|
| 1 | +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py |
| 2 | +; RUN: llc -verify-machineinstrs -mtriple=powerpc64-- -mattr=+popcntd < %s | FileCheck %s --check-prefixes=ANY,FAST |
| 3 | +; RUN: llc -verify-machineinstrs -mtriple=powerpc64-- -mattr=+slow-popcntd < %s | FileCheck %s --check-prefixes=ANY,SLOW |
| 4 | + |
| 5 | +define i16 @zpop_i8_i16(i8 %x) { |
| 6 | +; FAST-LABEL: zpop_i8_i16: |
| 7 | +; FAST: # %bb.0: |
| 8 | +; FAST-NEXT: rlwinm 3, 3, 0, 24, 31 |
| 9 | +; FAST-NEXT: popcntw 3, 3 |
| 10 | +; FAST-NEXT: blr |
| 11 | +; |
| 12 | +; SLOW-LABEL: zpop_i8_i16: |
| 13 | +; SLOW: # %bb.0: |
| 14 | +; SLOW-NEXT: clrlwi 5, 3, 24 |
| 15 | +; SLOW-NEXT: rlwinm 3, 3, 31, 0, 31 |
| 16 | +; SLOW-NEXT: andi. 3, 3, 85 |
| 17 | +; SLOW-NEXT: lis 4, 13107 |
| 18 | +; SLOW-NEXT: subf 3, 3, 5 |
| 19 | +; SLOW-NEXT: ori 4, 4, 13107 |
| 20 | +; SLOW-NEXT: rotlwi 5, 3, 30 |
| 21 | +; SLOW-NEXT: and 3, 3, 4 |
| 22 | +; SLOW-NEXT: andis. 4, 5, 13107 |
| 23 | +; SLOW-NEXT: andi. 5, 5, 13107 |
| 24 | +; SLOW-NEXT: or 4, 5, 4 |
| 25 | +; SLOW-NEXT: add 3, 3, 4 |
| 26 | +; SLOW-NEXT: lis 5, 3855 |
| 27 | +; SLOW-NEXT: srwi 4, 3, 4 |
| 28 | +; SLOW-NEXT: add 3, 3, 4 |
| 29 | +; SLOW-NEXT: lis 4, 257 |
| 30 | +; SLOW-NEXT: ori 5, 5, 3855 |
| 31 | +; SLOW-NEXT: and 3, 3, 5 |
| 32 | +; SLOW-NEXT: ori 4, 4, 257 |
| 33 | +; SLOW-NEXT: mullw 3, 3, 4 |
| 34 | +; SLOW-NEXT: srwi 3, 3, 24 |
| 35 | +; SLOW-NEXT: blr |
| 36 | + %z = zext i8 %x to i16 |
| 37 | + %pop = tail call i16 @llvm.ctpop.i16(i16 %z) |
| 38 | + ret i16 %pop |
| 39 | +} |
| 40 | + |
| 41 | +define i16 @popz_i8_i16(i8 %x) { |
| 42 | +; FAST-LABEL: popz_i8_i16: |
| 43 | +; FAST: # %bb.0: |
| 44 | +; FAST-NEXT: rlwinm 3, 3, 0, 24, 31 |
| 45 | +; FAST-NEXT: popcntw 3, 3 |
| 46 | +; FAST-NEXT: clrldi 3, 3, 32 |
| 47 | +; FAST-NEXT: blr |
| 48 | +; |
| 49 | +; SLOW-LABEL: popz_i8_i16: |
| 50 | +; SLOW: # %bb.0: |
| 51 | +; SLOW-NEXT: clrlwi 5, 3, 24 |
| 52 | +; SLOW-NEXT: rlwinm 3, 3, 31, 0, 31 |
| 53 | +; SLOW-NEXT: andi. 3, 3, 85 |
| 54 | +; SLOW-NEXT: lis 4, 13107 |
| 55 | +; SLOW-NEXT: subf 3, 3, 5 |
| 56 | +; SLOW-NEXT: ori 4, 4, 13107 |
| 57 | +; SLOW-NEXT: rotlwi 5, 3, 30 |
| 58 | +; SLOW-NEXT: and 3, 3, 4 |
