On Friday 07 June 2002 00:46, Sander Zuidema wrote:
> > About the speed test programs: could you publish those? I would like to
> > run them under openMSX and see how it scores.
>
> I'll mail them to you in private so not everyone can have a laugh at what I
> call benchmarks ;)
Thanks!
openMSX results:
CPU: 998.7
VDP: 979.4
In the CPU test, none of the subtest deviated more than one from 100. In the
VDP test, most subtests were pretty close to 100, except for one 107 and one
112.
Ofcourse these benchmarks only test part of the emulation, but as there is
already a lot of variation in the scores of the different emulators, it at
least shows that emulation accuracy is not optimal yet.
It would be interesting to have some kind of MSX emulator testkit, which runs
various tests and displays how close to real a MSX the emulator is
performing. Emulator programmers can use this to make their emulator more
accurate. Apart from timing tests, feature tests and display/sound tests are
interesting as well. The first two can easily be automated, display and sound
tests would be judged by a human.
By the way, did you make sure all emulators used the same (8250) ROMs?
Different ROMs may have slightly different routines, especially the interrupt
handler installed by the disk ROM.
> > And can you say something about the amount of CPU power that is required
> > by the emulators?
>
> It;s an interesting thought for the future, but hard to test. I think I
> will have to rely
> on the information the developers give me then.
You can get an approximation by looking at the CPU usage and comparing that to
the total CPU speed. For example, if CPU usage is 80%, then a much slower
machine has little chance of running the emulator at full speed.
However, what is usually the determining factor is how fast the screen is
rendered. So the drawing method is important (how much pixels are rewritten
every frame) and also AGP/PCI bus speed. So CPU power is certainly not
everything.
> And as soon as someone ported it, it will be in the comparison as well!
> (Be prepared) In the Linux-comparison, openMSX will be present as well.
> I am very curious in the results of this new approach.
Keep in mind that we're still in the 0.X version range: openMSX is not fully
functional yet. Some things work very well, other things are pretty broken.
Still, I think in it's current state a review is possible.
Bye,
Maarten
--
For info, see http://www.stack.nl/~wynke/MSX/listinfo.html
