Summary v0.9.26

Based on our detailed analysis, we have compiled polkadot version 0.9.26 with the following optimization options

rustup override set nightly
export RUSTFLAGS="-C target-cpu=native -C codegen=1"
cargo build --release --target=x86_64-unknown-linux-gnu --locked -Z unstable-options

In the table below, the benchmark scores are listed for a few machines with different CPU architectures. As explained here, we only focus on CPU scores and took the median value over 20 runs.

CPU arch	nb_build	BLAKE2-256	diff (%)	SR25519-Verify	diff (%)
	32	1410	5.2	1012.9	1.6
i7-12700	official	1340	0	997.1	0
	docker	1340	0	996.2	-0.1

CPU arch	nb_build	BLAKE2-256	diff (%)	SR25519-Verify	diff (%)
	28	1120	0.4	820.2	11.6
i9-9900K	official	1115	0	735.2	0
	docker	1100	-1.3	724.7	-1.4

The column diff is the relative improvement in percentage compared to the official polkadot binary on the same machine.

If you want to contribute to this list:

You first have to compile polkadot yourself using the options above for your machine.
Then you run our benchmark script when there are no other resource hungry tasks running.
Finally, send us the feather file generated by the parse script.