|
| 1 | +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 |
| 2 | +; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s |
| 3 | +; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve -aarch64-sve-vector-bits-min=128 -aarch64-sve-vector-bits-max=128 < %s | FileCheck %s --check-prefix=CHECK-128 |
| 4 | +; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve -aarch64-sve-vector-bits-min=256 -aarch64-sve-vector-bits-max=256 < %s | FileCheck %s --check-prefix=CHECK-256 |
| 5 | +; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve -aarch64-sve-vector-bits-min=512 -aarch64-sve-vector-bits-max=512 < %s | FileCheck %s --check-prefix=CHECK-512 |
| 6 | +; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve -aarch64-sve-vector-bits-min=1024 -aarch64-sve-vector-bits-max=1024 < %s | FileCheck %s --check-prefix=CHECK-1024 |
| 7 | +; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve -aarch64-sve-vector-bits-min=2048 -aarch64-sve-vector-bits-max=2048 < %s | FileCheck %s --check-prefix=CHECK-2048 |
| 8 | + |
| 9 | +define void @nxv16i8(ptr %ldptr, ptr %stptr) { |
| 10 | +; CHECK-LABEL: nxv16i8: |
| 11 | +; CHECK: // %bb.0: |
| 12 | +; CHECK-NEXT: ptrue p0.b |
| 13 | +; CHECK-NEXT: mov w8, #256 // =0x100 |
| 14 | +; CHECK-NEXT: ld1b { z0.b }, p0/z, [x0, x8] |
| 15 | +; CHECK-NEXT: st1b { z0.b }, p0, [x1, x8] |
| 16 | +; CHECK-NEXT: ret |
| 17 | +; |
| 18 | +; CHECK-128-LABEL: nxv16i8: |
| 19 | +; CHECK-128: // %bb.0: |
| 20 | +; CHECK-128-NEXT: ptrue p0.b |
| 21 | +; CHECK-128-NEXT: mov w8, #256 // =0x100 |
| 22 | +; CHECK-128-NEXT: ld1b { z0.b }, p0/z, [x0, x8] |
| 23 | +; CHECK-128-NEXT: st1b { z0.b }, p0, [x1, x8] |
| 24 | +; CHECK-128-NEXT: ret |
| 25 | +; |
| 26 | +; CHECK-256-LABEL: nxv16i8: |
| 27 | +; CHECK-256: // %bb.0: |
| 28 | +; CHECK-256-NEXT: ptrue p0.b |
| 29 | +; CHECK-256-NEXT: mov w8, #256 // =0x100 |
| 30 | +; CHECK-256-NEXT: ld1b { z0.b }, p0/z, [x0, x8] |
| 31 | +; CHECK-256-NEXT: st1b { z0.b }, p0, [x1, x8] |
| 32 | +; CHECK-256-NEXT: ret |
| 33 | +; |
| 34 | +; CHECK-512-LABEL: nxv16i8: |
| 35 | +; CHECK-512: // %bb.0: |
| 36 | +; CHECK-512-NEXT: ptrue p0.b |
| 37 | +; CHECK-512-NEXT: mov w8, #256 // =0x100 |
| 38 | +; CHECK-512-NEXT: ld1b { z0.b }, p0/z, [x0, x8] |
| 39 | +; CHECK-512-NEXT: st1b { z0.b }, p0, [x1, x8] |
| 40 | +; CHECK-512-NEXT: ret |
| 41 | +; |
| 42 | +; CHECK-1024-LABEL: nxv16i8: |
| 43 | +; CHECK-1024: // %bb.0: |
| 44 | +; CHECK-1024-NEXT: ptrue p0.b |
| 45 | +; CHECK-1024-NEXT: mov w8, #256 // =0x100 |
| 46 | +; CHECK-1024-NEXT: ld1b { z0.b }, p0/z, [x0, x8] |
| 47 | +; CHECK-1024-NEXT: st1b { z0.b }, p0, [x1, x8] |
| 48 | +; CHECK-1024-NEXT: ret |
| 49 | +; |
