Mike Klein | 267f507 | 2019-06-03 16:27:46 -0500 | [diff] [blame] | 1 | A8 over A8 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 2 | 14 values (originally 17): |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 3 | v0 = load8 arg(0) |
| 4 | v1 = to_f32 v0 |
| 5 | ↑ v2 = splat 3F800000 (1) |
| 6 | ↑ v3 = splat 3B808081 (0.0039215689) |
| 7 | v4 = fnma_f32 v1 v3 v2 |
| 8 | v5 = load8 arg(1) |
| 9 | v6 = to_f32 v5 |
| 10 | v7 = mul_f32 v6 v3 |
| 11 | v8 = mul_f32 v1 v3 |
| 12 | v9 = fma_f32 v7 v4 v8 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 13 | ↑ v10 = splat 437F0000 (255) |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 14 | v11 = mul_f32 v9 v10 |
| 15 | v12 = round v11 |
| 16 | store8 arg(1) v12 |
Mike Klein | 8c1e0ef | 2019-11-12 09:07:23 -0600 | [diff] [blame] | 17 | |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 18 | 6 registers, 14 instructions: |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 19 | 0 r0 = splat 3F800000 (1) |
| 20 | 1 r1 = splat 3B808081 (0.0039215689) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 21 | 2 r2 = splat 437F0000 (255) |
Mike Klein | 8c1e0ef | 2019-11-12 09:07:23 -0600 | [diff] [blame] | 22 | loop: |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 23 | 3 r3 = load8 arg(0) |
| 24 | 4 r3 = to_f32 r3 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 25 | 5 r4 = fnma_f32 r3 r1 r0 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 26 | 6 r5 = load8 arg(1) |
| 27 | 7 r5 = to_f32 r5 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 28 | 8 r5 = mul_f32 r5 r1 |
| 29 | 9 r3 = mul_f32 r3 r1 |
| 30 | 10 r3 = fma_f32 r5 r4 r3 |
| 31 | 11 r3 = mul_f32 r3 r2 |
| 32 | 12 r3 = round r3 |
| 33 | 13 store8 arg(1) r3 |
Mike Klein | 8c1e0ef | 2019-11-12 09:07:23 -0600 | [diff] [blame] | 34 | |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 35 | A8 over G8 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 36 | 19 values (originally 24): |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 37 | v0 = load8 arg(0) |
| 38 | v1 = to_f32 v0 |
| 39 | ↑ v2 = splat 3F800000 (1) |
| 40 | ↑ v3 = splat 3B808081 (0.0039215689) |
| 41 | v4 = fnma_f32 v1 v3 v2 |
| 42 | v5 = load8 arg(1) |
| 43 | v6 = to_f32 v5 |
| 44 | v7 = mul_f32 v6 v3 |
| 45 | v8 = mul_f32 v7 v4 |
| 46 | ↑ v9 = splat 3D93DD98 (0.0722) |
| 47 | v10 = mul_f32 v8 v9 |
| 48 | ↑ v11 = splat 3F371759 (0.71520001) |
| 49 | v12 = fma_f32 v8 v11 v10 |
| 50 | ↑ v13 = splat 3E59B3D0 (0.21259999) |
| 51 | v14 = fma_f32 v8 v13 v12 |
| 52 | ↑ v15 = splat 437F0000 (255) |
| 53 | v16 = mul_f32 v14 v15 |
| 54 | v17 = round v16 |
| 55 | store8 arg(1) v17 |
| 56 | |
| 57 | 8 registers, 19 instructions: |
| 58 | 0 r0 = splat 3F800000 (1) |
| 59 | 1 r1 = splat 3B808081 (0.0039215689) |
| 60 | 2 r2 = splat 3D93DD98 (0.0722) |
| 61 | 3 r3 = splat 3F371759 (0.71520001) |
| 62 | 4 r4 = splat 3E59B3D0 (0.21259999) |
| 63 | 5 r5 = splat 437F0000 (255) |
| 64 | loop: |
| 65 | 6 r6 = load8 arg(0) |
| 66 | 7 r6 = to_f32 r6 |
| 67 | 8 r6 = fnma_f32 r6 r1 r0 |
| 68 | 9 r7 = load8 arg(1) |
| 69 | 10 r7 = to_f32 r7 |
| 70 | 11 r7 = mul_f32 r7 r1 |
| 71 | 12 r6 = mul_f32 r7 r6 |
| 72 | 13 r7 = mul_f32 r6 r2 |
| 73 | 14 r7 = fma_f32 r6 r3 r7 |
| 74 | 15 r7 = fma_f32 r6 r4 r7 |
| 75 | 16 r7 = mul_f32 r7 r5 |
| 76 | 17 r7 = round r7 |
| 77 | 18 store8 arg(1) r7 |
| 78 | |
| 79 | A8 over RGBA_8888 |
| 80 | 39 values (originally 41): |
| 81 | v0 = load8 arg(0) |
| 82 | v1 = to_f32 v0 |
| 83 | ↑ v2 = splat 3F800000 (1) |
| 84 | ↑ v3 = splat 3B808081 (0.0039215689) |
| 85 | v4 = fnma_f32 v1 v3 v2 |
| 86 | v5 = load32 arg(1) |
| 87 | v6 = shr_i32 v5 24 |
| 88 | v7 = to_f32 v6 |
| 89 | v8 = mul_f32 v7 v3 |
| 90 | v9 = mul_f32 v1 v3 |
| 91 | v10 = fma_f32 v8 v4 v9 |
| 92 | ↑ v11 = splat 437F0000 (255) |
| 93 | v12 = mul_f32 v10 v11 |
| 94 | v13 = shr_i32 v5 16 |
| 95 | ↑ v14 = splat FF (3.5733111e-43) |
| 96 | v15 = bit_and v14 v13 |
| 97 | v16 = round v12 |
| 98 | v17 = to_f32 v15 |
| 99 | v18 = mul_f32 v17 v3 |
| 100 | v19 = mul_f32 v18 v4 |
| 101 | v20 = mul_f32 v19 v11 |
| 102 | v21 = round v20 |
| 103 | v22 = pack v21 v16 8 |
| 104 | v23 = shr_i32 v5 8 |
| 105 | v24 = bit_and v14 v23 |
