Latest results, statistics and graphs.
- WP2
- Explore new approaches to optimizing
TranslationBlockaccess.- Deferred to prioritize vle/vse optimization.
- Measure effectiveness of
vsetvloptimization and extend this work.- Deferred to prioritize vle/vse optimization.
- Continue optimization of the
vle.vandvse.vhelper functions.- In progress set up of comparison tests with ARM.
- In progress optimization of the vle/vse helper function loop for odd data sizes.
- Explore new approaches to optimizing
- WP2
- Enlarge testsuite with more memory benchmarks.
- Explore small block loads/stores optimizations.
It is useful to keep an idea of what we have achieved since 1 May 2024. We use ns/instruction as our metric.
| Metric | Start | Now | Speedup | Notes |
|---|---|---|---|---|
memcpy worst case |
6.84 | 5.06 | 35% | |
memcpy best case |
2,594.12 | 72.28 | 3,588% | 8,192 bytes |
| SPEC CPU 2017 average | TBC | 11.30 | - | |
| SPEC CPU 2017 best case | 12.35 | 7.42 | 66% | 628.pop2_s |
Percentage changes relative to "now". SPEC CPU 2017 results are from August, since we have stopped measuring this.
The SPEC CPU 2017 average needs checking, since the set of benchmarks which ran successfully is different.
- Paolo Savini will be on vacation 20-24 and 27 September.