    add sp, #8
 ; CHECK-BE-NEXT:    bx lr
 entry:
   %z = getelementptr inbounds i8, i8* %y, i32 4
   %0 = bitcast i8* %x to <4 x float>*
   %1 = load <4 x float>, <4 x float>* %0, align 4
   %2 = bitcast i8* %z to <4 x float>*
   %c = icmp sgt <4 x i32> %a, zeroinitializer
   call void @llvm.masked.store.v4f32.p0v4f32(<4 x float> %1, <4 x float>* %2, i32 4, <4 x i1> %c)
   ret i8* %z
 }
 
 define i8* @masked_v4f32_post(i8* %y, i8* %x, <4 x i32> %a) {
 ; CHECK-LE-LABEL: masked_v4f32_post:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    add r2, sp, #4
+; CHECK-LE-NEXT:    movs r2, #0
 ; CHECK-LE-NEXT:    vcmp.s32 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrb.w r2, [sp, #4]
+; CHECK-LE-NEXT:    vmrs r12, p0
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    and r2, r2, #15
 ; CHECK-LE-NEXT:    lsls r1, r2, #31
 ; CHECK-LE-NEXT:    it ne
 ; CHECK-LE-NEXT:    vstrne s0, [r0]
 ; CHECK-LE-NEXT:    lsls r1, r2, #30
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vstrmi s1, [r0, #4]
 ; CHECK-LE-NEXT:    adds r1, r0, #4
 ; CHECK-LE-NEXT:    lsls r3, r2, #29
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vstrmi s2, [r0, #8]
 ; CHECK-LE-NEXT:    lsls r2, r2, #28
 ; CHECK-LE-NEXT:    it mi
 ; CHECK-LE-NEXT:    vstrmi s3, [r0, #12]
 ; CHECK-LE-NEXT:    mov r0, r1
 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v4f32_post:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    add r2, sp, #4
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
 ; CHECK-BE-NEXT:    vcmp.s32 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp, #4]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
 ; CHECK-BE-NEXT:    vldrw.u32 q0, [r1]
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    and r2, r2, #15
 ; CHECK-BE-NEXT:    lsls r1, r2, #31
 ; CHECK-BE-NEXT:    it ne
 ; CHECK-BE-NEXT:    vstrne s0, [r0]
 ; CHECK-BE-NEXT:    lsls r1, r2, #30
 ; CHECK-BE-NEXT:    it mi
 ; CHECK-BE-NEXT:    vstrmi s1, [r0, #4]
 ; CHECK-BE-NEXT:    adds r1, r0, #4
 ; CHECK-BE-NEXT:    lsls r3, r2, #29

 }
 
 
 define arm_aapcs_vfpcc void @masked_v8f16(<8 x half> *%dest, <8 x half> %a, <8 x i16> %b) {
 ; CHECK-LE-LABEL: masked_v8f16:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
-; CHECK-LE-NEXT:    mov r1, sp
 ; CHECK-LE-NEXT:    vcmp.i16 ne, q1, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    bne .LBB15_9
 ; CHECK-LE-NEXT:  @ %bb.1: @ %else
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    bmi .LBB15_10
 ; CHECK-LE-NEXT:  .LBB15_2: @ %else2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    bmi .LBB15_11
 ; CHECK-LE-NEXT:  .LBB15_3: @ %else4

 ; CHECK-LE-NEXT:    add sp, #8
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8f16:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vrev64.16 q2, q1
-; CHECK-BE-NEXT:    mov r1, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.i16 ne, q2, zr
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    bne .LBB15_9
 ; CHECK-BE-NEXT:  @ %bb.1: @ %else
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    bmi .LBB15_10
 ; CHECK-BE-NEXT:  .LBB15_2: @ %else2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    bmi .LBB15_11
 ; CHECK-BE-NEXT:  .LBB15_3: @ %else4
entry:
   ret void
 }
 
 define arm_aapcs_vfpcc void @masked_v8f16_align1(<8 x half> *%dest, <8 x half> %a, <8 x i16> %b) {
 ; CHECK-LE-LABEL: masked_v8f16_align1:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #40
 ; CHECK-LE-NEXT:    sub sp, #40
-; CHECK-LE-NEXT:    add r1, sp, #32
 ; CHECK-LE-NEXT:    vcmp.i16 ne, q1, zr
-; CHECK-LE-NEXT:    vstr p0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp, #32]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    movs r2, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r3, r1, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    bne .LBB16_9
 ; CHECK-LE-NEXT:  @ %bb.1: @ %else
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    bmi .LBB16_10
 ; CHECK-LE-NEXT:  .LBB16_2: @ %else2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    bmi .LBB16_11
 ; CHECK-LE-NEXT:  .LBB16_3: @ %else4

 ; CHECK-LE-NEXT:    add sp, #40
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: masked_v8f16_align1:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #40
 ; CHECK-BE-NEXT:    sub sp, #40
 ; CHECK-BE-NEXT:    vrev64.16 q2, q1
-; CHECK-BE-NEXT:    add r1, sp, #32
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vcmp.i16 ne, q2, zr
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
-; CHECK-BE-NEXT:    vstr p0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp, #32]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    bne .LBB16_9
 ; CHECK-BE-NEXT:  @ %bb.1: @ %else
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    bmi .LBB16_10
 ; CHECK-BE-NEXT:  .LBB16_2: @ %else2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    bmi .LBB16_11
 ; CHECK-BE-NEXT:  .LBB16_3: @ %else4

 define i8* @masked_v8f16_pre(i8* %y, i8* %x, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8f16_pre:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-LE-NEXT:    adds r0, #4
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    mov r2, sp
+; CHECK-LE-NEXT:    movs r2, #0
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r2]
+; CHECK-LE-NEXT:    vmrs r12, p0
+; CHECK-LE-NEXT:    and r3, r12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-LE-NEXT:    ubfx r3, r12, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-LE-NEXT:    ldrb.w r1, [sp]
-; CHECK-LE-NEXT:    lsls r2, r1, #31
+; CHECK-LE-NEXT:    bfi r2, r3, #7, #1
+; CHECK-LE-NEXT:    uxtb r1, r2
+; CHECK-LE-NEXT:    lsls r2, r2, #31
 ; CHECK-LE-NEXT:    bne .LBB17_9
 ; CHECK-LE-NEXT:  @ %bb.1: @ %else
 ; CHECK-LE-NEXT:    lsls r2, r1, #30
 ; CHECK-LE-NEXT:    bmi .LBB17_10
 ; CHECK-LE-NEXT:  .LBB17_2: @ %else2
 ; CHECK-LE-NEXT:    lsls r2, r1, #29
 ; CHECK-LE-NEXT:    bmi .LBB17_11
 ; CHECK-LE-NEXT:  .LBB17_3: @ %else4

