RTG Support Picasso96 2 or 3 ?

[mahen]

Hi !

Just realized there was a new Picasso 96 3 update lately (3.2.1).
Is it any useful (& compatible with the MiSTer RTG implementation) to upgrade from Picasso 2 to Picasso 3 ? (in terms of performance / reliability)

Cheers

[mahen]

BTW, I've noticed that forgetting to delete the DEVS:Monitors/Native monitor from the Picasso96 distribution would create instability (regardless of whether the PAL or MiSTer is used)

[limi]

Just realized there was a new Picasso 96 3 update lately (3.2.1).
Is it any useful (& compatible with the MiSTer RTG implementation) to upgrade from Picasso 2 to Picasso 3 ? (in terms of performance / reliability)

As far as I know, the RTG support in Minimig/MiSTer was based around the Picasso 2 drivers, so that’s what’s been tested. So unless there’s a specific feature in Picasso 3 that you need, I assume it’s better to stick with 2 (which also happens to be freely available).

[mahen]

Okay, it's basically not supported then. Thanks for the reply !

[mahen]

Bumping this, just in case, someone has made the test meanwhile (upgraded to P96 3.x) ?

I tend to get RTG crashes (after modes switches / rebooting : gray screen needed a complete MiSTer reboot) - although I need to test in the long run with DCache disabled.

Would have liked to try P96 3.x to see if it makes a difference. (also, maybe v.3 supports dragging screen ? Don't remember).
cheers !

[docbobo]

After reading up a little bit on P96 v3 yesterday, I decided to give it a try. Given that the release notes were talking about some performance improvements, I decided to perform a few P96Speed tests on the old stack first, then doing them again on P96 v3.3.1, assuming that it comes up properly.

Updating from v2 to v3 was pretty painless as the installer allowed me to just perform an update operation of the existing install. Restart. Everything comes up as expected. General compatibility seems to be there.

Now for the test results. First of all, the tests that I did on v2 in 1080p are useless for comparison. When I ran them, P96Speed was only using the left half of the screen. With v3 it is using the fullscreen. So I guess that might be at least one bug fixed in v3.

I also ran test in 640x480x8 because that's the resolution that was used in the included Compare Database. In general, the performance of v3 is on the same level with a few notable exceptions:

- ScrollRaster() X and ScrollRaster() Y go up from 22 op/s to 28 op/s. That's a 25% performance increase, obviously outside of tolerance. Nice.
- Both BlitBitMap() and BlitBitMapRastPort() jump from ~800 op/s to ~1300 op/s. That's a 60% increase.

The rest of the operations in the low-level category are within measurement tolerances.

I believe as a result of the improvements above, some of the intuition performance tests of P96Speed also see improvements:

- MoveWindow() goes from ~90 op/s to ~200 op/s. So it's basically twice as fast.
- Same is true for CON-Output, which goes from 30 op/s to 60 op/s.

Hopefully this is useful.

[docbobo]

BTW, I also get those crashes you were talking about, @mahen. They are not gone with v3. So if that's the only thing you are looking for, v3 might not be what you are looking for.

[mahen]

Thanks so much for taking the time to carry out those tests ! Most appreciated & helpful !

Regarding the gray screens issues, do you get them only when using RTG ? Do you use a minimig_vadjust.dat file ?

Cheers !

[docbobo]

I can't say if this is limited to RTG. Yes, I have only seen those problems when RTG was being used, but I am not sure if that's sufficient evidence.

That said... I am using a minimig_vadjust.dat file and was seeing those problems when doing the performance tests from above. I've removed the file after your question, executed the tests a few more times and haven't seen the problem yet again. So maybe, this is related. I'll certainly keep an eye on that...

Some more context on the results from above: all off those were done with D-Cache OFF. Turning D-Cache on has to interesting effects: OpenWindow() will see almost constant 30 op/s, but MoveWindow() will drop to about 150 op/s - however, MoveWindow() seems to have a significantly wider range looking at the different numbers I've been seeing. So maybe that's just coincidence.

[mahen]

Yep I kinda have the feeling it's not only RTG related but maybe a combination of issues, as, when it occurs, it also affects AGA screens.
Would be super cool if you could try for a few days without the vadjust file !
Take care

[docbobo]

No need to wait that long - just had the problem occur. So it does not seem to be related to that.