| 106 | v25 = to_f32 v24 |
| 107 | v26 = mul_f32 v25 v3 |
| 108 | v27 = mul_f32 v26 v4 |
| 109 | v28 = mul_f32 v27 v11 |
| 110 | v29 = round v28 |
| 111 | v30 = bit_and v14 v5 |
| 112 | v31 = to_f32 v30 |
| 113 | v32 = mul_f32 v31 v3 |
| 114 | v33 = mul_f32 v32 v4 |
| 115 | v34 = mul_f32 v33 v11 |
| 116 | v35 = round v34 |
| 117 | v36 = pack v35 v29 8 |
| 118 | v37 = pack v36 v22 16 |
| 119 | store32 arg(1) v37 |
| 120 | |
| 121 | 8 registers, 39 instructions: |
| 122 | 0 r0 = splat 3F800000 (1) |
| 123 | 1 r1 = splat 3B808081 (0.0039215689) |
| 124 | 2 r2 = splat 437F0000 (255) |
| 125 | 3 r3 = splat FF (3.5733111e-43) |
| 126 | loop: |
| 127 | 4 r4 = load8 arg(0) |
| 128 | 5 r4 = to_f32 r4 |
| 129 | 6 r5 = fnma_f32 r4 r1 r0 |
| 130 | 7 r6 = load32 arg(1) |
| 131 | 8 r7 = shr_i32 r6 24 |
| 132 | 9 r7 = to_f32 r7 |
| 133 | 10 r7 = mul_f32 r7 r1 |
| 134 | 11 r4 = mul_f32 r4 r1 |
| 135 | 12 r4 = fma_f32 r7 r5 r4 |
| 136 | 13 r4 = mul_f32 r4 r2 |
| 137 | 14 r7 = shr_i32 r6 16 |
| 138 | 15 r7 = bit_and r3 r7 |
| 139 | 16 r4 = round r4 |
| 140 | 17 r7 = to_f32 r7 |
| 141 | 18 r7 = mul_f32 r7 r1 |
| 142 | 19 r7 = mul_f32 r7 r5 |
| 143 | 20 r7 = mul_f32 r7 r2 |
| 144 | 21 r7 = round r7 |
| 145 | 22 r4 = pack r7 r4 8 |
| 146 | 23 r7 = shr_i32 r6 8 |
| 147 | 24 r7 = bit_and r3 r7 |
| 148 | 25 r7 = to_f32 r7 |
| 149 | 26 r7 = mul_f32 r7 r1 |
| 150 | 27 r7 = mul_f32 r7 r5 |
| 151 | 28 r7 = mul_f32 r7 r2 |
| 152 | 29 r7 = round r7 |
| 153 | 30 r6 = bit_and r3 r6 |
| 154 | 31 r6 = to_f32 r6 |
| 155 | 32 r6 = mul_f32 r6 r1 |
| 156 | 33 r5 = mul_f32 r6 r5 |
| 157 | 34 r5 = mul_f32 r5 r2 |
| 158 | 35 r5 = round r5 |
| 159 | 36 r7 = pack r5 r7 8 |
| 160 | 37 r4 = pack r7 r4 16 |
| 161 | 38 store32 arg(1) r4 |
| 162 | |
| 163 | G8 over A8 |
| 164 | 11 values (originally 15): |
| 165 | v0 = load8 arg(1) |
| 166 | v1 = to_f32 v0 |
| 167 | ↑ v2 = splat 3B808081 (0.0039215689) |
| 168 | v3 = mul_f32 v1 v2 |
| 169 | ↑ v4 = splat 0 (0) |
| 170 | ↑ v5 = splat 3F800000 (1) |
| 171 | v6 = fma_f32 v3 v4 v5 |
| 172 | ↑ v7 = splat 437F0000 (255) |
| 173 | v8 = mul_f32 v6 v7 |
| 174 | v9 = round v8 |
| 175 | store8 arg(1) v9 |
| 176 | |
| 177 | 5 registers, 11 instructions: |
| 178 | 0 r0 = splat 3B808081 (0.0039215689) |
| 179 | 1 r1 = splat 0 (0) |
| 180 | 2 r2 = splat 3F800000 (1) |
| 181 | 3 r3 = splat 437F0000 (255) |
| 182 | loop: |
| 183 | 4 r4 = load8 arg(1) |
| 184 | 5 r4 = to_f32 r4 |
| 185 | 6 r4 = mul_f32 r4 r0 |
| 186 | 7 r4 = fma_f32 r4 r1 r2 |
| 187 | 8 r4 = mul_f32 r4 r3 |
| 188 | 9 r4 = round r4 |
| 189 | 10 store8 arg(1) r4 |
| 190 | |
| 191 | G8 over G8 |
| 192 | 19 values (originally 23): |
| 193 | v0 = load8 arg(1) |
| 194 | v1 = to_f32 v0 |
| 195 | ↑ v2 = splat 3B808081 (0.0039215689) |
| 196 | v3 = mul_f32 v1 v2 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 197 | v4 = load8 arg(0) |
| 198 | v5 = to_f32 v4 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 199 | v6 = mul_f32 v5 v2 |
| 200 | ↑ v7 = splat 0 (0) |
| 201 | v8 = fma_f32 v3 v7 v6 |
| 202 | ↑ v9 = splat 3D93DD98 (0.0722) |
| 203 | v10 = mul_f32 v8 v9 |
| 204 | ↑ v11 = splat 3F371759 (0.71520001) |
| 205 | v12 = fma_f32 v8 v11 v10 |
| 206 | ↑ v13 = splat 3E59B3D0 (0.21259999) |
| 207 | v14 = fma_f32 v8 v13 v12 |
Mike Klein | 7c0332c | 2020-03-05 14:18:04 -0600 | [diff] [blame] | 208 | ↑ v15 = splat 437F0000 (255) |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 209 | v16 = mul_f32 v14 v15 |
| 210 | v17 = round v16 |
| 211 | store8 arg(1) v17 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 212 | |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 213 | 8 registers, 19 instructions: |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 214 | 0 r0 = splat 3B808081 (0.0039215689) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 215 | 1 r1 = splat 0 (0) |
| 216 | 2 r2 = splat 3D93DD98 (0.0722) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 217 | 3 r3 = splat 3F371759 (0.71520001) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 218 | 4 r4 = splat 3E59B3D0 (0.21259999) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 219 | 5 r5 = splat 437F0000 (255) |
| 220 | loop: |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 221 | 6 r6 = load8 arg(1) |
