Fast CRC32 library, based on the work of Stephan Brumme (https://create.stephan-brumme.com/crc32/).
- C99 and C++11 support (C++17 recommended, eg std::string_view ...)
- CMake support (3.14 minimal or 3.19 with preset)
- The fastest algorithms need about 1 CPU cycle per byte
- Endian-aware
- Support for multi-threaded computation
- Multi OS support: Linux 32/64, MacOS, Windows 32/64, Android 32/64
- Cross-plateform support: AMD64, I386, ARMv8, ARMv7, ARMv6, ARMv5, RISC-V 32/64, PPC64le, Mips, m68k..
- Runs even on Arduino, Raspberry Pi, etc.
- More info https://create.stephan-brumme.com/crc32/
| Functions | version | Published by | CPU cycles/byte | Bits/iter | Table size | MB/s |
|---|---|---|---|---|---|---|
| bitwise_branch / crc32::bitwise_branch | 1.0.0 | unknown | 100 | 1 | - | 170 |
| bitwise / crc32::bitwise | 1.0.0 | unknown | 50 or 40 | 1 | - | 190 |
| half-byte / crc32::half-byte | 1.0.0 | unknown | 14 | 4 | 64 bytes | 263 |
| crc32_1byte / crc32::crc32_1byte | 1.0.0 | Hagai Gold | 8 | 8 | 1 KB | 522 |
| crc32_1byte_tableless / crc32::crc32_1byte_tableless | 1.0.0 | Henry S. Warren | 18 | 8 | - | 300 |
| crc32_1byte_tableless2 / crc32::crc32_1byte_tableless2 | 1.0.0 | Dilip V. Sarwate | 19 | 8 | - | 170 |
| crc32_4bytes / crc32::crc32_4bytes | 1.0.0 | Intel Corp. | 3 or 4 | 32 | 4 KB | 1458 |
| crc32_8bytes / crc32::crc32_8bytes | 1.0.0 | Intel Corp. | 3 or 4 | 64 | 8 KB | 2629 |
| crc32_4x8bytes / crc32::crc32_4x8bytes | 1.0.0 | ? | 1.75 or 2.25 | 256 | 8 KB | 3003 |
| crc32_16bytes / crc32::crc32_16bytes | 1.0.0 | Bulat Ziganshin | 1.1 or 1.5 | 128 | 16 KB | 6027 |
| crc32_16bytes_prefetch / crc32::crc32_16bytes_prefetch | 1.0.0 | Bulat Ziganshin | 1 or 1.5 | 512 | 16 KB | 6102 |
| crc32_fast / crc32::crc32_fast | 1.0.0 | None | varies | varies | varies | varies |
Tested on Lenovo Legion 5 Pro:
- Manjaro 21.1.6
- Clang 12.0.1
- Build type: Release
- CPU: AMD R7 5800H
- RAM: 32 GB DDR4 3200Mhz 22CL 8x
- Data: 1 GB
- crc32_combine() "merges" two indepedently computed CRC32 values which is the basis for even faster multi-threaded calculation
See my website https://create.stephan-brumme.com/crc32/ for documentation, code examples and a benchmark.
See the BUILDING document.
See the CONTRIBUTING document.
See the LICENSE document.