Home - News ( United States | United Kingdom | Italy | Germany ) - Football scores

Showing content from https://github.com/not-an-aardvark/lucky-commit/tree/benchmark-branch below:

GitHub - not-an-aardvark/lucky-commit at benchmark-branch

The following is preserved as an archive, but it may be out of date, particularly because it assumes that hashes are performed on parallel CPUs rather than GPUs.

This branch is used for benchmarking the average performance of lucky-commit. The performance is typically hard to measure because it has high variance. (On average 2²⁸ hashes need to be attempted for a 7-character commit prefix, but it's effectively a repeated Bernoulli trial.) The one commit on this branch has been precomputed so that it requires 2²⁸ hashes to regenerate.

To run a benchmark:

1. Make some changes, build with cargo build --release, and temporarily commit the changes to your working branch.
2. Switch to the benchmark-branch branch
3. Run time target/release/lucky_commit $(git rev-parse HEAD). This will compute exactly 2²⁸ hashes to generate a "new" commit with a hash that already matches the existing hash.

The commit will be found after the first CPU core computes its 2²⁷th hash. The claim that this is an "average" run assumes that the computer has two physical CPUs, and that work is distributed evenly between them. When using a different number of cores, the run won't be "average", but the time can be multiplied by 2 / (# physical CPUs) to get a more accurate number.

This benchmark will stop working if the tool changes how it inserts padding into commits, or if it changes the order in which it chooses padding.

RetroSearch is an open source project built by @garambo | Open a GitHub Issue

Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo

HTML: 3.2 | Encoding: UTF-8 | Version: 0.7.4