test/CodeGen/AArch64/sve-uunpklo-load-uzp1-store-combine.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s | FileCheck %s

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

; Check that we don't try and merge uunpklo/uzp1 with a load or store if we
; would end up creating a predicate that would be too large for the max VL.

; UUNPKLO + Load

define <vscale x 8 x i16> @uunpklo_i8_valid(ptr %b) #0 {
; CHECK-LABEL: uunpklo_i8_valid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl64
; CHECK-NEXT:    ld1b { z0.h }, p0/z, [x0]
; CHECK-NEXT:    ret
  %mask = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 11)
  %load = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8(ptr %b, i32 2, <vscale x 16 x i1> %mask, <vscale x 16 x i8> undef)
  %uzp = call <vscale x 8 x i16> @llvm.aarch64.sve.uunpklo.nxv8i16(<vscale x 16 x i8> %load)
  ret <vscale x 8 x i16> %uzp
}

define <vscale x 8 x i16> @uunpklo_i8_invalid(ptr %b) #0 {
; CHECK-LABEL: uunpklo_i8_invalid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.b, vl128
; CHECK-NEXT:    ld1b { z0.b }, p0/z, [x0]
; CHECK-NEXT:    uunpklo z0.h, z0.b
; CHECK-NEXT:    ret
  %mask = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 12)
  %load = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8(ptr %b, i32 2, <vscale x 16 x i1> %mask, <vscale x 16 x i8> undef)
  %uzp = call <vscale x 8 x i16> @llvm.aarch64.sve.uunpklo.nxv8i16(<vscale x 16 x i8> %load)
  ret <vscale x 8 x i16> %uzp
}

define <vscale x 4 x i32> @uunpklo_i16_valid(ptr %b) #0 {
; CHECK-LABEL: uunpklo_i16_valid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x0]
; CHECK-NEXT:    ret
  %mask = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 10)
  %load = call <vscale x 8 x i16> @llvm.masked.load.nxv8i16(ptr %b, i32 2, <vscale x 8 x i1> %mask, <vscale x 8 x i16> undef)
  %uzp = call <vscale x 4 x i32> @llvm.aarch64.sve.uunpklo.nxv4i32(<vscale x 8 x i16> %load)
  ret <vscale x 4 x i32> %uzp
}

define <vscale x 4 x i32> @uunpklo_i16_invalid(ptr %b) #0 {
; CHECK-LABEL: uunpklo_i16_invalid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl64
; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
; CHECK-NEXT:    uunpklo z0.s, z0.h
; CHECK-NEXT:    ret
  %mask = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 11)
  %load = call <vscale x 8 x i16> @llvm.masked.load.nxv8i16(ptr %b, i32 2, <vscale x 8 x i1> %mask, <vscale x 8 x i16> undef)
  %uzp = call <vscale x 4 x i32> @llvm.aarch64.sve.uunpklo.nxv4i32(<vscale x 8 x i16> %load)
  ret <vscale x 4 x i32> %uzp
}

define <vscale x 2 x i64> @uunpklo_i32_valid(ptr %b) #0 {
; CHECK-LABEL: uunpklo_i32_valid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl16
; CHECK-NEXT:    ld1w { z0.d }, p0/z, [x0]
; CHECK-NEXT:    ret
  %mask = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 9)
  %load = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32(ptr %b, i32 2, <vscale x 4 x i1> %mask, <vscale x 4 x i32> undef)
  %uzp = call <vscale x 2 x i64> @llvm.aarch64.sve.uunpklo.nxv2i64(<vscale x 4 x i32> %load)
  ret <vscale x 2 x i64> %uzp
}

define <vscale x 2 x i64> @uunpklo_i32_invalid(ptr %b) #0 {
; CHECK-LABEL: uunpklo_i32_invalid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    uunpklo z0.d, z0.s
; CHECK-NEXT:    ret
  %mask = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 10)
  %load = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32(ptr %b, i32 2, <vscale x 4 x i1> %mask, <vscale x 4 x i32> undef)
  %uzp = call <vscale x 2 x i64> @llvm.aarch64.sve.uunpklo.nxv2i64(<vscale x 4 x i32> %load)
  ret <vscale x 2 x i64> %uzp
}

