Coming soon! Steem SSE 3.8

[Cyprian]

+1 for a new shortcuts

Elluigi I see you use old 3.6.4 I would recommend to switch to current one 3.7.3 or beta 3.8.0

[elluigi]

Not compatible with DemoBaseST but version will normaly be fixed with 3.8. I'll test with beta. THX Cyprian.

[Steven Seagal]

Hi Steven,

Many thanks for your program and great support on it

This is probably not the place but I would like to ask you if it will be possible to add a shortcut which allow to swap ST Model (very useful with DBST and many other games/progs ). New features ?

Actually there's an apart thread for requests: http://www.atari-forum.com/viewtopic.php?f=94&t=26998

But I notice a mistake in the screenshot. In fact, ST model change has immediate effect at your risk.

[elluigi]

Thanks Steven

[Steven Seagal]

closure_txt2-min.png

It is time for a disclosure: Steem SSE v3.8 will support a new demo by Sync shown first at STNICCC 2015, called Closure.
Thanks to video timing refactoring, and the fact that wake-up states were already handled in Steem SSE.
Despite this new progress, emulation of wake-up states is still tentative (many hacks...).
Important: You must select a wake-up state on the Machine option page for it to work, and accurate disk access times in the Disk manager for it not to crash later.
STE: WS1; STF: not WS1

BTW I'm surprised there's no thread for this very interesting demo, or did I miss it?

[Cyprian]

cool.
in STE DL6 WU1 WS1 mode demo looks wrongly . It seems screen have shifted bitplanes

[Steven Seagal]

in STE DL6 WU1 WS1 mode demo looks wrongly . It seems screen have shifted bitplanes

I forgot to say: option 'Hacks' for this mode + STF WS2 (he he)


EDIT: also, don't check the code for those hacks yet, if you're curious. It's nothing special, but unfortunately the source lags a week or two behind, because it's much work isolating and uploading changes.
EDIT: code uploaded R457

[troed]

It is time for a disclosure: Steem SSE v3.8 will support a new demo by Sync shown first at STNICCC 2015, called Closure

Nice! I did a brief test with the latest released version a while back and I'm impressed you've managed to sort out all the small glitches already! ... There were a bunch ...

Btw, there are no (GLUE-CPU) wakestates on STE so you can simplify settings by not even offering it as an option. It behaves like WS1 - except for the differences in the state machine, of course.

/Troed

[Steven Seagal]

Nice! I did a brief test with the latest released version a while back and I'm impressed you've managed to sort out all the small glitches already! ... There were a bunch ...

Apparently GLU/Shifter refactoring helped beforehand, like for mono demo Time Slices.
Some glitches were bugs revealed by the demo.
At the same time, like all that concerns wake-up states, emulation is very pragmatic. I don't have a real model yet. Just adding cases to a collection of hacks, without breaking the rest (eg Overdrive/Dragon). But the demo does display correctly.

Btw, there are no (GLUE-CPU) wakestates on STE so you can simplify settings by not even offering it as an option. It behaves like WS1 - except for the differences in the state machine, of course.

/Troed

Actually STE WS2 is used to emulate the pixel shift in some Spectrum-like pics. It's not GLUE-MMU (you don't mean CPU?) but it's more convenient than yet another option.
Still it's somehow a relieve that WS1 is the one that should work.

[troed]

Btw, there are no (GLUE-CPU) wakestates on STE so you can simplify settings by not even offering it as an option. It behaves like WS1 - except for the differences in the state machine, of course.

Actually STE WS2 is used to emulate the pixel shift in some Spectrum-like pics. It's not GLUE-MMU (you don't mean CPU?) but it's more convenient than yet another option.
Still it's somehow a relieve that WS1 is the one that should work.

"Spectrum 512 pixels" is caused by Shifter-MMU offsets - and is so far undetectable via software. It's AFAIK completely independent of GLUE-CPU wakestates. Only ST has GLUE-CPU wakestates, but both ST and STE can have the "Spectrum 512 pixels".

(Which is pretty reasonable considering the Shifter is still a separate chip in the STE)

/Troed

[Steven Seagal]

I see you did mean GLUE-CPU...
Yes, the option name is not very accurate, just a way to show a curiosity, like "Shifter panic". Who wants broken demos?

