Codebase Browser
llvm
Go to App

llvm/llvm/test/CodeGen/Thumb2/mve-gather-ind32-unscaled.ll 

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp %s -o - | FileCheck %s

define arm_aapcs_vfpcc <4 x i32> @zext_unscaled_i8_i32(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_unscaled_i8_i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q1, [r1]
; CHECK-NEXT:    vldrb.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i32>, ptr %offptr, align 4
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.zext = zext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_unscaled_i8_i32(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_unscaled_i8_i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q1, [r1]
; CHECK-NEXT:    vldrb.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i32>, ptr %offptr, align 4
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.sext = sext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

define arm_aapcs_vfpcc <4 x i32> @zext_unscaled_i16_i32(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_unscaled_i16_i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q1, [r1]
; CHECK-NEXT:    vldrh.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i32>, ptr %offptr, align 4
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.zext = zext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_unscaled_i16_i32(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_unscaled_i16_i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q1, [r1]
; CHECK-NEXT:    vldrh.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i32>, ptr %offptr, align 4
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.sext = sext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

define arm_aapcs_vfpcc <4 x i32> @unscaled_i32_i32(ptr %base, ptr %offptr) {
; CHECK-LABEL: unscaled_i32_i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i32>, ptr %offptr, align 4
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef)
  ret <4 x i32> %gather
}

define arm_aapcs_vfpcc <4 x float> @unscaled_f32_i32(ptr %base, ptr %offptr) {
; CHECK-LABEL: unscaled_f32_i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i32>, ptr %offptr, align 4
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x float> @llvm.masked.gather.v4f32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
  ret <4 x float> %gather
}

define arm_aapcs_vfpcc <4 x i32> @unsigned_unscaled_b_i32_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: unsigned_unscaled_b_i32_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.zext = zext <4 x i16> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef)
  ret <4 x i32> %gather
}

define arm_aapcs_vfpcc <4 x i32> @signed_unscaled_i32_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: signed_unscaled_i32_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.s32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.sext = sext <4 x i16> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef)
  ret <4 x i32> %gather
}

define arm_aapcs_vfpcc <4 x float> @a_unsigned_unscaled_f32_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: a_unsigned_unscaled_f32_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.zext = zext <4 x i16> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x float> @llvm.masked.gather.v4f32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
  ret <4 x float> %gather
}

define arm_aapcs_vfpcc <4 x float> @b_signed_unscaled_f32_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: b_signed_unscaled_f32_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.s32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.sext = sext <4 x i16> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x float> @llvm.masked.gather.v4f32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
  ret <4 x float> %gather
}

define arm_aapcs_vfpcc <4 x i32> @zext_signed_unscaled_i16_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_signed_unscaled_i16_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.s32 q1, [r1]
; CHECK-NEXT:    vldrh.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.sext = sext <4 x i16> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.zext = zext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_signed_unscaled_i16_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_signed_unscaled_i16_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.s32 q1, [r1]
; CHECK-NEXT:    vldrh.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.sext = sext <4 x i16> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.sext = sext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

define arm_aapcs_vfpcc <4 x i32> @zext_unsigned_unscaled_i16_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_unsigned_unscaled_i16_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u32 q1, [r1]
; CHECK-NEXT:    vldrh.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.zext = zext <4 x i16> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.zext = zext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_unsigned_unscaled_i16_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_unsigned_unscaled_i16_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u32 q1, [r1]
; CHECK-NEXT:    vldrh.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.zext = zext <4 x i16> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.sext = sext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

define arm_aapcs_vfpcc <4 x i32> @zext_signed_unscaled_i8_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_signed_unscaled_i8_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.s32 q1, [r1]
; CHECK-NEXT:    vldrb.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.sext = sext <4 x i16> %offs to <4 x i32>
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.zext = zext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_signed_unscaled_i8_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_signed_unscaled_i8_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.s32 q1, [r1]
; CHECK-NEXT:    vldrb.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.sext = sext <4 x i16> %offs to <4 x i32>
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.sext = sext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

define arm_aapcs_vfpcc <4 x i32> @zext_unsigned_unscaled_i8_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_unsigned_unscaled_i8_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u32 q1, [r1]
; CHECK-NEXT:    vldrb.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.zext = zext <4 x i16> %offs to <4 x i32>
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.zext = zext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_unsigned_unscaled_i8_i16(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_unsigned_unscaled_i8_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u32 q1, [r1]
; CHECK-NEXT:    vldrb.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i16>, ptr %offptr, align 2
  %offs.zext = zext <4 x i16> %offs to <4 x i32>
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.sext = sext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

define arm_aapcs_vfpcc <4 x i32> @unsigned_unscaled_b_i32_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: unsigned_unscaled_b_i32_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.zext = zext <4 x i8> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef)
  ret <4 x i32> %gather
}

define arm_aapcs_vfpcc <4 x i32> @signed_unscaled_i32_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: signed_unscaled_i32_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.s32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.sext = sext <4 x i8> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef)
  ret <4 x i32> %gather
}

define arm_aapcs_vfpcc <4 x float> @a_unsigned_unscaled_f32_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: a_unsigned_unscaled_f32_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.zext = zext <4 x i8> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x float> @llvm.masked.gather.v4f32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
  ret <4 x float> %gather
}

