Skip to content

buffer: add SIMD Neon optimization for `byteLength`

Rodrigo Muino Tomonari requested to merge simd-bytelength into main

Ref https://github.com/nodejs/performance/issues/52

Since, benchmark CI does not have ARM processor, here's the benchmark results from my local:

node benchmark/compare.js --old ./node-main --new ./out/Release/node --filter buffer-bytelength buffers > bytelength.csv && node-benchmark-compare bytelength.csv
[00:18:29|% 100| 2/2 files | 60/60 runs | 32/32 configs]: Done
                                                                                               confidence improvement accuracy (*)   (**)  (***)
buffers/buffer-bytelength-buffer.js n=4000000 len=16                                                          0.39 %       ±0.77% ±1.03% ±1.35%
buffers/buffer-bytelength-buffer.js n=4000000 len=2                                                          -0.02 %       ±0.45% ±0.60% ±0.78%
buffers/buffer-bytelength-buffer.js n=4000000 len=256                                                         1.97 %       ±3.41% ±4.60% ±6.10%
buffers/buffer-bytelength-string.js n=1000000 repeat=1 encoding='base64' type='four_bytes'                    1.71 %       ±3.63% ±4.84% ±6.31%
buffers/buffer-bytelength-string.js n=1000000 repeat=1 encoding='base64' type='one_byte'                      0.03 %       ±0.68% ±0.90% ±1.17%
buffers/buffer-bytelength-string.js n=1000000 repeat=1 encoding='base64' type='three_bytes'                   2.16 %       ±2.97% ±3.98% ±5.24%
buffers/buffer-bytelength-string.js n=1000000 repeat=1 encoding='base64' type='two_bytes'                    -1.08 %       ±2.05% ±2.76% ±3.65%
buffers/buffer-bytelength-string.js n=1000000 repeat=1 encoding='utf8' type='four_bytes'                     -0.13 %       ±0.96% ±1.28% ±1.67%
buffers/buffer-bytelength-string.js n=1000000 repeat=1 encoding='utf8' type='one_byte'               ***      4.59 %       ±0.77% ±1.02% ±1.33%
buffers/buffer-bytelength-string.js n=1000000 repeat=1 encoding='utf8' type='three_bytes'                     0.21 %       ±0.61% ±0.82% ±1.07%
buffers/buffer-bytelength-string.js n=1000000 repeat=1 encoding='utf8' type='two_bytes'                      -0.30 %       ±0.43% ±0.57% ±0.74%
buffers/buffer-bytelength-string.js n=1000000 repeat=16 encoding='base64' type='four_bytes'                  -0.68 %       ±3.18% ±4.24% ±5.53%
buffers/buffer-bytelength-string.js n=1000000 repeat=16 encoding='base64' type='one_byte'                     0.65 %       ±2.05% ±2.75% ±3.61%
buffers/buffer-bytelength-string.js n=1000000 repeat=16 encoding='base64' type='three_bytes'                 -0.89 %       ±1.98% ±2.66% ±3.51%
buffers/buffer-bytelength-string.js n=1000000 repeat=16 encoding='base64' type='two_bytes'                    0.10 %       ±1.44% ±1.92% ±2.50%
buffers/buffer-bytelength-string.js n=1000000 repeat=16 encoding='utf8' type='four_bytes'                    -1.39 %       ±2.75% ±3.70% ±4.91%
buffers/buffer-bytelength-string.js n=1000000 repeat=16 encoding='utf8' type='one_byte'                      -0.97 %       ±1.35% ±1.82% ±2.40%
buffers/buffer-bytelength-string.js n=1000000 repeat=16 encoding='utf8' type='three_bytes'                    0.94 %       ±1.67% ±2.25% ±2.98%
buffers/buffer-bytelength-string.js n=1000000 repeat=16 encoding='utf8' type='two_bytes'                      0.19 %       ±0.47% ±0.63% ±0.83%
buffers/buffer-bytelength-string.js n=1000000 repeat=2 encoding='base64' type='four_bytes'                   -1.67 %       ±2.60% ±3.50% ±4.63%
buffers/buffer-bytelength-string.js n=1000000 repeat=2 encoding='base64' type='one_byte'                      0.77 %       ±1.95% ±2.63% ±3.48%
buffers/buffer-bytelength-string.js n=1000000 repeat=2 encoding='base64' type='three_bytes'                   2.70 %       ±3.23% ±4.34% ±5.73%
buffers/buffer-bytelength-string.js n=1000000 repeat=2 encoding='base64' type='two_bytes'                     0.02 %       ±1.06% ±1.41% ±1.83%
buffers/buffer-bytelength-string.js n=1000000 repeat=2 encoding='utf8' type='four_bytes'                      0.24 %       ±0.61% ±0.81% ±1.06%
buffers/buffer-bytelength-string.js n=1000000 repeat=2 encoding='utf8' type='one_byte'                        0.04 %       ±2.03% ±2.70% ±3.53%
buffers/buffer-bytelength-string.js n=1000000 repeat=2 encoding='utf8' type='three_bytes'                     1.33 %       ±2.73% ±3.67% ±4.87%
buffers/buffer-bytelength-string.js n=1000000 repeat=2 encoding='utf8' type='two_bytes'                       0.10 %       ±0.40% ±0.53% ±0.69%
buffers/buffer-bytelength-string.js n=1000000 repeat=256 encoding='base64' type='four_bytes'                 -0.16 %       ±3.65% ±4.85% ±6.32%
buffers/buffer-bytelength-string.js n=1000000 repeat=256 encoding='base64' type='one_byte'                    0.47 %       ±2.66% ±3.54% ±4.60%
buffers/buffer-bytelength-string.js n=1000000 repeat=256 encoding='base64' type='three_bytes'                -0.51 %       ±0.78% ±1.04% ±1.37%
buffers/buffer-bytelength-string.js n=1000000 repeat=256 encoding='base64' type='two_bytes'                  -1.66 %       ±2.60% ±3.49% ±4.62%
buffers/buffer-bytelength-string.js n=1000000 repeat=256 encoding='utf8' type='four_bytes'                   -0.42 %       ±0.54% ±0.72% ±0.94%
buffers/buffer-bytelength-string.js n=1000000 repeat=256 encoding='utf8' type='one_byte'                     -0.07 %       ±0.32% ±0.43% ±0.56%
buffers/buffer-bytelength-string.js n=1000000 repeat=256 encoding='utf8' type='three_bytes'                  -0.49 %       ±0.69% ±0.93% ±1.21%
buffers/buffer-bytelength-string.js n=1000000 repeat=256 encoding='utf8' type='two_bytes'              *     -0.53 %       ±0.49% ±0.65% ±0.85%

Be aware that when doing many comparisons the risk of a false-positive result increases.
In this case, there are 35 comparisons, you can thus expect the following amount of false-positive results:
  1.75 false positives, when considering a   5% risk acceptance (*, **, ***),
  0.35 false positives, when considering a   1% risk acceptance (**, ***),
  0.04 false positives, when considering a 0.1% risk acceptance (***)

Merge request reports