1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=aarch64-- | FileCheck %s
4 declare i8 @llvm.fshl.i8(i8, i8, i8)
5 declare i16 @llvm.fshl.i16(i16, i16, i16)
6 declare i32 @llvm.fshl.i32(i32, i32, i32)
7 declare i64 @llvm.fshl.i64(i64, i64, i64)
8 declare <4 x i32> @llvm.fshl.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
10 declare i8 @llvm.fshr.i8(i8, i8, i8)
11 declare i16 @llvm.fshr.i16(i16, i16, i16)
12 declare i32 @llvm.fshr.i32(i32, i32, i32)
13 declare i64 @llvm.fshr.i64(i64, i64, i64)
14 declare <4 x i32> @llvm.fshr.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
16 ; When first 2 operands match, it's a rotate.
18 define i8 @rotl_i8_const_shift(i8 %x) {
19 ; CHECK-LABEL: rotl_i8_const_shift:
21 ; CHECK-NEXT: ubfx w8, w0, #5, #3
22 ; CHECK-NEXT: orr w0, w8, w0, lsl #3
24 %f = call i8 @llvm.fshl.i8(i8 %x, i8 %x, i8 3)
28 define i64 @rotl_i64_const_shift(i64 %x) {
29 ; CHECK-LABEL: rotl_i64_const_shift:
31 ; CHECK-NEXT: ror x0, x0, #61
33 %f = call i64 @llvm.fshl.i64(i64 %x, i64 %x, i64 3)
37 ; When first 2 operands match, it's a rotate (by variable amount).
39 define i16 @rotl_i16(i16 %x, i16 %z) {
40 ; CHECK-LABEL: rotl_i16:
42 ; CHECK-NEXT: neg w8, w1
43 ; CHECK-NEXT: and w9, w0, #0xffff
44 ; CHECK-NEXT: and w10, w1, #0xf
45 ; CHECK-NEXT: and w8, w8, #0xf
46 ; CHECK-NEXT: lsl w10, w0, w10
47 ; CHECK-NEXT: lsr w8, w9, w8
48 ; CHECK-NEXT: orr w0, w10, w8
50 %f = call i16 @llvm.fshl.i16(i16 %x, i16 %x, i16 %z)
54 define i32 @rotl_i32(i32 %x, i32 %z) {
55 ; CHECK-LABEL: rotl_i32:
57 ; CHECK-NEXT: neg w8, w1
58 ; CHECK-NEXT: ror w0, w0, w8
60 %f = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 %z)
64 define i64 @rotl_i64(i64 %x, i64 %z) {
65 ; CHECK-LABEL: rotl_i64:
67 ; CHECK-NEXT: neg w8, w1
68 ; CHECK-NEXT: ror x0, x0, x8
70 %f = call i64 @llvm.fshl.i64(i64 %x, i64 %x, i64 %z)
76 define <4 x i32> @rotl_v4i32(<4 x i32> %x, <4 x i32> %z) {
77 ; CHECK-LABEL: rotl_v4i32:
79 ; CHECK-NEXT: movi v2.4s, #31
80 ; CHECK-NEXT: neg v3.4s, v1.4s
81 ; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
82 ; CHECK-NEXT: and v2.16b, v3.16b, v2.16b
83 ; CHECK-NEXT: neg v2.4s, v2.4s
84 ; CHECK-NEXT: ushl v1.4s, v0.4s, v1.4s
85 ; CHECK-NEXT: ushl v0.4s, v0.4s, v2.4s
86 ; CHECK-NEXT: orr v0.16b, v1.16b, v0.16b
88 %f = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %x, <4 x i32> %x, <4 x i32> %z)
92 ; Vector rotate by constant splat amount.
94 define <4 x i32> @rotl_v4i32_rotl_const_shift(<4 x i32> %x) {
95 ; CHECK-LABEL: rotl_v4i32_rotl_const_shift:
97 ; CHECK-NEXT: shl v1.4s, v0.4s, #3
98 ; CHECK-NEXT: usra v1.4s, v0.4s, #29
99 ; CHECK-NEXT: mov v0.16b, v1.16b
101 %f = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %x, <4 x i32> %x, <4 x i32> <i32 3, i32 3, i32 3, i32 3>)
105 ; Repeat everything for funnel shift right.
107 ; When first 2 operands match, it's a rotate.