[Steven Seagal]

More wake-up state madness...

A good tester reported a bad scroller in World of Music by New Power Generation in STE mode.
I saw that on the real STE, the scroller staid smooth, only changing colour.

NPG_WOM1-min.png

NPG_WOM2-min.png

This is how it looks on a real STE. The clown picture is shifted.

On an STF this should look like this:

NPG_WOM_STF-min.png

The scroller staying in the same colour.
On the STE, the problem is a "line +2" triggered by a late switch back to 50hz after the lower border has been removed by a switch to 60hz, because timings are different on an STE, which is emulated in Steem SSE (if it looks better in other emulators, maybe they don't emulate this difference). It's quite frequent.

But a "line +2" by itself will destabilise the Shifter, there's one more word (2 bytes) loaded in it. This explains the shift of the picture above: the Shifter is destabilised at the lower overscan limit, and stays in that condition during the next frame.


The beauty of this is that it also fixes the Nordlicht case in STE mode.

nordlicht-min.png

I'm not one to brag, but If you want to check STNICCC 2015 demos, you know which emulator to use.

[Steven Seagal]

Kryos-min.png

A quite confusing STE screen to get to correctly display. There's horizontal scrolling, on the fly change of the video counter and unstable overscan. So, let's use some hacks. In fact, some were already in use for XMAS 2004, there's some generality in those.

[Steven Seagal]

lazer_insane-min.png

The insane programmer of this demo inadvertently told the STE to start and stop a DMA sound sample at the same address. For the STE, it actually means it will never stop instead. Steem was more civilised, but that was wrong.

[ijor]

Only ST has GLUE-CPU wakestates, but both ST and STE can have the "Spectrum 512 pixels".
(Which is pretty reasonable considering the Shifter is still a separate chip in the STE)

I was always very curious about the exact reason why there are no GLUE wakeups in the STe. Of course, the fact that GLUE and MMU are combined has much to do with it, but not necessary. Depending how both chips were combined (and I would really like to open one and see this), wakes ups could still exist, even when being more rare.

But my main question is if this was incidental or by design. It is very possible, that wakeups, while harmless (from Atari engineers point of view) on the ST, would have been fatal in the STe. May be they found that when developing the STe. And may be, one of the reasons to combine the chips was actually for the very purpose to avoid the wakeup. It is not trivial to avoid the wakeups otherwise.

Btw, I agree with Steven. Even when it's just semantics, I think it's more natural to talk about GLUE-MMU than GLUE-CPU wakeups. The CPU point of view is just a consequence. MMU (together with GLUE) is the reason.

Sorry Steven if I'm digressing

[troed]

With "GLUE-CPU" I mean that there's an offset 0-3 cycles between those two chips. Why would GLUE-MMU be better?

/Troed

[ijor]

With "GLUE-CPU" I mean that there's an offset 0-3 cycles between those two chips. Why would GLUE-MMU be better?

Well, again, it is probably just semantics. But I do think GLUE-MMU is a better name.

The wakeups are produced by logic inside GLUE and MMU, and because of how exactly they both interface which each other. The CPU shift is just a consequence of MMU introducing wait states to some CPU accesses. In turn, those wait states are produced depending on the current wakeup. You can say the CPU shift is just collateral damage

Moreover, the CPU-GLUE shift depends on the specific code is running, and even the location of the code (code access in ROM is not aligned). While GLUE-MMU wakeup relation is constant and permanent since power-up to turn off. You could even visualize it by scoping the clocks.

I don't know. May be it just an issue of the point of view. From a software point of view it might sound more intuitive as CPU shift. From the actual hardware, it seems to me MMU reflects the behaviour better.

[Steven Seagal]

Steem SSE being a low-level emulator, it naturally uses "GLUE-MMU".

The clock is running and the end of the month is close. A promise was made.
There's still some suspense, 2016 being a leap year, will Steem SSE v3.8 be released on 29/2?
Who knows? What's for sure is that giant leaps will be accomplished again in the field of emulation.