| 50 | +; CHECK-2048-LABEL: nxv16i8: |
| 51 | +; CHECK-2048: // %bb.0: |
| 52 | +; CHECK-2048-NEXT: ptrue p0.b |
| 53 | +; CHECK-2048-NEXT: mov w8, #256 // =0x100 |
| 54 | +; CHECK-2048-NEXT: ld1b { z0.b }, p0/z, [x0, x8] |
| 55 | +; CHECK-2048-NEXT: st1b { z0.b }, p0, [x1, x8] |
| 56 | +; CHECK-2048-NEXT: ret |
| 57 | + %ldoff = getelementptr inbounds nuw i8, ptr %ldptr, i64 256 |
| 58 | + %stoff = getelementptr inbounds nuw i8, ptr %stptr, i64 256 |
| 59 | + %x = load <vscale x 16 x i8>, ptr %ldoff, align 1 |
| 60 | + store <vscale x 16 x i8> %x, ptr %stoff, align 1 |
| 61 | + ret void |
| 62 | +} |
| 63 | + |
| 64 | +define void @nxv8i16(ptr %ldptr, ptr %stptr) { |
| 65 | +; CHECK-LABEL: nxv8i16: |
| 66 | +; CHECK: // %bb.0: |
| 67 | +; CHECK-NEXT: ptrue p0.h |
| 68 | +; CHECK-NEXT: mov x8, #128 // =0x80 |
| 69 | +; CHECK-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1] |
| 70 | +; CHECK-NEXT: st1h { z0.h }, p0, [x1, x8, lsl #1] |
| 71 | +; CHECK-NEXT: ret |
| 72 | +; |
| 73 | +; CHECK-128-LABEL: nxv8i16: |
| 74 | +; CHECK-128: // %bb.0: |
| 75 | +; CHECK-128-NEXT: ptrue p0.h |
| 76 | +; CHECK-128-NEXT: mov x8, #128 // =0x80 |
| 77 | +; CHECK-128-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1] |
| 78 | +; CHECK-128-NEXT: st1h { z0.h }, p0, [x1, x8, lsl #1] |
| 79 | +; CHECK-128-NEXT: ret |
| 80 | +; |
| 81 | +; CHECK-256-LABEL: nxv8i16: |
| 82 | +; CHECK-256: // %bb.0: |
| 83 | +; CHECK-256-NEXT: ptrue p0.h |
| 84 | +; CHECK-256-NEXT: mov x8, #128 // =0x80 |
| 85 | +; CHECK-256-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1] |
| 86 | +; CHECK-256-NEXT: st1h { z0.h }, p0, [x1, x8, lsl #1] |
| 87 | +; CHECK-256-NEXT: ret |
| 88 | +; |
| 89 | +; CHECK-512-LABEL: nxv8i16: |
| 90 | +; CHECK-512: // %bb.0: |
| 91 | +; CHECK-512-NEXT: ptrue p0.h |
| 92 | +; CHECK-512-NEXT: mov x8, #128 // =0x80 |
| 93 | +; CHECK-512-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1] |
| 94 | +; CHECK-512-NEXT: st1h { z0.h }, p0, [x1, x8, lsl #1] |
| 95 | +; CHECK-512-NEXT: ret |
| 96 | +; |
| 97 | +; CHECK-1024-LABEL: nxv8i16: |
| 98 | +; CHECK-1024: // %bb.0: |
| 99 | +; CHECK-1024-NEXT: ptrue p0.h |
| 100 | +; CHECK-1024-NEXT: mov x8, #128 // =0x80 |
| 101 | +; CHECK-1024-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1] |
| 102 | +; CHECK-1024-NEXT: st1h { z0.h }, p0, [x1, x8, lsl #1] |
| 103 | +; CHECK-1024-NEXT: ret |
| 104 | +; |
| 105 | +; CHECK-2048-LABEL: nxv8i16: |
| 106 | +; CHECK-2048: // %bb.0: |
| 107 | +; CHECK-2048-NEXT: ptrue p0.h |
| 108 | +; CHECK-2048-NEXT: mov x8, #128 // =0x80 |
| 109 | +; CHECK-2048-NEXT: ld1h { z0.h }, p0/z, [x0, x8, lsl #1] |
| 110 | +; CHECK-2048-NEXT: st1h { z0.h }, p0, [x1, x8, lsl #1] |
