It's no more a MiSTery

[RealLarry]

Hmmm...the mist-vhi.prg (Viking/SM194) seem to doesn't work anymore. I'm getting a white screen on all normal modes and black/white stripes an STEroids.

Did you enable 14M RAM?

YesNo. If I setup 14MB RAM the menu Viking/SM194 greys out, turning to off. This menu is availabe/selectable from 512KB to 8MB only, yet.

[slingshot]

Ok, it's fixed. Btw, with 14 MB, Viking is available only in STeroids mode. The reason: Viking RAM is at c00000, but 14MB ends at DFFFFF. In STeroids, the VRAM is relocated to E80000.
Did I say I was lazy to re-test everything before release?

[alberchi]

Hi
, Thank you very much for this core. It works great, except the sound in some games. For example in "the light corridor" there is like a noise. In the previous core, this happens only if I choose Stereo sound, but if I choose Mono sound, it sounds perfectly.

thanx

[RealLarry]

Ok, it's fixed. Btw, with 14 MB, Viking is available only in STeroids mode. The reason: Viking RAM is at c00000, but 14MB ends at DFFFFF. In STeroids, the VRAM is relocated to E80000.

Great, many thanks for this! This mode was really fantasic and helpful on the old mist core.
More (hardware) test coming soon...

Did I say I was lazy to re-test everything before release?

No. But that's okay, as we, the community, are takin' over the tests. You gave us the toy, we'll telling you the children's diseases

[jamesrc]

This one looks similarly broken to the STrange robots demo, which is a heavy blitter user.

Okay, sounds like I'll chalk that one up to the blitter for now, then.

I'm planning to work my way through the POV disks. Would some consolidated lists of what doesn't work be helpful?

[MasterOfGizmo]

My personal point of view is that this new core is to perfectly re-implement the ST series. The previous MIST core will still be there with STEroids, ethernet etc... but they may stay as two separate cores. Imho that doesn't hurt.

This will also make things easier as you can have both cores with different configs on the same sd card.

But that's just my point of view. If we'll be able to combine everything into one core then i'll be happy with that.

Both cores also differ technically. The old core had a lot of custom support routines in the Arm controller. Currently the new core still uses some of these but not all. In the end the new core is supposed to not have any special demands and to use the same infrastructure on arm side which 99% of the other cores also use. The first big step into that direction was the implementation of the ikbd keyboard controller in the fpga. Next will be a major rewrite of the floppy which will e.g. be slower than the one in the old core. Exactly as slow as a real floppy ...

The goal is to be able to run any ST software incl. demos 100% perfectly on the new core. This is exactly what I initially intended to achieve with the MIST board in the first place. So I am very happy that we are finally approaching that goal.

[Chris23235]

I tried some more games, the sound in Star Dust is broken, all the samples are very noisy.

[sebdel]

Am I the only one that has no sound in Crickey wot a scorcher prologue? I can't hear the quartet tune with the starfield and zooming text.

[slingshot]

I've fixed Stardust (and the DMA sound in general), is Crickey using it? Or is it the good old YM2149? Is it worked in the original core? (the YM2149 was replaced with the one in the CPC and ZX cores).

[slingshot]

Next will be a major rewrite of the floppy which will e.g. be slower than the one in the old core. Exactly as slow as a real floppy

The slowness already there But sure it would be good to have .stx or .scp support.

[sebdel]

I've fixed Stardust (and the DMA sound in general), is Crickey using it? Or is it the good old YM2149? Is it worked in the original core? (the YM2149 was replaced with the one in the CPC and ZX cores).

The original core was not even passing the boot loader iirc. If it's using the DMA, it's wrongly detecting a STe as I am in ST mode. I'll double check tonight.

Regarding my VGA problems: Do you think it would be possible to blank extreme borders like it's done on the NES core? It would hide artifacts that were invisible on CRTs and fix the backporch problem.

[slingshot]

Usually you can hear the difference between DMA digi sound and YM
I saw something in the CPC thread about missing sounds in AY, maybe it's the same issue.

