It's no more a MiSTery

[Total Eclipse]

Are there any compiled versions of the latest updates?

[sebdel]

Not sure and this one is a mist-firmware update, so you would need the cross compiling environment, unless...
I found out recently that the firmware builds just fine with a single docker command:

docker run --rm -v $(pwd):/build st3ph4n/gcc-arm-none-eabi make

I don't know if it runs, I couldn't test it. If it does, it's like magic compared to the wiki version of how to compile the firmware

[slingshot]

...and by one you mean you

Feel free to join the debugging club

[sebdel]

...and by one you mean you

Feel free to join the debugging club

I'm actually considering it but there are a lot of skills involved and I don't know where to start honestly. First things first, I'm acquiring a USB blaster (I need to find one that works with linux). But then what? Could you describe your workflow roughly?

[slingshot]

...and by one you mean you

Feel free to join the debugging club

I'm actually considering it but there are a lot of skills involved and I don't know where to start honestly. First things first, I'm acquiring a USB blaster (I need to find one that works with linux). But then what? Could you describe your workflow roughly?

- Get a demo which doesn't work.
- Run in an emulator with debugger (both Hatari and Steem are good for it), trace what it does.
- Compare what happens in the FPGA.

The tricky thing is what to watch: CPU flow, FDC behavior, MFP, ...
Also in SignalTap II I found useful the sequential trigger ordering - then based on the trace generated by the emulator, it's possible to watch "checkpoints".
Debugging the European demos loader took 2 days - and the size of the fix doesn't reflects the time spent on that. But I'm a novice in 68000 assembly, so it's always a learning experience.

[vebxenon]

This core will be shown tomorrow ay AtariMad event Madrid, Spain by me. Hope people will enjoy MiST and the Atari cores

[slingshot]

Good luck! Spread the MiSTery

[vebxenon]

Good luck! Spread the MiSTery

Sure People enjoyed It!

[vebxenon]

Are there any compiled versions of the latest updates?

Same question

[slingshot]

There are not enough improvements for a new version.

[vebxenon]

There are not enough improvements for a new version.

No problem

[sebdel]

- Get a demo which doesn't work.
−> I'm going to try to get 0913 to work. It's the best music disk and you can't change tracks for some reason.

- Run in an emulator with debugger (both Hatari and Steem are good for it), trace what it does.
-> check, 68K is still my preferred ASM by far

- Compare what happens in the FPGA.
-> I'm stuck. How do you do that ? I guess some demos can be run under MON but...

I was surprised how flawless The european demos are. Also surprised at the ST connexion screen from Punish Your Machine. One of the very few 4 bit hardscrollers out there and it works.
Not so happy: the ventura demos that share a lot of infrastructure code with the european demos are still broken Looks like a FDC problem.

[Total Eclipse]

There are not enough improvements for a new version.

Fair enough.

I'm looking forward to wandering around the Skid Row demos once a new release is issued in the future

[slingshot]

I thought I fixed Ventura and Skid Row, will re-check.

[slingshot]

- Get a demo which doesn't work.
−> I'm going to try to get 0913 to work. It's the best music disk and you can't change tracks for some reason.

Maybe ikbd? Is there any clue in e.g. in Hatari if is it using a custom ikbd routine?

- Run in an emulator with debugger (both Hatari and Steem are good for it), trace what it does.
-> check, 68K is still my preferred ASM by far

Personal taste

- Compare what happens in the FPGA.
-> I'm stuck. How do you do that ? I guess some demos can be run under MON but...

Simulation or SignalTap II, however reproducing what the demo does in simulation is very tricky, as there's no sim framework for the CPU/MFP/etc. IKBD is pretty much isolated, and GSTMCU also, but that's bug-free (hehe, of course not, at least ST mode needs some adjusting in the horizontal events section).

[Total Eclipse]

I thought I fixed Ventura and Skid Row, will re-check.

I believe you did, however the only binaries I've seen are from August, from the link in the first post.

Have you uploaded any more recent ones?

[sebdel]

I thought I fixed Ventura and Skid Row, will re-check.

Skid row works fine.
Ventura, you go to load a screen (mine was sphere and checkerboard), and the loader hangs with the estimated time 'inf'. props to the OVR for handling the 'inf' case

Maybe ikbd? Is there any clue in e.g. in Hatari if is it using a custom ikbd routine?

No, I could see the cmp.b #1, $ffffc02.w, standard stuff. And you can use the keyboard but when it loads/unpack the selected track there seems to be memory corruption of some sort. Sometimes the screen display vertical lines, sometimes the scroller is corrupted. I see a couple of relative address modes access to spare registers. I wonder if it doesn't exploit 'ghost' memory mapping. Maybe a MMU problem?

