| --- Log opened Sat Sep 19 00:00:50 2015 | ||
| poke53281 | Got an email from Fabrice Bellard. Told me, that he ported his code to asm.js code and that I need to revise my benchmark :) | 17:51 |
|---|---|---|
| dalias | :) | 17:52 |
| dalias | which code? | 17:52 |
| dalias | linux.js or something? | 17:52 |
| dalias | (his x86 emulator) | 17:52 |
| dalias | or does he have an or1k one? | 17:52 |
| poke53281 | linux.js | 17:53 |
| poke53281 | But he admitted, that my emulator is still faster :) | 17:54 |
| poke53281 | https://github.com/s-macke/jor1k/wiki/Benchmark-with-other-emulators | 17:58 |
| poke53281 | The benchmark is not perfectly valid. Because I might compare different kernel images with different urandom generators. | 17:58 |
| poke53281 | Also gzip and bzip2 might be from busybox or the standalone one. | 17:58 |
| dalias | *nod* | 17:59 |
| dalias | to make it more meaningful you'd want to build at least the same userspace for both | 17:59 |
| dalias | e.g. using buildroot or something | 18:00 |
| poke53281 | Additional, benchmarks with Javascript are difficult. The variation is pretty high. | 18:01 |
| poke53281 | Yeah, but this is too much effort. Maybe something with awk. | 18:02 |
| poke53281 | There might be also assembler fastpaths for x86 at some places. For example in gzip and bzip2. | 18:03 |
| poke53281 | Anyhow, the difference between the emulators is/was pretty high. So, the main statement of the benchmark was valid. | 18:04 |
| poke53281 | Any other well known bash commands, which can be used this way for benchmarks? | 18:05 |
| poke53281 | Good, wrote him, that I don't think, that he can beat jor1k ..... ever ..... . | 19:07 |
| poke53281 | And if he accepts the challenge ;-) | 19:07 |
| dalias | poke53281, :) | 22:20 |
| --- Log closed Sun Sep 20 00:00:51 2015 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!