 ;
 ; CHECK-BE-LABEL: masked_v8f16_pre:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-BE-NEXT:    adds r0, #4
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r2, #0
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r3, r12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #2, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #1, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #6, #1
+; CHECK-BE-NEXT:    ubfx r3, r12, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
 ; CHECK-BE-NEXT:    vldrh.u16 q0, [r1]
-; CHECK-BE-NEXT:    ldrb.w r1, [sp]
-; CHECK-BE-NEXT:    lsls r2, r1, #31
+; CHECK-BE-NEXT:    bfi r2, r3, #7, #1
+; CHECK-BE-NEXT:    uxtb r1, r2
+; CHECK-BE-NEXT:    lsls r2, r2, #31
 ; CHECK-BE-NEXT:    bne .LBB17_9
 ; CHECK-BE-NEXT:  @ %bb.1: @ %else
 ; CHECK-BE-NEXT:    lsls r2, r1, #30
 ; CHECK-BE-NEXT:    bmi .LBB17_10
 ; CHECK-BE-NEXT:  .LBB17_2: @ %else2
 ; CHECK-BE-NEXT:    lsls r2, r1, #29
 ; CHECK-BE-NEXT:    bmi .LBB17_11
 ; CHECK-BE-NEXT:  .LBB17_3: @ %else4

 
 define i8* @masked_v8f16_post(i8* %y, i8* %x, <8 x i16> %a) {
 ; CHECK-LE-LABEL: masked_v8f16_post:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    .pad #8
 ; CHECK-LE-NEXT:    sub sp, #8
 ; CHECK-LE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-LE-NEXT:    vmov d0, r2, r3
-; CHECK-LE-NEXT:    mov r2, sp
+; CHECK-LE-NEXT:    movs r3, #0
 ; CHECK-LE-NEXT:    vcmp.s16 gt, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r2]
-; CHECK-LE-NEXT:    ldrb.w r2, [sp]
+; CHECK-LE-NEXT:    vmrs r12, p0
+; CHECK-LE-NEXT:    and r2, r12, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #4, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #6, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #8, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #10, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #12, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-LE-NEXT:    ubfx r2, r12, #14, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
 ; CHECK-LE-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-LE-NEXT:    lsls r1, r2, #31
+; CHECK-LE-NEXT:    bfi r3, r2, #7, #1
+; CHECK-LE-NEXT:    uxtb r2, r3
+; CHECK-LE-NEXT:    lsls r1, r3, #31
 ; CHECK-LE-NEXT:    bne .LBB18_12
 ; CHECK-LE-NEXT:  @ %bb.1: @ %else
 ; CHECK-LE-NEXT:    lsls r1, r2, #30
 ; CHECK-LE-NEXT:    bmi .LBB18_13
 ; CHECK-LE-NEXT:  .LBB18_2: @ %else2
 ; CHECK-LE-NEXT:    lsls r1, r2, #29
 ; CHECK-LE-NEXT:    bmi .LBB18_14
 ; CHECK-LE-NEXT:  .LBB18_3: @ %else4

 ; CHECK-LE-NEXT:    b .LBB18_7
 ;
 ; CHECK-BE-LABEL: masked_v8f16_post:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .pad #8
 ; CHECK-BE-NEXT:    sub sp, #8
 ; CHECK-BE-NEXT:    vldr d1, [sp, #8]
 ; CHECK-BE-NEXT:    vmov d0, r3, r2
-; CHECK-BE-NEXT:    mov r2, sp
+; CHECK-BE-NEXT:    movs r3, #0
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
 ; CHECK-BE-NEXT:    vcmp.s16 gt, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r2]
-; CHECK-BE-NEXT:    ldrb.w r2, [sp]
+; CHECK-BE-NEXT:    vmrs r12, p0
+; CHECK-BE-NEXT:    and r2, r12, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #1, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #4, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #2, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #6, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #3, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #8, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #4, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #10, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #5, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #12, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #6, #1
+; CHECK-BE-NEXT:    ubfx r2, r12, #14, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
 ; CHECK-BE-NEXT:    vldrh.u16 q0, [r1]
-; CHECK-BE-NEXT:    lsls r1, r2, #31
+; CHECK-BE-NEXT:    bfi r3, r2, #7, #1
+; CHECK-BE-NEXT:    uxtb r2, r3
+; CHECK-BE-NEXT:    lsls r1, r3, #31
 ; CHECK-BE-NEXT:    bne .LBB18_12
 ; CHECK-BE-NEXT:  @ %bb.1: @ %else
 ; CHECK-BE-NEXT:    lsls r1, r2, #30
 ; CHECK-BE-NEXT:    bmi .LBB18_13
 ; CHECK-BE-NEXT:  .LBB18_2: @ %else2
 ; CHECK-BE-NEXT:    lsls r1, r2, #29
 ; CHECK-BE-NEXT:    bmi .LBB18_14
 ; CHECK-BE-NEXT:  .LBB18_3: @ %else4

llvm/trunk/test/CodeGen/Thumb2/mve-pred-bitcast.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK
+; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-LE
+; RUN: llc -mtriple=thumbebv8.1m.main-arm-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-BE
 