| 59 | +; SLOW-NEXT: andis. 4, 5, 13107 |
| 60 | +; SLOW-NEXT: andi. 5, 5, 13107 |
| 61 | +; SLOW-NEXT: or 4, 5, 4 |
| 62 | +; SLOW-NEXT: add 3, 3, 4 |
| 63 | +; SLOW-NEXT: lis 5, 3855 |
| 64 | +; SLOW-NEXT: srwi 4, 3, 4 |
| 65 | +; SLOW-NEXT: add 3, 3, 4 |
| 66 | +; SLOW-NEXT: lis 4, 257 |
| 67 | +; SLOW-NEXT: ori 5, 5, 3855 |
| 68 | +; SLOW-NEXT: and 3, 3, 5 |
| 69 | +; SLOW-NEXT: ori 4, 4, 257 |
| 70 | +; SLOW-NEXT: mullw 3, 3, 4 |
| 71 | +; SLOW-NEXT: rlwinm 3, 3, 8, 24, 31 |
| 72 | +; SLOW-NEXT: blr |
| 73 | + %pop = tail call i8 @llvm.ctpop.i8(i8 %x) |
| 74 | + %z = zext i8 %pop to i16 |
| 75 | + ret i16 %z |
| 76 | +} |
| 77 | + |
| 78 | +define i32 @zpop_i8_i32(i8 %x) { |
| 79 | +; FAST-LABEL: zpop_i8_i32: |
| 80 | +; FAST: # %bb.0: |
| 81 | +; FAST-NEXT: rlwinm 3, 3, 0, 24, 31 |
| 82 | +; FAST-NEXT: popcntw 3, 3 |
| 83 | +; FAST-NEXT: blr |
| 84 | +; |
| 85 | +; SLOW-LABEL: zpop_i8_i32: |
| 86 | +; SLOW: # %bb.0: |
| 87 | +; SLOW-NEXT: clrlwi 5, 3, 24 |
| 88 | +; SLOW-NEXT: rlwinm 3, 3, 31, 0, 31 |
| 89 | +; SLOW-NEXT: andi. 3, 3, 85 |
| 90 | +; SLOW-NEXT: lis 4, 13107 |
| 91 | +; SLOW-NEXT: subf 3, 3, 5 |
| 92 | +; SLOW-NEXT: ori 4, 4, 13107 |
| 93 | +; SLOW-NEXT: rotlwi 5, 3, 30 |
| 94 | +; SLOW-NEXT: and 3, 3, 4 |
| 95 | +; SLOW-NEXT: andis. 4, 5, 13107 |
| 96 | +; SLOW-NEXT: andi. 5, 5, 13107 |
| 97 | +; SLOW-NEXT: or 4, 5, 4 |
| 98 | +; SLOW-NEXT: add 3, 3, 4 |
| 99 | +; SLOW-NEXT: lis 5, 3855 |
| 100 | +; SLOW-NEXT: srwi 4, 3, 4 |
| 101 | +; SLOW-NEXT: add 3, 3, 4 |
| 102 | +; SLOW-NEXT: lis 4, 257 |
| 103 | +; SLOW-NEXT: ori 5, 5, 3855 |
| 104 | +; SLOW-NEXT: and 3, 3, 5 |
| 105 | +; SLOW-NEXT: ori 4, 4, 257 |
| 106 | +; SLOW-NEXT: mullw 3, 3, 4 |
| 107 | +; SLOW-NEXT: srwi 3, 3, 24 |
| 108 | +; SLOW-NEXT: blr |
| 109 | + %z = zext i8 %x to i32 |
| 110 | + %pop = tail call i32 @llvm.ctpop.i32(i32 %z) |
| 111 | + ret i32 %pop |
| 112 | +} |
| 113 | + |
| 114 | +define i32 @popz_i8_32(i8 %x) { |
| 115 | +; FAST-LABEL: popz_i8_32: |
| 116 | +; FAST: # %bb.0: |
| 117 | +; FAST-NEXT: rlwinm 3, 3, 0, 24, 31 |
| 118 | +; FAST-NEXT: popcntw 3, 3 |
| 119 | +; FAST-NEXT: clrldi 3, 3, 32 |
| 120 | +; FAST-NEXT: blr |
| 121 | +; |
| 122 | +; SLOW-LABEL: popz_i8_32: |
