llvm/test/CodeGen/AArch64/sve-fixed-length-masked-gather.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -aarch64-sve-vector-bits-min=256  < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_256
; RUN: llc -aarch64-sve-vector-bits-min=512  < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512
; RUN: llc -aarch64-sve-vector-bits-min=2048 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512

target triple = "aarch64-unknown-linux-gnu"

;
; LD1B
;

define void @masked_gather_v2i8(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v2i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldrb w8, [x0]
; CHECK-NEXT:    ldrb w9, [x0, #1]
; CHECK-NEXT:    ptrue p0.d, vl2
; CHECK-NEXT:    fmov s0, w8
; CHECK-NEXT:    mov v0.s[1], w9
; CHECK-NEXT:    cmeq v0.2s, v0.2s, #0
; CHECK-NEXT:    sshll v0.2d, v0.2s, #0
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ldr q0, [x1]
; CHECK-NEXT:    ld1b { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    ptrue p0.s, vl2
; CHECK-NEXT:    xtn v0.2s, v0.2d
; CHECK-NEXT:    st1b { z0.s }, p0, [x0]
; CHECK-NEXT:    ret
  %cval = load <2 x i8>, ptr %a
  %ptrs = load <2 x ptr>, ptr %b
  %mask = icmp eq <2 x i8> %cval, zeroinitializer
  %vals = call <2 x i8> @llvm.masked.gather.v2i8(<2 x ptr> %ptrs, i32 8, <2 x i1> %mask, <2 x i8> undef)
  store <2 x i8> %vals, ptr %a
  ret void
}

define void @masked_gather_v4i8(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v4i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr s0, [x0]
; CHECK-NEXT:    ptrue p0.d, vl4
; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
; CHECK-NEXT:    cmeq v0.4h, v0.4h, #0
; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
; CHECK-NEXT:    sunpklo z0.s, z0.h
; CHECK-NEXT:    sunpklo z0.d, z0.s
; CHECK-NEXT:    cmpne p1.d, p0/z, z0.d, #0
; CHECK-NEXT:    ld1b { z0.d }, p1/z, [z1.d]
; CHECK-NEXT:    st1b { z0.d }, p0, [x0]
; CHECK-NEXT:    ret
  %cval = load <4 x i8>, ptr %a
  %ptrs = load <4 x ptr>, ptr %b
  %mask = icmp eq <4 x i8> %cval, zeroinitializer
  %vals = call <4 x i8> @llvm.masked.gather.v4i8(<4 x ptr> %ptrs, i32 8, <4 x i1> %mask, <4 x i8> undef)
  store <4 x i8> %vals, ptr %a
  ret void
}

define void @masked_gather_v8i8(ptr %a, ptr %b) #0 {
; VBITS_GE_256-LABEL: masked_gather_v8i8:
; VBITS_GE_256:       // %bb.0:
; VBITS_GE_256-NEXT:    ldr d0, [x0]
; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
; VBITS_GE_256-NEXT:    cmeq v0.8b, v0.8b, #0
; VBITS_GE_256-NEXT:    zip2 v1.8b, v0.8b, v0.8b
; VBITS_GE_256-NEXT:    zip1 v0.8b, v0.8b, v0.8b
; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x1]
; VBITS_GE_256-NEXT:    shl v1.4h, v1.4h, #8
; VBITS_GE_256-NEXT:    shl v0.4h, v0.4h, #8
; VBITS_GE_256-NEXT:    sshr v1.4h, v1.4h, #8
; VBITS_GE_256-NEXT:    sshr v0.4h, v0.4h, #8
; VBITS_GE_256-NEXT:    sunpklo z1.s, z1.h
; VBITS_GE_256-NEXT:    sunpklo z0.s, z0.h
; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z1.d, #0
; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; VBITS_GE_256-NEXT:    ld1b { z0.d }, p1/z, [z2.d]
; VBITS_GE_256-NEXT:    ld1b { z1.d }, p0/z, [z3.d]
; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
; VBITS_GE_256-NEXT:    uzp1 v0.8b, v1.8b, v0.8b
; VBITS_GE_256-NEXT:    str d0, [x0]
; VBITS_GE_256-NEXT:    ret
;
; VBITS_GE_512-LABEL: masked_gather_v8i8:
; VBITS_GE_512:       // %bb.0:
; VBITS_GE_512-NEXT:    ldr d0, [x0]
; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
; VBITS_GE_512-NEXT:    cmeq v0.8b, v0.8b, #0
; VBITS_GE_512-NEXT:    sunpklo z0.h, z0.b
; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
; VBITS_GE_512-NEXT:    sunpklo z0.s, z0.h
; VBITS_GE_512-NEXT:    sunpklo z0.d, z0.s
; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; VBITS_GE_512-NEXT:    ld1b { z0.d }, p0/z, [z1.d]
; VBITS_GE_512-NEXT:    uzp1 z0.s, z0.s, z0.s
; VBITS_GE_512-NEXT:    uzp1 z0.h, z0.h, z0.h
; VBITS_GE_512-NEXT:    uzp1 z0.b, z0.b, z0.b
; VBITS_GE_512-NEXT:    str d0, [x0]
; VBITS_GE_512-NEXT:    ret
  %cval = load <8 x i8>, ptr %a
  %ptrs = load <8 x ptr>, ptr %b
  %mask = icmp eq <8 x i8> %cval, zeroinitializer
  %vals = call <8 x i8> @llvm.masked.gather.v8i8(<8 x ptr> %ptrs, i32 8, <8 x i1> %mask, <8 x i8> undef)
  store <8 x i8> %vals, ptr %a
  ret void
}

define void @masked_gather_v16i8(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_v16i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    ptrue p0.d, vl16
; CHECK-NEXT:    cmeq v0.16b, v0.16b, #0
; CHECK-NEXT:    sunpklo z0.h, z0.b
; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
; CHECK-NEXT:    sunpklo z0.s, z0.h
; CHECK-NEXT:    sunpklo z0.d, z0.s
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ld1b { z0.d }, p0/z, [z1.d]
; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT:    uzp1 z0.b, z0.b, z0.b
; CHECK-NEXT:    str q0, [x0]
; CHECK-NEXT:    ret
  %cval = load <16 x i8>, ptr %a
  %ptrs = load <16 x ptr>, ptr %b
  %mask = icmp eq <16 x i8> %cval, zeroinitializer
  %vals = call <16 x i8> @llvm.masked.gather.v16i8(<16 x ptr> %ptrs, i32 8, <16 x i1> %mask, <16 x i8> undef)
  store <16 x i8> %vals, ptr %a
  ret void
}

define void @masked_gather_v32i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_v32i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.b, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1b { z0.b }, p0/z, [x0]
; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1b { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1b { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cval = load <32 x i8>, ptr %a
  %ptrs = load <32 x ptr>, ptr %b
  %mask = icmp eq <32 x i8> %cval, zeroinitializer
  %vals = call <32 x i8> @llvm.masked.gather.v32i8(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x i8> undef)
  store <32 x i8> %vals, ptr %a
  ret void
}

