BLITTER RE-CREATION THOUGHTS

[exxos]

how to buy that glitter
my ST is waiting for it

Parts shortages :(

Was just looking back at this post viewtopic.php?p=51449#p51449 Seems we never fully solved the odd timing issues :( I guess we need a more "readable" program of tests ;) If we knew exactly what timings was wrong we could adjust that in the FPGA core.

But its is pretty much there really by the looks of it. At last in GB6 tests ;)

IMG_5817.jpeg

[ijor]

Was just looking back at this post viewtopic.php?p=51449#p51449 Seems we never fully solved the odd timing issues :( I guess we need a more "readable" program of tests ;) If we knew exactly what timings was wrong we could adjust that in the FPGA core.

Any reason why you don't use my core? It is cycle exact, even in cases more subtle and complicated than the ones on that thread.

[exxos]

Any reason why you don't use my core? It is cycle exact, even in cases more subtle and complicated than the ones on that thread.

We did do originally right at the start, EG viewtopic.php?p=50888#p50888 . But I think there was odd timing issues with your core , but different odd issues to what we currently have with @Icky's modified suska core. But I think we went with the suska core as I had permission from Wolfgang to use it in various projects. But I don't remember exactly anymore as it was a few years ago now and @Icky did much of the testing there.

[Icky]

Any reason why you don't use my core? It is cycle exact, even in cases more subtle and complicated than the ones on that thread.

We did do originally right at the start, EG viewtopic.php?p=50888#p50888 . But I think there was odd timing issues with your core , but different odd issues to what we currently have with @Icky's modified suska core. But I think we went with the suska core as I had permission from Wolfgang to use it in various projects. But I don't remember exactly anymore as it was a few years ago now and @Icky did much of the testing there.

Yep was over 2 years ago and can't really remember the details.

Am looking to get the BLITTER board back on the bench to do some more testing so am happy to load your core @ijor to see the state and issue.

[exxos]

Yep was over 2 years ago and cant really remember the details.

I can't see the results posted in this thread. But IIRC When you tested @Cyprian's program @ijor's core it had issues with a lot of the tests. The suska core had issues also but you fixed most of them and got slightly better results than Ijor's core in GB6. Later you fixed some other tests but it messed up with other tests. Again IIRC, we suspected some odd bus arb issue as @Cyprian mentions we took 5 cycles not the usual 4. I don't think we ever looked into it.

Not sure if it was around that time we started to port the original blitter schematics over in Quartus to try and get proper cycle accuracy that way. But I think it ended up taking a lot longer than I thought it would. I remember those differential logic things getting confusing in some places as I was trying to remove them all. But think I just didn't have time to sit and figure it all out. Think we was leaning towards just trying to fix the Suska core more as it was 99% there anyway.

[ijor]

We did do originally right at the start, EG viewtopic.php?p=50888#p50888 . But I think there was odd timing issues with your core , but different odd issues to what we currently have with @Icky's modified suska core.

Are you sure? That would be very surprising. Do you have any samples about timing issues with my core?

I can't see the results posted in this thread. But IIRC When you tested @Cyprian's program @ijor's core it had issues with a lot of the tests

As far as I know my core was tested with Cyprian's and other programs and didn't have any issues. But I can test it again, if you found it otherwise.

But I think we went with the suska core as I had permission from Wolfgang to use it in various projects.

Well, Wolfgang's cores were released much earlier. So probably my core didn't even exist then :) But you are more than welcome to use mine.

... and got slightly better results than Ijors core in GB6.

It is possible to get a slightly, just slightly, better performance. But then you loose cycle accuracy, it is one or the other.

[exxos]

Are you sure? That would be very surprising.

Yep 100% sure. If it had matched a original blitter we would have stuck with it and not even bothered with the suska core.

Do you have any samples about timing issues with my core?

There are likely images *somewhere* but cannot see them on the forum. I will look around see if I can find them.. but its like 2 years or so worth of emails to look though. I'm a little slow with Covid at the moment so it could take me a while. But @Icky might be able to take new screenshots depending when he gets up and running again.

Well, Wolfgang's cores were released much earlier. So probably my core didn't even exist then :)

Looks like your core was first mentioned here. Indeed we tested your core and suska at the same time.

viewtopic.php?p=36229#p36229

But you are more than welcome to use mine.

Awesome thanks.

It is possible to get a slightly, just slightly, better performance. But then you loose cycle accuracy, it is one or the other.

But performance should match the original and be cycle accurate yes ?

We basically had these results with your core.

IMG_5776.jpeg

Some tests were running slower than expected. Basically 98% speed.

With @Icky's modified suska core we got this.

IMG_5817.jpeg

But we was still off 100% for some reason.

[ijor]

But performance should match the original and be cycle accurate yes ?

Yes, of course. I misunderstood what you were saying. Sorry ...

Some tests were running slower than expected. Basically 98% speed.

I just tested my core with Gembench. As you can see, timing is perfect, 100%. May be the timing issues you had were, perhaps, due to porting and adapting to your hardware?

FX Blitter core. TOS 2.06. GEMBENCH 6.31

20221016_021753.jpg

[Icky]

I just tested my core with Gembench. As you can see, timing is perfect, 100%. May be the timing issues you had were, perhaps, due to porting and adapting to your hardware?

FX Blitter core. TOS 2.06. GEMBENCH 6.31

@ijor was this core on real hardware or was it on something like the MiST / MiSTer etc?

I will hopefully get my test bench up tonight so will re-run all the tests I performed previously and post the details.

[exxos]

@ijor was this core on real hardware or was it on something like the MiST / MiSTer etc?

Was just wondering that as well. GB6 is showing as a STE , IIRC @ijor yours wasn't a STE chip set ? which is one reason I was using suska cores as well.

I guess we could be running into a terminology issue here as well. As cycle accurate may well be true for mister etc but when I mention cycle accurate, it mimics a real chip, which to my knowleage nobody has tested the cores in a real machine @ijor ? If they have been tested, how did you create the boards ? Can your cores deal with wakeup states from a real chip set or are they all fixed for mister ?

Wolfgang did everything on separate boards originally for suska, one chip at a time. Myself and @Icky have the standalone blitter board to test the accuracy of the cores. Possible we might have screwed up something, but what ?