| 123 | +; SLOW: # %bb.0: |
| 124 | +; SLOW-NEXT: clrlwi 5, 3, 24 |
| 125 | +; SLOW-NEXT: rlwinm 3, 3, 31, 0, 31 |
| 126 | +; SLOW-NEXT: andi. 3, 3, 85 |
| 127 | +; SLOW-NEXT: lis 4, 13107 |
| 128 | +; SLOW-NEXT: subf 3, 3, 5 |
| 129 | +; SLOW-NEXT: ori 4, 4, 13107 |
| 130 | +; SLOW-NEXT: rotlwi 5, 3, 30 |
| 131 | +; SLOW-NEXT: and 3, 3, 4 |
| 132 | +; SLOW-NEXT: andis. 4, 5, 13107 |
| 133 | +; SLOW-NEXT: andi. 5, 5, 13107 |
| 134 | +; SLOW-NEXT: or 4, 5, 4 |
| 135 | +; SLOW-NEXT: add 3, 3, 4 |
| 136 | +; SLOW-NEXT: lis 5, 3855 |
| 137 | +; SLOW-NEXT: srwi 4, 3, 4 |
| 138 | +; SLOW-NEXT: add 3, 3, 4 |
| 139 | +; SLOW-NEXT: lis 4, 257 |
| 140 | +; SLOW-NEXT: ori 5, 5, 3855 |
| 141 | +; SLOW-NEXT: and 3, 3, 5 |
| 142 | +; SLOW-NEXT: ori 4, 4, 257 |
| 143 | +; SLOW-NEXT: mullw 3, 3, 4 |
| 144 | +; SLOW-NEXT: rlwinm 3, 3, 8, 24, 31 |
| 145 | +; SLOW-NEXT: blr |
| 146 | + %pop = tail call i8 @llvm.ctpop.i8(i8 %x) |
| 147 | + %z = zext i8 %pop to i32 |
| 148 | + ret i32 %z |
| 149 | +} |
| 150 | + |
| 151 | +define i32 @zpop_i16_i32(i16 %x) { |
| 152 | +; FAST-LABEL: zpop_i16_i32: |
| 153 | +; FAST: # %bb.0: |
| 154 | +; FAST-NEXT: rlwinm 3, 3, 0, 16, 31 |
| 155 | +; FAST-NEXT: popcntw 3, 3 |
| 156 | +; FAST-NEXT: blr |
| 157 | +; |
| 158 | +; SLOW-LABEL: zpop_i16_i32: |
| 159 | +; SLOW: # %bb.0: |
| 160 | +; SLOW-NEXT: clrlwi 5, 3, 16 |
| 161 | +; SLOW-NEXT: rlwinm 3, 3, 31, 0, 31 |
| 162 | +; SLOW-NEXT: andi. 3, 3, 21845 |
| 163 | +; SLOW-NEXT: lis 4, 13107 |
| 164 | +; SLOW-NEXT: subf 3, 3, 5 |
| 165 | +; SLOW-NEXT: ori 4, 4, 13107 |
| 166 | +; SLOW-NEXT: rotlwi 5, 3, 30 |
| 167 | +; SLOW-NEXT: and 3, 3, 4 |
| 168 | +; SLOW-NEXT: andis. 4, 5, 13107 |
| 169 | +; SLOW-NEXT: andi. 5, 5, 13107 |
| 170 | +; SLOW-NEXT: or 4, 5, 4 |
| 171 | +; SLOW-NEXT: add 3, 3, 4 |
| 172 | +; SLOW-NEXT: lis 5, 3855 |
| 173 | +; SLOW-NEXT: srwi 4, 3, 4 |
| 174 | +; SLOW-NEXT: add 3, 3, 4 |
| 175 | +; SLOW-NEXT: lis 4, 257 |
| 176 | +; SLOW-NEXT: ori 5, 5, 3855 |
| 177 | +; SLOW-NEXT: and 3, 3, 5 |
| 178 | +; SLOW-NEXT: ori 4, 4, 257 |
| 179 | +; SLOW-NEXT: mullw 3, 3, 4 |
| 180 | +; SLOW-NEXT: srwi 3, 3, 24 |
| 181 | +; SLOW-NEXT: blr |
| 182 | + %z = zext i16 %x to i32 |