Of course it's possible to invent some artificial blanking, just I wanted to remain as close as possible to the original.

[slingshot]

But that's just my point of view. If we'll be able to combine everything into one core then i'll be happy with that.

I'm sure it's possible to add everything to this core, too, just TG68k is a big question - it needs some vaseline to fit next to the FX68K. Or there can be one codebase, and compile time selection from the two CPUs.
I left out ethernec simply because I don't have such hardware to test it. Same for RTC (but maybe I'll build one, it's really easy using your tinyusb firmware).
Cleaning up the firmware from the "old" stuff would have an advantage of freeing up some kbytes of RAM in the ARM.

[joska]

Personally I'd like two ST cores - one with maximum hardware compatibility for games and demos, and one with maximum speed for "work". Combining both into a single core might work but I suspect it will be easier to develop these cores individually.

A core developed for speed does not need very detailed reimplementations of all the ST hardware. It needs a fast CPU (030 with MMU would be absolutely perfect), as much RAM as possible and extended resolutions (but still with support for standard resolutions).

[sebdel]

Usually you can hear the difference between DMA digi sound and YM
I saw something in the CPC thread about missing sounds in AY, maybe it's the same issue.

Difficult to distinguish between silence from the DMA and silence from the YM in this particular case
Wasn't there a difference between AY and YM ? something about envelope volume resolution, I don't know if it's audible, internet says no.

Of course it's possible to invent some artificial blanking, just I wanted to remain as close as possible to the original.

Ok, how about a protection of the backporch ? Blank the first line maybe ?

Oh and if you are going after perfection: another thing I remember from the original ST is that the FDD activity LED could be driven by some demos. It was blinking in rythm with the music. I swear it was connected to the YM I/O port but I can't find this information anywhere and it makes more sense that it was driven by the ikbd since the led is on the keyboard. I don't know if you can drive a MiST led from the FPGA though...

[slingshot]

But probably you can tell from running it in an emulator
In the core, the volume table is different only between AY and YM. I never investigated the AY/YM chip myself - and I don't really want to do it - just used the already available code.

For the back porch, not first line, but first column: maybe blank some cycles after hsync unconditionally in the scandoubler, or something like that should do the job.

The LED is driven currently by the fdd select signal (which is driven by the AY chip).

[Thenesis]

Personally I'd like two ST cores - one with maximum hardware compatibility for games and demos, and one with maximum speed for "work". Combining both into a single core might work but I suspect it will be easier to develop these cores individually.

A core developed for speed does not need very detailed reimplementations of all the ST hardware. It needs a fast CPU (030 with MMU would be absolutely perfect), as much RAM as possible and extended resolutions (but still with support for standard resolutions).

On the Mist it is possible to have at least 4 cores running at roughly 100 MHz (with 1 cycle per instruction on average) together with other circuits (sound, video, other I/Os). But if used together with the new Atari ST core, there might be room for 1 or 2 cores of this kind (I have not checked this).

For example you can use f32c (https://github.com/f32c) which is a very simple, yet efficient, MIPS cpu.
I'm not the author but I'm using it for a project. It works very well. I can run it from 75 MHz to 95 MHz, depending of the compilation, on a DE0-Nano (which is not faster than a Mist).

The BRAM (internal SRAM used by the CPU to store the ROM, instructions and data) is on the critical path and is probably the reason why I cannot cross the 100 MHz. But it can be easily removed.

Since the core supports little endian, the main thing to add is a 68K decoder (and remove the MIPS decoder). The decoder would generate several RISC instructions for one complex 68K instruction or just jump to microcode. To achieve this, additional pipeline stages might be needed (but maybe would it fit in the current decoding stage timings). To achieve a high frequency on FPGA, I think it is preferable to have a simple decoder and jump to microcode most of the time, the core is single-issue anyway.

[slingshot]

Difficult to distinguish between silence from the DMA and silence from the YM in this particular case
Wasn't there a difference between AY and YM ? something about envelope volume resolution, I don't know if it's audible, internet says no.

