1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
4 ; For all the functions below should the operation is a nop
5 define <vscale x 8 x i8> @trunc_i16toi8(<vscale x 8 x i16> %in) {
6 ; CHECK-LABEL: trunc_i16toi8:
7 ; CHECK: // %bb.0: // %entry
10 %out = trunc <vscale x 8 x i16> %in to <vscale x 8 x i8>
11 ret <vscale x 8 x i8> %out
14 define <vscale x 4 x i8> @trunc_i32toi8(<vscale x 4 x i32> %in) {
15 ; CHECK-LABEL: trunc_i32toi8:
16 ; CHECK: // %bb.0: // %entry
19 %out = trunc <vscale x 4 x i32> %in to <vscale x 4 x i8>
20 ret <vscale x 4 x i8> %out
23 define <vscale x 2 x i8> @trunc_i64toi8(<vscale x 2 x i64> %in) {
24 ; CHECK-LABEL: trunc_i64toi8:
25 ; CHECK: // %bb.0: // %entry
28 %out = trunc <vscale x 2 x i64> %in to <vscale x 2 x i8>
29 ret <vscale x 2 x i8> %out
32 define <vscale x 4 x i16> @trunc_i32toi16(<vscale x 4 x i32> %in) {
33 ; CHECK-LABEL: trunc_i32toi16:
34 ; CHECK: // %bb.0: // %entry
37 %out = trunc <vscale x 4 x i32> %in to <vscale x 4 x i16>
38 ret <vscale x 4 x i16> %out
41 define <vscale x 2 x i16> @trunc_i64toi16(<vscale x 2 x i64> %in) {
42 ; CHECK-LABEL: trunc_i64toi16:
43 ; CHECK: // %bb.0: // %entry
46 %out = trunc <vscale x 2 x i64> %in to <vscale x 2 x i16>
47 ret <vscale x 2 x i16> %out
50 define <vscale x 2 x i32> @trunc_i64toi32(<vscale x 2 x i64> %in) {
51 ; CHECK-LABEL: trunc_i64toi32:
52 ; CHECK: // %bb.0: // %entry
55 %out = trunc <vscale x 2 x i64> %in to <vscale x 2 x i32>
56 ret <vscale x 2 x i32> %out
59 ; Truncating to i1 requires convert it to a cmp
61 define <vscale x 2 x i1> @trunc_i64toi1(<vscale x 2 x i64> %in) {
62 ; CHECK-LABEL: trunc_i64toi1:
63 ; CHECK: // %bb.0: // %entry
64 ; CHECK-NEXT: ptrue p0.d
65 ; CHECK-NEXT: and z0.d, z0.d, #0x1
66 ; CHECK-NEXT: cmpne p0.d, p0/z, z0.d, #0
69 %out = trunc <vscale x 2 x i64> %in to <vscale x 2 x i1>
70 ret <vscale x 2 x i1> %out
73 define <vscale x 4 x i1> @trunc_i64toi1_split(<vscale x 4 x i64> %in) {
74 ; CHECK-LABEL: trunc_i64toi1_split:
75 ; CHECK: // %bb.0: // %entry
76 ; CHECK-NEXT: ptrue p0.d
77 ; CHECK-NEXT: and z1.d, z1.d, #0x1
78 ; CHECK-NEXT: and z0.d, z0.d, #0x1
79 ; CHECK-NEXT: cmpne p1.d, p0/z, z1.d, #0
80 ; CHECK-NEXT: cmpne p0.d, p0/z, z0.d, #0
81 ; CHECK-NEXT: uzp1 p0.s, p0.s, p1.s
84 %out = trunc <vscale x 4 x i64> %in to <vscale x 4 x i1>
85 ret <vscale x 4 x i1> %out
88 define <vscale x 8 x i1> @trunc_i64toi1_split2(<vscale x 8 x i64> %in) {
89 ; CHECK-LABEL: trunc_i64toi1_split2:
90 ; CHECK: // %bb.0: // %entry
91 ; CHECK-NEXT: ptrue p0.d
92 ; CHECK-NEXT: and z3.d, z3.d, #0x1
93 ; CHECK-NEXT: and z2.d, z2.d, #0x1
94 ; CHECK-NEXT: and z1.d, z1.d, #0x1
95 ; CHECK-NEXT: and z0.d, z0.d, #0x1
96 ; CHECK-NEXT: cmpne p1.d, p0/z, z3.d, #0
97 ; CHECK-NEXT: cmpne p2.d, p0/z, z2.d, #0
98 ; CHECK-NEXT: uzp1 p1.s, p2.s, p1.s
99 ; CHECK-NEXT: cmpne p2.d, p0/z, z1.d, #0
100 ; CHECK-NEXT: cmpne p0.d, p0/z, z0.d, #0
101 ; CHECK-NEXT: uzp1 p0.s, p0.s, p2.s
102 ; CHECK-NEXT: uzp1 p0.h, p0.h, p1.h
105 %out = trunc <vscale x 8 x i64> %in to <vscale x 8 x i1>
106 ret <vscale x 8 x i1> %out
109 define <vscale x 16 x i1> @trunc_i64toi1_split3(<vscale x 16 x i64> %in) {
