1 ; RUN: llc < %s -asm-verbose=false -verify-machineinstrs -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -mattr=+unimplemented-simd128 | FileCheck %s
3 ; Test that the logic to choose between v128.const vector
4 ; initialization and splat vector initialization and to optimize the
5 ; choice of splat value works correctly.
7 target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
8 target triple = "wasm32-unknown-unknown"
10 ; CHECK-LABEL: same_const_one_replaced_i8x16:
11 ; CHECK-NEXT: .functype same_const_one_replaced_i8x16 (i32) -> (v128)
12 ; CHECK-NEXT: i32.const $push[[L0:[0-9]+]]=, 42
13 ; CHECK-NEXT: i16x8.splat $push[[L1:[0-9]+]]=, $pop[[L0]]
14 ; CHECK-NEXT: i16x8.replace_lane $push[[L2:[0-9]+]]=, $pop[[L1]], 5, $0
15 ; CHECK-NEXT: return $pop[[L2]]
16 define <8 x i16> @same_const_one_replaced_i8x16(i16 %x) {
18 <8 x i16> <i16 42, i16 42, i16 42, i16 42, i16 42, i16 42, i16 42, i16 42>,
24 ; CHECK-LABEL: different_const_one_replaced_i8x16:
25 ; CHECK-NEXT: .functype different_const_one_replaced_i8x16 (i32) -> (v128)
26 ; CHECK-NEXT: v128.const $push[[L0:[0-9]+]]=, 1, -2, 3, -4, 5, 0, 7, -8
27 ; CHECK-NEXT: i16x8.replace_lane $push[[L1:[0-9]+]]=, $pop[[L0]], 5, $0
28 ; CHECK-NEXT: return $pop[[L1]]
29 define <8 x i16> @different_const_one_replaced_i8x16(i16 %x) {
31 <8 x i16> <i16 1, i16 -2, i16 3, i16 -4, i16 5, i16 -6, i16 7, i16 -8>,
37 ; CHECK-LABEL: same_const_one_replaced_f32x4:
38 ; CHECK-NEXT: .functype same_const_one_replaced_f32x4 (f32) -> (v128)
39 ; CHECK-NEXT: f32.const $push[[L0:[0-9]+]]=, 0x1.5p5
40 ; CHECK-NEXT: f32x4.splat $push[[L1:[0-9]+]]=, $pop[[L0]]
41 ; CHECK-NEXT: f32x4.replace_lane $push[[L2:[0-9]+]]=, $pop[[L1]], 2, $0
42 ; CHECK-NEXT: return $pop[[L2]]
43 define <4 x float> @same_const_one_replaced_f32x4(float %x) {
45 <4 x float> <float 42., float 42., float 42., float 42.>,
51 ; CHECK-LABEL: different_const_one_replaced_f32x4:
52 ; CHECK-NEXT: .functype different_const_one_replaced_f32x4 (f32) -> (v128)
53 ; CHECK-NEXT: v128.const $push[[L0:[0-9]+]]=, 0x1p0, 0x1p1, 0x0p0, 0x1p2
54 ; CHECK-NEXT: f32x4.replace_lane $push[[L1:[0-9]+]]=, $pop[[L0]], 2, $0
55 ; CHECK-NEXT: return $pop[[L1]]
56 define <4 x float> @different_const_one_replaced_f32x4(float %x) {
58 <4 x float> <float 1., float 2., float 3., float 4.>,
64 ; CHECK-LABEL: splat_common_const_i32x4:
65 ; CHECK-NEXT: .functype splat_common_const_i32x4 () -> (v128)
66 ; CHECK-NEXT: i32.const $push[[L0:[0-9]+]]=, 3
67 ; CHECK-NEXT: i32x4.splat $push[[L1:[0-9]+]]=, $pop[[L0]]
68 ; CHECK-NEXT: i32.const $push[[L2:[0-9]+]]=, 1
69 ; CHECK-NEXT: i32x4.replace_lane $push[[L3:[0-9]+]]=, $pop[[L1]], 3, $pop[[L2]]
70 ; CHECK-NEXT: return $pop[[L3]]
71 define <4 x i32> @splat_common_const_i32x4() {
72 ret <4 x i32> <i32 undef, i32 3, i32 3, i32 1>
75 ; CHECK-LABEL: splat_common_arg_i16x8:
76 ; CHECK-NEXT: .functype splat_common_arg_i16x8 (i32, i32, i32) -> (v128)
77 ; CHECK-NEXT: i16x8.splat $push[[L0:[0-9]+]]=, $2
78 ; CHECK-NEXT: i16x8.replace_lane $push[[L1:[0-9]+]]=, $pop[[L0]], 0, $1
79 ; CHECK-NEXT: i16x8.replace_lane $push[[L2:[0-9]+]]=, $pop[[L1]], 2, $0
80 ; CHECK-NEXT: i16x8.replace_lane $push[[L3:[0-9]+]]=, $pop[[L2]], 4, $1
81 ; CHECK-NEXT: i16x8.replace_lane $push[[L4:[0-9]+]]=, $pop[[L3]], 7, $1
82 ; CHECK-NEXT: return $pop[[L4]]
83 define <8 x i16> @splat_common_arg_i16x8(i16 %a, i16 %b, i16 %c) {
84 %v0 = insertelement <8 x i16> undef, i16 %b, i32 0
85 %v1 = insertelement <8 x i16> %v0, i16 %c, i32 1
86 %v2 = insertelement <8 x i16> %v1, i16 %a, i32 2
87 %v3 = insertelement <8 x i16> %v2, i16 %c, i32 3
88 %v4 = insertelement <8 x i16> %v3, i16 %b, i32 4
89 %v5 = insertelement <8 x i16> %v4, i16 %c, i32 5
90 %v6 = insertelement <8 x i16> %v5, i16 %c, i32 6
91 %v7 = insertelement <8 x i16> %v6, i16 %b, i32 7
95 ; CHECK-LABEL: undef_const_insert_f32x4:
96 ; CHECK-NEXT: .functype undef_const_insert_f32x4 () -> (v128)
97 ; CHECK-NEXT: f32.const $push[[L0:[0-9]+]]=, 0x1.5p5
98 ; CHECK-NEXT: f32x4.splat $push[[L1:[0-9]+]]=, $pop[[L0]]
99 ; CHECK-NEXT: return $pop[[L1]]
100 define <4 x float> @undef_const_insert_f32x4() {
101 %v = insertelement <4 x float> undef, float 42., i32 1
105 ; CHECK-LABEL: undef_arg_insert_i32x4:
106 ; CHECK-NEXT: .functype undef_arg_insert_i32x4 (i32) -> (v128)
107 ; CHECK-NEXT: i32x4.splat $push[[L0:[0-9]+]]=, $0
108 ; CHECK-NEXT: return $pop[[L0]]
109 define <4 x i32> @undef_arg_insert_i32x4(i32 %x) {
110 %v = insertelement <4 x i32> undef, i32 %x, i32 3
114 ; CHECK-LABEL: all_undef_i8x16:
115 ; CHECK-NEXT: .functype all_undef_i8x16 () -> (v128)
116 ; CHECK-NEXT: return $0
117 define <16 x i8> @all_undef_i8x16() {
118 %v = insertelement <16 x i8> undef, i8 undef, i32 4
122 ; CHECK-LABEL: all_undef_f64x2:
123 ; CHECK-NEXT: .functype all_undef_f64x2 () -> (v128)
124 ; CHECK-NEXT: return $0
125 define <2 x double> @all_undef_f64x2() {
126 ret <2 x double> undef