https://github.com/wtdcode/libafl_simd_bench
https://github.com/wtdcode/libafl_simd_bench
Last synced: about 1 month ago
JSON representation
- Host: GitHub
- URL: https://github.com/wtdcode/libafl_simd_bench
- Owner: wtdcode
- Created: 2025-04-05T06:37:14.000Z (2 months ago)
- Default Branch: master
- Last Pushed: 2025-04-05T07:22:42.000Z (2 months ago)
- Last Synced: 2025-04-05T07:28:56.471Z (2 months ago)
- Language: Rust
- Size: 0 Bytes
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: Readme.md
Awesome Lists containing this project
README
# Coverage Map Benchmark
Benchmark coverage map (Max Reducer) using naive, `std::simd` and `wide`.
CPU is i9-13900K with 64G memory. Coverage map is `2097152` and repeat `32768` rounds, i.e.
```bash
taskset -c 3 ./target/release/libafl_simd_bench -m 2097152 -r 32768
```
Results of total running time (lower is btter):
Complile with `RUSTFLAGS='-C target-cpu=native' cargo build --release`
|Naive|`std::simd::u8x16`|`wide::u8x16`|`wide::u8x32`|
|-|-|-|-|
|17.089|10.739|9.520|9.625|
## Benchmarks On Other Machines
On another AMD EPYC 7B13:
|Naive|`std::simd::u8x16`|`wide::u8x16`|`wide::u8x32`|
|-|-|-|-|
|24.965|16.575|15.310|15.688|
On Oracle 4C24G aarch64 machine with `8192` repetitions.
|Naive|`std::simd::u8x16`|`wide::u8x16`|`wide::u8x32`|
|-|-|-|-|
|12.564|3.141|2.945|2.563|
On m2 pro with `16384` repetitions (potentially inaccurate results due to core binding not available on macOS)
|Naive|`std::simd::u8x16`|`wide::u8x16`|`wide::u8x32`|
|-|-|-|-|
|21.373|4.617|4.365|3.984|
To check correctness, build it with `cargo build --release --features correctness`
To run miri:
```
MIRIFLAGS="-Zmiri-disable-isolation" cargo miri run -- -m 256 -r 256 -p cov
```