 define arm_aapcs_vfpcc <4 x i32> @bitcast_to_v4i1(i4 %b, <4 x i32> %a) {
-; CHECK-LABEL: bitcast_to_v4i1:
-; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .pad #4
-; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    and r0, r0, #15
-; CHECK-NEXT:    strb.w r0, [sp]
-; CHECK-NEXT:    mov r0, sp
-; CHECK-NEXT:    vmov.i32 q1, #0x0
-; CHECK-NEXT:    vldr p0, [r0]
-; CHECK-NEXT:    vpsel q0, q0, q1
-; CHECK-NEXT:    add sp, #4
-; CHECK-NEXT:    bx lr
+; CHECK-LE-LABEL: bitcast_to_v4i1:
+; CHECK-LE:       @ %bb.0: @ %entry
+; CHECK-LE-NEXT:    .pad #4
+; CHECK-LE-NEXT:    sub sp, #4
+; CHECK-LE-NEXT:    and r0, r0, #15
+; CHECK-LE-NEXT:    vmov.i8 q1, #0x0
+; CHECK-LE-NEXT:    vmov.i8 q2, #0xff
+; CHECK-LE-NEXT:    vmsr p0, r0
+; CHECK-LE-NEXT:    vpsel q1, q2, q1
+; CHECK-LE-NEXT:    vmov.u8 r0, q1[0]
+; CHECK-LE-NEXT:    vmov.32 q2[0], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q1[1]
+; CHECK-LE-NEXT:    vmov.32 q2[1], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q1[2]
+; CHECK-LE-NEXT:    vmov.32 q2[2], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q1[3]
+; CHECK-LE-NEXT:    vmov.32 q2[3], r0
+; CHECK-LE-NEXT:    vmov.i32 q1, #0x0
+; CHECK-LE-NEXT:    vcmp.i32 ne, q2, zr
+; CHECK-LE-NEXT:    vpsel q0, q0, q1
+; CHECK-LE-NEXT:    add sp, #4
+; CHECK-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: bitcast_to_v4i1:
+; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    .pad #4
+; CHECK-BE-NEXT:    sub sp, #4
+; CHECK-BE-NEXT:    and r0, r0, #15
+; CHECK-BE-NEXT:    vmov.i8 q1, #0x0
+; CHECK-BE-NEXT:    vmov.i8 q2, #0xff
+; CHECK-BE-NEXT:    vmsr p0, r0
+; CHECK-BE-NEXT:    vpsel q1, q2, q1
+; CHECK-BE-NEXT:    vmov.u8 r0, q1[0]
+; CHECK-BE-NEXT:    vmov.32 q2[0], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q1[1]
+; CHECK-BE-NEXT:    vmov.32 q2[1], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q1[2]
+; CHECK-BE-NEXT:    vmov.32 q2[2], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q1[3]
+; CHECK-BE-NEXT:    vmov.32 q2[3], r0
+; CHECK-BE-NEXT:    vrev64.32 q1, q0
+; CHECK-BE-NEXT:    vcmp.i32 ne, q2, zr
+; CHECK-BE-NEXT:    vmov.i32 q0, #0x0
+; CHECK-BE-NEXT:    vpsel q1, q1, q0
+; CHECK-BE-NEXT:    vrev64.32 q0, q1
+; CHECK-BE-NEXT:    add sp, #4
+; CHECK-BE-NEXT:    bx lr
 entry:
   %c = bitcast i4 %b to <4 x i1>
   %s = select <4 x i1> %c, <4 x i32> %a, <4 x i32> zeroinitializer
   ret <4 x i32> %s
 }
 
 define arm_aapcs_vfpcc <8 x i16> @bitcast_to_v8i1(i8 %b, <8 x i16> %a) {
-; CHECK-LABEL: bitcast_to_v8i1:
-; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .pad #8
-; CHECK-NEXT:    sub sp, #8
-; CHECK-NEXT:    strb.w r0, [sp]
-; CHECK-NEXT:    mov r0, sp
-; CHECK-NEXT:    vldr p0, [r0]
-; CHECK-NEXT:    vmov.i32 q1, #0x0
-; CHECK-NEXT:    vpsel q0, q0, q1
-; CHECK-NEXT:    add sp, #8
-; CHECK-NEXT:    bx lr
+; CHECK-LE-LABEL: bitcast_to_v8i1:
+; CHECK-LE:       @ %bb.0: @ %entry
+; CHECK-LE-NEXT:    .pad #8
+; CHECK-LE-NEXT:    sub sp, #8
+; CHECK-LE-NEXT:    uxtb r0, r0
+; CHECK-LE-NEXT:    vmov.i8 q1, #0x0
+; CHECK-LE-NEXT:    vmov.i8 q2, #0xff
+; CHECK-LE-NEXT:    vmsr p0, r0
+; CHECK-LE-NEXT:    vpsel q2, q2, q1
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[0]
+; CHECK-LE-NEXT:    vmov.16 q1[0], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[1]
+; CHECK-LE-NEXT:    vmov.16 q1[1], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[2]
+; CHECK-LE-NEXT:    vmov.16 q1[2], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[3]
+; CHECK-LE-NEXT:    vmov.16 q1[3], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[4]
+; CHECK-LE-NEXT:    vmov.16 q1[4], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[5]
+; CHECK-LE-NEXT:    vmov.16 q1[5], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[6]
+; CHECK-LE-NEXT:    vmov.16 q1[6], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[7]
+; CHECK-LE-NEXT:    vmov.16 q1[7], r0
+; CHECK-LE-NEXT:    vcmp.i16 ne, q1, zr
+; CHECK-LE-NEXT:    vmov.i32 q1, #0x0
+; CHECK-LE-NEXT:    vpsel q0, q0, q1
+; CHECK-LE-NEXT:    add sp, #8
+; CHECK-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: bitcast_to_v8i1:
+; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    .pad #8
+; CHECK-BE-NEXT:    sub sp, #8
+; CHECK-BE-NEXT:    uxtb r0, r0
+; CHECK-BE-NEXT:    vmov.i8 q1, #0x0
+; CHECK-BE-NEXT:    vmov.i8 q2, #0xff
+; CHECK-BE-NEXT:    vmsr p0, r0
+; CHECK-BE-NEXT:    vpsel q2, q2, q1
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[0]
+; CHECK-BE-NEXT:    vmov.16 q1[0], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[1]
+; CHECK-BE-NEXT:    vmov.16 q1[1], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[2]
+; CHECK-BE-NEXT:    vmov.16 q1[2], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[3]
+; CHECK-BE-NEXT:    vmov.16 q1[3], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[4]
+; CHECK-BE-NEXT:    vmov.16 q1[4], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[5]
+; CHECK-BE-NEXT:    vmov.16 q1[5], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[6]
+; CHECK-BE-NEXT:    vmov.16 q1[6], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[7]
+; CHECK-BE-NEXT:    vmov.16 q1[7], r0
+; CHECK-BE-NEXT:    vcmp.i16 ne, q1, zr
+; CHECK-BE-NEXT:    vrev64.16 q1, q0
+; CHECK-BE-NEXT:    vmov.i32 q0, #0x0
+; CHECK-BE-NEXT:    vrev32.16 q0, q0
+; CHECK-BE-NEXT:    vpsel q1, q1, q0
+; CHECK-BE-NEXT:    vrev64.16 q0, q1
+; CHECK-BE-NEXT:    add sp, #8
+; CHECK-BE-NEXT:    bx lr
 entry:
   %c = bitcast i8 %b to <8 x i1>
   %s = select <8 x i1> %c, <8 x i16> %a, <8 x i16> zeroinitializer
   ret <8 x i16> %s
 }
 
