Emulator/FPGA accuracy tests? | MSX Resource Center (Página 1/2)

Ascended (10771)

17-01-2022, 21:38

I think for some basic testing of like screensplits and timing one would check demos like e.g. Unknown Reality and IO.

But for the real details which tells emulator accuracy apart, there are many different tools and test programs spread out in e.g. openMSX bug report tickets. E.g. the screensplit in Psycho World is one example. In openMSX the timing of this one got fixed fairly recently. In WebMSX the timing of it is not correct, however due to its lesser rendering accuracy (line rather than pixel accuracy), it actually looks correct there too! Just as an example that you can’t just take such a simple test at face value.

Honestly there are so many details that can be implemented inaccurately, there is no single or small set of test programs which cover it. Nobody has written and maintains a test suite of that kind, though useful, it would be a huge task.

Anyway, my ratings are:

openMSX - The most accurate.
blueMSX - Reasonably accurate. Not updated anymore so it has fallen behind.
WebMSX - Somewhat accurate. Accuracy is not a main goal.
1chipMSX - Reasonably accurate.

There are other emulators like Clock Signal, Emulicious, fMSX Android, MSX PLAYer, but I don’t really use them. I assume they probably fall in the “somewhat accurate” to “reasonably accurate” range, since I find it hard to imagine they can top the amount of research and continued maintenance that has gone into openMSX.

I develop in openMSX (as I would recommend to everybody), but I also fairly regularly test in WebMSX and on the 1chipMSX. If I encounter a bug on one of those three I usually report it to their developers.

Por ducasp

Paladin (680)

17-01-2022, 21:53

Another problem is, accurate related to what? Some demos require certain hardware constraints, so, even demos are a though proposition... They could work on PAL machine with certain z80 speed and not having certain cartridges expansion but having something else could wreck havock on the demo

Best bet for a really accurate emulation would be a decap of a V9958 and implementing it on FPGA based on that, but then I'm not sure how much FPGA resources would be needed for this kind of implementation, and probably the same would apply to 9918 / 9928, decap and implementation (as 9918 and 9928 have quirks/stuff in a way 9938 and 9958 don't implement or treat a little different, even though compatible in most things), and surely, if a good FPGA developer would be willing to look at those decaps and doing it (it is way different than the black box / data sheet /tweak it to get closer to real life stuff approach), and finally, those decaps and good microscope pictures costs a lot, so, who would pay for it?

The good stuff if that route would be taken is that what would be learned on the FPGA 99X8 after decaping and redoing it based on the real circuit could be used on software emulators too, but that is a scenario that would take quite some effort and money and don't think is going to happen anytime soon, if ever....

Por zadokallen

Resident (45)

17-01-2022, 22:42

ducasp wrote:

Another problem is, accurate related to what?

Just generic, most common hardware configs. Of course there will always be some edge cases, especially in an ecosystem as varied as the MSX standard, but these can be noted and ticked off.

I'm watching a lot of development on MiSTer, with MSX being one of my faves and I thought it could be useful to know about some tools since I'm new to this scene (and a new, rather great MSX1 core has just appeared). They did help a lot on other platforms, since otherwise you're left with only games as the yardstick, and people don't always spot glitches in them because often they can be very subtle.

Por Grauw

Ascended (10771)

17-01-2022, 23:35

The new MSX1 core is based on a Philips MSX1 right? I think the IO demo would be a good starting place.

Por zadokallen

Resident (45)

18-01-2022, 00:16

Yes, VG 8010, I believe. Today it had a major update adding a lot of Konami mappers. I will give IO a spin, thanks for the tip!

Por zadokallen

Resident (45)

18-01-2022, 11:44

This is a "VDP Timing Test" from one of the MiSTer forum regulars.

Por Grauw

Ascended (10771)

18-01-2022, 13:02

Ah, from pgimeno. If you’re looking for test cases for specific issues, it can be worthwhile to browse the list of open & closed openMSX issues as well.

Por ducasp

Paladin (680)

18-01-2022, 15:26

zadokallen wrote:

ducasp wrote:

Another problem is, accurate related to what?

Just generic, most common hardware configs. Of course there will always be some edge cases, especially in an ecosystem as varied as the MSX standard, but these can be noted and ticked off.

I'm watching a lot of development on MiSTer, with MSX being one of my faves and I thought it could be useful to know about some tools since I'm new to this scene (and a new, rather great MSX1 core has just appeared). They did help a lot on other platforms, since otherwise you're left with only games as the yardstick, and people don't always spot glitches in them because often they can be very subtle.

One of the cores is based n the ocm Tha KdL and Hara-san works, since you are being so kind wanting to help, I would recommend talking to KdL, he has a list of know stuff that needs to improve on OCM and he surely will appreciate the help, and any change there will benefit Zemmix, OCM, SX?, SM-X users as well... KdL email is available on his website, Google got kdl ocm gnogni and you should find it

Por Fabf

Champion (266)

18-01-2022, 15:55

zadokallen wrote:

(and a new, rather great MSX1 core has just appeared).

A new MSX1 core for MiSTer
Do you have a link please ?

Por Grauw

Ascended (10771)

18-01-2022, 15:58

@Fabf See New MSX1 MiSTer core (based on VG-8010).