| 111 | +; CHECK-2048-NEXT: ret |
| 112 | + %ldoff = getelementptr inbounds nuw i16, ptr %ldptr, i64 128 |
| 113 | + %stoff = getelementptr inbounds nuw i16, ptr %stptr, i64 128 |
| 114 | + %x = load <vscale x 8 x i16>, ptr %ldoff, align 2 |
| 115 | + store <vscale x 8 x i16> %x, ptr %stoff, align 2 |
| 116 | + ret void |
| 117 | +} |
| 118 | + |
| 119 | +define void @nxv4i32(ptr %ldptr, ptr %stptr) { |
| 120 | +; CHECK-LABEL: nxv4i32: |
| 121 | +; CHECK: // %bb.0: |
| 122 | +; CHECK-NEXT: ptrue p0.s |
| 123 | +; CHECK-NEXT: mov x8, #64 // =0x40 |
| 124 | +; CHECK-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2] |
| 125 | +; CHECK-NEXT: st1w { z0.s }, p0, [x1, x8, lsl #2] |
| 126 | +; CHECK-NEXT: ret |
| 127 | +; |
| 128 | +; CHECK-128-LABEL: nxv4i32: |
| 129 | +; CHECK-128: // %bb.0: |
| 130 | +; CHECK-128-NEXT: ptrue p0.s |
| 131 | +; CHECK-128-NEXT: mov x8, #64 // =0x40 |
| 132 | +; CHECK-128-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2] |
| 133 | +; CHECK-128-NEXT: st1w { z0.s }, p0, [x1, x8, lsl #2] |
| 134 | +; CHECK-128-NEXT: ret |
| 135 | +; |
| 136 | +; CHECK-256-LABEL: nxv4i32: |
| 137 | +; CHECK-256: // %bb.0: |
| 138 | +; CHECK-256-NEXT: ptrue p0.s |
| 139 | +; CHECK-256-NEXT: mov x8, #64 // =0x40 |
| 140 | +; CHECK-256-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2] |
| 141 | +; CHECK-256-NEXT: st1w { z0.s }, p0, [x1, x8, lsl #2] |
| 142 | +; CHECK-256-NEXT: ret |
| 143 | +; |
| 144 | +; CHECK-512-LABEL: nxv4i32: |
| 145 | +; CHECK-512: // %bb.0: |
| 146 | +; CHECK-512-NEXT: ptrue p0.s |
| 147 | +; CHECK-512-NEXT: mov x8, #64 // =0x40 |
| 148 | +; CHECK-512-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2] |
| 149 | +; CHECK-512-NEXT: st1w { z0.s }, p0, [x1, x8, lsl #2] |
| 150 | +; CHECK-512-NEXT: ret |
| 151 | +; |
| 152 | +; CHECK-1024-LABEL: nxv4i32: |
| 153 | +; CHECK-1024: // %bb.0: |
| 154 | +; CHECK-1024-NEXT: ptrue p0.s |
| 155 | +; CHECK-1024-NEXT: mov x8, #64 // =0x40 |
| 156 | +; CHECK-1024-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2] |
| 157 | +; CHECK-1024-NEXT: st1w { z0.s }, p0, [x1, x8, lsl #2] |
| 158 | +; CHECK-1024-NEXT: ret |
| 159 | +; |
| 160 | +; CHECK-2048-LABEL: nxv4i32: |
| 161 | +; CHECK-2048: // %bb.0: |
| 162 | +; CHECK-2048-NEXT: ptrue p0.s |
| 163 | +; CHECK-2048-NEXT: mov x8, #64 // =0x40 |
| 164 | +; CHECK-2048-NEXT: ld1w { z0.s }, p0/z, [x0, x8, lsl #2] |
| 165 | +; CHECK-2048-NEXT: st1w { z0.s }, p0, [x1, x8, lsl #2] |
| 166 | +; CHECK-2048-NEXT: ret |
| 167 | + %ldoff = getelementptr inbounds nuw i32, ptr %ldptr, i64 64 |
| 168 | + %stoff = getelementptr inbounds nuw i32, ptr %stptr, i64 64 |