 define arm_aapcs_vfpcc <16 x i8> @bitcast_to_v16i1(i16 %b, <16 x i8> %a) {
-; CHECK-LABEL: bitcast_to_v16i1:
-; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r6, r7, lr}
-; CHECK-NEXT:    push {r4, r6, r7, lr}
-; CHECK-NEXT:    .setfp r7, sp, #8
-; CHECK-NEXT:    add r7, sp, #8
-; CHECK-NEXT:    .pad #16
-; CHECK-NEXT:    sub sp, #16
-; CHECK-NEXT:    mov r4, sp
-; CHECK-NEXT:    bfc r4, #0, #4
-; CHECK-NEXT:    mov sp, r4
-; CHECK-NEXT:    strh.w r0, [sp]
-; CHECK-NEXT:    mov r0, sp
-; CHECK-NEXT:    sub.w r4, r7, #8
-; CHECK-NEXT:    vldr p0, [r0]
-; CHECK-NEXT:    vmov.i32 q1, #0x0
-; CHECK-NEXT:    vpsel q0, q0, q1
-; CHECK-NEXT:    mov sp, r4
-; CHECK-NEXT:    pop {r4, r6, r7, pc}
+; CHECK-LE-LABEL: bitcast_to_v16i1:
+; CHECK-LE:       @ %bb.0: @ %entry
+; CHECK-LE-NEXT:    .save {r4, r6, r7, lr}
+; CHECK-LE-NEXT:    push {r4, r6, r7, lr}
+; CHECK-LE-NEXT:    .setfp r7, sp, #8
+; CHECK-LE-NEXT:    add r7, sp, #8
+; CHECK-LE-NEXT:    .pad #16
+; CHECK-LE-NEXT:    sub sp, #16
+; CHECK-LE-NEXT:    mov r4, sp
+; CHECK-LE-NEXT:    bfc r4, #0, #4
+; CHECK-LE-NEXT:    mov sp, r4
+; CHECK-LE-NEXT:    uxth r0, r0
+; CHECK-LE-NEXT:    sub.w r4, r7, #8
+; CHECK-LE-NEXT:    vmov.i32 q1, #0x0
+; CHECK-LE-NEXT:    vmsr p0, r0
+; CHECK-LE-NEXT:    vpsel q0, q0, q1
+; CHECK-LE-NEXT:    mov sp, r4
+; CHECK-LE-NEXT:    pop {r4, r6, r7, pc}
+;
+; CHECK-BE-LABEL: bitcast_to_v16i1:
+; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    .save {r4, r6, r7, lr}
+; CHECK-BE-NEXT:    push {r4, r6, r7, lr}
+; CHECK-BE-NEXT:    .setfp r7, sp, #8
+; CHECK-BE-NEXT:    add r7, sp, #8
+; CHECK-BE-NEXT:    .pad #16
+; CHECK-BE-NEXT:    sub sp, #16
+; CHECK-BE-NEXT:    mov r4, sp
+; CHECK-BE-NEXT:    bfc r4, #0, #4
+; CHECK-BE-NEXT:    mov sp, r4
+; CHECK-BE-NEXT:    vrev64.8 q1, q0
+; CHECK-BE-NEXT:    vmov.i32 q0, #0x0
+; CHECK-BE-NEXT:    uxth r0, r0
+; CHECK-BE-NEXT:    sub.w r4, r7, #8
+; CHECK-BE-NEXT:    vrev32.8 q0, q0
+; CHECK-BE-NEXT:    vmsr p0, r0
+; CHECK-BE-NEXT:    vpsel q1, q1, q0
+; CHECK-BE-NEXT:    vrev64.8 q0, q1
+; CHECK-BE-NEXT:    mov sp, r4
+; CHECK-BE-NEXT:    pop {r4, r6, r7, pc}
 entry:
   %c = bitcast i16 %b to <16 x i1>
   %s = select <16 x i1> %c, <16 x i8> %a, <16 x i8> zeroinitializer
   ret <16 x i8> %s
 }
 
 define arm_aapcs_vfpcc <2 x i64> @bitcast_to_v2i1(i2 %b, <2 x i64> %a) {
-; CHECK-LABEL: bitcast_to_v2i1:
-; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .pad #4
-; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    and r0, r0, #3
-; CHECK-NEXT:    sbfx r1, r0, #0, #1
-; CHECK-NEXT:    sbfx r0, r0, #1, #1
-; CHECK-NEXT:    vmov.32 q1[0], r1
-; CHECK-NEXT:    vmov.32 q1[1], r1
-; CHECK-NEXT:    vmov.32 q1[2], r0
-; CHECK-NEXT:    vmov.32 q1[3], r0
-; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    add sp, #4
-; CHECK-NEXT:    bx lr
+; CHECK-LE-LABEL: bitcast_to_v2i1:
+; CHECK-LE:       @ %bb.0: @ %entry
+; CHECK-LE-NEXT:    .pad #4
+; CHECK-LE-NEXT:    sub sp, #4
+; CHECK-LE-NEXT:    and r0, r0, #3
+; CHECK-LE-NEXT:    sbfx r1, r0, #0, #1
+; CHECK-LE-NEXT:    sbfx r0, r0, #1, #1
+; CHECK-LE-NEXT:    vmov.32 q1[0], r1
+; CHECK-LE-NEXT:    vmov.32 q1[1], r1
+; CHECK-LE-NEXT:    vmov.32 q1[2], r0
+; CHECK-LE-NEXT:    vmov.32 q1[3], r0
+; CHECK-LE-NEXT:    vand q0, q0, q1
+; CHECK-LE-NEXT:    add sp, #4
+; CHECK-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: bitcast_to_v2i1:
+; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    .pad #4
+; CHECK-BE-NEXT:    sub sp, #4
+; CHECK-BE-NEXT:    and r0, r0, #3
+; CHECK-BE-NEXT:    sbfx r1, r0, #0, #1
+; CHECK-BE-NEXT:    sbfx r0, r0, #1, #1
+; CHECK-BE-NEXT:    vmov.32 q1[0], r1
+; CHECK-BE-NEXT:    vmov.32 q1[1], r1
+; CHECK-BE-NEXT:    vmov.32 q1[2], r0
+; CHECK-BE-NEXT:    vmov.32 q1[3], r0
+; CHECK-BE-NEXT:    vrev64.32 q2, q1
+; CHECK-BE-NEXT:    vand q0, q0, q2
+; CHECK-BE-NEXT:    add sp, #4
+; CHECK-BE-NEXT:    bx lr
 entry:
   %c = bitcast i2 %b to <2 x i1>
   %s = select <2 x i1> %c, <2 x i64> %a, <2 x i64> zeroinitializer
   ret <2 x i64> %s
 }
 