| 222 | 7 r6 = to_f32 r6 |
| 223 | 8 r6 = mul_f32 r6 r0 |
| 224 | 9 r7 = load8 arg(0) |
| 225 | 10 r7 = to_f32 r7 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 226 | 11 r7 = mul_f32 r7 r0 |
| 227 | 12 r7 = fma_f32 r6 r1 r7 |
| 228 | 13 r6 = mul_f32 r7 r2 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 229 | 14 r6 = fma_f32 r7 r3 r6 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 230 | 15 r6 = fma_f32 r7 r4 r6 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 231 | 16 r6 = mul_f32 r6 r5 |
| 232 | 17 r6 = round r6 |
| 233 | 18 store8 arg(1) r6 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 234 | |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 235 | G8 over RGBA_8888 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 236 | 39 values (originally 43): |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 237 | v0 = load32 arg(1) |
| 238 | v1 = shr_i32 v0 24 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 239 | v2 = to_f32 v1 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 240 | ↑ v3 = splat 3B808081 (0.0039215689) |
| 241 | v4 = mul_f32 v2 v3 |
| 242 | ↑ v5 = splat 0 (0) |
| 243 | ↑ v6 = splat 3F800000 (1) |
| 244 | v7 = fma_f32 v4 v5 v6 |
| 245 | ↑ v8 = splat 437F0000 (255) |
| 246 | v9 = mul_f32 v7 v8 |
| 247 | v10 = shr_i32 v0 16 |
| 248 | ↑ v11 = splat FF (3.5733111e-43) |
| 249 | v12 = bit_and v11 v10 |
| 250 | v13 = to_f32 v12 |
| 251 | v14 = mul_f32 v13 v3 |
| 252 | v15 = load8 arg(0) |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 253 | v16 = to_f32 v15 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 254 | v17 = mul_f32 v16 v3 |
| 255 | v18 = fma_f32 v14 v5 v17 |
| 256 | v19 = round v9 |
| 257 | v20 = mul_f32 v18 v8 |
| 258 | v21 = round v20 |
| 259 | v22 = pack v21 v19 8 |
| 260 | v23 = shr_i32 v0 8 |
| 261 | v24 = bit_and v11 v23 |
| 262 | v25 = to_f32 v24 |
| 263 | v26 = mul_f32 v25 v3 |
| 264 | v27 = fma_f32 v26 v5 v17 |
| 265 | v28 = mul_f32 v27 v8 |
| 266 | v29 = round v28 |
| 267 | v30 = bit_and v11 v0 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 268 | v31 = to_f32 v30 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 269 | v32 = mul_f32 v31 v3 |
| 270 | v33 = fma_f32 v32 v5 v17 |
| 271 | v34 = mul_f32 v33 v8 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 272 | v35 = round v34 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 273 | v36 = pack v35 v29 8 |
| 274 | v37 = pack v36 v22 16 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 275 | store32 arg(1) v37 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 276 | |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 277 | 9 registers, 39 instructions: |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 278 | 0 r0 = splat 3B808081 (0.0039215689) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 279 | 1 r1 = splat 0 (0) |
| 280 | 2 r2 = splat 3F800000 (1) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 281 | 3 r3 = splat 437F0000 (255) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 282 | 4 r4 = splat FF (3.5733111e-43) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 283 | loop: |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 284 | 5 r5 = load32 arg(1) |
| 285 | 6 r6 = shr_i32 r5 24 |
| 286 | 7 r6 = to_f32 r6 |
| 287 | 8 r6 = mul_f32 r6 r0 |
| 288 | 9 r6 = fma_f32 r6 r1 r2 |
| 289 | 10 r6 = mul_f32 r6 r3 |
| 290 | 11 r7 = shr_i32 r5 16 |
| 291 | 12 r7 = bit_and r4 r7 |
| 292 | 13 r7 = to_f32 r7 |
| 293 | 14 r7 = mul_f32 r7 r0 |
| 294 | 15 r8 = load8 arg(0) |
| 295 | 16 r8 = to_f32 r8 |
| 296 | 17 r8 = mul_f32 r8 r0 |
| 297 | 18 r7 = fma_f32 r7 r1 r8 |
| 298 | 19 r6 = round r6 |
| 299 | 20 r7 = mul_f32 r7 r3 |
| 300 | 21 r7 = round r7 |
| 301 | 22 r6 = pack r7 r6 8 |
| 302 | 23 r7 = shr_i32 r5 8 |
| 303 | 24 r7 = bit_and r4 r7 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 304 | 25 r7 = to_f32 r7 |
| 305 | 26 r7 = mul_f32 r7 r0 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 306 | 27 r7 = fma_f32 r7 r1 r8 |
| 307 | 28 r7 = mul_f32 r7 r3 |
| 308 | 29 r7 = round r7 |
| 309 | 30 r5 = bit_and r4 r5 |
| 310 | 31 r5 = to_f32 r5 |
| 311 | 32 r5 = mul_f32 r5 r0 |