;
; LD1H
;

define void @masked_gather_v2i16(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v2i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldrh w8, [x0]
; CHECK-NEXT:    ldrh w9, [x0, #2]
; CHECK-NEXT:    ptrue p0.d, vl2
; CHECK-NEXT:    fmov s0, w8
; CHECK-NEXT:    mov v0.s[1], w9
; CHECK-NEXT:    cmeq v0.2s, v0.2s, #0
; CHECK-NEXT:    sshll v0.2d, v0.2s, #0
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ldr q0, [x1]
; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    ptrue p0.s, vl2
; CHECK-NEXT:    xtn v0.2s, v0.2d
; CHECK-NEXT:    st1h { z0.s }, p0, [x0]
; CHECK-NEXT:    ret
  %cval = load <2 x i16>, ptr %a
  %ptrs = load <2 x ptr>, ptr %b
  %mask = icmp eq <2 x i16> %cval, zeroinitializer
  %vals = call <2 x i16> @llvm.masked.gather.v2i16(<2 x ptr> %ptrs, i32 8, <2 x i1> %mask, <2 x i16> undef)
  store <2 x i16> %vals, ptr %a
  ret void
}

define void @masked_gather_v4i16(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v4i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    ptrue p0.d, vl4
; CHECK-NEXT:    cmeq v0.4h, v0.4h, #0
; CHECK-NEXT:    sunpklo z0.s, z0.h
; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
; CHECK-NEXT:    sunpklo z0.d, z0.s
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z1.d]
; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT:    str d0, [x0]
; CHECK-NEXT:    ret
  %cval = load <4 x i16>, ptr %a
  %ptrs = load <4 x ptr>, ptr %b
  %mask = icmp eq <4 x i16> %cval, zeroinitializer
  %vals = call <4 x i16> @llvm.masked.gather.v4i16(<4 x ptr> %ptrs, i32 8, <4 x i1> %mask, <4 x i16> undef)
  store <4 x i16> %vals, ptr %a
  ret void
}

define void @masked_gather_v8i16(ptr %a, ptr %b) #0 {
; VBITS_GE_256-LABEL: masked_gather_v8i16:
; VBITS_GE_256:       // %bb.0:
; VBITS_GE_256-NEXT:    ldr q0, [x0]
; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
; VBITS_GE_256-NEXT:    cmeq v0.8h, v0.8h, #0
; VBITS_GE_256-NEXT:    ext v1.16b, v0.16b, v0.16b, #8
; VBITS_GE_256-NEXT:    sunpklo z0.s, z0.h
; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x1]
; VBITS_GE_256-NEXT:    sunpklo z1.s, z1.h
; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z0.d, #0
; VBITS_GE_256-NEXT:    ld1h { z0.d }, p1/z, [z3.d]
; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z1.d, #0
; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
; VBITS_GE_256-NEXT:    ld1h { z1.d }, p0/z, [z2.d]
; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
; VBITS_GE_256-NEXT:    mov v0.d[1], v1.d[0]
; VBITS_GE_256-NEXT:    str q0, [x0]
; VBITS_GE_256-NEXT:    ret
;
; VBITS_GE_512-LABEL: masked_gather_v8i16:
; VBITS_GE_512:       // %bb.0:
; VBITS_GE_512-NEXT:    ldr q0, [x0]
; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
; VBITS_GE_512-NEXT:    cmeq v0.8h, v0.8h, #0
; VBITS_GE_512-NEXT:    sunpklo z0.s, z0.h
; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
; VBITS_GE_512-NEXT:    sunpklo z0.d, z0.s
; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; VBITS_GE_512-NEXT:    ld1h { z0.d }, p0/z, [z1.d]
; VBITS_GE_512-NEXT:    uzp1 z0.s, z0.s, z0.s
; VBITS_GE_512-NEXT:    uzp1 z0.h, z0.h, z0.h
; VBITS_GE_512-NEXT:    str q0, [x0]
; VBITS_GE_512-NEXT:    ret
  %cval = load <8 x i16>, ptr %a
  %ptrs = load <8 x ptr>, ptr %b
  %mask = icmp eq <8 x i16> %cval, zeroinitializer
  %vals = call <8 x i16> @llvm.masked.gather.v8i16(<8 x ptr> %ptrs, i32 8, <8 x i1> %mask, <8 x i16> undef)
  store <8 x i16> %vals, ptr %a
  ret void
}

define void @masked_gather_v16i16(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_v16i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl16
; CHECK-NEXT:    ptrue p1.d, vl16
; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1h { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cval = load <16 x i16>, ptr %a
  %ptrs = load <16 x ptr>, ptr %b
  %mask = icmp eq <16 x i16> %cval, zeroinitializer
  %vals = call <16 x i16> @llvm.masked.gather.v16i16(<16 x ptr> %ptrs, i32 8, <16 x i1> %mask, <16 x i16> undef)
  store <16 x i16> %vals, ptr %a
  ret void
}

define void @masked_gather_v32i16(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_v32i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1h { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cval = load <32 x i16>, ptr %a
  %ptrs = load <32 x ptr>, ptr %b
  %mask = icmp eq <32 x i16> %cval, zeroinitializer
  %vals = call <32 x i16> @llvm.masked.gather.v32i16(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x i16> undef)
  store <32 x i16> %vals, ptr %a
  ret void
}