| 183 | + %pop = tail call i32 @llvm.ctpop.i32(i32 %z) |
| 184 | + ret i32 %pop |
| 185 | +} |
| 186 | + |
| 187 | +define i32 @popz_i16_32(i16 %x) { |
| 188 | +; FAST-LABEL: popz_i16_32: |
| 189 | +; FAST: # %bb.0: |
| 190 | +; FAST-NEXT: rlwinm 3, 3, 0, 16, 31 |
| 191 | +; FAST-NEXT: popcntw 3, 3 |
| 192 | +; FAST-NEXT: clrldi 3, 3, 32 |
| 193 | +; FAST-NEXT: blr |
| 194 | +; |
| 195 | +; SLOW-LABEL: popz_i16_32: |
| 196 | +; SLOW: # %bb.0: |
| 197 | +; SLOW-NEXT: clrlwi 5, 3, 16 |
| 198 | +; SLOW-NEXT: rlwinm 3, 3, 31, 0, 31 |
| 199 | +; SLOW-NEXT: andi. 3, 3, 21845 |
| 200 | +; SLOW-NEXT: lis 4, 13107 |
| 201 | +; SLOW-NEXT: subf 3, 3, 5 |
| 202 | +; SLOW-NEXT: ori 4, 4, 13107 |
| 203 | +; SLOW-NEXT: rotlwi 5, 3, 30 |
| 204 | +; SLOW-NEXT: and 3, 3, 4 |
| 205 | +; SLOW-NEXT: andis. 4, 5, 13107 |
| 206 | +; SLOW-NEXT: andi. 5, 5, 13107 |
| 207 | +; SLOW-NEXT: or 4, 5, 4 |
| 208 | +; SLOW-NEXT: add 3, 3, 4 |
| 209 | +; SLOW-NEXT: lis 5, 3855 |
| 210 | +; SLOW-NEXT: srwi 4, 3, 4 |
| 211 | +; SLOW-NEXT: add 3, 3, 4 |
| 212 | +; SLOW-NEXT: lis 4, 257 |
| 213 | +; SLOW-NEXT: ori 5, 5, 3855 |
| 214 | +; SLOW-NEXT: and 3, 3, 5 |
| 215 | +; SLOW-NEXT: ori 4, 4, 257 |
| 216 | +; SLOW-NEXT: mullw 3, 3, 4 |
| 217 | +; SLOW-NEXT: rlwinm 3, 3, 8, 24, 31 |
| 218 | +; SLOW-NEXT: blr |
| 219 | + %pop = tail call i16 @llvm.ctpop.i16(i16 %x) |
| 220 | + %z = zext i16 %pop to i32 |
| 221 | + ret i32 %z |
| 222 | +} |
| 223 | + |
| 224 | +define i64 @zpop_i32_i64(i32 %x) { |
| 225 | +; FAST-LABEL: zpop_i32_i64: |
| 226 | +; FAST: # %bb.0: |
| 227 | +; FAST-NEXT: clrldi 3, 3, 32 |
| 228 | +; FAST-NEXT: popcntd 3, 3 |
| 229 | +; FAST-NEXT: blr |
| 230 | +; |
| 231 | +; SLOW-LABEL: zpop_i32_i64: |
| 232 | +; SLOW: # %bb.0: |
| 233 | +; SLOW-NEXT: rlwinm 5, 3, 31, 1, 0 |
| 234 | +; SLOW-NEXT: lis 4, 13107 |
| 235 | +; SLOW-NEXT: andis. 6, 5, 21845 |
| 236 | +; SLOW-NEXT: andi. 5, 5, 21845 |
| 237 | +; SLOW-NEXT: ori 4, 4, 13107 |
| 238 | +; SLOW-NEXT: or 5, 5, 6 |
| 239 | +; SLOW-NEXT: clrldi 3, 3, 32 |
| 240 | +; SLOW-NEXT: rldimi 4, 4, 32, 0 |
| 241 | +; SLOW-NEXT: sub 3, 3, 5 |
| 242 | +; SLOW-NEXT: and 5, 3, 4 |
| 243 | +; SLOW-NEXT: rotldi 3, 3, 62 |
| 244 | +; SLOW-NEXT: and 3, 3, 4 |
| 245 | +; SLOW-NEXT: add 3, 5, 3 |