define <vscale x 2 x i64> @uunpklo_invalid_all(ptr %b) #0 {
; CHECK-LABEL: uunpklo_invalid_all:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
; CHECK-NEXT:    uunpklo z0.d, z0.s
; CHECK-NEXT:    ret
  %mask = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
  %load = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32(ptr %b, i32 2, <vscale x 4 x i1> %mask, <vscale x 4 x i32> undef)
  %uzp = call <vscale x 2 x i64> @llvm.aarch64.sve.uunpklo.nxv2i64(<vscale x 4 x i32> %load)
  ret <vscale x 2 x i64> %uzp
}

; UZP1 + Store

define void @uzp1_i8_valid(<vscale x 8 x i16> %a, ptr %b) #0 {
; CHECK-LABEL: uzp1_i8_valid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl64
; CHECK-NEXT:    st1b { z0.h }, p0, [x0]
; CHECK-NEXT:    ret
  %a.bc = bitcast <vscale x 8 x i16> %a to <vscale x 16 x i8>
  %uzp = call <vscale x 16 x i8> @llvm.aarch64.sve.uzp1.nxv16i8(<vscale x 16 x i8> %a.bc, <vscale x 16 x i8> %a.bc)
  %mask = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 11)
  call void @llvm.masked.store.nxv16i8(<vscale x 16 x i8> %uzp, ptr %b, i32 2, <vscale x 16 x i1> %mask)
  ret void
}

define void @uzp1_i8_invalid(<vscale x 8 x i16> %a, ptr %b) #0 {
; CHECK-LABEL: uzp1_i8_invalid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.b, vl128
; CHECK-NEXT:    uzp1 z0.b, z0.b, z0.b
; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
; CHECK-NEXT:    ret
  %a.bc = bitcast <vscale x 8 x i16> %a to <vscale x 16 x i8>
  %uzp = call <vscale x 16 x i8> @llvm.aarch64.sve.uzp1.nxv16i8(<vscale x 16 x i8> %a.bc, <vscale x 16 x i8> %a.bc)
  %mask = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 12)
  call void @llvm.masked.store.nxv16i8(<vscale x 16 x i8> %uzp, ptr %b, i32 2, <vscale x 16 x i1> %mask)
  ret void
}

define void @uzp1_i16_valid(<vscale x 4 x i32> %a, ptr %b) #0 {
; CHECK-LABEL: uzp1_i16_valid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    st1h { z0.s }, p0, [x0]
; CHECK-NEXT:    ret
  %a.bc = bitcast <vscale x 4 x i32> %a to <vscale x 8 x i16>
  %uzp = call <vscale x 8 x i16> @llvm.aarch64.sve.uzp1.nxv8i16(<vscale x 8 x i16> %a.bc, <vscale x 8 x i16> %a.bc)
  %mask = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 10)
  call void @llvm.masked.store.nxv8i16(<vscale x 8 x i16> %uzp, ptr %b, i32 2, <vscale x 8 x i1> %mask)
  ret void
}

define void @uzp1_i16_invalid(<vscale x 4 x i32> %a, ptr %b) #0 {
; CHECK-LABEL: uzp1_i16_invalid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.h, vl64
; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
; CHECK-NEXT:    ret
  %a.bc = bitcast <vscale x 4 x i32> %a to <vscale x 8 x i16>
  %uzp = call <vscale x 8 x i16> @llvm.aarch64.sve.uzp1.nxv8i16(<vscale x 8 x i16> %a.bc, <vscale x 8 x i16> %a.bc)
  %mask = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 11)
  call void @llvm.masked.store.nxv8i16(<vscale x 8 x i16> %uzp, ptr %b, i32 2, <vscale x 8 x i1> %mask)
  ret void
}