;
; LD1W
;

define void @masked_gather_v2i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v2i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    ptrue p0.d, vl2
; CHECK-NEXT:    cmeq v0.2s, v0.2s, #0
; CHECK-NEXT:    sshll v0.2d, v0.2s, #0
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ldr q0, [x1]
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    xtn v0.2s, v0.2d
; CHECK-NEXT:    str d0, [x0]
; CHECK-NEXT:    ret
  %cval = load <2 x i32>, ptr %a
  %ptrs = load <2 x ptr>, ptr %b
  %mask = icmp eq <2 x i32> %cval, zeroinitializer
  %vals = call <2 x i32> @llvm.masked.gather.v2i32(<2 x ptr> %ptrs, i32 8, <2 x i1> %mask, <2 x i32> undef)
  store <2 x i32> %vals, ptr %a
  ret void
}

define void @masked_gather_v4i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v4i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    ptrue p0.d, vl4
; CHECK-NEXT:    cmeq v0.4s, v0.4s, #0
; CHECK-NEXT:    sunpklo z0.d, z0.s
; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z1.d]
; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
; CHECK-NEXT:    str q0, [x0]
; CHECK-NEXT:    ret
  %cval = load <4 x i32>, ptr %a
  %ptrs = load <4 x ptr>, ptr %b
  %mask = icmp eq <4 x i32> %cval, zeroinitializer
  %vals = call <4 x i32> @llvm.masked.gather.v4i32(<4 x ptr> %ptrs, i32 8, <4 x i1> %mask, <4 x i32> undef)
  store <4 x i32> %vals, ptr %a
  ret void
}

define void @masked_gather_v8i32(ptr %a, ptr %b) #0 {
; VBITS_GE_256-LABEL: masked_gather_v8i32:
; VBITS_GE_256:       // %bb.0:
; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
; VBITS_GE_256-NEXT:    ptrue p2.d, vl4
; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0]
; VBITS_GE_256-NEXT:    ld1d { z1.d }, p2/z, [x1, x8, lsl #3]
; VBITS_GE_256-NEXT:    ld1d { z2.d }, p2/z, [x1]
; VBITS_GE_256-NEXT:    cmpeq p1.s, p0/z, z0.s, #0
; VBITS_GE_256-NEXT:    mov z0.s, p1/z, #-1 // =0xffffffffffffffff
; VBITS_GE_256-NEXT:    punpklo p1.h, p1.b
; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
; VBITS_GE_256-NEXT:    and p1.b, p1/z, p1.b, p2.b
; VBITS_GE_256-NEXT:    cmpne p2.d, p2/z, z0.d, #0
; VBITS_GE_256-NEXT:    ld1w { z0.d }, p1/z, [z2.d]
; VBITS_GE_256-NEXT:    ptrue p1.s, vl4
; VBITS_GE_256-NEXT:    ld1w { z1.d }, p2/z, [z1.d]
; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
; VBITS_GE_256-NEXT:    splice z0.s, p1, z0.s, z1.s
; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0]
; VBITS_GE_256-NEXT:    ret
;
; VBITS_GE_512-LABEL: masked_gather_v8i32:
; VBITS_GE_512:       // %bb.0:
; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
; VBITS_GE_512-NEXT:    ptrue p1.d, vl8
; VBITS_GE_512-NEXT:    ld1w { z0.s }, p0/z, [x0]
; VBITS_GE_512-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
; VBITS_GE_512-NEXT:    ld1d { z0.d }, p1/z, [x1]
; VBITS_GE_512-NEXT:    punpklo p0.h, p0.b
; VBITS_GE_512-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
; VBITS_GE_512-NEXT:    st1w { z0.d }, p1, [x0]
; VBITS_GE_512-NEXT:    ret
  %cval = load <8 x i32>, ptr %a
  %ptrs = load <8 x ptr>, ptr %b
  %mask = icmp eq <8 x i32> %cval, zeroinitializer
  %vals = call <8 x i32> @llvm.masked.gather.v8i32(<8 x ptr> %ptrs, i32 8, <8 x i1> %mask, <8 x i32> undef)
  store <8 x i32> %vals, ptr %a
  ret void
}

define void @masked_gather_v16i32(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_v16i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl16
; CHECK-NEXT:    ptrue p1.d, vl16
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1w { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cval = load <16 x i32>, ptr %a
  %ptrs = load <16 x ptr>, ptr %b
  %mask = icmp eq <16 x i32> %cval, zeroinitializer
  %vals = call <16 x i32> @llvm.masked.gather.v16i32(<16 x ptr> %ptrs, i32 8, <16 x i1> %mask, <16 x i32> undef)
  store <16 x i32> %vals, ptr %a
  ret void
}

define void @masked_gather_v32i32(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_v32i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1w { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cval = load <32 x i32>, ptr %a
  %ptrs = load <32 x ptr>, ptr %b
  %mask = icmp eq <32 x i32> %cval, zeroinitializer
  %vals = call <32 x i32> @llvm.masked.gather.v32i32(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x i32> undef)
  store <32 x i32> %vals, ptr %a
  ret void
}

;
; LD1D
;

; Scalarize 1 x i64 gathers
define void @masked_gather_v1i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v1i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    fmov x8, d0
; CHECK-NEXT:    // implicit-def: $d0
; CHECK-NEXT:    cbnz x8, .LBB15_2
; CHECK-NEXT:  // %bb.1: // %cond.load
; CHECK-NEXT:    ldr d0, [x1]
; CHECK-NEXT:    fmov x8, d0
; CHECK-NEXT:    ldr d0, [x8]
; CHECK-NEXT:  .LBB15_2: // %else
; CHECK-NEXT:    str d0, [x0]
; CHECK-NEXT:    ret
  %cval = load <1 x i64>, ptr %a
  %ptrs = load <1 x ptr>, ptr %b
  %mask = icmp eq <1 x i64> %cval, zeroinitializer
  %vals = call <1 x i64> @llvm.masked.gather.v1i64(<1 x ptr> %ptrs, i32 8, <1 x i1> %mask, <1 x i64> undef)
  store <1 x i64> %vals, ptr %a
  ret void
}

define void @masked_gather_v2i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v2i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl2
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    cmeq v0.2d, v0.2d, #0
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ldr q0, [x1]
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    str q0, [x0]
; CHECK-NEXT:    ret
  %cval = load <2 x i64>, ptr %a
  %ptrs = load <2 x ptr>, ptr %b
  %mask = icmp eq <2 x i64> %cval, zeroinitializer
  %vals = call <2 x i64> @llvm.masked.gather.v2i64(<2 x ptr> %ptrs, i32 8, <2 x i1> %mask, <2 x i64> undef)
  store <2 x i64> %vals, ptr %a
  ret void
}

