llvm/test/CodeGen/WebAssembly/extend-narrow.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc -mtriple=wasm32 -mattr=+simd128 %s -o - | FileCheck %s

target datalayout = "e-m:e-p:32:32-p10:8:8-p20:8:8-i64:64-i128:128-n32:64-S128-ni:1:10:20"

define hidden <4 x i32> @i8_narrow_i8(<4 x i32> noundef %0, <4 x i32> noundef %1) {
; CHECK-LABEL: i8_narrow_i8:
; CHECK:         .functype i8_narrow_i8 (v128, v128) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i16x8.extend_low_i8x16_s
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i16x8.extend_high_i8x16_s
; CHECK-NEXT:    i8x16.narrow_i16x8_s
; CHECK-NEXT:    # fallthrough-return
  %3 = bitcast <4 x i32> %0 to <16 x i8>
  %4 = shufflevector <16 x i8> %3, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %5 = sext <8 x i8> %4 to <8 x i16>
  %6 = shufflevector <16 x i8> %3, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %7 = sext <8 x i8> %6 to <8 x i16>
  %8 = tail call <16 x i8> @llvm.wasm.narrow.signed.v16i8.v8i16(<8 x i16> %5, <8 x i16> %7)
  %9 = bitcast <16 x i8> %8 to <4 x i32>
  ret <4 x i32> %9
}

define hidden <4 x i32> @i16_narrow_i16(<4 x i32> noundef %0, <4 x i32> noundef %1) {
; CHECK-LABEL: i16_narrow_i16:
; CHECK:         .functype i16_narrow_i16 (v128, v128) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32x4.extend_low_i16x8_s
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32x4.extend_high_i16x8_s
; CHECK-NEXT:    i16x8.narrow_i32x4_s
; CHECK-NEXT:    # fallthrough-return
  %3 = bitcast <4 x i32> %0 to <8 x i16>
  %4 = shufflevector <8 x i16> %3, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %5 = sext <4 x i16> %4 to <4 x i32>
  %6 = shufflevector <8 x i16> %3, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %7 = sext <4 x i16> %6 to <4 x i32>
  %8 = tail call <8 x i16> @llvm.wasm.narrow.signed.v8i16.v4i32(<4 x i32> range(i32 -32768, 65536) %5, <4 x i32> range(i32 -32768, 65536) %7)
  %9 = bitcast <8 x i16> %8 to <4 x i32>
  ret <4 x i32> %9
}

define hidden <4 x i32> @i8_narrow_u8(<4 x i32> noundef %0, <4 x i32> noundef %1) {
; CHECK-LABEL: i8_narrow_u8:
; CHECK:         .functype i8_narrow_u8 (v128, v128) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i16x8.extend_low_i8x16_u
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i16x8.extend_high_i8x16_u
; CHECK-NEXT:    i8x16.narrow_i16x8_s
; CHECK-NEXT:    # fallthrough-return
  %3 = bitcast <4 x i32> %0 to <16 x i8>
  %4 = shufflevector <16 x i8> %3, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %5 = zext <8 x i8> %4 to <8 x i16>
  %6 = shufflevector <16 x i8> %3, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %7 = zext <8 x i8> %6 to <8 x i16>
  %8 = tail call <16 x i8> @llvm.wasm.narrow.signed.v16i8.v8i16(<8 x i16> %5, <8 x i16> %7)
  %9 = bitcast <16 x i8> %8 to <4 x i32>
  ret <4 x i32> %9
}

define hidden <4 x i32> @i16_narrow_u16(<4 x i32> noundef %0, <4 x i32> noundef %1) {
; CHECK-LABEL: i16_narrow_u16:
; CHECK:         .functype i16_narrow_u16 (v128, v128) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32x4.extend_low_i16x8_u
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32x4.extend_high_i16x8_u
; CHECK-NEXT:    i16x8.narrow_i32x4_s
; CHECK-NEXT:    # fallthrough-return
  %3 = bitcast <4 x i32> %0 to <8 x i16>
  %4 = shufflevector <8 x i16> %3, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %5 = zext <4 x i16> %4 to <4 x i32>
  %6 = shufflevector <8 x i16> %3, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %7 = zext <4 x i16> %6 to <4 x i32>
  %8 = tail call <8 x i16> @llvm.wasm.narrow.signed.v8i16.v4i32(<4 x i32> range(i32 -32768, 65536) %5, <4 x i32> range(i32 -32768, 65536) %7)
  %9 = bitcast <8 x i16> %8 to <4 x i32>
  ret <4 x i32> %9
}

