MaikKlein / ecs_bench

Benchmarks of various Rust Entity Component Systems

Benchmarks

Benchmarks are run on Travis CI.

Benchmarks are located in benches/[bench_name]_[ecs_crate_name].rs.

Library	pos_vel build	pos_vel update	parallel build	parallel update
calx-ecs	{pos_vel_build_calx_ecs}	{pos_vel_update_calx_ecs}	{parallel_build_calx_ecs}	{parallel_update_calx_ecs}
constellation	2,257 µs/iter (+/- 344)	77 µs/iter (+/- 0)	{parallel_build_constellation}	{parallel_update_constellation}
ecs	15,876 µs/iter (+/- 279)	2,211 µs/iter (+/- 13)	{parallel_build_ecs}	{parallel_update_ecs}
froggy	5,376 µs/iter (+/- 117)	76 µs/iter (+/- 0)	{parallel_build_froggy}	{parallel_update_froggy}
specs	3,359 µs/iter (+/- 191)	47 µs/iter (+/- 0)	{parallel_build_specs}	{parallel_update_specs}
trex	11,674 µs/iter (+/- 641)	2,179 µs/iter (+/- 8)	{parallel_build_trex}	{parallel_update_trex}

Visualization of benchmarks, smaller is better.

pos_vel

• 1000 entities with position and velocity components
• 9000 entities with position components only
• stub render system
• physics system: position += velocity

parallel

• 10000 entities with 3 simple components R, W1 and W2
• w1 system reads R and writes to W1
• w2 system reads R and writes to W2
• systems could be run in parallel

Notes

• the benchmarks explore a limited subset of ECS use-cases and do not necessarily reflect the peformance of large-scale applications
• froggy is technically not an ECS, but a Component Graph System (CGS)