define void @masked_gather_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v4i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl4
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z1.d]
; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
; CHECK-NEXT:    ret
  %cval = load <4 x i64>, ptr %a
  %ptrs = load <4 x ptr>, ptr %b
  %mask = icmp eq <4 x i64> %cval, zeroinitializer
  %vals = call <4 x i64> @llvm.masked.gather.v4i64(<4 x ptr> %ptrs, i32 8, <4 x i1> %mask, <4 x i64> undef)
  store <4 x i64> %vals, ptr %a
  ret void
}

define void @masked_gather_v8i64(ptr %a, ptr %b) #0 {
; VBITS_GE_256-LABEL: masked_gather_v8i64:
; VBITS_GE_256:       // %bb.0:
; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
; VBITS_GE_256-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x1]
; VBITS_GE_256-NEXT:    cmpeq p2.d, p0/z, z1.d, #0
; VBITS_GE_256-NEXT:    ld1d { z1.d }, p1/z, [z2.d]
; VBITS_GE_256-NEXT:    ld1d { z0.d }, p2/z, [z0.d]
; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0, x8, lsl #3]
; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0]
; VBITS_GE_256-NEXT:    ret
;
; VBITS_GE_512-LABEL: masked_gather_v8i64:
; VBITS_GE_512:       // %bb.0:
; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x0]
; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
; VBITS_GE_512-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
; VBITS_GE_512-NEXT:    ld1d { z0.d }, p1/z, [z1.d]
; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [x0]
; VBITS_GE_512-NEXT:    ret
  %cval = load <8 x i64>, ptr %a
  %ptrs = load <8 x ptr>, ptr %b
  %mask = icmp eq <8 x i64> %cval, zeroinitializer
  %vals = call <8 x i64> @llvm.masked.gather.v8i64(<8 x ptr> %ptrs, i32 8, <8 x i1> %mask, <8 x i64> undef)
  store <8 x i64> %vals, ptr %a
  ret void
}

define void @masked_gather_v16i64(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_v16i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl16
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z1.d]
; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
; CHECK-NEXT:    ret
  %cval = load <16 x i64>, ptr %a
  %ptrs = load <16 x ptr>, ptr %b
  %mask = icmp eq <16 x i64> %cval, zeroinitializer
  %vals = call <16 x i64> @llvm.masked.gather.v16i64(<16 x ptr> %ptrs, i32 8, <16 x i1> %mask, <16 x i64> undef)
  store <16 x i64> %vals, ptr %a
  ret void
}

define void @masked_gather_v32i64(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_v32i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl32
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z1.d]
; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
; CHECK-NEXT:    ret
  %cval = load <32 x i64>, ptr %a
  %ptrs = load <32 x ptr>, ptr %b
  %mask = icmp eq <32 x i64> %cval, zeroinitializer
  %vals = call <32 x i64> @llvm.masked.gather.v32i64(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x i64> undef)
  store <32 x i64> %vals, ptr %a
  ret void
}

;
; LD1H (float)
;

define void @masked_gather_v2f16(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v2f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr s1, [x0]
; CHECK-NEXT:    movi v0.2d, #0000000000000000
; CHECK-NEXT:    ptrue p0.d, vl4
; CHECK-NEXT:    fcmeq v1.4h, v1.4h, #0.0
; CHECK-NEXT:    sshll v1.4s, v1.4h, #0
; CHECK-NEXT:    mov v0.h[0], v1.h[0]
; CHECK-NEXT:    mov w8, v1.s[1]
; CHECK-NEXT:    mov v0.h[1], w8
; CHECK-NEXT:    sunpklo z0.s, z0.h
; CHECK-NEXT:    sunpklo z0.d, z0.s
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ldr q0, [x1]
; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT:    str s0, [x0]
; CHECK-NEXT:    ret
  %cval = load <2 x half>, ptr %a
  %ptrs = load <2 x ptr>, ptr %b
  %mask = fcmp oeq <2 x half> %cval, zeroinitializer
  %vals = call <2 x half> @llvm.masked.gather.v2f16(<2 x ptr> %ptrs, i32 8, <2 x i1> %mask, <2 x half> undef)
  store <2 x half> %vals, ptr %a
  ret void
}

define void @masked_gather_v4f16(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v4f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    ptrue p0.d, vl4
; CHECK-NEXT:    fcmeq v0.4h, v0.4h, #0.0
; CHECK-NEXT:    sunpklo z0.s, z0.h
; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
; CHECK-NEXT:    sunpklo z0.d, z0.s
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z1.d]
; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT:    str d0, [x0]
; CHECK-NEXT:    ret
  %cval = load <4 x half>, ptr %a
  %ptrs = load <4 x ptr>, ptr %b
  %mask = fcmp oeq <4 x half> %cval, zeroinitializer
  %vals = call <4 x half> @llvm.masked.gather.v4f16(<4 x ptr> %ptrs, i32 8, <4 x i1> %mask, <4 x half> undef)
  store <4 x half> %vals, ptr %a
  ret void
}

define void @masked_gather_v8f16(ptr %a, ptr %b) #0 {
; VBITS_GE_256-LABEL: masked_gather_v8f16:
; VBITS_GE_256:       // %bb.0:
; VBITS_GE_256-NEXT:    ldr q0, [x0]
; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
; VBITS_GE_256-NEXT:    fcmeq v0.8h, v0.8h, #0.0
; VBITS_GE_256-NEXT:    ext v1.16b, v0.16b, v0.16b, #8
; VBITS_GE_256-NEXT:    sunpklo z0.s, z0.h
; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x1]
; VBITS_GE_256-NEXT:    sunpklo z1.s, z1.h
; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z0.d, #0
; VBITS_GE_256-NEXT:    ld1h { z0.d }, p1/z, [z3.d]
; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z1.d, #0
; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
; VBITS_GE_256-NEXT:    ld1h { z1.d }, p0/z, [z2.d]
; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
; VBITS_GE_256-NEXT:    mov v0.d[1], v1.d[0]
; VBITS_GE_256-NEXT:    str q0, [x0]
; VBITS_GE_256-NEXT:    ret
;
; VBITS_GE_512-LABEL: masked_gather_v8f16:
; VBITS_GE_512:       // %bb.0:
; VBITS_GE_512-NEXT:    ldr q0, [x0]
; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
; VBITS_GE_512-NEXT:    fcmeq v0.8h, v0.8h, #0.0
; VBITS_GE_512-NEXT:    sunpklo z0.s, z0.h
; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
; VBITS_GE_512-NEXT:    sunpklo z0.d, z0.s
; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; VBITS_GE_512-NEXT:    ld1h { z0.d }, p0/z, [z1.d]
; VBITS_GE_512-NEXT:    uzp1 z0.s, z0.s, z0.s
; VBITS_GE_512-NEXT:    uzp1 z0.h, z0.h, z0.h
; VBITS_GE_512-NEXT:    str q0, [x0]
; VBITS_GE_512-NEXT:    ret
  %cval = load <8 x half>, ptr %a
  %ptrs = load <8 x ptr>, ptr %b
  %mask = fcmp oeq <8 x half> %cval, zeroinitializer
  %vals = call <8 x half> @llvm.masked.gather.v8f16(<8 x ptr> %ptrs, i32 8, <8 x i1> %mask, <8 x half> undef)
  store <8 x half> %vals, ptr %a
  ret void
}

