Mister SNES port to MIST possible?

[slingshot]

Try this please:
https://github.com/gyurco/SNES_MiSTer/r ... tag/test15
(and watch out especially for sound issues - there's a big DSP optimization, hopefully not buggy)

[ijor]

Code: Select all

Flow Status	Successful - Wed Sep 11 21:47:40 2019
Quartus II 64-Bit Version	13.1.4 Build 182 03/12/2014 SJ Web Edition
Revision Name	snes_mist
Top-level Entity Name	SNES_MIST_TOP
Family	Cyclone III
Device	EP3C25E144C8
Total logic elements	24,429 / 24,624 ( 99 % )[/quote]

 8O  8O  8O
 
 [quote="alexh"][quote]Total logic elements   24,429 / 24,624 ( 99 % )
Total combinational functions   24,150 / 24,624 ( 98 % )[/quote]
How did you manage to get that to route? With Xilinx FPGA's I'm lucky if I can get 70% full and still route.[/quote]

Really? With modern FPGA families? FPGA's usually  have spare routing resources, on purpose, to avoid that kind of problem.

[Televicious]

Try this please:
https://github.com/gyurco/SNES_MiSTer/r ... tag/test15
(and watch out especially for sound issues - there's a big DSP optimization, hopefully not buggy)

Hmm still same. Green text on the field and freezing. Actually gets more graphic corruption than before as it plays too. Don't think Tecmo SB uses DSP, just Regular ROM with Battery.

[slingshot]

Try this please:
https://github.com/gyurco/SNES_MiSTer/r ... tag/test15
(and watch out especially for sound issues - there's a big DSP optimization, hopefully not buggy)

Hmm still same. Green text on the field and freezing. Actually gets more graphic corruption than before as it plays too. Don't think Tecmo SB uses DSP, just Regular ROM with Battery.

What version? I played about 5 minutes of Super Bowl I, US.

[alexh]

Total logic elements 24,429 / 24,624 ( 99 % )
Total combinational functions 24,150 / 24,624 ( 98 % )

How did you manage to get that to route? With Xilinx FPGA's I'm lucky if I can get 70% full and still route.

Really? With modern FPGA families? FPGA's usually have spare routing resources, on purpose, to avoid that kind of problem.

It was the case with Virtex6 760, Virtex7 2000T and Ultrascale XCVU440. I've not been MrFPGA for 1-2 years so it may have improved. Maybe the problem is with big Xilinx devices? I know they are not a monolithic die. They are 2 or sometimes 4 die and perhaps when your logic and clocks cross die the tools just don't optimise well?

[arty]

Try this please:
https://github.com/gyurco/SNES_MiSTer/r ... tag/test15
(and watch out especially for sound issues - there's a big DSP optimization, hopefully not buggy)

I recorded audio tracks from Chrono Trigger Demo Mode for visual comparison of DSP optimization with previous|original version of DSP implementation.
On next screenshots: upper stereo track is audio from Test Core Ver10, lower stereo track from Test Core Ver15. Both tracks look(sound) almost identically with visual difference in noise transition parts(but sound the same). As for highest possible level of comparison, it seems we need real SNES for recording with same hardware in addition to audio tracks from MiST.
So, well done. Thank you, Slingshot! This additional 5% of FPGA will be useful for other features.

[slingshot]

Thanks for testing!
However about extra features: as GSUand SDD1 could fit now, the limit is the extra memory bandwidth they're using.

[arty]

Thanks for testing!
However about extra features: as GSUand SDD1 could fit now, the limit is the extra memory bandwidth they're using.

You're welcome!
Needless to say it would be great to see these extra chips for SNES in action on MiST.

[Televicious]

Try this please:
https://github.com/gyurco/SNES_MiSTer/r ... tag/test15
(and watch out especially for sound issues - there's a big DSP optimization, hopefully not buggy)

Hmm still same. Green text on the field and freezing. Actually gets more graphic corruption than before as it plays too. Don't think Tecmo SB uses DSP, just Regular ROM with Battery.

What version? I played about 5 minutes of Super Bowl I, US.

That's the one. I tested 2 and 3 also and did not see any of the same issues. Madden works perfect and other tecmo games are good too. I did try different roms also and the issues persisted. Sometimes it takes a bit before it starts acting up. During plays the current down and yards to first pop up on the field and they are bright green instead of silver. Occassionally they come up silver. Then as I play through the game I get graphic corruption in the playbook on that latest version. I haven't had any interceptions happen on the newest core yet, but I did have it freeze regardless randomly and on the previous at least twice in a row the ball was intercepted and it froze.

Also, curious, which GSU chip were you thinking to implement?

[gtv3000]

Thanks for testing!
However about extra features: as GSUand SDD1 could fit now, the limit is the extra memory bandwidth they're using.

thank's for your job!!

thanks to the core of the snes and the new fixes for the core minimig aga the old mist has a new life!
I tested the core v15 on Super Mario World USA, Super Metroid and Knights Of The Round without freeze for over 2 hours:

Link of Video from my Youtube Account: https://youtu.be/4-IUAKJoT1k




:

[slingshot]

That's the one. I tested 2 and 3 also and did not see any of the same issues.