define void @uzp1_i32_valid(<vscale x 2 x i64> %a, ptr %b) #0 {
; CHECK-LABEL: uzp1_i32_valid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d, vl16
; CHECK-NEXT:    st1w { z0.d }, p0, [x0]
; CHECK-NEXT:    ret
  %a.bc = bitcast <vscale x 2 x i64> %a to <vscale x 4 x i32>
  %uzp = call <vscale x 4 x i32> @llvm.aarch64.sve.uzp1.nxv4i32(<vscale x 4 x i32> %a.bc, <vscale x 4 x i32> %a.bc)
  %mask = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 9)
  call void @llvm.masked.store.nxv4i32(<vscale x 4 x i32> %uzp, ptr %b, i32 2, <vscale x 4 x i1> %mask)
  ret void
}

define void @uzp1_i32_invalid(<vscale x 2 x i64> %a, ptr %b) #0 {
; CHECK-LABEL: uzp1_i32_invalid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s, vl32
; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
; CHECK-NEXT:    ret
  %a.bc = bitcast <vscale x 2 x i64> %a to <vscale x 4 x i32>
  %uzp = call <vscale x 4 x i32> @llvm.aarch64.sve.uzp1.nxv4i32(<vscale x 4 x i32> %a.bc, <vscale x 4 x i32> %a.bc)
  %mask = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 10)
  call void @llvm.masked.store.nxv4i32(<vscale x 4 x i32> %uzp, ptr %b, i32 2, <vscale x 4 x i1> %mask)
  ret void
}

define void @uzp1_invalid_all(<vscale x 2 x i64> %a, ptr %b) #0 {
; CHECK-LABEL: uzp1_invalid_all:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s
; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
; CHECK-NEXT:    ret
  %a.bc = bitcast <vscale x 2 x i64> %a to <vscale x 4 x i32>
  %uzp = call <vscale x 4 x i32> @llvm.aarch64.sve.uzp1.nxv4i32(<vscale x 4 x i32> %a.bc, <vscale x 4 x i32> %a.bc)
  %mask = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
  call void @llvm.masked.store.nxv4i32(<vscale x 4 x i32> %uzp, ptr %b, i32 2, <vscale x 4 x i1> %mask)
  ret void
}

declare <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 %pattern)
declare <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 %pattern)
declare <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 %pattern)

declare <vscale x 8 x i16> @llvm.aarch64.sve.uunpklo.nxv8i16(<vscale x 16 x i8>)
declare <vscale x 4 x i32> @llvm.aarch64.sve.uunpklo.nxv4i32(<vscale x 8 x i16>)
declare <vscale x 2 x i64> @llvm.aarch64.sve.uunpklo.nxv2i64(<vscale x 4 x i32>)

declare <vscale x 16 x i8> @llvm.aarch64.sve.uzp1.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
declare <vscale x 8 x i16> @llvm.aarch64.sve.uzp1.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
declare <vscale x 4 x i32> @llvm.aarch64.sve.uzp1.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)

declare <vscale x 16 x i8> @llvm.masked.load.nxv16i8(<vscale x 16 x i8>*, i32, <vscale x 16 x i1>, <vscale x 16 x i8>)
declare <vscale x 8 x i16> @llvm.masked.load.nxv8i16(<vscale x 8 x i16>*, i32, <vscale x 8 x i1>, <vscale x 8 x i16>)
declare <vscale x 4 x i32> @llvm.masked.load.nxv4i32(<vscale x 4 x i32>*, i32, <vscale x 4 x i1>, <vscale x 4 x i32>)

declare void @llvm.masked.store.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>*, i32, <vscale x 16 x i1>)
declare void @llvm.masked.store.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>*, i32, <vscale x 8 x i1>)
declare void @llvm.masked.store.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>*, i32, <vscale x 4 x i1>)

attributes #0 = { "target-features"="+sve" vscale_range(8,0) }