Instruction latency & throughput profiler for AArch64

Run

$ git clone --recursive https://github.com/ocxtal/insn_bench_aarch64.git && cd insn_bench_aarch64
$ make -j                                         # needs gcc or clang
$ ./insn_bench_aarch64 > result.txt               # dump in tsv, -m for dump in markdown

Results

• Apple M1 Firestorm
• Snapdragon 8cx Gen3 (Cortex-X1)
• aws Graviton 3
• aws Graviton 2
• GCP Ampere Altra
• NXP LX2160A (Cortex-A72)

My brief notes on M1 code tuning is here.

Copyright and License

Hajime Suzuki (2021-2022), all the codes in this repository are licensed under MIT (not applied to submodules).