110 ; CHECK-LABEL: trunc_i64toi1_split3:
111 ; CHECK: // %bb.0: // %entry
112 ; CHECK-NEXT: str x29, [sp, #-16]! // 8-byte Folded Spill
113 ; CHECK-NEXT: addvl sp, sp, #-1
114 ; CHECK-NEXT: str p4, [sp, #7, mul vl] // 2-byte Folded Spill
115 ; CHECK-NEXT: .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
116 ; CHECK-NEXT: .cfi_offset w29, -16
117 ; CHECK-NEXT: ptrue p0.d
118 ; CHECK-NEXT: and z7.d, z7.d, #0x1
119 ; CHECK-NEXT: and z6.d, z6.d, #0x1
120 ; CHECK-NEXT: and z5.d, z5.d, #0x1
121 ; CHECK-NEXT: and z4.d, z4.d, #0x1
122 ; CHECK-NEXT: and z3.d, z3.d, #0x1
123 ; CHECK-NEXT: and z2.d, z2.d, #0x1
124 ; CHECK-NEXT: cmpne p1.d, p0/z, z7.d, #0
125 ; CHECK-NEXT: cmpne p2.d, p0/z, z6.d, #0
126 ; CHECK-NEXT: cmpne p3.d, p0/z, z5.d, #0
127 ; CHECK-NEXT: cmpne p4.d, p0/z, z4.d, #0
128 ; CHECK-NEXT: and z1.d, z1.d, #0x1
129 ; CHECK-NEXT: and z0.d, z0.d, #0x1
130 ; CHECK-NEXT: uzp1 p1.s, p2.s, p1.s
131 ; CHECK-NEXT: cmpne p2.d, p0/z, z3.d, #0
132 ; CHECK-NEXT: uzp1 p3.s, p4.s, p3.s
133 ; CHECK-NEXT: cmpne p4.d, p0/z, z2.d, #0
134 ; CHECK-NEXT: uzp1 p2.s, p4.s, p2.s
135 ; CHECK-NEXT: cmpne p4.d, p0/z, z1.d, #0
136 ; CHECK-NEXT: cmpne p0.d, p0/z, z0.d, #0
137 ; CHECK-NEXT: uzp1 p0.s, p0.s, p4.s
138 ; CHECK-NEXT: ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
139 ; CHECK-NEXT: uzp1 p1.h, p3.h, p1.h
140 ; CHECK-NEXT: uzp1 p0.h, p0.h, p2.h
141 ; CHECK-NEXT: uzp1 p0.b, p0.b, p1.b
142 ; CHECK-NEXT: addvl sp, sp, #1
143 ; CHECK-NEXT: ldr x29, [sp], #16 // 8-byte Folded Reload
146 %out = trunc <vscale x 16 x i64> %in to <vscale x 16 x i1>
147 ret <vscale x 16 x i1> %out
151 define <vscale x 4 x i1> @trunc_i32toi1(<vscale x 4 x i32> %in) {
152 ; CHECK-LABEL: trunc_i32toi1:
153 ; CHECK: // %bb.0: // %entry
154 ; CHECK-NEXT: ptrue p0.s
155 ; CHECK-NEXT: and z0.s, z0.s, #0x1
156 ; CHECK-NEXT: cmpne p0.s, p0/z, z0.s, #0
159 %out = trunc <vscale x 4 x i32> %in to <vscale x 4 x i1>
160 ret <vscale x 4 x i1> %out
163 define <vscale x 8 x i1> @trunc_i16toi1(<vscale x 8 x i16> %in) {
164 ; CHECK-LABEL: trunc_i16toi1:
165 ; CHECK: // %bb.0: // %entry
166 ; CHECK-NEXT: ptrue p0.h
167 ; CHECK-NEXT: and z0.h, z0.h, #0x1
168 ; CHECK-NEXT: cmpne p0.h, p0/z, z0.h, #0
171 %out = trunc <vscale x 8 x i16> %in to <vscale x 8 x i1>
172 ret <vscale x 8 x i1> %out
175 define <vscale x 16 x i1> @trunc_i8toi1(<vscale x 16 x i8> %in) {
176 ; CHECK-LABEL: trunc_i8toi1:
177 ; CHECK: // %bb.0: // %entry
178 ; CHECK-NEXT: ptrue p0.b
179 ; CHECK-NEXT: and z0.b, z0.b, #0x1
180 ; CHECK-NEXT: cmpne p0.b, p0/z, z0.b, #0
183 %out = trunc <vscale x 16 x i8> %in to <vscale x 16 x i1>
184 ret <vscale x 16 x i1> %out
187 define void @trunc_promoteIntRes(<vscale x 4 x i64> %0, i16* %ptr) {
188 ; CHECK-LABEL: trunc_promoteIntRes:
189 ; CHECK: // %bb.0: // %entry
190 ; CHECK-NEXT: uzp1 z0.s, z0.s, z1.s
191 ; CHECK-NEXT: ptrue p0.s
192 ; CHECK-NEXT: st1h { z0.s }, p0, [x0]
195 %1 = trunc <vscale x 4 x i64> %0 to <vscale x 4 x i16>
196 %2 = bitcast i16* %ptr to <vscale x 4 x i16>*
197 store <vscale x 4 x i16> %1, <vscale x 4 x i16>* %2, align 2