| 169 | + %x = load <vscale x 4 x i32>, ptr %ldoff, align 4 |
| 170 | + store <vscale x 4 x i32> %x, ptr %stoff, align 4 |
| 171 | + ret void |
| 172 | +} |
| 173 | + |
| 174 | +define void @nxv2i64(ptr %ldptr, ptr %stptr) { |
| 175 | +; CHECK-LABEL: nxv2i64: |
| 176 | +; CHECK: // %bb.0: |
| 177 | +; CHECK-NEXT: ptrue p0.d |
| 178 | +; CHECK-NEXT: mov x8, #32 // =0x20 |
| 179 | +; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3] |
| 180 | +; CHECK-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3] |
| 181 | +; CHECK-NEXT: ret |
| 182 | +; |
| 183 | +; CHECK-128-LABEL: nxv2i64: |
| 184 | +; CHECK-128: // %bb.0: |
| 185 | +; CHECK-128-NEXT: ptrue p0.d |
| 186 | +; CHECK-128-NEXT: mov x8, #32 // =0x20 |
| 187 | +; CHECK-128-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3] |
| 188 | +; CHECK-128-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3] |
| 189 | +; CHECK-128-NEXT: ret |
| 190 | +; |
| 191 | +; CHECK-256-LABEL: nxv2i64: |
| 192 | +; CHECK-256: // %bb.0: |
| 193 | +; CHECK-256-NEXT: ptrue p0.d |
| 194 | +; CHECK-256-NEXT: mov x8, #32 // =0x20 |
| 195 | +; CHECK-256-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3] |
| 196 | +; CHECK-256-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3] |
| 197 | +; CHECK-256-NEXT: ret |
| 198 | +; |
| 199 | +; CHECK-512-LABEL: nxv2i64: |
| 200 | +; CHECK-512: // %bb.0: |
| 201 | +; CHECK-512-NEXT: ptrue p0.d |
| 202 | +; CHECK-512-NEXT: mov x8, #32 // =0x20 |
| 203 | +; CHECK-512-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3] |
| 204 | +; CHECK-512-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3] |
| 205 | +; CHECK-512-NEXT: ret |
| 206 | +; |
| 207 | +; CHECK-1024-LABEL: nxv2i64: |
| 208 | +; CHECK-1024: // %bb.0: |
| 209 | +; CHECK-1024-NEXT: ptrue p0.d |
| 210 | +; CHECK-1024-NEXT: mov x8, #32 // =0x20 |
| 211 | +; CHECK-1024-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3] |
| 212 | +; CHECK-1024-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3] |
| 213 | +; CHECK-1024-NEXT: ret |
| 214 | +; |
| 215 | +; CHECK-2048-LABEL: nxv2i64: |
| 216 | +; CHECK-2048: // %bb.0: |
| 217 | +; CHECK-2048-NEXT: ptrue p0.d |
| 218 | +; CHECK-2048-NEXT: mov x8, #32 // =0x20 |
| 219 | +; CHECK-2048-NEXT: ld1d { z0.d }, p0/z, [x0, x8, lsl #3] |
| 220 | +; CHECK-2048-NEXT: st1d { z0.d }, p0, [x1, x8, lsl #3] |
| 221 | +; CHECK-2048-NEXT: ret |
| 222 | + %ldoff = getelementptr inbounds nuw i64, ptr %ldptr, i64 32 |
| 223 | + %stoff = getelementptr inbounds nuw i64, ptr %stptr, i64 32 |
| 224 | + %x = load <vscale x 2 x i64>, ptr %ldoff, align 8 |
| 225 | + store <vscale x 2 x i64> %x, ptr %stoff, align 8 |
| 226 | + ret void |
| 227 | +} |
0 commit comments