[slingshot]

No, I could see the cmp.b #1, $ffffc02.w, standard stuff. And you can use the keyboard but when it loads/unpack the selected track there seems to be memory corruption of some sort. Sometimes the screen display vertical lines, sometimes the scroller is corrupted. I see a couple of relative address modes access to spare registers. I wonder if it doesn't exploit 'ghost' memory mapping. Maybe a MMU problem?

Is it using the Blitter (as it's an STe demo)? Disk DMA and blitter DMA together is buggy, I think.

[sebdel]

No, I could see the cmp.b #1, $ffffc02.w, standard stuff. And you can use the keyboard but when it loads/unpack the selected track there seems to be memory corruption of some sort. Sometimes the screen display vertical lines, sometimes the scroller is corrupted. I see a couple of relative address modes access to spare registers. I wonder if it doesn't exploit 'ghost' memory mapping. Maybe a MMU problem?

Is it using the Blitter (as it's an STe demo)? Disk DMA and blitter DMA together is buggy, I think.

I'll have a look tonight. It's back to javascript for the time being, why do I have to work

[slingshot]

I'll have a look tonight. It's back to javascript for the time being, why do I have to work

Ah, javascript doesn't look that interesting as 68k assembly

[slingshot]

Skid row works fine.
Ventura, you go to load a screen (mine was sphere and checkerboard), and the loader hangs with the estimated time 'inf'. props to the OVR for handling the 'inf' case

Inf because you didn't put the second disk to drive B: (there was an another demo, where I sucked this).

[sebdel]

Skid row works fine.
Ventura, you go to load a screen (mine was sphere and checkerboard), and the loader hangs with the estimated time 'inf'. props to the OVR for handling the 'inf' case

Inf because you didn't put the second disk to drive B: (there was an another demo, where I sucked this).

Ah, sorry about that, I swear it didn't tell me to insert disk 2 (or dit it ?)

Re: 0913, it's not working anytime soon It uses the ACSI DMA and the blitter, tests FDC interrupt bit on the MFP, writes to a bunch of Falcon video registers and write to $ffff860e which as far as I can tell is an unassigned I/O right after the DMA controller. I don't get it.

[slingshot]

Ah, sorry about that, I swear it didn't tell me to insert disk 2 (or dit it ?)

No, that was just luck I tried it.

Re: 0913, it's not working anytime soon It uses the ACSI DMA and the blitter, tests FDC interrupt bit on the MFP, writes to a bunch of Falcon video registers and write to $ffff860e which as far as I can tell is an unassigned I/O right after the DMA controller. I don't get it.

Most probably it's because of the blitter. Would need to implement its bus arbitration part correctly. $ffff860e AFAIK is the MegaSTE HD floppy mode selector (no use here, the FDC automatically supports HD images without any switch, nor the demo requires a HD sized image).

[sebdel]

Ah, sorry about that, I swear it didn't tell me to insert disk 2 (or dit it ?)

No, that was just luck I tried it.

Re: 0913, it's not working anytime soon It uses the ACSI DMA and the blitter, tests FDC interrupt bit on the MFP, writes to a bunch of Falcon video registers and write to $ffff860e which as far as I can tell is an unassigned I/O right after the DMA controller. I don't get it.

Most probably it's because of the blitter. Would need to implement its bus arbitration part correctly. $ffff860e AFAIK is the MegaSTE HD floppy mode selector (no use here, the FDC automatically supports HD images without any switch, nor the demo requires a HD sized image).

I think this is way out of my league but thanks to your super clean coding style I almost understand what I'm reading... almost. Could you elaborate on what you think needs to be done? What kind of docs do you have to replicate the blitter? I found a bunch of articles that explain pretty well how it's connected to the MCU and the roles of BR, BG and BGACK but that doesn't tell much about the internals of the blitter. I mean, do you think the internal state machine would need to be rewritten?

[slingshot]

I found a bunch of articles that explain pretty well how it's connected to the MCU and the roles of BR, BG and BGACK but that doesn't tell much about the internals of the blitter. I mean, do you think the internal state machine would need to be rewritten?

The blitter is not my work, it's the original from Till Harbaum. I just made it somewhat compatible with the FX68K, and the 3-way handshaking of the bus arbitration, which is not in TG68K. That didn't make it more compatible with demos. I think the best would be to use the original schematics (which worked pretty well for the GSTMCU). But it's not a trivial work.