 
 define arm_aapcs_vfpcc i4 @bitcast_from_v4i1(<4 x i32> %a) {
-; CHECK-LABEL: bitcast_from_v4i1:
-; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .pad #4
-; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    vcmp.i32 eq, q0, zr
-; CHECK-NEXT:    mov r0, sp
-; CHECK-NEXT:    vstr p0, [r0]
-; CHECK-NEXT:    ldrb.w r0, [sp]
-; CHECK-NEXT:    add sp, #4
-; CHECK-NEXT:    bx lr
+; CHECK-LE-LABEL: bitcast_from_v4i1:
+; CHECK-LE:       @ %bb.0: @ %entry
+; CHECK-LE-NEXT:    .pad #4
+; CHECK-LE-NEXT:    sub sp, #4
+; CHECK-LE-NEXT:    vcmp.i32 eq, q0, zr
+; CHECK-LE-NEXT:    movs r0, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r2, r1, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #0, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #1, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #8, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #12, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #2, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r0, r1, #3, #1
+; CHECK-LE-NEXT:    and r0, r0, #15
+; CHECK-LE-NEXT:    add sp, #4
+; CHECK-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: bitcast_from_v4i1:
+; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    .pad #4
+; CHECK-BE-NEXT:    sub sp, #4
+; CHECK-BE-NEXT:    vrev64.32 q1, q0
+; CHECK-BE-NEXT:    movs r3, #0
+; CHECK-BE-NEXT:    vcmp.i32 eq, q1, zr
+; CHECK-BE-NEXT:    vmrs r0, p0
+; CHECK-BE-NEXT:    and r2, r0, #1
+; CHECK-BE-NEXT:    ubfx r1, r0, #4, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r3, r2, #0, #1
+; CHECK-BE-NEXT:    bfi r3, r1, #1, #1
+; CHECK-BE-NEXT:    ubfx r1, r0, #8, #1
+; CHECK-BE-NEXT:    ubfx r0, r0, #12, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r3, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r0, r0, #0
+; CHECK-BE-NEXT:    bfi r3, r0, #3, #1
+; CHECK-BE-NEXT:    and r0, r3, #15
+; CHECK-BE-NEXT:    add sp, #4
+; CHECK-BE-NEXT:    bx lr
 entry:
   %c = icmp eq <4 x i32> %a, zeroinitializer
   %b = bitcast <4 x i1> %c to i4
   ret i4 %b
 }
 
 define arm_aapcs_vfpcc i8 @bitcast_from_v8i1(<8 x i16> %a) {
-; CHECK-LABEL: bitcast_from_v8i1:
-; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .pad #8
-; CHECK-NEXT:    sub sp, #8
-; CHECK-NEXT:    vcmp.i16 eq, q0, zr
-; CHECK-NEXT:    mov r0, sp
-; CHECK-NEXT:    vstr p0, [r0]
-; CHECK-NEXT:    ldrb.w r0, [sp]
-; CHECK-NEXT:    add sp, #8
-; CHECK-NEXT:    bx lr
+; CHECK-LE-LABEL: bitcast_from_v8i1:
+; CHECK-LE:       @ %bb.0: @ %entry
+; CHECK-LE-NEXT:    .pad #8
+; CHECK-LE-NEXT:    sub sp, #8
+; CHECK-LE-NEXT:    vcmp.i16 eq, q0, zr
+; CHECK-LE-NEXT:    movs r0, #0
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    and r2, r1, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #0, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #1, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #4, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #2, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #6, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #3, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #8, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #4, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #10, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #5, #1
+; CHECK-LE-NEXT:    ubfx r2, r1, #12, #1
+; CHECK-LE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r0, r2, #6, #1
+; CHECK-LE-NEXT:    rsbs r1, r1, #0
+; CHECK-LE-NEXT:    bfi r0, r1, #7, #1
+; CHECK-LE-NEXT:    uxtb r0, r0
+; CHECK-LE-NEXT:    add sp, #8
+; CHECK-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: bitcast_from_v8i1:
+; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    .pad #8
+; CHECK-BE-NEXT:    sub sp, #8
+; CHECK-BE-NEXT:    vrev64.16 q1, q0
+; CHECK-BE-NEXT:    vcmp.i16 eq, q1, zr
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    ubfx r0, r1, #2, #1
+; CHECK-BE-NEXT:    rsbs r2, r0, #0
+; CHECK-BE-NEXT:    and r0, r1, #1
+; CHECK-BE-NEXT:    rsbs r3, r0, #0
+; CHECK-BE-NEXT:    movs r0, #0
+; CHECK-BE-NEXT:    bfi r0, r3, #0, #1
+; CHECK-BE-NEXT:    bfi r0, r2, #1, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r0, r2, #2, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #6, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r0, r2, #3, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #8, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r0, r2, #4, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #10, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r0, r2, #5, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #12, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #14, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r0, r2, #6, #1
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r0, r1, #7, #1
+; CHECK-BE-NEXT:    uxtb r0, r0
+; CHECK-BE-NEXT:    add sp, #8
+; CHECK-BE-NEXT:    bx lr
 entry:
   %c = icmp eq <8 x i16> %a, zeroinitializer
   %b = bitcast <8 x i1> %c to i8
   ret i8 %b
 }
 
