test/CodeGen/AArch64/sve-extract-subvector.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s

; Test that DAGCombiner doesn't drop the scalable flag when it tries to fold:
;   extract_subv (bitcast X), Index --> bitcast (extract_subv X, Index')
define <vscale x 16 x i8> @extract_nxv16i8_nxv4i64(<vscale x 4 x i64> %z0_z1) {
; CHECK-LABEL: extract_nxv16i8_nxv4i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z0.d, z1.d
; CHECK-NEXT:    ret
  %z0_z1_bc = bitcast <vscale x 4 x i64> %z0_z1 to <vscale x 32 x i8>
  %ext = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv32i8(<vscale x 32 x i8> %z0_z1_bc, i32 1)
  ret <vscale x 16 x i8> %ext
}


define <vscale x 2 x i64> @extract_nxv2i64_nxv32i8(<vscale x 32 x i8> %z0_z1) {
; CHECK-LABEL: extract_nxv2i64_nxv32i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z0.d, z1.d
; CHECK-NEXT:    ret
  %z0_z1_bc = bitcast <vscale x 32 x i8> %z0_z1 to <vscale x 4 x i64>
  %ext = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv4i64(<vscale x 4 x i64> %z0_z1_bc, i32 1)
  ret <vscale x 2 x i64> %ext
}

define <vscale x 4 x half> @extract_lo_nxv4f16_nxv8f16(<vscale x 8 x half> %z0) {
; CHECK-LABEL: extract_lo_nxv4f16_nxv8f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    uunpklo z0.s, z0.h
; CHECK-NEXT:    ret
  %ext = call <vscale x 4 x half> @llvm.aarch64.sve.tuple.get.nxv8f16(<vscale x 8 x half> %z0, i32 0)
  ret <vscale x 4 x half> %ext
}

define <vscale x 4 x half> @extract_hi_nxv4f16_nxv8f16(<vscale x 8 x half> %z0) {
; CHECK-LABEL: extract_hi_nxv4f16_nxv8f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    uunpkhi z0.s, z0.h
; CHECK-NEXT:    ret
  %ext = call <vscale x 4 x half> @llvm.aarch64.sve.tuple.get.nxv8f16(<vscale x 8 x half> %z0, i32 1)
  ret <vscale x 4 x half> %ext
}

define <vscale x 2 x float> @extract_lo_nxv2f32_nxv4f32(<vscale x 4 x float> %z0) {
; CHECK-LABEL: extract_lo_nxv2f32_nxv4f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    uunpklo z0.d, z0.s
; CHECK-NEXT:    ret
  %ext = call <vscale x 2 x float> @llvm.aarch64.sve.tuple.get.nxv4f32(<vscale x 4 x float> %z0, i32 0)
  ret <vscale x 2 x float> %ext
}

define <vscale x 2 x float> @extract_hi_nxv2f32_nxv4f32(<vscale x 4 x float> %z0) {
; CHECK-LABEL: extract_hi_nxv2f32_nxv4f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    uunpkhi z0.d, z0.s
; CHECK-NEXT:    ret
  %ext = call <vscale x 2 x float> @llvm.aarch64.sve.tuple.get.nxv4f32(<vscale x 4 x float> %z0, i32 1)
  ret <vscale x 2 x float> %ext
}

define <vscale x 4 x float> @load_extract_nxv4f32_nxv8f32(<vscale x 8 x float>* %p) {
; CHECK-LABEL: load_extract_nxv4f32_nxv8f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.s
; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0, #1, mul vl]
; CHECK-NEXT:    ret
  %tmp1 = load <vscale x 8 x float>, <vscale x 8 x float>* %p, align 16
  %tmp2 = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv8f32(<vscale x 8 x float> %tmp1, i32 1)
  ret <vscale x 4 x float> %tmp2
}

declare <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv4i64(<vscale x 4 x i64>, i32)
declare <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv32i8(<vscale x 32 x i8>, i32)
declare <vscale x 2 x float> @llvm.aarch64.sve.tuple.get.nxv4f32(<vscale x 4 x float>, i32)
declare <vscale x 4 x half> @llvm.aarch64.sve.tuple.get.nxv8f16(<vscale x 8 x half>, i32)
declare <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv8f32(<vscale x 8 x float>, i32)