define arm_aapcs_vfpcc <4 x float> @b_signed_unscaled_f32_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: b_signed_unscaled_f32_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.s32 q1, [r1]
; CHECK-NEXT:    vldrw.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.sext = sext <4 x i8> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x float> @llvm.masked.gather.v4f32.v4p0(<4 x ptr> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
  ret <4 x float> %gather
}

define arm_aapcs_vfpcc <4 x i32> @zext_signed_unscaled_i16_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_signed_unscaled_i16_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.s32 q1, [r1]
; CHECK-NEXT:    vldrh.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.sext = sext <4 x i8> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.zext = zext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_signed_unscaled_i16_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_signed_unscaled_i16_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.s32 q1, [r1]
; CHECK-NEXT:    vldrh.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.sext = sext <4 x i8> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.sext = sext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

define arm_aapcs_vfpcc <4 x i32> @zext_unsigned_unscaled_i16_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_unsigned_unscaled_i16_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u32 q1, [r1]
; CHECK-NEXT:    vldrh.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.zext = zext <4 x i8> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.zext = zext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_unsigned_unscaled_i16_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_unsigned_unscaled_i16_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u32 q1, [r1]
; CHECK-NEXT:    vldrh.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.zext = zext <4 x i8> %offs to <4 x i32>
  %byte_ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %ptrs = bitcast <4 x ptr> %byte_ptrs to <4 x ptr>
  %gather = call <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr> %ptrs, i32 2, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i16> undef)
  %gather.sext = sext <4 x i16> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

define arm_aapcs_vfpcc <4 x i32> @zext_signed_unscaled_i8_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_signed_unscaled_i8_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.s32 q1, [r1]
; CHECK-NEXT:    vldrb.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.sext = sext <4 x i8> %offs to <4 x i32>
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.zext = zext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_signed_unscaled_i8_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_signed_unscaled_i8_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.s32 q1, [r1]
; CHECK-NEXT:    vldrb.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.sext = sext <4 x i8> %offs to <4 x i32>
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.sext
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.sext = sext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

define arm_aapcs_vfpcc <4 x i32> @zext_unsigned_unscaled_i8_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: zext_unsigned_unscaled_i8_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u32 q1, [r1]
; CHECK-NEXT:    vldrb.u32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.zext = zext <4 x i8> %offs to <4 x i32>
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.zext = zext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.zext
}

define arm_aapcs_vfpcc <4 x i32> @sext_unsigned_unscaled_i8_i8(ptr %base, ptr %offptr) {
; CHECK-LABEL: sext_unsigned_unscaled_i8_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u32 q1, [r1]
; CHECK-NEXT:    vldrb.s32 q0, [r0, q1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <4 x i8>, ptr %offptr, align 1
  %offs.zext = zext <4 x i8> %offs to <4 x i32>
  %ptrs = getelementptr inbounds i8, ptr %base, <4 x i32> %offs.zext
  %gather = call <4 x i8> @llvm.masked.gather.v4i8.v4p0(<4 x ptr> %ptrs, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i8> undef)
  %gather.sext = sext <4 x i8> %gather to <4 x i32>
  ret <4 x i32> %gather.sext
}

; VLDRW.u32 Qd, [P, 4]
define arm_aapcs_vfpcc <4 x i32> @qi4(<4 x ptr> %p) {
; CHECK-LABEL: qi4:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    movs r0, #16
; CHECK-NEXT:    vadd.i32 q1, q0, r0
; CHECK-NEXT:    vldrw.u32 q0, [q1]
; CHECK-NEXT:    bx lr
entry:
  %g = getelementptr inbounds i32, <4 x ptr> %p, i32 4
  %gather = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> %g, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef)
  ret <4 x i32> %gather
}

define arm_aapcs_vfpcc <4 x i32> @qi4_unaligned(<4 x ptr> %p) {
; CHECK-LABEL: qi4_unaligned:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    movs r0, #16
; CHECK-NEXT:    vadd.i32 q0, q0, r0
; CHECK-NEXT:    vmov r0, r1, d1
; CHECK-NEXT:    vmov r2, r3, d0
; CHECK-NEXT:    ldr r0, [r0]
; CHECK-NEXT:    ldr r2, [r2]
; CHECK-NEXT:    ldr r1, [r1]
; CHECK-NEXT:    ldr r3, [r3]
; CHECK-NEXT:    vmov q0[2], q0[0], r2, r0
; CHECK-NEXT:    vmov q0[3], q0[1], r3, r1
; CHECK-NEXT:    bx lr
entry:
  %g = getelementptr inbounds i32, <4 x ptr> %p, i32 4
  %gather = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> %g, i32 1, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef)
  ret <4 x i32> %gather
}

declare <4 x i8>  @llvm.masked.gather.v4i8.v4p0(<4 x ptr>, i32, <4 x i1>, <4 x i8>)
declare <4 x i16> @llvm.masked.gather.v4i16.v4p0(<4 x ptr>, i32, <4 x i1>, <4 x i16>)
declare <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr>, i32, <4 x i1>, <4 x i32>)
declare <4 x half> @llvm.masked.gather.v4f16.v4p0(<4 x ptr>, i32, <4 x i1>, <4 x half>)
declare <4 x float> @llvm.masked.gather.v4f32.v4p0(<4 x ptr>, i32, <4 x i1>, <4 x float>)