 define arm_aapcs_vfpcc i16 @bitcast_from_v16i1(<16 x i8> %a) {
-; CHECK-LABEL: bitcast_from_v16i1:
-; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r6, r7, lr}
-; CHECK-NEXT:    push {r4, r6, r7, lr}
-; CHECK-NEXT:    .setfp r7, sp, #8
-; CHECK-NEXT:    add r7, sp, #8
-; CHECK-NEXT:    .pad #16
-; CHECK-NEXT:    sub sp, #16
-; CHECK-NEXT:    mov r4, sp
-; CHECK-NEXT:    bfc r4, #0, #4
-; CHECK-NEXT:    mov sp, r4
-; CHECK-NEXT:    sub.w r4, r7, #8
-; CHECK-NEXT:    vcmp.i8 eq, q0, zr
-; CHECK-NEXT:    mov r0, sp
-; CHECK-NEXT:    vstr p0, [r0]
-; CHECK-NEXT:    ldrh.w r0, [sp]
-; CHECK-NEXT:    mov sp, r4
-; CHECK-NEXT:    pop {r4, r6, r7, pc}
+; CHECK-LE-LABEL: bitcast_from_v16i1:
+; CHECK-LE:       @ %bb.0: @ %entry
+; CHECK-LE-NEXT:    .save {r4, r6, r7, lr}
+; CHECK-LE-NEXT:    push {r4, r6, r7, lr}
+; CHECK-LE-NEXT:    .setfp r7, sp, #8
+; CHECK-LE-NEXT:    add r7, sp, #8
+; CHECK-LE-NEXT:    .pad #16
+; CHECK-LE-NEXT:    sub sp, #16
+; CHECK-LE-NEXT:    mov r4, sp
+; CHECK-LE-NEXT:    bfc r4, #0, #4
+; CHECK-LE-NEXT:    mov sp, r4
+; CHECK-LE-NEXT:    vcmp.i8 eq, q0, zr
+; CHECK-LE-NEXT:    sub.w r4, r7, #8
+; CHECK-LE-NEXT:    vmrs r0, p0
+; CHECK-LE-NEXT:    uxth r0, r0
+; CHECK-LE-NEXT:    mov sp, r4
+; CHECK-LE-NEXT:    pop {r4, r6, r7, pc}
+;
+; CHECK-BE-LABEL: bitcast_from_v16i1:
+; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    .save {r4, r6, r7, lr}
+; CHECK-BE-NEXT:    push {r4, r6, r7, lr}
+; CHECK-BE-NEXT:    .setfp r7, sp, #8
+; CHECK-BE-NEXT:    add r7, sp, #8
+; CHECK-BE-NEXT:    .pad #16
+; CHECK-BE-NEXT:    sub sp, #16
+; CHECK-BE-NEXT:    mov r4, sp
+; CHECK-BE-NEXT:    bfc r4, #0, #4
+; CHECK-BE-NEXT:    mov sp, r4
+; CHECK-BE-NEXT:    vrev64.8 q1, q0
+; CHECK-BE-NEXT:    sub.w r4, r7, #8
+; CHECK-BE-NEXT:    vcmp.i8 eq, q1, zr
+; CHECK-BE-NEXT:    vmrs r0, p0
+; CHECK-BE-NEXT:    uxth r0, r0
+; CHECK-BE-NEXT:    mov sp, r4
+; CHECK-BE-NEXT:    pop {r4, r6, r7, pc}
 entry:
   %c = icmp eq <16 x i8> %a, zeroinitializer
   %b = bitcast <16 x i1> %c to i16
   ret i16 %b
 }
 
 define arm_aapcs_vfpcc i2 @bitcast_from_v2i1(<2 x i64> %a) {
-; CHECK-LABEL: bitcast_from_v2i1:
-; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .pad #4
-; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vmov r1, s0
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    orrs r0, r1
-; CHECK-NEXT:    vmov r1, s3
-; CHECK-NEXT:    cset r0, eq
-; CHECK-NEXT:    orrs r1, r2
-; CHECK-NEXT:    cset r1, eq
-; CHECK-NEXT:    ands r1, r1, #1
-; CHECK-NEXT:    it ne
-; CHECK-NEXT:    mvnne r1, #1
-; CHECK-NEXT:    bfi r1, r0, #0, #1
-; CHECK-NEXT:    and r0, r1, #3
-; CHECK-NEXT:    add sp, #4
-; CHECK-NEXT:    bx lr
+; CHECK-LE-LABEL: bitcast_from_v2i1:
+; CHECK-LE:       @ %bb.0: @ %entry
+; CHECK-LE-NEXT:    .pad #4
+; CHECK-LE-NEXT:    sub sp, #4
+; CHECK-LE-NEXT:    vmov r0, s1
+; CHECK-LE-NEXT:    vmov r1, s0
+; CHECK-LE-NEXT:    vmov r2, s2
+; CHECK-LE-NEXT:    orrs r0, r1
+; CHECK-LE-NEXT:    vmov r1, s3
+; CHECK-LE-NEXT:    cset r0, eq
+; CHECK-LE-NEXT:    orrs r1, r2
+; CHECK-LE-NEXT:    cset r1, eq
+; CHECK-LE-NEXT:    ands r1, r1, #1
+; CHECK-LE-NEXT:    it ne
+; CHECK-LE-NEXT:    mvnne r1, #1
+; CHECK-LE-NEXT:    bfi r1, r0, #0, #1
+; CHECK-LE-NEXT:    and r0, r1, #3
+; CHECK-LE-NEXT:    add sp, #4
+; CHECK-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: bitcast_from_v2i1:
+; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    .pad #4
+; CHECK-BE-NEXT:    sub sp, #4
+; CHECK-BE-NEXT:    vrev64.32 q1, q0
+; CHECK-BE-NEXT:    vmov r0, s6
+; CHECK-BE-NEXT:    vmov r1, s7
+; CHECK-BE-NEXT:    vmov r2, s5
+; CHECK-BE-NEXT:    orrs r0, r1
+; CHECK-BE-NEXT:    vmov r1, s4
+; CHECK-BE-NEXT:    cset r0, eq
+; CHECK-BE-NEXT:    orrs r1, r2
+; CHECK-BE-NEXT:    cset r1, eq
+; CHECK-BE-NEXT:    ands r1, r1, #1
+; CHECK-BE-NEXT:    it ne
+; CHECK-BE-NEXT:    mvnne r1, #1
+; CHECK-BE-NEXT:    bfi r1, r0, #0, #1
+; CHECK-BE-NEXT:    and r0, r1, #3
+; CHECK-BE-NEXT:    add sp, #4
+; CHECK-BE-NEXT:    bx lr
 entry:
   %c = icmp eq <2 x i64> %a, zeroinitializer
   %b = bitcast <2 x i1> %c to i2
   ret i2 %b
 }

llvm/trunk/test/CodeGen/Thumb2/mve-pred-loadstore.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-LE
 ; RUN: llc -mtriple=thumbebv8.1m.main-arm-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-BE
 