| 312 | 33 r8 = fma_f32 r5 r1 r8 |
| 313 | 34 r8 = mul_f32 r8 r3 |
| 314 | 35 r8 = round r8 |
| 315 | 36 r7 = pack r8 r7 8 |
| 316 | 37 r6 = pack r7 r6 16 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 317 | 38 store32 arg(1) r6 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 318 | |
| 319 | RGBA_8888 over A8 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 320 | 15 values (originally 33): |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 321 | v0 = load32 arg(0) |
| 322 | v1 = shr_i32 v0 24 |
| 323 | v2 = to_f32 v1 |
| 324 | ↑ v3 = splat 3F800000 (1) |
| 325 | ↑ v4 = splat 3B808081 (0.0039215689) |
| 326 | v5 = fnma_f32 v2 v4 v3 |
| 327 | v6 = load8 arg(1) |
| 328 | v7 = to_f32 v6 |
| 329 | v8 = mul_f32 v7 v4 |
| 330 | v9 = mul_f32 v2 v4 |
| 331 | v10 = fma_f32 v8 v5 v9 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 332 | ↑ v11 = splat 437F0000 (255) |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 333 | v12 = mul_f32 v10 v11 |
| 334 | v13 = round v12 |
| 335 | store8 arg(1) v13 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 336 | |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 337 | 6 registers, 15 instructions: |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 338 | 0 r0 = splat 3F800000 (1) |
| 339 | 1 r1 = splat 3B808081 (0.0039215689) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 340 | 2 r2 = splat 437F0000 (255) |
Mike Klein | a630732 | 2019-06-07 15:44:26 -0500 | [diff] [blame] | 341 | loop: |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 342 | 3 r3 = load32 arg(0) |
| 343 | 4 r3 = shr_i32 r3 24 |
| 344 | 5 r3 = to_f32 r3 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 345 | 6 r4 = fnma_f32 r3 r1 r0 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 346 | 7 r5 = load8 arg(1) |
| 347 | 8 r5 = to_f32 r5 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 348 | 9 r5 = mul_f32 r5 r1 |
| 349 | 10 r3 = mul_f32 r3 r1 |
| 350 | 11 r3 = fma_f32 r5 r4 r3 |
| 351 | 12 r3 = mul_f32 r3 r2 |
| 352 | 13 r3 = round r3 |
| 353 | 14 store8 arg(1) r3 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 354 | |
| 355 | RGBA_8888 over G8 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 356 | 34 values (originally 39): |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 357 | v0 = load32 arg(0) |
| 358 | v1 = shr_i32 v0 24 |
| 359 | v2 = to_f32 v1 |
| 360 | ↑ v3 = splat 3F800000 (1) |
| 361 | ↑ v4 = splat 3B808081 (0.0039215689) |
| 362 | v5 = fnma_f32 v2 v4 v3 |
| 363 | v6 = shr_i32 v0 16 |
| 364 | ↑ v7 = splat FF (3.5733111e-43) |
| 365 | v8 = bit_and v7 v6 |
| 366 | v9 = load8 arg(1) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 367 | v10 = to_f32 v9 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 368 | v11 = mul_f32 v10 v4 |
| 369 | v12 = to_f32 v8 |
| 370 | v13 = mul_f32 v12 v4 |
| 371 | v14 = fma_f32 v11 v5 v13 |
| 372 | ↑ v15 = splat 3D93DD98 (0.0722) |
| 373 | v16 = mul_f32 v14 v15 |
| 374 | v17 = shr_i32 v0 8 |
| 375 | v18 = bit_and v7 v17 |
| 376 | v19 = to_f32 v18 |
| 377 | v20 = mul_f32 v19 v4 |
| 378 | v21 = fma_f32 v11 v5 v20 |
| 379 | ↑ v22 = splat 3F371759 (0.71520001) |
| 380 | v23 = fma_f32 v21 v22 v16 |
| 381 | v24 = bit_and v7 v0 |
| 382 | v25 = to_f32 v24 |
| 383 | v26 = mul_f32 v25 v4 |
| 384 | v27 = fma_f32 v11 v5 v26 |
| 385 | ↑ v28 = splat 3E59B3D0 (0.21259999) |
| 386 | v29 = fma_f32 v27 v28 v23 |
Mike Klein | 7c0332c | 2020-03-05 14:18:04 -0600 | [diff] [blame] | 387 | ↑ v30 = splat 437F0000 (255) |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 388 | v31 = mul_f32 v29 v30 |
| 389 | v32 = round v31 |
| 390 | store8 arg(1) v32 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 391 | |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 392 | 12 registers, 34 instructions: |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 393 | 0 r0 = splat 3F800000 (1) |
| 394 | 1 r1 = splat 3B808081 (0.0039215689) |
| 395 | 2 r2 = splat FF (3.5733111e-43) |
| 396 | 3 r3 = splat 3D93DD98 (0.0722) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 397 | 4 r4 = splat 3F371759 (0.71520001) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 398 | 5 r5 = splat 3E59B3D0 (0.21259999) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 399 | 6 r6 = splat 437F0000 (255) |