| 246 | +; SLOW-NEXT: lis 4, 3855 |
| 247 | +; SLOW-NEXT: rldicl 5, 3, 60, 4 |
| 248 | +; SLOW-NEXT: ori 4, 4, 3855 |
| 249 | +; SLOW-NEXT: add 3, 3, 5 |
| 250 | +; SLOW-NEXT: lis 5, 257 |
| 251 | +; SLOW-NEXT: rldimi 4, 4, 32, 0 |
| 252 | +; SLOW-NEXT: ori 5, 5, 257 |
| 253 | +; SLOW-NEXT: and 3, 3, 4 |
| 254 | +; SLOW-NEXT: rldimi 5, 5, 32, 0 |
| 255 | +; SLOW-NEXT: mulld 3, 3, 5 |
| 256 | +; SLOW-NEXT: rldicl 3, 3, 8, 56 |
| 257 | +; SLOW-NEXT: blr |
| 258 | + %z = zext i32 %x to i64 |
| 259 | + %pop = tail call i64 @llvm.ctpop.i64(i64 %z) |
| 260 | + ret i64 %pop |
| 261 | +} |
| 262 | + |
| 263 | +define i64 @popz_i32_i64(i32 %x) { |
| 264 | +; FAST-LABEL: popz_i32_i64: |
| 265 | +; FAST: # %bb.0: |
| 266 | +; FAST-NEXT: popcntw 3, 3 |
| 267 | +; FAST-NEXT: clrldi 3, 3, 32 |
| 268 | +; FAST-NEXT: blr |
| 269 | +; |
| 270 | +; SLOW-LABEL: popz_i32_i64: |
| 271 | +; SLOW: # %bb.0: |
| 272 | +; SLOW-NEXT: rotlwi 5, 3, 31 |
| 273 | +; SLOW-NEXT: andis. 6, 5, 21845 |
| 274 | +; SLOW-NEXT: andi. 5, 5, 21845 |
| 275 | +; SLOW-NEXT: or 5, 5, 6 |
| 276 | +; SLOW-NEXT: lis 4, 13107 |
| 277 | +; SLOW-NEXT: subf 3, 5, 3 |
| 278 | +; SLOW-NEXT: ori 4, 4, 13107 |
| 279 | +; SLOW-NEXT: rotlwi 5, 3, 30 |
| 280 | +; SLOW-NEXT: and 3, 3, 4 |
| 281 | +; SLOW-NEXT: andis. 4, 5, 13107 |
| 282 | +; SLOW-NEXT: andi. 5, 5, 13107 |
| 283 | +; SLOW-NEXT: or 4, 5, 4 |
| 284 | +; SLOW-NEXT: add 3, 3, 4 |
| 285 | +; SLOW-NEXT: lis 5, 3855 |
| 286 | +; SLOW-NEXT: srwi 4, 3, 4 |
| 287 | +; SLOW-NEXT: add 3, 3, 4 |
| 288 | +; SLOW-NEXT: lis 4, 257 |
| 289 | +; SLOW-NEXT: ori 5, 5, 3855 |
| 290 | +; SLOW-NEXT: and 3, 3, 5 |
| 291 | +; SLOW-NEXT: ori 4, 4, 257 |
| 292 | +; SLOW-NEXT: mullw 3, 3, 4 |
| 293 | +; SLOW-NEXT: rlwinm 3, 3, 8, 24, 31 |
| 294 | +; SLOW-NEXT: blr |
| 295 | + %pop = tail call i32 @llvm.ctpop.i32(i32 %x) |
| 296 | + %z = zext i32 %pop to i64 |
| 297 | + ret i64 %z |
| 298 | +} |
| 299 | + |
| 300 | +declare i8 @llvm.ctpop.i8(i8) nounwind readnone |
| 301 | +declare i16 @llvm.ctpop.i16(i16) nounwind readnone |
| 302 | +declare i32 @llvm.ctpop.i32(i32) nounwind readnone |
| 303 | +declare i64 @llvm.ctpop.i64(i64) nounwind readnone |
| 304 | + |
0 commit comments