define void @masked_gather_v16f16(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_v16f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl16
; CHECK-NEXT:    ptrue p1.d, vl16
; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1h { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cval = load <16 x half>, ptr %a
  %ptrs = load <16 x ptr>, ptr %b
  %mask = fcmp oeq <16 x half> %cval, zeroinitializer
  %vals = call <16 x half> @llvm.masked.gather.v16f16(<16 x ptr> %ptrs, i32 8, <16 x i1> %mask, <16 x half> undef)
  store <16 x half> %vals, ptr %a
  ret void
}

define void @masked_gather_v32f16(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_v32f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1h { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cval = load <32 x half>, ptr %a
  %ptrs = load <32 x ptr>, ptr %b
  %mask = fcmp oeq <32 x half> %cval, zeroinitializer
  %vals = call <32 x half> @llvm.masked.gather.v32f16(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x half> undef)
  store <32 x half> %vals, ptr %a
  ret void
}

;
; LD1W (float)
;

define void @masked_gather_v2f32(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v2f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    ptrue p0.d, vl2
; CHECK-NEXT:    fcmeq v0.2s, v0.2s, #0.0
; CHECK-NEXT:    sshll v0.2d, v0.2s, #0
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ldr q0, [x1]
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    xtn v0.2s, v0.2d
; CHECK-NEXT:    str d0, [x0]
; CHECK-NEXT:    ret
  %cval = load <2 x float>, ptr %a
  %ptrs = load <2 x ptr>, ptr %b
  %mask = fcmp oeq <2 x float> %cval, zeroinitializer
  %vals = call <2 x float> @llvm.masked.gather.v2f32(<2 x ptr> %ptrs, i32 8, <2 x i1> %mask, <2 x float> undef)
  store <2 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_v4f32(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v4f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    ptrue p0.d, vl4
; CHECK-NEXT:    fcmeq v0.4s, v0.4s, #0.0
; CHECK-NEXT:    sunpklo z0.d, z0.s
; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z1.d]
; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
; CHECK-NEXT:    str q0, [x0]
; CHECK-NEXT:    ret
  %cval = load <4 x float>, ptr %a
  %ptrs = load <4 x ptr>, ptr %b
  %mask = fcmp oeq <4 x float> %cval, zeroinitializer
  %vals = call <4 x float> @llvm.masked.gather.v4f32(<4 x ptr> %ptrs, i32 8, <4 x i1> %mask, <4 x float> undef)
  store <4 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_v8f32(ptr %a, ptr %b) #0 {
; VBITS_GE_256-LABEL: masked_gather_v8f32:
; VBITS_GE_256:       // %bb.0:
; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
; VBITS_GE_256-NEXT:    ptrue p2.d, vl4
; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0]
; VBITS_GE_256-NEXT:    ld1d { z1.d }, p2/z, [x1, x8, lsl #3]
; VBITS_GE_256-NEXT:    ld1d { z2.d }, p2/z, [x1]
; VBITS_GE_256-NEXT:    fcmeq p1.s, p0/z, z0.s, #0.0
; VBITS_GE_256-NEXT:    mov z0.s, p1/z, #-1 // =0xffffffffffffffff
; VBITS_GE_256-NEXT:    punpklo p1.h, p1.b
; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
; VBITS_GE_256-NEXT:    and p1.b, p1/z, p1.b, p2.b
; VBITS_GE_256-NEXT:    cmpne p2.d, p2/z, z0.d, #0
; VBITS_GE_256-NEXT:    ld1w { z0.d }, p1/z, [z2.d]
; VBITS_GE_256-NEXT:    ptrue p1.s, vl4
; VBITS_GE_256-NEXT:    ld1w { z1.d }, p2/z, [z1.d]
; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
; VBITS_GE_256-NEXT:    splice z0.s, p1, z0.s, z1.s
; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0]
; VBITS_GE_256-NEXT:    ret
;
; VBITS_GE_512-LABEL: masked_gather_v8f32:
; VBITS_GE_512:       // %bb.0:
; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
; VBITS_GE_512-NEXT:    ptrue p1.d, vl8
; VBITS_GE_512-NEXT:    ld1w { z0.s }, p0/z, [x0]
; VBITS_GE_512-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
; VBITS_GE_512-NEXT:    ld1d { z0.d }, p1/z, [x1]
; VBITS_GE_512-NEXT:    punpklo p0.h, p0.b
; VBITS_GE_512-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
; VBITS_GE_512-NEXT:    st1w { z0.d }, p1, [x0]
; VBITS_GE_512-NEXT:    ret
  %cval = load <8 x float>, ptr %a
  %ptrs = load <8 x ptr>, ptr %b
  %mask = fcmp oeq <8 x float> %cval, zeroinitializer
  %vals = call <8 x float> @llvm.masked.gather.v8f32(<8 x ptr> %ptrs, i32 8, <8 x i1> %mask, <8 x float> undef)
  store <8 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_v16f32(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_v16f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl16
; CHECK-NEXT:    ptrue p1.d, vl16
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1w { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cval = load <16 x float>, ptr %a
  %ptrs = load <16 x ptr>, ptr %b
  %mask = fcmp oeq <16 x float> %cval, zeroinitializer
  %vals = call <16 x float> @llvm.masked.gather.v16f32(<16 x ptr> %ptrs, i32 8, <16 x i1> %mask, <16 x float> undef)
  store <16 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_v32f32(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_v32f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1w { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cval = load <32 x float>, ptr %a
  %ptrs = load <32 x ptr>, ptr %b
  %mask = fcmp oeq <32 x float> %cval, zeroinitializer
  %vals = call <32 x float> @llvm.masked.gather.v32f32(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x float> undef)
  store <32 x float> %vals, ptr %a
  ret void
}