define hidden <4 x i32> @u8_narrow_i8(<4 x i32> noundef %0, <4 x i32> noundef %1) {
; CHECK-LABEL: u8_narrow_i8:
; CHECK:         .functype u8_narrow_i8 (v128, v128) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i16x8.extend_low_i8x16_s
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i16x8.extend_high_i8x16_s
; CHECK-NEXT:    i8x16.narrow_i16x8_u
; CHECK-NEXT:    # fallthrough-return
  %3 = bitcast <4 x i32> %0 to <16 x i8>
  %4 = shufflevector <16 x i8> %3, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %5 = sext <8 x i8> %4 to <8 x i16>
  %6 = shufflevector <16 x i8> %3, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %7 = sext <8 x i8> %6 to <8 x i16>
  %8 = tail call <16 x i8> @llvm.wasm.narrow.unsigned.v16i8.v8i16(<8 x i16> %5, <8 x i16> %7)
  %9 = bitcast <16 x i8> %8 to <4 x i32>
  ret <4 x i32> %9
}

define hidden <4 x i32> @u16_narrow_i16(<4 x i32> noundef %0, <4 x i32> noundef %1) {
; CHECK-LABEL: u16_narrow_i16:
; CHECK:         .functype u16_narrow_i16 (v128, v128) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32x4.extend_low_i16x8_s
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32x4.extend_high_i16x8_s
; CHECK-NEXT:    i16x8.narrow_i32x4_u
; CHECK-NEXT:    # fallthrough-return
  %3 = bitcast <4 x i32> %0 to <8 x i16>
  %4 = shufflevector <8 x i16> %3, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %5 = sext <4 x i16> %4 to <4 x i32>
  %6 = shufflevector <8 x i16> %3, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %7 = sext <4 x i16> %6 to <4 x i32>
  %8 = tail call <8 x i16> @llvm.wasm.narrow.unsigned.v8i16.v4i32(<4 x i32> range(i32 -32768, 65536) %5, <4 x i32> range(i32 -32768, 65536) %7)
  %9 = bitcast <8 x i16> %8 to <4 x i32>
  ret <4 x i32> %9
}

define hidden <4 x i32> @u8_narrow_u8(<4 x i32> noundef %0, <4 x i32> noundef %1) {
; CHECK-LABEL: u8_narrow_u8:
; CHECK:         .functype u8_narrow_u8 (v128, v128) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i16x8.extend_low_i8x16_u
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i16x8.extend_high_i8x16_u
; CHECK-NEXT:    i8x16.narrow_i16x8_u
; CHECK-NEXT:    # fallthrough-return
  %3 = bitcast <4 x i32> %0 to <16 x i8>
  %4 = shufflevector <16 x i8> %3, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %5 = zext <8 x i8> %4 to <8 x i16>
  %6 = shufflevector <16 x i8> %3, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %7 = zext <8 x i8> %6 to <8 x i16>
  %8 = tail call <16 x i8> @llvm.wasm.narrow.unsigned.v16i8.v8i16(<8 x i16> %5, <8 x i16> %7)
  %9 = bitcast <16 x i8> %8 to <4 x i32>
  ret <4 x i32> %9
}

define hidden <4 x i32> @u16_narrow_u16(<4 x i32> noundef %0, <4 x i32> noundef %1) {
; CHECK-LABEL: u16_narrow_u16:
; CHECK:         .functype u16_narrow_u16 (v128, v128) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32x4.extend_low_i16x8_u
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32x4.extend_high_i16x8_u
; CHECK-NEXT:    i16x8.narrow_i32x4_u
; CHECK-NEXT:    # fallthrough-return
  %3 = bitcast <4 x i32> %0 to <8 x i16>
  %4 = shufflevector <8 x i16> %3, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %5 = zext <4 x i16> %4 to <4 x i32>
  %6 = shufflevector <8 x i16> %3, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %7 = zext <4 x i16> %6 to <4 x i32>
  %8 = tail call <8 x i16> @llvm.wasm.narrow.unsigned.v8i16.v4i32(<4 x i32> range(i32 -32768, 65536) %5, <4 x i32> range(i32 -32768, 65536) %7)
  %9 = bitcast <8 x i16> %8 to <4 x i32>
  ret <4 x i32> %9
}

declare <16 x i8> @llvm.wasm.narrow.signed.v16i8.v8i16(<8 x i16>, <8 x i16>)
declare <8 x i16> @llvm.wasm.narrow.signed.v8i16.v4i32(<4 x i32>, <4 x i32>)
declare <16 x i8> @llvm.wasm.narrow.unsigned.v16i8.v8i16(<8 x i16>, <8 x i16>)
declare <8 x i16> @llvm.wasm.narrow.unsigned.v8i16.v4i32(<4 x i32>, <4 x i32>)