| 400 | loop: |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 401 | 7 r7 = load32 arg(0) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 402 | 8 r8 = shr_i32 r7 24 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 403 | 9 r8 = to_f32 r8 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 404 | 10 r8 = fnma_f32 r8 r1 r0 |
| 405 | 11 r9 = shr_i32 r7 16 |
| 406 | 12 r9 = bit_and r2 r9 |
| 407 | 13 r10 = load8 arg(1) |
| 408 | 14 r10 = to_f32 r10 |
| 409 | 15 r10 = mul_f32 r10 r1 |
| 410 | 16 r9 = to_f32 r9 |
| 411 | 17 r9 = mul_f32 r9 r1 |
| 412 | 18 r9 = fma_f32 r10 r8 r9 |
| 413 | 19 r9 = mul_f32 r9 r3 |
| 414 | 20 r11 = shr_i32 r7 8 |
| 415 | 21 r11 = bit_and r2 r11 |
| 416 | 22 r11 = to_f32 r11 |
| 417 | 23 r11 = mul_f32 r11 r1 |
| 418 | 24 r11 = fma_f32 r10 r8 r11 |
| 419 | 25 r9 = fma_f32 r11 r4 r9 |
| 420 | 26 r7 = bit_and r2 r7 |
| 421 | 27 r7 = to_f32 r7 |
| 422 | 28 r7 = mul_f32 r7 r1 |
| 423 | 29 r7 = fma_f32 r10 r8 r7 |
| 424 | 30 r9 = fma_f32 r7 r5 r9 |
| 425 | 31 r9 = mul_f32 r9 r6 |
| 426 | 32 r9 = round r9 |
| 427 | 33 store8 arg(1) r9 |
Mike Klein | 8c1e0ef | 2019-11-12 09:07:23 -0600 | [diff] [blame] | 428 | |
| 429 | RGBA_8888 over RGBA_8888 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 430 | 51 values (originally 55): |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 431 | v0 = load32 arg(0) |
| 432 | v1 = shr_i32 v0 24 |
| 433 | v2 = to_f32 v1 |
| 434 | ↑ v3 = splat 3F800000 (1) |
| 435 | ↑ v4 = splat 3B808081 (0.0039215689) |
| 436 | v5 = fnma_f32 v2 v4 v3 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 437 | v6 = load32 arg(1) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 438 | v7 = shr_i32 v6 24 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 439 | v8 = to_f32 v7 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 440 | v9 = mul_f32 v8 v4 |
| 441 | v10 = mul_f32 v2 v4 |
| 442 | v11 = fma_f32 v9 v5 v10 |
| 443 | ↑ v12 = splat 437F0000 (255) |
| 444 | v13 = mul_f32 v11 v12 |
| 445 | v14 = shr_i32 v6 16 |
| 446 | ↑ v15 = splat FF (3.5733111e-43) |
| 447 | v16 = bit_and v15 v14 |
| 448 | v17 = to_f32 v16 |
| 449 | v18 = mul_f32 v17 v4 |
| 450 | v19 = shr_i32 v0 16 |
| 451 | v20 = bit_and v15 v19 |
| 452 | v21 = to_f32 v20 |
| 453 | v22 = mul_f32 v21 v4 |
| 454 | v23 = fma_f32 v18 v5 v22 |
| 455 | v24 = round v13 |
| 456 | v25 = mul_f32 v23 v12 |
| 457 | v26 = round v25 |
| 458 | v27 = pack v26 v24 8 |
| 459 | v28 = shr_i32 v6 8 |
| 460 | v29 = bit_and v15 v28 |
| 461 | v30 = to_f32 v29 |
| 462 | v31 = mul_f32 v30 v4 |
| 463 | v32 = shr_i32 v0 8 |
| 464 | v33 = bit_and v15 v32 |
| 465 | v34 = to_f32 v33 |
| 466 | v35 = mul_f32 v34 v4 |
| 467 | v36 = fma_f32 v31 v5 v35 |
| 468 | v37 = bit_and v15 v6 |
| 469 | v38 = to_f32 v37 |
| 470 | v39 = mul_f32 v38 v4 |
| 471 | v40 = bit_and v15 v0 |
| 472 | v41 = to_f32 v40 |
| 473 | v42 = mul_f32 v41 v4 |
| 474 | v43 = fma_f32 v39 v5 v42 |
| 475 | v44 = mul_f32 v36 v12 |
| 476 | v45 = round v44 |
| 477 | v46 = mul_f32 v43 v12 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 478 | v47 = round v46 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 479 | v48 = pack v47 v45 8 |
| 480 | v49 = pack v48 v27 16 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 481 | store32 arg(1) v49 |
Mike Klein | 8c1e0ef | 2019-11-12 09:07:23 -0600 | [diff] [blame] | 482 | |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 483 | 10 registers, 51 instructions: |
| 484 | 0 r0 = splat 3F800000 (1) |
| 485 | 1 r1 = splat 3B808081 (0.0039215689) |
| 486 | 2 r2 = splat 437F0000 (255) |
| 487 | 3 r3 = splat FF (3.5733111e-43) |
Mike Klein | 8c1e0ef | 2019-11-12 09:07:23 -0600 | [diff] [blame] | 488 | loop: |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 489 | 4 r4 = load32 arg(0) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 490 | 5 r5 = shr_i32 r4 24 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 491 | 6 r5 = to_f32 r5 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 492 | 7 r6 = fnma_f32 r5 r1 r0 |
| 493 | 8 r7 = load32 arg(1) |
| 494 | 9 r8 = shr_i32 r7 24 |
| 495 | 10 r8 = to_f32 r8 |
