asmbench

A benchmark toolkit for assembly instructions using the LLVM JIT.

Usage

To benchmark latency and throughput of a 64bit integer add use the following command:

python -m asmbench 'add {src:i64:r}, {srcdst:i64:r}'

To benchmark two instructions interleaved use this:

python -m asmbench 'add {src:i64:r}, {srcdst:i64:r}' 'sub {src:i64:r}, {srcdst:i64:r}'

To find out more add -h for help and -v for verbose mode.