 define arm_aapcs_vfpcc <4 x i32> @load_v4i1(<4 x i1> *%src, <4 x i32> %a) {
 ; CHECK-LE-LABEL: load_v4i1:
 ; CHECK-LE:       @ %bb.0: @ %entry
-; CHECK-LE-NEXT:    vldr p0, [r0]
+; CHECK-LE-NEXT:    ldrb r0, [r0]
+; CHECK-LE-NEXT:    vmov.i8 q1, #0x0
+; CHECK-LE-NEXT:    vmov.i8 q2, #0xff
+; CHECK-LE-NEXT:    vmsr p0, r0
+; CHECK-LE-NEXT:    vpsel q1, q2, q1
+; CHECK-LE-NEXT:    vmov.u8 r0, q1[0]
+; CHECK-LE-NEXT:    vmov.32 q2[0], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q1[1]
+; CHECK-LE-NEXT:    vmov.32 q2[1], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q1[2]
+; CHECK-LE-NEXT:    vmov.32 q2[2], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q1[3]
+; CHECK-LE-NEXT:    vmov.32 q2[3], r0
 ; CHECK-LE-NEXT:    vmov.i32 q1, #0x0
+; CHECK-LE-NEXT:    vcmp.i32 ne, q2, zr
 ; CHECK-LE-NEXT:    vpsel q0, q0, q1
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: load_v4i1:
 ; CHECK-BE:       @ %bb.0: @ %entry
-; CHECK-BE-NEXT:    vldr p0, [r0]
+; CHECK-BE-NEXT:    ldrb r0, [r0]
+; CHECK-BE-NEXT:    vmov.i8 q1, #0x0
+; CHECK-BE-NEXT:    vmov.i8 q2, #0xff
+; CHECK-BE-NEXT:    vmsr p0, r0
+; CHECK-BE-NEXT:    vpsel q1, q2, q1
+; CHECK-BE-NEXT:    vmov.u8 r0, q1[0]
+; CHECK-BE-NEXT:    vmov.32 q2[0], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q1[1]
+; CHECK-BE-NEXT:    vmov.32 q2[1], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q1[2]
+; CHECK-BE-NEXT:    vmov.32 q2[2], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q1[3]
+; CHECK-BE-NEXT:    vmov.32 q2[3], r0
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
+; CHECK-BE-NEXT:    vcmp.i32 ne, q2, zr
 ; CHECK-BE-NEXT:    vmov.i32 q0, #0x0
 ; CHECK-BE-NEXT:    vpsel q1, q1, q0
 ; CHECK-BE-NEXT:    vrev64.32 q0, q1
 ; CHECK-BE-NEXT:    bx lr
 entry:
   %c = load <4 x i1>, <4 x i1>* %src
   %s = select <4 x i1> %c, <4 x i32> %a, <4 x i32> zeroinitializer
   ret <4 x i32> %s
 }
 
 define arm_aapcs_vfpcc <8 x i16> @load_v8i1(<8 x i1> *%src, <8 x i16> %a) {
 ; CHECK-LE-LABEL: load_v8i1:
 ; CHECK-LE:       @ %bb.0: @ %entry
-; CHECK-LE-NEXT:    vldr p0, [r0]
+; CHECK-LE-NEXT:    ldrb r0, [r0]
+; CHECK-LE-NEXT:    vmov.i8 q1, #0x0
+; CHECK-LE-NEXT:    vmov.i8 q2, #0xff
+; CHECK-LE-NEXT:    vmsr p0, r0
+; CHECK-LE-NEXT:    vpsel q2, q2, q1
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[0]
+; CHECK-LE-NEXT:    vmov.16 q1[0], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[1]
+; CHECK-LE-NEXT:    vmov.16 q1[1], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[2]
+; CHECK-LE-NEXT:    vmov.16 q1[2], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[3]
+; CHECK-LE-NEXT:    vmov.16 q1[3], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[4]
+; CHECK-LE-NEXT:    vmov.16 q1[4], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[5]
+; CHECK-LE-NEXT:    vmov.16 q1[5], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[6]
+; CHECK-LE-NEXT:    vmov.16 q1[6], r0
+; CHECK-LE-NEXT:    vmov.u8 r0, q2[7]
+; CHECK-LE-NEXT:    vmov.16 q1[7], r0
+; CHECK-LE-NEXT:    vcmp.i16 ne, q1, zr
 ; CHECK-LE-NEXT:    vmov.i32 q1, #0x0
 ; CHECK-LE-NEXT:    vpsel q0, q0, q1
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: load_v8i1:
 ; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    ldrb r0, [r0]
+; CHECK-BE-NEXT:    vmov.i8 q1, #0x0
+; CHECK-BE-NEXT:    vmov.i8 q2, #0xff
+; CHECK-BE-NEXT:    vmsr p0, r0
+; CHECK-BE-NEXT:    vpsel q2, q2, q1
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[0]
+; CHECK-BE-NEXT:    vmov.16 q1[0], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[1]
+; CHECK-BE-NEXT:    vmov.16 q1[1], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[2]
+; CHECK-BE-NEXT:    vmov.16 q1[2], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[3]
+; CHECK-BE-NEXT:    vmov.16 q1[3], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[4]
+; CHECK-BE-NEXT:    vmov.16 q1[4], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[5]
+; CHECK-BE-NEXT:    vmov.16 q1[5], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[6]
+; CHECK-BE-NEXT:    vmov.16 q1[6], r0
+; CHECK-BE-NEXT:    vmov.u8 r0, q2[7]
+; CHECK-BE-NEXT:    vmov.16 q1[7], r0
+; CHECK-BE-NEXT:    vcmp.i16 ne, q1, zr
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
 ; CHECK-BE-NEXT:    vmov.i32 q0, #0x0
-; CHECK-BE-NEXT:    vldr p0, [r0]
 ; CHECK-BE-NEXT:    vrev32.16 q0, q0
 ; CHECK-BE-NEXT:    vpsel q1, q1, q0
 ; CHECK-BE-NEXT:    vrev64.16 q0, q1
 ; CHECK-BE-NEXT:    bx lr
 entry:
   %c = load <8 x i1>, <8 x i1>* %src
   %s = select <8 x i1> %c, <8 x i16> %a, <8 x i16> zeroinitializer
   ret <8 x i16> %s
 }
 
