Skip to content

Conversation

jangko
Copy link
Contributor

@jangko jangko commented Oct 6, 2025

Linux amd64, with and without avx512 instruction:

Operation Nim-bn254 mcl-x64 mcl-avx512 nim->mcl-x64(%) nim->mcl-avx512(%)
G1 + G1 511.494.241 4.714.393 4.834.672 faster 99,08 faster 99,05
G1 * FR 718.740.303 70.999.718 71.017.095 faster 90,12 faster 90,12
G1 G2 pairing 25.122.616.596 5.310.392.374 4.661.134.042 faster 78,86 faster 81,45

No data available for arm64 yet.

TODO:

  • nightly/release build for arm64
  • docker build for arm64

fix #3378
fix #2713

|Operation    |      Nim-bn254 |    mcl-x64    | mcl-avx512    | nim->mcl-x64(%) | nim->mcl-avx512(%) |
|:------------|---------------:|--------------:|--------------:|----------------:|-------------------:|
|G1 + G1      |    511.494.241 |     4.714.393 |     4.834.672 | 99,08           | 99,05 |
|G1 * FR	    |    718.740.303 |    70.999.718 |    71.017.095 | 90,12           | 90,12 |
|G1 G2 pairing|	25.122.616.596 | 5.310.392.374 | 4.661.134.042 | 78,86           | 81,45 |
@jangko jangko marked this pull request as ready for review October 7, 2025 11:01
@jangko jangko merged commit 552755a into master Oct 8, 2025
18 checks passed
@jangko jangko deleted the mcl branch October 8, 2025 05:32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Perf: EcAdd precompile extremely slow Add mcl (and possibly constantine) backends for alt_bn128 precompiles
2 participants