It was the PSG chip (YM), seems if you set the tone frequency to 0, then it should output a constant "1", not "0". And you can modulate this with rapidly changing the volume registers. I wonder what frequency the other digi demos are using, maybe 1?

[sebdel]

I did some more tests yesterday but I think I rebuilt just before you commited your latest fix for the digisound using the PSG. So, I guess you can ignore the ones about sound:

Source: Atari_demos_v3 archive, some msa converted to st with msa converter.
Config: Mistery@16001c8, 1MB ST, TOS1.04fr, VGA monitor
Test: boot, go into a screen or 2. I'll test extensively once there are no more problems in the first 5 min

- 1984/ACF: freezes on 1984 logo
- just buggin/ACF: ok, the 512KB detection works too
- Hifi/Aura: freezes on black screen
- Punish your machine/Party: freezes on black screen
- Ohh Crickey/TLB: no digi sound in prologue
- Big demo/TEX: ok; incl. the digital departement
- read_me.prg/TEX: ok
- Syntax terror/Delta Force: freezes on black screen
- Electra demo/Electra: intro ok, garbage on main menu (and music too fast ?)
- Things not to do/Inner circle: no digi sound
- Eat my b*llocks/Equinox: ok
- Virtual escape/Equinox: ok
- My socks are weapon/Legacy: distorted digi sound
- Overscan demos/Paulo Simoes: ok, like really ok, pixel perfect on VGA, no artifacts, no dark screen, nothing. Amazing.
- Overdrive/Phalanx: ok
- Cuddly demos/TCB: ok
- Life is a bitch/TLB: miss digisound / distortion
- Phaleon gigademo/Next: ok, fast loader ko, needs more tests
- Ventura/OVR: ok to main menu, then loader fails, needs more tests
- Nostalgic-O-demo/Oxygen: ok
- Decade demo/Inner circle: ok
- Sowatt/TCB: ok with graphical glitches (maybe only on VGA?)
- Wattaheck/TCB: ok
- Swedish New Year 1/Sync: ok
- Swedish New Year 2/Sync: ok
- Union demo/The Union: boots ok , I need a joystick to go any further. That's the original Union demo experience

Random speculation:
I believe some of the early freezes have to do with custom FDC routines. Regarding digisound: the fact that some demos have digisound or digisound with distortion makes me think that you could have a MFP problem too, on top of the PSG one. I'll test your fix tonight.

[slingshot]

Added a small fix to the FDC's multi-sector transfer mode (first time I encountered with this usage of the FDC). Also replaced the channel mixer of the YM chip using a mixing table from Hatari, most of the digi sound demos I tried has much cleaner sound now (especially in mono mode). Please report if something still wrong.

[solskogen]

Where can I get the fix? I don't see a new binary on https://github.com/mist-devel/mist-binaries

[slingshot]

It's only the source code in the git repo, no new binary released yet. If you want to test, prepare for compiling yourself

[GoingDown]

I have strange problem with this core. On my TV (Plasma tv, Mist connected via scart cable) the picture jumps litlle bit left and then back right about once per second... Looks just if the timing is bit off somewhere?

Edit: After playing around bit more, this problems suddently stopped happening. Not sure what I did.

[MasterOfGizmo]

Added a small fix to the FDC's multi-sector transfer mode

Slingshot, you are my hero. Yesterday i planned to work on the floppy but i ended up watching demos until 2am ....

Btw: what's the best self running demo? I'd like to do some demo setup that runs unattended and shows the ST's capabilities.

[sebdel]

Added a small fix to the FDC's multi-sector transfer mode

Slingshot, you are my hero. Yesterday i planned to work on the floppy but i ended up watching demos until 2am ....

Btw: what's the best self running demo? I'd like to do some demo setup that runs unattended and shows the ST's capabilities.

Cuddly demos have a kiosk mode iirc. Just leave it unattended at the main menu.