 define arm_aapcs_vfpcc <16 x i8> @load_v16i1(<16 x i1> *%src, <16 x i8> %a) {
 ; CHECK-LE-LABEL: load_v16i1:
 ; CHECK-LE:       @ %bb.0: @ %entry
-; CHECK-LE-NEXT:    vldr p0, [r0]
+; CHECK-LE-NEXT:    ldrh r0, [r0]
 ; CHECK-LE-NEXT:    vmov.i32 q1, #0x0
+; CHECK-LE-NEXT:    vmsr p0, r0
 ; CHECK-LE-NEXT:    vpsel q0, q0, q1
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: load_v16i1:
 ; CHECK-BE:       @ %bb.0: @ %entry
+; CHECK-BE-NEXT:    ldrh r0, [r0]
 ; CHECK-BE-NEXT:    vrev64.8 q1, q0
 ; CHECK-BE-NEXT:    vmov.i32 q0, #0x0
-; CHECK-BE-NEXT:    vldr p0, [r0]
 ; CHECK-BE-NEXT:    vrev32.8 q0, q0
+; CHECK-BE-NEXT:    vmsr p0, r0
 ; CHECK-BE-NEXT:    vpsel q1, q1, q0
 ; CHECK-BE-NEXT:    vrev64.8 q0, q1
 ; CHECK-BE-NEXT:    bx lr
 entry:
   %c = load <16 x i1>, <16 x i1>* %src
   %s = select <16 x i1> %c, <16 x i8> %a, <16 x i8> zeroinitializer
   ret <16 x i8> %s
 }
entry:
   ret <2 x i64> %s
 }
 
 
 define arm_aapcs_vfpcc void @store_v4i1(<4 x i1> *%dst, <4 x i32> %a) {
 ; CHECK-LE-LABEL: store_v4i1:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    vcmp.i32 eq, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r0]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #3, #1
+; CHECK-LE-NEXT:    and r1, r1, #15
+; CHECK-LE-NEXT:    strb r1, [r0]
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: store_v4i1:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    vrev64.32 q1, q0
 ; CHECK-BE-NEXT:    vcmp.i32 eq, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r0]
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    and r3, r1, #1
+; CHECK-BE-NEXT:    ubfx r2, r1, #4, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    rsb.w r12, r2, #0
+; CHECK-BE-NEXT:    movs r2, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r1, #8, #1
+; CHECK-BE-NEXT:    ubfx r1, r1, #12, #1
+; CHECK-BE-NEXT:    bfi r2, r12, #1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    rsbs r1, r1, #0
+; CHECK-BE-NEXT:    bfi r2, r3, #2, #1
+; CHECK-BE-NEXT:    bfi r2, r1, #3, #1
+; CHECK-BE-NEXT:    and r1, r2, #15
+; CHECK-BE-NEXT:    strb r1, [r0]
 ; CHECK-BE-NEXT:    bx lr
 entry:
   %c = icmp eq <4 x i32> %a, zeroinitializer
   store <4 x i1> %c, <4 x i1>* %dst
   ret void
 }
 
 define arm_aapcs_vfpcc void @store_v8i1(<8 x i1> *%dst, <8 x i16> %a) {
 ; CHECK-LE-LABEL: store_v8i1:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    vcmp.i16 eq, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r0]
+; CHECK-LE-NEXT:    movs r1, #0
+; CHECK-LE-NEXT:    vmrs r2, p0
+; CHECK-LE-NEXT:    and r3, r2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #2, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #1, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #6, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #3, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #4, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #10, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #5, #1
+; CHECK-LE-NEXT:    ubfx r3, r2, #12, #1
+; CHECK-LE-NEXT:    ubfx r2, r2, #14, #1
+; CHECK-LE-NEXT:    rsbs r3, r3, #0
+; CHECK-LE-NEXT:    bfi r1, r3, #6, #1
+; CHECK-LE-NEXT:    rsbs r2, r2, #0
+; CHECK-LE-NEXT:    bfi r1, r2, #7, #1
+; CHECK-LE-NEXT:    strb r1, [r0]
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: store_v8i1:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    vrev64.16 q1, q0
 ; CHECK-BE-NEXT:    vcmp.i16 eq, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r0]
+; CHECK-BE-NEXT:    vmrs r2, p0
+; CHECK-BE-NEXT:    ubfx r1, r2, #2, #1
+; CHECK-BE-NEXT:    rsb.w r12, r1, #0
+; CHECK-BE-NEXT:    and r1, r2, #1
+; CHECK-BE-NEXT:    rsbs r3, r1, #0
+; CHECK-BE-NEXT:    movs r1, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #0, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-BE-NEXT:    bfi r1, r12, #1, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #2, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #6, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #3, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #4, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #10, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #5, #1
+; CHECK-BE-NEXT:    ubfx r3, r2, #12, #1
+; CHECK-BE-NEXT:    ubfx r2, r2, #14, #1
+; CHECK-BE-NEXT:    rsbs r3, r3, #0
+; CHECK-BE-NEXT:    bfi r1, r3, #6, #1
+; CHECK-BE-NEXT:    rsbs r2, r2, #0
+; CHECK-BE-NEXT:    bfi r1, r2, #7, #1
+; CHECK-BE-NEXT:    strb r1, [r0]
 ; CHECK-BE-NEXT:    bx lr
 entry:
   %c = icmp eq <8 x i16> %a, zeroinitializer
   store <8 x i1> %c, <8 x i1>* %dst
   ret void
 }
 
 define arm_aapcs_vfpcc void @store_v16i1(<16 x i1> *%dst, <16 x i8> %a) {
 ; CHECK-LE-LABEL: store_v16i1:
 ; CHECK-LE:       @ %bb.0: @ %entry
 ; CHECK-LE-NEXT:    vcmp.i8 eq, q0, zr
-; CHECK-LE-NEXT:    vstr p0, [r0]
+; CHECK-LE-NEXT:    vmrs r1, p0
+; CHECK-LE-NEXT:    strh r1, [r0]
 ; CHECK-LE-NEXT:    bx lr
 ;
 ; CHECK-BE-LABEL: store_v16i1:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    vrev64.8 q1, q0
 ; CHECK-BE-NEXT:    vcmp.i8 eq, q1, zr
-; CHECK-BE-NEXT:    vstr p0, [r0]
+; CHECK-BE-NEXT:    vmrs r1, p0
+; CHECK-BE-NEXT:    strh r1, [r0]
 ; CHECK-BE-NEXT:    bx lr
 entry:
   %c = icmp eq <16 x i8> %a, zeroinitializer
   store <16 x i1> %c, <16 x i1>* %dst
   ret void
 }
 
 define arm_aapcs_vfpcc void @store_v2i1(<2 x i1> *%dst, <2 x i64> %a) {