| 496 | 11 r8 = mul_f32 r8 r1 |
| 497 | 12 r5 = mul_f32 r5 r1 |
| 498 | 13 r5 = fma_f32 r8 r6 r5 |
| 499 | 14 r5 = mul_f32 r5 r2 |
| 500 | 15 r8 = shr_i32 r7 16 |
| 501 | 16 r8 = bit_and r3 r8 |
| 502 | 17 r8 = to_f32 r8 |
| 503 | 18 r8 = mul_f32 r8 r1 |
| 504 | 19 r9 = shr_i32 r4 16 |
| 505 | 20 r9 = bit_and r3 r9 |
| 506 | 21 r9 = to_f32 r9 |
| 507 | 22 r9 = mul_f32 r9 r1 |
| 508 | 23 r9 = fma_f32 r8 r6 r9 |
| 509 | 24 r5 = round r5 |
| 510 | 25 r9 = mul_f32 r9 r2 |
| 511 | 26 r9 = round r9 |
| 512 | 27 r5 = pack r9 r5 8 |
| 513 | 28 r9 = shr_i32 r7 8 |
| 514 | 29 r9 = bit_and r3 r9 |
| 515 | 30 r9 = to_f32 r9 |
| 516 | 31 r9 = mul_f32 r9 r1 |
| 517 | 32 r8 = shr_i32 r4 8 |
| 518 | 33 r8 = bit_and r3 r8 |
| 519 | 34 r8 = to_f32 r8 |
| 520 | 35 r8 = mul_f32 r8 r1 |
| 521 | 36 r8 = fma_f32 r9 r6 r8 |
| 522 | 37 r7 = bit_and r3 r7 |
| 523 | 38 r7 = to_f32 r7 |
| 524 | 39 r7 = mul_f32 r7 r1 |
| 525 | 40 r4 = bit_and r3 r4 |
| 526 | 41 r4 = to_f32 r4 |
| 527 | 42 r4 = mul_f32 r4 r1 |
| 528 | 43 r4 = fma_f32 r7 r6 r4 |
| 529 | 44 r8 = mul_f32 r8 r2 |
| 530 | 45 r8 = round r8 |
| 531 | 46 r4 = mul_f32 r4 r2 |
| 532 | 47 r4 = round r4 |
Mike Klein | 5caf7de | 2020-03-12 11:05:46 -0500 | [diff] [blame] | 533 | 48 r8 = pack r4 r8 8 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 534 | 49 r5 = pack r8 r5 16 |
| 535 | 50 store32 arg(1) r5 |
Mike Klein | 267f507 | 2019-06-03 16:27:46 -0500 | [diff] [blame] | 536 | |
Mike Klein | 397fc88 | 2019-06-20 11:37:10 -0500 | [diff] [blame] | 537 | I32 (Naive) 8888 over 8888 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 538 | 33 values (originally 33): |
Mike Klein | c2fb3b4 | 2019-07-17 12:09:09 -0500 | [diff] [blame] | 539 | v0 = load32 arg(0) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 540 | v1 = shr_i32 v0 24 |
| 541 | ↑ v2 = splat 100 (3.5873241e-43) |
| 542 | v3 = sub_i32 v2 v1 |
| 543 | v4 = load32 arg(1) |
| 544 | v5 = shr_i32 v4 16 |
| 545 | ↑ v6 = splat FF (3.5733111e-43) |
| 546 | v7 = bit_and v6 v5 |
| 547 | v8 = mul_i32 v7 v3 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 548 | v9 = shr_i32 v8 8 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 549 | v10 = shr_i32 v0 16 |
| 550 | v11 = bit_and v6 v10 |
| 551 | v12 = add_i32 v11 v9 |
| 552 | v13 = shr_i32 v4 24 |
| 553 | v14 = mul_i32 v13 v3 |
| 554 | v15 = shr_i32 v14 8 |
| 555 | v16 = add_i32 v1 v15 |
| 556 | v17 = pack v12 v16 8 |
| 557 | v18 = shr_i32 v4 8 |
| 558 | v19 = bit_and v6 v18 |
| 559 | v20 = mul_i32 v19 v3 |
| 560 | v21 = shr_i32 v20 8 |
| 561 | v22 = shr_i32 v0 8 |
| 562 | v23 = bit_and v6 v22 |
| 563 | v24 = add_i32 v23 v21 |
| 564 | v25 = bit_and v6 v4 |
| 565 | v26 = mul_i32 v25 v3 |
| 566 | v27 = shr_i32 v26 8 |
| 567 | v28 = bit_and v6 v0 |
| 568 | v29 = add_i32 v28 v27 |
| 569 | v30 = pack v29 v24 8 |
| 570 | v31 = pack v30 v17 16 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 571 | store32 arg(1) v31 |
Mike Klein | aab45b5 | 2019-07-02 15:39:23 -0500 | [diff] [blame] | 572 | |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 573 | 8 registers, 33 instructions: |
| 574 | 0 r0 = splat 100 (3.5873241e-43) |
| 575 | 1 r1 = splat FF (3.5733111e-43) |
Mike Klein | 397fc88 | 2019-06-20 11:37:10 -0500 | [diff] [blame] | 576 | loop: |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 577 | 2 r2 = load32 arg(0) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 578 | 3 r3 = shr_i32 r2 24 |
| 579 | 4 r4 = sub_i32 r0 r3 |
| 580 | 5 r5 = load32 arg(1) |
| 581 | 6 r6 = shr_i32 r5 16 |
| 582 | 7 r6 = bit_and r1 r6 |
| 583 | 8 r6 = mul_i32 r6 r4 |
| 584 | 9 r6 = shr_i32 r6 8 |
| 585 | 10 r7 = shr_i32 r2 16 |
| 586 | 11 r7 = bit_and r1 r7 |
| 587 | 12 r6 = add_i32 r7 r6 |
| 588 | 13 r7 = shr_i32 r5 24 |
| 589 | 14 r7 = mul_i32 r7 r4 |
| 590 | 15 r7 = shr_i32 r7 8 |
| 591 | 16 r7 = add_i32 r3 r7 |
| 592 | 17 r7 = pack r6 r7 8 |
| 593 | 18 r6 = shr_i32 r5 8 |
| 594 | 19 r6 = bit_and r1 r6 |
| 595 | 20 r6 = mul_i32 r6 r4 |
| 596 | 21 r6 = shr_i32 r6 8 |
| 597 | 22 r3 = shr_i32 r2 8 |
| 598 | 23 r3 = bit_and r1 r3 |
| 599 | 24 r6 = add_i32 r3 r6 |
| 600 | 25 r5 = bit_and r1 r5 |
| 601 | 26 r4 = mul_i32 r5 r4 |
| 602 | 27 r4 = shr_i32 r4 8 |