109 define i8 @rotr_i8_const_shift(i8 %x) {
110 ; CHECK-LABEL: rotr_i8_const_shift:
112 ; CHECK-NEXT: lsl w8, w0, #5
113 ; CHECK-NEXT: bfxil w8, w0, #3, #5
114 ; CHECK-NEXT: mov w0, w8
116 %f = call i8 @llvm.fshr.i8(i8 %x, i8 %x, i8 3)
120 define i32 @rotr_i32_const_shift(i32 %x) {
121 ; CHECK-LABEL: rotr_i32_const_shift:
123 ; CHECK-NEXT: ror w0, w0, #3
125 %f = call i32 @llvm.fshr.i32(i32 %x, i32 %x, i32 3)
129 ; When first 2 operands match, it's a rotate (by variable amount).
131 define i16 @rotr_i16(i16 %x, i16 %z) {
132 ; CHECK-LABEL: rotr_i16:
134 ; CHECK-NEXT: neg w8, w1
135 ; CHECK-NEXT: and w9, w0, #0xffff
136 ; CHECK-NEXT: and w10, w1, #0xf
137 ; CHECK-NEXT: and w8, w8, #0xf
138 ; CHECK-NEXT: lsr w9, w9, w10
139 ; CHECK-NEXT: lsl w8, w0, w8
140 ; CHECK-NEXT: orr w0, w9, w8
142 %f = call i16 @llvm.fshr.i16(i16 %x, i16 %x, i16 %z)
146 define i32 @rotr_i32(i32 %x, i32 %z) {
147 ; CHECK-LABEL: rotr_i32:
149 ; CHECK-NEXT: ror w0, w0, w1
151 %f = call i32 @llvm.fshr.i32(i32 %x, i32 %x, i32 %z)
155 define i64 @rotr_i64(i64 %x, i64 %z) {
156 ; CHECK-LABEL: rotr_i64:
158 ; CHECK-NEXT: ror x0, x0, x1
160 %f = call i64 @llvm.fshr.i64(i64 %x, i64 %x, i64 %z)
166 define <4 x i32> @rotr_v4i32(<4 x i32> %x, <4 x i32> %z) {
167 ; CHECK-LABEL: rotr_v4i32:
169 ; CHECK-NEXT: movi v2.4s, #31
170 ; CHECK-NEXT: neg v3.4s, v1.4s
171 ; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
172 ; CHECK-NEXT: and v2.16b, v3.16b, v2.16b
173 ; CHECK-NEXT: neg v1.4s, v1.4s
174 ; CHECK-NEXT: ushl v2.4s, v0.4s, v2.4s
175 ; CHECK-NEXT: ushl v0.4s, v0.4s, v1.4s
176 ; CHECK-NEXT: orr v0.16b, v0.16b, v2.16b
178 %f = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %x, <4 x i32> %x, <4 x i32> %z)
182 ; Vector rotate by constant splat amount.
184 define <4 x i32> @rotr_v4i32_const_shift(<4 x i32> %x) {
185 ; CHECK-LABEL: rotr_v4i32_const_shift:
187 ; CHECK-NEXT: shl v1.4s, v0.4s, #29
188 ; CHECK-NEXT: usra v1.4s, v0.4s, #3
189 ; CHECK-NEXT: mov v0.16b, v1.16b
191 %f = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %x, <4 x i32> %x, <4 x i32> <i32 3, i32 3, i32 3, i32 3>)
195 define i32 @rotl_i32_shift_by_bitwidth(i32 %x) {
196 ; CHECK-LABEL: rotl_i32_shift_by_bitwidth:
199 %f = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 32)
203 define i32 @rotr_i32_shift_by_bitwidth(i32 %x) {
204 ; CHECK-LABEL: rotr_i32_shift_by_bitwidth:
207 %f = call i32 @llvm.fshr.i32(i32 %x, i32 %x, i32 32)
211 define <4 x i32> @rotl_v4i32_shift_by_bitwidth(<4 x i32> %x) {
212 ; CHECK-LABEL: rotl_v4i32_shift_by_bitwidth:
215 %f = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %x, <4 x i32> %x, <4 x i32> <i32 32, i32 32, i32 32, i32 32>)
219 define <4 x i32> @rotr_v4i32_shift_by_bitwidth(<4 x i32> %x) {
220 ; CHECK-LABEL: rotr_v4i32_shift_by_bitwidth:
223 %f = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %x, <4 x i32> %x, <4 x i32> <i32 32, i32 32, i32 32, i32 32>)