;
; LD1D (float)
;

; Scalarize 1 x double gathers
define void @masked_gather_v1f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v1f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    fcmp d0, #0.0
; CHECK-NEXT:    // implicit-def: $d0
; CHECK-NEXT:    b.ne .LBB31_2
; CHECK-NEXT:  // %bb.1: // %cond.load
; CHECK-NEXT:    ldr d0, [x1]
; CHECK-NEXT:    fmov x8, d0
; CHECK-NEXT:    ldr d0, [x8]
; CHECK-NEXT:  .LBB31_2: // %else
; CHECK-NEXT:    str d0, [x0]
; CHECK-NEXT:    ret
  %cval = load <1 x double>, ptr %a
  %ptrs = load <1 x ptr>, ptr %b
  %mask = fcmp oeq <1 x double> %cval, zeroinitializer
  %vals = call <1 x double> @llvm.masked.gather.v1f64(<1 x ptr> %ptrs, i32 8, <1 x i1> %mask, <1 x double> undef)
  store <1 x double> %vals, ptr %a
  ret void
}

define void @masked_gather_v2f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v2f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl2
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    fcmeq v0.2d, v0.2d, #0.0
; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT:    ldr q0, [x1]
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    str q0, [x0]
; CHECK-NEXT:    ret
  %cval = load <2 x double>, ptr %a
  %ptrs = load <2 x ptr>, ptr %b
  %mask = fcmp oeq <2 x double> %cval, zeroinitializer
  %vals = call <2 x double> @llvm.masked.gather.v2f64(<2 x ptr> %ptrs, i32 8, <2 x i1> %mask, <2 x double> undef)
  store <2 x double> %vals, ptr %a
  ret void
}

define void @masked_gather_v4f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
; CHECK-LABEL: masked_gather_v4f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl4
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p1.d, p0/z, z0.d, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
; CHECK-NEXT:    ret
  %cval = load <4 x double>, ptr %a
  %ptrs = load <4 x ptr>, ptr %b
  %mask = fcmp oeq <4 x double> %cval, zeroinitializer
  %vals = call <4 x double> @llvm.masked.gather.v4f64(<4 x ptr> %ptrs, i32 8, <4 x i1> %mask, <4 x double> undef)
  store <4 x double> %vals, ptr %a
  ret void
}

define void @masked_gather_v8f64(ptr %a, ptr %b) #0 {
; VBITS_GE_256-LABEL: masked_gather_v8f64:
; VBITS_GE_256:       // %bb.0:
; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
; VBITS_GE_256-NEXT:    fcmeq p1.d, p0/z, z0.d, #0.0
; VBITS_GE_256-NEXT:    fcmeq p2.d, p0/z, z1.d, #0.0
; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x1, x8, lsl #3]
; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1]
; VBITS_GE_256-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
; VBITS_GE_256-NEXT:    ld1d { z1.d }, p2/z, [z1.d]
; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
; VBITS_GE_256-NEXT:    ret
;
; VBITS_GE_512-LABEL: masked_gather_v8f64:
; VBITS_GE_512:       // %bb.0:
; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x0]
; VBITS_GE_512-NEXT:    fcmeq p1.d, p0/z, z0.d, #0.0
; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x1]
; VBITS_GE_512-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [x0]
; VBITS_GE_512-NEXT:    ret
  %cval = load <8 x double>, ptr %a
  %ptrs = load <8 x ptr>, ptr %b
  %mask = fcmp oeq <8 x double> %cval, zeroinitializer
  %vals = call <8 x double> @llvm.masked.gather.v8f64(<8 x ptr> %ptrs, i32 8, <8 x i1> %mask, <8 x double> undef)
  store <8 x double> %vals, ptr %a
  ret void
}

define void @masked_gather_v16f64(ptr %a, ptr %b) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_v16f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl16
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p1.d, p0/z, z0.d, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
; CHECK-NEXT:    ret
  %cval = load <16 x double>, ptr %a
  %ptrs = load <16 x ptr>, ptr %b
  %mask = fcmp oeq <16 x double> %cval, zeroinitializer
  %vals = call <16 x double> @llvm.masked.gather.v16f64(<16 x ptr> %ptrs, i32 8, <16 x i1> %mask, <16 x double> undef)
  store <16 x double> %vals, ptr %a
  ret void
}

define void @masked_gather_v32f64(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_v32f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl32
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p1.d, p0/z, z0.d, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
; CHECK-NEXT:    ret
  %cval = load <32 x double>, ptr %a
  %ptrs = load <32 x ptr>, ptr %b
  %mask = fcmp oeq <32 x double> %cval, zeroinitializer
  %vals = call <32 x double> @llvm.masked.gather.v32f64(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x double> undef)
  store <32 x double> %vals, ptr %a
  ret void
}

; The above tests test the types, the below tests check that the addressing
; modes still function

define void @masked_gather_32b_scaled_sext_f16(ptr %a, ptr %b, ptr %base) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_32b_scaled_sext_f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl32
; CHECK-NEXT:    ptrue p1.s, vl32
; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x2, z0.s, sxtw #1]
; CHECK-NEXT:    st1h { z0.s }, p1, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x half>, ptr %a
  %idxs = load <32 x i32>, ptr %b
  %ext = sext <32 x i32> %idxs to <32 x i64>
  %ptrs = getelementptr half, ptr %base, <32 x i64> %ext
  %mask = fcmp oeq <32 x half> %cvals, zeroinitializer
  %vals = call <32 x half> @llvm.masked.gather.v32f16(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x half> undef)
  store <32 x half> %vals, ptr %a
  ret void
}