And did you use the save/load feature before the freezes, or just happened without that?

Also, curious, which GSU chip were you thinking to implement?

Well, neither of them at the moment, I don't think I can find a way to satisfy the memory bandwidth requirement.

[slingshot]

thanks to the core of the snes and the new fixes for the core minimig aga the old mist has a new life!

Old, but not obsolete - as the Terminator said.

[gtv3000]

That's the one. I tested 2 and 3 also and did not see any of the same issues.

And did you use the save/load feature before the freezes, or just happened without that?

I used the savings with a 2 x 64k file size for Super Mario World USA and Super Metroid JAP and they work properly, after 1 hour of play I played another 40 minutes at Knights Of The Round with no freeze, I have a MIST rev1.3 with Hinix memory

[gtv3000]

And did you use the save/load feature before the freezes, or just happened without that?

I used the savings with a 2 x 64k file size for Super Mario World USA and Super Metroid JAP and they work properly, after 1 hour of play I played another 40 minutes at Knights Of The Round with no freeze, I have a MIST rev1.3 with Hinix memory[/quote]

[gtv3000]

slingshot wrote: Old, but not obsolete - as the Terminator said



I had put the MIST on sale to buy a MISTER or wait for an FPGA all in one more powerful but with the core of the SNES and that of the AMIGA fixed I decided to keep it

[slingshot]

I had put the MIST on sale to buy a MISTER or wait for an FPGA all in one more powerful but with the core of the SNES and that of the AMIGA fixed I decided to keep it

Good decision
Actually the problem is with the "more powerful" FPGAs that even they have more internal RAM/more logical elements, in this price category, the speed grade is not higher. So even if there's place in them, you still won't get any CPU core running at 100MHz for example.

[Televicious]

That's the one. I tested 2 and 3 also and did not see any of the same issues.

And did you use the save/load feature before the freezes, or just happened without that?

Also, curious, which GSU chip were you thinking to implement?

Well, neither of them at the moment, I don't think I can find a way to satisfy the memory bandwidth requirement.

I have tried both with and without .sav mounted, but had not written the file. Usually to recreate the interceptions I play as a terrible team against the packers or vikings preseason man vs com in the snow since that was the first time I noticed it, but I've had it happen in all sorts of different settings and games and com vs. com is the easiest way to test. Sometimes it will play through whole games with minor corruption. Only have 2 roms of this, I'll try some different ones, but this is the only game I could find that still has significant issues. Also, a fave. I have tried some hacked roms and had issues too, but I wasn't expecting total compatibility with those. Some of the hacked ones actually run really well though.

[sardine]

Please help..

tried every version of snes rbf

tried 3 usb pads, cheap chinese snes pad, psx classic usb and I just got the official licenced megadrive pro 6 button usb pad

every pad in windows works. mapper.exe used for example megadrive pro 6 button pad
joystick_remap=0F0D,00C1,1,2,4,8,200,20,10,100,0,0,400,800,40,80

in the mist menu works, in joystick test every button on the pad i press shows up on the mist with the correct description (A or select etc)
while in a snes game I can press start + Select and the mist menu appears.

NOTHING will work in a snes game.

I've tried joystick swap (in early snes rbf) latest snes rbf its not there, i'm using VGA output.

thanks.

[sardine]

Never mind got it working.

no idea why i just formatted the sd card and copied it all over again.

now works

[DanyPPC]

This Core works great !

But I need help to rename HI, Lo and ExHi roms.
Is There a specific table online or can I use a tool to recognize and rename all my roms ?

Thanks for your help

[slingshot]

This Core works great !

But I need help to rename HI, Lo and ExHi roms.
Is There a specific table online or can I use a tool to recognize and rename all my roms ?

Thanks for your help

Most of the ROMs are Lo, there are only one or two ExHi, the rest is Hi

[DanyPPC]

Ok, thanks Slingshot

[MasterOfGizmo]

Most of the ROMs are Lo, there are only one or two ExHi, the rest is Hi

How do emulators distinguish these automatically? I'd assume that this is e.g. stored in the header of smc files. Couldn't we just parse the header during upload? Like we do with the load address of c64 cartridges?

[slingshot]

Most of the ROMs are Lo, there are only one or two ExHi, the rest is Hi

How do emulators distinguish these automatically? I'd assume that this is e.g. stored in the header of smc files. Couldn't we just parse the header during upload? Like we do with the load address of c64 cartridges?

AFAIK they're using some heuristics to look at instructions at the possible entry points, and use what's more plausible. On MiSTer, it's implemented on the ARM side.

[squid4]

Is There a specific table online or can I use a tool to recognize and rename all my roms ?

https://github.com/mist-devel/mist-boar ... oreDocSNES "ROM Validation and SAV File Generation" section.