[Steven Seagal]

Steem SSE 3.8.0 WIN32 has been released on 28 February 2016.
Another milestone in emulation history!


v3.8.0 Features
_________________

- Configuration files (option sets may be loaded or saved as ini files)

- Some icons instead of text in the status bar

- Keyboard click (or not!) is now a persistent option

- Direct handling of ACSI hard disk images, in addition to existing
GEMDOS emulation.
ACSI uses an image file and is lower level, GEMDOS uses your PC hard
disk and is more practical.
Right-click on respective icon to toggle emulation. It's even possible to
combine ACSI and GEMDOS to edit your files.

- PAL aspect ratio also available for window "stretch" mode

- Big display sizes are now 400 x 275, 413 x 275, 416 x 281

- STE default memory now 4MB, many demos require as much

- High-speed emulation up to 2GHz (no guarantee)

- New ini-only options TosDefaultCountry and MonochromeDisableBorder

- Emulation improvements
68000: Better cycle rounding in some instructions (Bird Mad Girl Show)
No rounding for ROM or IO access, except Shifter (based on
original hardware specification)
Start horzontal blank interrupt before prefetch for some
intructions: only TST for now (Phantom)
Blitter: Hog bit, some timings
Sound: Some DMA/Microwire improvements/hacks (A Little Bit Insane,
Light megademo, Sabotage)
GLU: Line +20 with HSCROLL (E605, Sommerhack 2010, ...)
Shorter lower overscan at 60hz (It's a girl 2)
MMU: Fixes (hacks) for writes to videocounter (Sommarhack 2010 greets,
Ooh Crikey hidden screen #2)
Shifter: Monochrome raster effects (Time Slices)
More wake-up state-sensitive tricks (Closure, World of Music-NPG,
STNICCC2015/Nordlicht...)

All features, old and new, are explained in the included manual.


v3.8.0 Internal
_________________

- Video timings refactoring (GLU/Shifter)
Steem acts more like the real GLU chip to decide video timings on the fly.
It's a big internal change, but it shouldn't alter program support much,
or else for the best (hopefully!)

- "Cycle-precise" ACIA/6301 transmission (before, there was a 512 cycle
imprecision).

- MFP emulation refactoring
IACK cycle and latch delay are now handled in "real time".


v3.8.0 bugfixes
_________________

- Launching Steem with a memory snapshot as argument (broken v3.7)
- Shifter rendering when HSCROLL is set at start of scanline (We Were)
- Shifter HSCROLL prefetch in medium res when address $ff8264 is used
instead of $ff8265 (Desktop Center) (broken v3.4)
- 6301 emu hack can write TDRE (Pothole 2) (broken v3.7.3)
- Floppy DMA FIFO filled also if invalid address (Sabotage) (broken v3.7)
- Some large display modes corrections (Appendix, Gobliins II-ICS, MOLZ...)
- Blitter blit mode timings (AC2011)

Thx to Amilo for all his bug reports.


Download
__________

http://sourceforge.net/projects/steemsse/
http://ataristeven.exxoshost.co.uk/Steem.htm


Bugs
______

I usually claim that there are no bugs in the release, but this time I won't
because due to all those reports many changes were carried out during the last
weeks before release.

Please report bugs here:

viewtopic.php?f=94&t=29412

[TheNameOfTheGame]

Fantastic! I am very interested in the and can't wait to try it out:

Code: Select all

- Direct handling of ACSI hard disk images, in addition to existing
GEMDOS emulation.
ACSI uses an image file and is lower level, GEMDOS uses your PC hard
disk and is more practical.
Right-click on respective icon to toggle emulation. It's even possible to
combine ACSI and GEMDOS to edit your files.

[Steven Seagal]

Already changed the file, icons for "chipset" 1 and 2 failed to appear.

[bugs]

Thanks for all your hard work on the improvements steven seagal, or whoever you are

[axelfoley]

Hi Steven. Thank's for 3.8.
I still have issue with fullscreen, no border d3d with interpolated scanlines. Ths picture hangs low, missing last part of the bottom.
Without interpolated scanlines fullscreen works perfect.

[Cogweasel]

Excellent news! Looking forward to trying it out! Steven! You are awesome, both at coding and hand to hand combat!!

[Maartau]

Huge thanks Steven !

** Any boiler/debugger in view too ? . **