# `wgpu-sig-ops` Benchmarks

[TOC]

## Multiple-shader benchmarks

### Linux + Nvidia A1000

secp256k1 signature recovery benchmarks (multiple shaders):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 32 | 297 |
| 2048 | 70 | 165 |
| 4096 | 128 | 162 |
| 8192 | 256 | 198 |
| 16384 | 513 | 288 |

GPU timings include data transfer.

secp256r1 signature recovery benchmarks (multiple shaders):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 256 | 125 | 258 |
| 512 | 252 | 164 |
| 1024 | 503 | 158 |
| 2048 | 1006 | 162 |
| 4096 | 2010 | 181 |
| 8192 | 4027 | 255 |

GPU timings include data transfer.

ed25519 signature verification benchmarks (multiple shaders):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 89 | 270 |
| 2048 | 219 | 149 |
| 4096 | 433 | 149 |
| 8192 | 866 | 195 |
| 16384 | 1424 | 270 |

### Apple M1 Mini

system_profiler SPHardwareDataType
Hardware:

    Hardware Overview:

      Model Name: Mac mini
      Model Identifier: Macmini9,1
      Model Number: MGNR3FN/A
      Chip: Apple M1
      Total Number of Cores: 8 (4 performance and 4 efficiency)
      Memory: 8 GB
      System Firmware Version: 10151.101.3
      OS Loader Version: 10151.101.3
      Serial Number (system): C07F9BT8Q6NV
      Hardware UUID: CA426CDA-9D8C-581C-8DDC-EFAF0B849996
      Provisioning UDID: 00008103-000C58C13AF2001E
      Activation Lock Status: Disabled

secp256k1 signature recovery benchmarks (multiple shaders):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 33 | 734 |
| 2048 | 103 | 934 |
| 4096 | 167 | 2164 |
| 8192 | 297 | 3923 |
| 16384 | 555 | 8107 |

GPU timings include data transfer.

secp256r1 signature recovery benchmarks (multiple shaders):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 256 | 127 | 851 |
| 512 | 293 | 893 |
| 1024 | 546 | 986 |
| 2048 | 1059 | 1321 |
| 4096 | 2079 | 2616 |
| 8192 | 4122 | 5044 |

GPU timings include data transfer.

ed25519 signature verification benchmarks (multiple shaders):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 103 | 489 |
| 2048 | 241 | 598 |
| 4096 | 449 | 1074 |
| 8192 | 864 | 2328 |
| 16384 | 1687 | 4228 |

GPU timings include data transfer.

### Apple M3 Pro

system_profiler SPHardwareDataType

Hardware:

    Hardware Overview:

      Model Name: MacBook Pro
      Model Identifier: Mac15,7
      Model Number: MRW23LL/A
      Chip: Apple M3 Pro
      Total Number of Cores: 12 (6 performance and 6 efficiency)
      Memory: 36 GB
      System Firmware Version: 10151.101.3
      OS Loader Version: 10151.101.3
      Serial Number (system): H6XY12XXV4
      Hardware UUID: 61A10CB7-1CA9-52DA-BEE0-F40D25126908
      Provisioning UDID: 00006030-000624583628001C
      Activation Lock Status: Disabled

secp256k1 signature recovery benchmarks (multiple shaders):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 25 | 359 |
| 2048 | 46 | 378 |
| 4096 | 98 | 456 |
| 8192 | 185 | 920 |
| 16384 | 393 | 1814 |
| 32768 | 759 | 3658 |
| 65536 | 1573 | 7421 |

GPU timings include data transfer.

secp256r1 signature recovery benchmarks (multiple shaders):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 256 | 97 | 435 |
| 512 | 193 | 445 |
| 1024 | 389 | 483 |
| 2048 | 781 | 531 |
| 4096 | 1567 | 627 |
| 8192 | 3139 | 1270 |
| 16384 | 6326 | 2526 |

GPU timings include data transfer.

ed25519 signature verification benchmarks (multiple shaders):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 81 | 177 |
| 2048 | 162 | 204 |
| 4096 | 324 | 309 |
| 8192 | 649 | 586 |
| 16384 | 1298 | 1133 |
| 32768 | 2597 | 2288 |
| 65536 | 5194 | 4547 |

GPU timings include data transfer.

## Single-shader benchmarks

### Linux + Nvidia A1000

On the Linux machine with an Nvidia A1000 GPU, we found that performing the
whole computation using a single shader had a very slight performance
advantage over splitting up the computation into multiple shaders.

secp256k1 signature recovery benchmarks (single shader):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 32 | 692 |
| 2048 | 65 | 657 |
| 4096 | 129 | 919 |
| 8192 | 257 | 3997 |

GPU timings include data transfer.

secp256r1 signature recovery benchmarks (single shader):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 256 | 125 | 253 |
| 512 | 251 | 171 |
| 1024 | 516 | 151 |
| 2048 | 1007 | 140 |
| 4096 | 2014 | 193 |
| 8192 | 4026 | 288 |

GPU timings include data transfer.

ed25519 signature verification benchmarks (single shader):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 89 | 530 |
| 2048 | 178 | 446 |
| 4096 | 429 | 592 |
| 8192 | 712 | 1115 |
| 16384 | 1424 | 2246 |

GPU timings include data transfer.

### Apple M1 Mini

Internal error: new_compute_pipeline_state: "Compute function exceeds available stack space" for all 3 benchmarks.

### Apple M3 Pro

secp256k1 signature recovery benchmarks (single shader):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 25 | 498 |
| 2048 | 49 | 550 |
| 4096 | 98 | 640 |
| 8192 | 197 | 1243 |
| 16384 | 394 | 2407 |
| 32768 | 788 | 4825 |
| 65536 | 1575 | 9773 |

GPU timings include data transfer.

secp256r1 signature recovery benchmarks (single shader):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 256 | 95 | 979 |
| 512 | 198 | 658 |
| 1024 | 396 | 719 |
| 2048 | 791 | 780 |
| 4096 | 1581 | 896 |
| 8192 | 3163 | 1677 |
| 16384 | 6328 | 3286 |

GPU timings include data transfer.

ed25519 signature verification benchmarks (single shader):
| Num. signatures | CPU, serial (ms) | GPU, parallel (ms) |
| ------------------ | ------------------ | ------------------ |
| 1024 | 81 | 300 |
| 2048 | 162 | 333 |
| 4096 | 324 | 380 |
| 8192 | 651 | 741 |
| 16384 | 1297 | 1418 |
| 32768 | 2597 | 2832 |
| 65536 | 5195 | 5818 |

GPU timings include data transfer.