fast-crc32c is a CRC-32C algorithm implementation for node.js, which uses hardware acceleration (via voxer/sse4_crc32 by Anand Suresh), and fallback to software implementation when hardware acceleration fails.
CRC-32C is faster then CRC-32, which could take advantage of full CPU operand register width (64bit) instead of CRC-32's 8bit.
When using hardware acceleration, CRC-32C is about 7x ~ 9x faster than software implemented CRC-32C.
Benchmark
The 3 tested implementations are:
- sse4_crc32c Hardware accelerated CRC-32C from sse4_crc32
- js_crc32c Javascript implemented CRC-32C
- js_crc32 Javascript implemented CRC-32 from buffer-crc32
$ npm run-script benchmark
> fast-crc32c@1.0.2 benchmark /Users/xshi/Projects/node-fast-crc32c
> node benchmark
sse4_crc32c_hw for inputs 1024B x 1,837,024 ops/sec ±1.76% (83 runs sampled)
sse4_crc32c_hw for inputs 16856835B, avg 2057B x 203 ops/sec ±2.37% (78 runs sampled)
sse4_crc32c_sw for inputs 1024B x 911,096 ops/sec ±1.07% (85 runs sampled)
sse4_crc32c_sw for inputs 16856835B, avg 2057B x 72.74 ops/sec ±1.29% (72 runs sampled)
js_crc32c for inputs 1024B x 356,253 ops/sec ±1.07% (86 runs sampled)
js_crc32c for inputs 16856835B, avg 2057B x 22.40 ops/sec ±1.09% (41 runs sampled)
js_crc32 for inputs 1024B x 351,668 ops/sec ±0.94% (87 runs sampled)
js_crc32 for inputs 16856835B, avg 2057B x 22.25 ops/sec ±1.11% (40 runs sampled)
| 1024B | 16856835B, avg 2057B | |
|---|---|---|
| sse4_crc32c_hw | 1,837,024 ops/sec | 203 ops/sec |
| sse4_crc32c_sw | 911,096 ops/sec | 72.74 ops/sec |
| js_crc32c | 356,253 ops/sec | 22.40 ops/sec |
| js_crc32 | 351,668 ops/sec | 22.25 ops/sec |
npm install fast-crc32c --savevar crc32 = require('fast-crc32c');
var result = crc32.calculate(data, initial);- data required, String|Buffer, data to digest
- initial optional, Number, initial CRC-32C digest