define void @masked_gather_32b_scaled_sext_f32(ptr %a, ptr %b, ptr %base) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_32b_scaled_sext_f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p1.s, p0/z, z0.s, #0.0
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x1]
; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x2, z0.s, sxtw #2]
; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x float>, ptr %a
  %idxs = load <32 x i32>, ptr %b
  %ext = sext <32 x i32> %idxs to <32 x i64>
  %ptrs = getelementptr float, ptr %base, <32 x i64> %ext
  %mask = fcmp oeq <32 x float> %cvals, zeroinitializer
  %vals = call <32 x float> @llvm.masked.gather.v32f32(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x float> undef)
  store <32 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_32b_scaled_sext_f64(ptr %a, ptr %b, ptr %base) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_32b_scaled_sext_f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl32
; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p1.d, p0/z, z0.d, #0.0
; CHECK-NEXT:    ld1sw { z0.d }, p0/z, [x1]
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x2, z0.d, lsl #3]
; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x double>, ptr %a
  %idxs = load <32 x i32>, ptr %b
  %ext = sext <32 x i32> %idxs to <32 x i64>
  %ptrs = getelementptr double, ptr %base, <32 x i64> %ext
  %mask = fcmp oeq <32 x double> %cvals, zeroinitializer
  %vals = call <32 x double> @llvm.masked.gather.v32f64(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x double> undef)
  store <32 x double> %vals, ptr %a
  ret void
}

define void @masked_gather_32b_scaled_zext(ptr %a, ptr %b, ptr %base) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_32b_scaled_zext:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl32
; CHECK-NEXT:    ptrue p1.s, vl32
; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x2, z0.s, uxtw #1]
; CHECK-NEXT:    st1h { z0.s }, p1, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x half>, ptr %a
  %idxs = load <32 x i32>, ptr %b
  %ext = zext <32 x i32> %idxs to <32 x i64>
  %ptrs = getelementptr half, ptr %base, <32 x i64> %ext
  %mask = fcmp oeq <32 x half> %cvals, zeroinitializer
  %vals = call <32 x half> @llvm.masked.gather.v32f16(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x half> undef)
  store <32 x half> %vals, ptr %a
  ret void
}

define void @masked_gather_32b_unscaled_sext(ptr %a, ptr %b, ptr %base) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_32b_unscaled_sext:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl32
; CHECK-NEXT:    ptrue p1.s, vl32
; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x2, z0.s, sxtw]
; CHECK-NEXT:    st1h { z0.s }, p1, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x half>, ptr %a
  %idxs = load <32 x i32>, ptr %b
  %ext = sext <32 x i32> %idxs to <32 x i64>
  %byte_ptrs = getelementptr i8, ptr %base, <32 x i64> %ext
  %ptrs = bitcast <32 x ptr> %byte_ptrs to <32 x ptr>
  %mask = fcmp oeq <32 x half> %cvals, zeroinitializer
  %vals = call <32 x half> @llvm.masked.gather.v32f16(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x half> undef)
  store <32 x half> %vals, ptr %a
  ret void
}

define void @masked_gather_32b_unscaled_zext(ptr %a, ptr %b, ptr %base) vscale_range(8,0) #0 {
; CHECK-LABEL: masked_gather_32b_unscaled_zext:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl32
; CHECK-NEXT:    ptrue p1.s, vl32
; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x2, z0.s, uxtw]
; CHECK-NEXT:    st1h { z0.s }, p1, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x half>, ptr %a
  %idxs = load <32 x i32>, ptr %b
  %ext = zext <32 x i32> %idxs to <32 x i64>
  %byte_ptrs = getelementptr i8, ptr %base, <32 x i64> %ext
  %ptrs = bitcast <32 x ptr> %byte_ptrs to <32 x ptr>
  %mask = fcmp oeq <32 x half> %cvals, zeroinitializer
  %vals = call <32 x half> @llvm.masked.gather.v32f16(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x half> undef)
  store <32 x half> %vals, ptr %a
  ret void
}

define void @masked_gather_64b_scaled(ptr %a, ptr %b, ptr %base) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_64b_scaled:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [x2, z0.d, lsl #2]
; CHECK-NEXT:    st1w { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x float>, ptr %a
  %idxs = load <32 x i64>, ptr %b
  %ptrs = getelementptr float, ptr %base, <32 x i64> %idxs
  %mask = fcmp oeq <32 x float> %cvals, zeroinitializer
  %vals = call <32 x float> @llvm.masked.gather.v32f32(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x float> undef)
  store <32 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_64b_unscaled(ptr %a, ptr %b, ptr %base) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_64b_unscaled:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [x2, z0.d]
; CHECK-NEXT:    st1w { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x float>, ptr %a
  %idxs = load <32 x i64>, ptr %b
  %byte_ptrs = getelementptr i8, ptr %base, <32 x i64> %idxs
  %ptrs = bitcast <32 x ptr> %byte_ptrs to <32 x ptr>
  %mask = fcmp oeq <32 x float> %cvals, zeroinitializer
  %vals = call <32 x float> @llvm.masked.gather.v32f32(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x float> undef)
  store <32 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_vec_plus_reg(ptr %a, ptr %b, i64 %off) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_vec_plus_reg:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [x2, z0.d]
; CHECK-NEXT:    st1w { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x float>, ptr %a
  %bases = load <32 x ptr>, ptr %b
  %byte_ptrs = getelementptr i8, <32 x ptr> %bases, i64 %off
  %ptrs = bitcast <32 x ptr> %byte_ptrs to <32 x ptr>
  %mask = fcmp oeq <32 x float> %cvals, zeroinitializer
  %vals = call <32 x float> @llvm.masked.gather.v32f32(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x float> undef)
  store <32 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_vec_plus_imm(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_vec_plus_imm:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z0.d, #4]
; CHECK-NEXT:    st1w { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x float>, ptr %a
  %bases = load <32 x ptr>, ptr %b
  %byte_ptrs = getelementptr i8, <32 x ptr> %bases, i64 4
  %ptrs = bitcast <32 x ptr> %byte_ptrs to <32 x ptr>
  %mask = fcmp oeq <32 x float> %cvals, zeroinitializer
  %vals = call <32 x float> @llvm.masked.gather.v32f32(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x float> undef)
  store <32 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_passthru(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_passthru:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ptrue p2.d, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x2]
; CHECK-NEXT:    fcmeq p1.s, p0/z, z0.s, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p2/z, [x1]
; CHECK-NEXT:    punpklo p3.h, p1.b
; CHECK-NEXT:    ld1w { z0.d }, p3/z, [z0.d]
; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
; CHECK-NEXT:    sel z0.s, p1, z0.s, z1.s
; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x float>, ptr %a
  %ptrs = load <32 x ptr>, ptr %b
  %passthru = load <32 x float>, ptr %c
  %mask = fcmp oeq <32 x float> %cvals, zeroinitializer
  %vals = call <32 x float> @llvm.masked.gather.v32f32(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x float> %passthru)
  store <32 x float> %vals, ptr %a
  ret void
}