| 603 | 28 r2 = bit_and r1 r2 |
| 604 | 29 r4 = add_i32 r2 r4 |
| 605 | 30 r6 = pack r4 r6 8 |
| 606 | 31 r7 = pack r6 r7 16 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 607 | 32 store32 arg(1) r7 |
Mike Klein | 397fc88 | 2019-06-20 11:37:10 -0500 | [diff] [blame] | 608 | |
Mike Klein | 7b7077c | 2019-06-03 17:10:59 -0500 | [diff] [blame] | 609 | I32 8888 over 8888 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 610 | 29 values (originally 29): |
Mike Klein | c2fb3b4 | 2019-07-17 12:09:09 -0500 | [diff] [blame] | 611 | v0 = load32 arg(0) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 612 | v1 = shr_i32 v0 24 |
| 613 | ↑ v2 = splat 100 (3.5873241e-43) |
| 614 | v3 = sub_i32 v2 v1 |
| 615 | v4 = load32 arg(1) |
| 616 | v5 = bytes v4 3 |
| 617 | v6 = mul_i16x2 v5 v3 |
| 618 | v7 = shr_i32 v6 8 |
| 619 | v8 = bytes v0 3 |
| 620 | v9 = add_i32 v8 v7 |
| 621 | v10 = shr_i32 v4 24 |
| 622 | v11 = mul_i16x2 v10 v3 |
| 623 | v12 = shr_i32 v11 8 |
| 624 | v13 = add_i32 v1 v12 |
| 625 | v14 = pack v9 v13 8 |
| 626 | v15 = bytes v4 2 |
| 627 | v16 = mul_i16x2 v15 v3 |
| 628 | v17 = shr_i32 v16 8 |
| 629 | v18 = bytes v0 2 |
| 630 | v19 = add_i32 v18 v17 |
| 631 | ↑ v20 = splat FF (3.5733111e-43) |
| 632 | v21 = bit_and v4 v20 |
| 633 | v22 = mul_i16x2 v21 v3 |
| 634 | v23 = shr_i32 v22 8 |
| 635 | v24 = bit_and v0 v20 |
| 636 | v25 = add_i32 v24 v23 |
| 637 | v26 = pack v25 v19 8 |
| 638 | v27 = pack v26 v14 16 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 639 | store32 arg(1) v27 |
Mike Klein | aab45b5 | 2019-07-02 15:39:23 -0500 | [diff] [blame] | 640 | |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 641 | 8 registers, 29 instructions: |
| 642 | 0 r0 = splat 100 (3.5873241e-43) |
| 643 | 1 r1 = splat FF (3.5733111e-43) |
Mike Klein | 754bad3 | 2019-06-05 10:47:46 -0500 | [diff] [blame] | 644 | loop: |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 645 | 2 r2 = load32 arg(0) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 646 | 3 r3 = shr_i32 r2 24 |
| 647 | 4 r4 = sub_i32 r0 r3 |
| 648 | 5 r5 = load32 arg(1) |
| 649 | 6 r6 = bytes r5 3 |
| 650 | 7 r6 = mul_i16x2 r6 r4 |
| 651 | 8 r6 = shr_i32 r6 8 |
| 652 | 9 r7 = bytes r2 3 |
| 653 | 10 r6 = add_i32 r7 r6 |
| 654 | 11 r7 = shr_i32 r5 24 |
| 655 | 12 r7 = mul_i16x2 r7 r4 |
| 656 | 13 r7 = shr_i32 r7 8 |
| 657 | 14 r7 = add_i32 r3 r7 |
| 658 | 15 r7 = pack r6 r7 8 |
| 659 | 16 r6 = bytes r5 2 |
| 660 | 17 r6 = mul_i16x2 r6 r4 |
| 661 | 18 r6 = shr_i32 r6 8 |
| 662 | 19 r3 = bytes r2 2 |
| 663 | 20 r6 = add_i32 r3 r6 |
| 664 | 21 r5 = bit_and r5 r1 |
| 665 | 22 r4 = mul_i16x2 r5 r4 |
| 666 | 23 r4 = shr_i32 r4 8 |
| 667 | 24 r2 = bit_and r2 r1 |
| 668 | 25 r4 = add_i32 r2 r4 |
| 669 | 26 r6 = pack r4 r6 8 |
| 670 | 27 r7 = pack r6 r7 16 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 671 | 28 store32 arg(1) r7 |
Mike Klein | 821f5e8 | 2019-06-13 10:56:51 -0500 | [diff] [blame] | 672 | |
| 673 | I32 (SWAR) 8888 over 8888 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 674 | 15 values (originally 15): |
Mike Klein | c2fb3b4 | 2019-07-17 12:09:09 -0500 | [diff] [blame] | 675 | v0 = load32 arg(0) |
| 676 | v1 = bytes v0 404 |
Mike Klein | 5e533c9 | 2019-07-22 13:44:54 -0500 | [diff] [blame] | 677 | ↑ v2 = splat 1000100 (2.3510604e-38) |
Mike Klein | c2fb3b4 | 2019-07-17 12:09:09 -0500 | [diff] [blame] | 678 | v3 = sub_i16x2 v2 v1 |
| 679 | v4 = load32 arg(1) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 680 | v5 = shr_i16x2 v4 8 |
| 681 | v6 = mul_i16x2 v5 v3 |
| 682 | ↑ v7 = splat FF00FF (2.3418409e-38) |
| 683 | v8 = bit_clear v6 v7 |
| 684 | v9 = bit_and v4 v7 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 685 | v10 = mul_i16x2 v9 v3 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 686 | v11 = shr_i16x2 v10 8 |
| 687 | v12 = bit_or v11 v8 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 688 | v13 = add_i32 v0 v12 |
| 689 | store32 arg(1) v13 |
Mike Klein | aab45b5 | 2019-07-02 15:39:23 -0500 | [diff] [blame] | 690 | |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 691 | 6 registers, 15 instructions: |