define void @masked_gather_passthru_0(ptr %a, ptr %b) vscale_range(16,0) #0 {
; CHECK-LABEL: masked_gather_passthru_0:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ptrue p1.d, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x1]
; CHECK-NEXT:    punpklo p0.h, p0.b
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
; CHECK-NEXT:    st1w { z0.d }, p1, [x0]
; CHECK-NEXT:    ret
  %cvals = load <32 x float>, ptr %a
  %ptrs = load <32 x ptr>, ptr %b
  %mask = fcmp oeq <32 x float> %cvals, zeroinitializer
  %vals = call <32 x float> @llvm.masked.gather.v32f32(<32 x ptr> %ptrs, i32 8, <32 x i1> %mask, <32 x float> zeroinitializer)
  store <32 x float> %vals, ptr %a
  ret void
}

declare <2 x i8> @llvm.masked.gather.v2i8(<2 x ptr>, i32, <2 x i1>, <2 x i8>)
declare <4 x i8> @llvm.masked.gather.v4i8(<4 x ptr>, i32, <4 x i1>, <4 x i8>)
declare <8 x i8> @llvm.masked.gather.v8i8(<8 x ptr>, i32, <8 x i1>, <8 x i8>)
declare <16 x i8> @llvm.masked.gather.v16i8(<16 x ptr>, i32, <16 x i1>, <16 x i8>)
declare <32 x i8> @llvm.masked.gather.v32i8(<32 x ptr>, i32, <32 x i1>, <32 x i8>)

declare <2 x i16> @llvm.masked.gather.v2i16(<2 x ptr>, i32, <2 x i1>, <2 x i16>)
declare <4 x i16> @llvm.masked.gather.v4i16(<4 x ptr>, i32, <4 x i1>, <4 x i16>)
declare <8 x i16> @llvm.masked.gather.v8i16(<8 x ptr>, i32, <8 x i1>, <8 x i16>)
declare <16 x i16> @llvm.masked.gather.v16i16(<16 x ptr>, i32, <16 x i1>, <16 x i16>)
declare <32 x i16> @llvm.masked.gather.v32i16(<32 x ptr>, i32, <32 x i1>, <32 x i16>)

declare <2 x i32> @llvm.masked.gather.v2i32(<2 x ptr>, i32, <2 x i1>, <2 x i32>)
declare <4 x i32> @llvm.masked.gather.v4i32(<4 x ptr>, i32, <4 x i1>, <4 x i32>)
declare <8 x i32> @llvm.masked.gather.v8i32(<8 x ptr>, i32, <8 x i1>, <8 x i32>)
declare <16 x i32> @llvm.masked.gather.v16i32(<16 x ptr>, i32, <16 x i1>, <16 x i32>)
declare <32 x i32> @llvm.masked.gather.v32i32(<32 x ptr>, i32, <32 x i1>, <32 x i32>)

declare <1 x i64> @llvm.masked.gather.v1i64(<1 x ptr>, i32, <1 x i1>, <1 x i64>)
declare <2 x i64> @llvm.masked.gather.v2i64(<2 x ptr>, i32, <2 x i1>, <2 x i64>)
declare <4 x i64> @llvm.masked.gather.v4i64(<4 x ptr>, i32, <4 x i1>, <4 x i64>)
declare <8 x i64> @llvm.masked.gather.v8i64(<8 x ptr>, i32, <8 x i1>, <8 x i64>)
declare <16 x i64> @llvm.masked.gather.v16i64(<16 x ptr>, i32, <16 x i1>, <16 x i64>)
declare <32 x i64> @llvm.masked.gather.v32i64(<32 x ptr>, i32, <32 x i1>, <32 x i64>)

declare <2 x half> @llvm.masked.gather.v2f16(<2 x ptr>, i32, <2 x i1>, <2 x half>)
declare <4 x half> @llvm.masked.gather.v4f16(<4 x ptr>, i32, <4 x i1>, <4 x half>)
declare <8 x half> @llvm.masked.gather.v8f16(<8 x ptr>, i32, <8 x i1>, <8 x half>)
declare <16 x half> @llvm.masked.gather.v16f16(<16 x ptr>, i32, <16 x i1>, <16 x half>)
declare <32 x half> @llvm.masked.gather.v32f16(<32 x ptr>, i32, <32 x i1>, <32 x half>)

declare <2 x float> @llvm.masked.gather.v2f32(<2 x ptr>, i32, <2 x i1>, <2 x float>)
declare <4 x float> @llvm.masked.gather.v4f32(<4 x ptr>, i32, <4 x i1>, <4 x float>)
declare <8 x float> @llvm.masked.gather.v8f32(<8 x ptr>, i32, <8 x i1>, <8 x float>)
declare <16 x float> @llvm.masked.gather.v16f32(<16 x ptr>, i32, <16 x i1>, <16 x float>)
declare <32 x float> @llvm.masked.gather.v32f32(<32 x ptr>, i32, <32 x i1>, <32 x float>)

declare <1 x double> @llvm.masked.gather.v1f64(<1 x ptr>, i32, <1 x i1>, <1 x double>)
declare <2 x double> @llvm.masked.gather.v2f64(<2 x ptr>, i32, <2 x i1>, <2 x double>)
declare <4 x double> @llvm.masked.gather.v4f64(<4 x ptr>, i32, <4 x i1>, <4 x double>)
declare <8 x double> @llvm.masked.gather.v8f64(<8 x ptr>, i32, <8 x i1>, <8 x double>)
declare <16 x double> @llvm.masked.gather.v16f64(<16 x ptr>, i32, <16 x i1>, <16 x double>)
declare <32 x double> @llvm.masked.gather.v32f64(<32 x ptr>, i32, <32 x i1>, <32 x double>)

attributes #0 = { "target-features"="+sve" }