Mike Klein | b5c4355 | 2020-01-07 11:39:30 -0600 | [diff] [blame] | 692 | 0 r0 = splat 1000100 (2.3510604e-38) |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 693 | 1 r1 = splat FF00FF (2.3418409e-38) |
Mike Klein | 821f5e8 | 2019-06-13 10:56:51 -0500 | [diff] [blame] | 694 | loop: |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 695 | 2 r2 = load32 arg(0) |
| 696 | 3 r3 = bytes r2 404 |
| 697 | 4 r3 = sub_i16x2 r0 r3 |
| 698 | 5 r4 = load32 arg(1) |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 699 | 6 r5 = shr_i16x2 r4 8 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 700 | 7 r5 = mul_i16x2 r5 r3 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 701 | 8 r5 = bit_clear r5 r1 |
| 702 | 9 r4 = bit_and r4 r1 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 703 | 10 r3 = mul_i16x2 r4 r3 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 704 | 11 r3 = shr_i16x2 r3 8 |
| 705 | 12 r5 = bit_or r3 r5 |
| 706 | 13 r5 = add_i32 r2 r5 |
| 707 | 14 store32 arg(1) r5 |
Mike Klein | 7b7077c | 2019-06-03 17:10:59 -0500 | [diff] [blame] | 708 | |
Mike Klein | ed9b1f1 | 2020-02-06 13:02:32 -0600 | [diff] [blame] | 709 | 6 values (originally 6): |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 710 | ↟ v0 = splat 2 (2.8025969e-45) |
| 711 | ↟ v1 = splat 1 (1.4012985e-45) |
| 712 | ↑ v2 = add_i32 v1 v0 |
Mike Klein | f996311 | 2019-08-08 15:13:25 -0400 | [diff] [blame] | 713 | v3 = load32 arg(0) |
| 714 | v4 = mul_i32 v3 v2 |
| 715 | store32 arg(0) v4 |
| 716 | |
| 717 | 2 registers, 6 instructions: |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 718 | 0 r0 = splat 2 (2.8025969e-45) |
| 719 | 1 r1 = splat 1 (1.4012985e-45) |
| 720 | 2 r0 = add_i32 r1 r0 |
Mike Klein | f996311 | 2019-08-08 15:13:25 -0400 | [diff] [blame] | 721 | loop: |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 722 | 3 r1 = load32 arg(0) |
| 723 | 4 r1 = mul_i32 r1 r0 |
| 724 | 5 store32 arg(0) r1 |
Mike Klein | f996311 | 2019-08-08 15:13:25 -0400 | [diff] [blame] | 725 | |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 726 | 23 values (originally 23): |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 727 | v0 = load32 arg(1) |
| 728 | v1 = shr_i32 v0 24 |
| 729 | v2 = load32 arg(0) |
| 730 | v3 = shr_i32 v2 24 |
| 731 | v4 = add_i32 v3 v1 |
| 732 | v5 = shr_i32 v0 16 |
| 733 | ↑ v6 = splat FF (3.5733111e-43) |
| 734 | v7 = bit_and v6 v5 |
| 735 | v8 = shr_i32 v2 16 |
| 736 | v9 = bit_and v6 v8 |
| 737 | v10 = add_i32 v9 v7 |
| 738 | v11 = pack v10 v4 8 |
| 739 | v12 = shr_i32 v0 8 |
| 740 | v13 = bit_and v6 v12 |
| 741 | v14 = shr_i32 v2 8 |
| 742 | v15 = bit_and v6 v14 |
| 743 | v16 = add_i32 v15 v13 |
| 744 | v17 = bit_and v6 v0 |
| 745 | v18 = bit_and v6 v2 |
| 746 | v19 = add_i32 v18 v17 |
| 747 | v20 = pack v19 v16 8 |
| 748 | v21 = pack v20 v11 16 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 749 | store32 arg(1) v21 |
Mike Klein | d48488b | 2019-10-22 12:27:58 -0500 | [diff] [blame] | 750 | |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 751 | 6 registers, 23 instructions: |
| 752 | 0 r0 = splat FF (3.5733111e-43) |
Mike Klein | d48488b | 2019-10-22 12:27:58 -0500 | [diff] [blame] | 753 | loop: |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 754 | 1 r1 = load32 arg(1) |
| 755 | 2 r2 = shr_i32 r1 24 |
| 756 | 3 r3 = load32 arg(0) |
| 757 | 4 r4 = shr_i32 r3 24 |
| 758 | 5 r2 = add_i32 r4 r2 |
| 759 | 6 r4 = shr_i32 r1 16 |
| 760 | 7 r4 = bit_and r0 r4 |
| 761 | 8 r5 = shr_i32 r3 16 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 762 | 9 r5 = bit_and r0 r5 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 763 | 10 r4 = add_i32 r5 r4 |
| 764 | 11 r2 = pack r4 r2 8 |
| 765 | 12 r4 = shr_i32 r1 8 |
Mike Klein | 5cdeb39 | 2020-02-10 12:10:36 -0600 | [diff] [blame] | 766 | 13 r4 = bit_and r0 r4 |
Herb Derby | 43f7641 | 2020-03-11 16:54:35 -0400 | [diff] [blame^] | 767 | 14 r5 = shr_i32 r3 8 |
| 768 | 15 r5 = bit_and r0 r5 |
| 769 | 16 r4 = add_i32 r5 r4 |
| 770 | 17 r1 = bit_and r0 r1 |
| 771 | 18 r3 = bit_and r0 r3 |
| 772 | 19 r1 = add_i32 r3 r1 |
| 773 | 20 r4 = pack r1 r4 8 |
| 774 | 21 r2 = pack r4 r2 16 |
| 775 | 22 store32 arg(1) r2 |
Mike Klein | d48488b | 2019-10-22 12:27:58 -0500 | [diff] [blame] | 776 | |