PhotoChrome v6

GFA, ASM, STOS, ...

Moderators: simonsunnyboy, Mug UK, Zorro 2, Moderator Team

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Fri Dec 06, 2013 11:28 am

I should add that Zerkman's MPP displayrouts are very, very good and I've been recommending that system privately to people for using in demos etc. while I work on the display support side of PCS.

My original plan was to sidestep doing new modes completely and provide conversion-to-MPP as part of the PCS upgrade and IMO that is still worthwhile. However there's always a temptation to try some new modes to see if anything else can be done.


I think things must be nearing the limits already though for the displaymode part. In most routines used now, most cycles are consumed with CCs. I only found 'free cycles' by combining blitter with overscan in a fiddly pattern (lots of tiny blits interleaved vs one big blit + no overscan). It won't take long for that trick to hit a wall.

The greatest variation I think is in the convertors now - any room left in the future will be found there.

User avatar
FedePede04
Atari God
Atari God
Posts: 1211
Joined: Fri Feb 04, 2011 12:14 am
Location: Denmark
Contact:

Re: PhotoChrome v6

Postby FedePede04 » Fri Dec 06, 2013 12:23 pm

it is looking really good doug :thumbs:
Atari will rule the world, long after man has disappeared

sometime my English is a little weird, Google translate is my best friend :)

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Fri Dec 06, 2013 1:37 pm

FedePede04 wrote:it is looking really good doug :thumbs:


Thanks.

I think the best way to unlock the last few colours would be to combine experience from a few people - overscan, wakeup issues, conversion rules etc. One person can do it but it takes a lot more experimentation and time.

I can see exactly what needs done but TBH the trial-and-error process with machines, emulators and weird special cases is so far is quite painful.

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Fri Dec 06, 2013 3:50 pm

I did a quick calculation over lunch and I think now it may be possible to get 56 colours per line, using 400/408 pixel overscan on STE. So a few border pixels can be traded for 4 more colour changes.

Since it's just a back-of-envelope calculation it might not work in practice - but it looks achievable.

If there is time over the weekend I'll try using one of the newer stabilizer-free STE overscan routs (evil, Cyprian) as a basis to test timings at least. One of the benefits of doing that is that I may not have to adjust any overscan timings at all, which means less change of 'weird stuff' in wakeup modes and other machines. Modifying the STF timings and having to add a stabilizer to the end of the first line really troubled me.

I think I have corrected the problem I saw in the last build but overall it seems like a better idea to spend the time on the improved solution with multiple benefits.

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Fri Dec 06, 2013 7:10 pm

I tried just now to implement the same blitter pattern using a STE-specific overscan routine without the stabilizer toggle on every line.

e.g.

Code: Select all

  <toggle>
  nops:90
  <toggle>
  nops:26


This works ok for the first section (90 nops) but there isn't enough time in the second section to blit again, or reconfigure the blitter to blit again usefully. So new colours can only be loaded with the CPU in that short space.

I tried a modified version of this which Cyprian had documented e.g.

Code: Select all

  <toggle>
  nops:90
  <toggle>
  nops:30


This might be viable for blitting - but I could not get a sensible overscan pattern. The overscanned lines are separated by a non-constant linewidth (or they are just corrupt, I was unable to figure out exactly whats going on there). It probably needs more time to diagnose the actual pattern.

I then tried another variant using the CPU to load more colours at the end:

Code: Select all

  <toggle>
  nops:91
  <toggle P1>
  nops:5
  <toggle P2>
  nops:24


...but found there was no gain over the method I already tested - I could load a max of 10 colours near the right border, with a few nops left over - not enough for one more colour.

I did mange to pull 2 pixels worth of an existing colour change in from the right border but that's all. Any other benefits would just be overscan stability / compatibility issues from using a simpler and better exposed overscan pattern.

So currently I can only get 52 per line with overscan, and no more. Somebody else might find a way to improve on this but I'm stuck for now.

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 12:24 am

Didn't get anywhere new with improved overscan stuff, so I spent the last half our fiddling with stabilizing sequences for a quad-field image.

Here's an output image in quad-field mode. This renders from a palette of 226,981 colours - approaching 18-bit colour. It's using the same overscan mode from the last test but uses 4 fields and a different interlace sequence.

It's not as stable as the 3-field version but its close, and probably still worth playing with a bit more. I'll upload a test binary for this too but it's late now - so another time...
You do not have the required permissions to view the files attached to this post.

User avatar
Xerus
Moderator
Moderator
Posts: 1233
Joined: Fri Dec 13, 2002 9:31 pm
Location: France

Re: PhotoChrome v6

Postby Xerus » Sat Dec 07, 2013 12:42 am

I tried on my 520STE and CRT Monitor and sadly the sync is bad.

I will try on an other STE later !
You do not have the required permissions to view the files attached to this post.

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 10:06 am

This looks like one of the two problems I saw here on my own STE. It showed as a wakeup state - sometimes it happens persistently when the machine is cold, but not when it warms up and is re-powered on. I have been testing a fix but it will take a few days of 'cold starts' to see if this one is fixed for that overscan mode.

The other overscan mode doesn't seem to cause it, so I can migrate it if the fix is not effective.

[EDIT]

BTW the test executable uses a lot of extra space - I don't know if it definitely works on a 512k machine. I'll test it here to be sure.

User avatar
Xerus
Moderator
Moderator
Posts: 1233
Joined: Fri Dec 13, 2002 9:31 pm
Location: France

Re: PhotoChrome v6

Postby Xerus » Sat Dec 07, 2013 4:24 pm

It's not a problem of memory because my 520STE has 4MB ;)
Yes there is a difference when the computer is cold or 'hot', it's a "little" better at cold.
I'll dust off another STE for compare.

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 5:33 pm

Xerus wrote:It's not a problem of memory because my 520STE has 4MB ;)


Heh!

Xerus wrote:Yes there is a difference when the computer is cold or 'hot', it's a "little" better at cold.
I'll dust off another STE for compare.


Ok, great - thanks.

On this machine I get a correct image when the machine is hot, but usually an identical 'mess' from cold/overnight sleep.

I'm migrating the colourchange pattern to the second overscan technique and I think there's still perhaps a chance for further improvements but at least the wakeup thing should go away if nothing else.

User avatar
Xerus
Moderator
Moderator
Posts: 1233
Joined: Fri Dec 13, 2002 9:31 pm
Location: France

Re: PhotoChrome v6

Postby Xerus » Sat Dec 07, 2013 5:54 pm

Ok, new test with another STE and it's almost the same result.
I'm sure that the next version will better, courage :wink:

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 6:03 pm

Xerus wrote:Ok, new test with another STE and it's almost the same result.
I'm sure that the next version will better, courage :wink:


So you don't get the image at all, even when the machine is hot (either machine)? It requires a full power cycle for a few seconds - reset button won't do. If you can get an image when warmed up and power cycled, then the machines are all behaving the same way at least.

If it's not just a wakeup thing but a general incompatibility then I won't bother retesting the first technique and switch to the second.

User avatar
Xerus
Moderator
Moderator
Posts: 1233
Joined: Fri Dec 13, 2002 9:31 pm
Location: France

Re: PhotoChrome v6

Postby Xerus » Sat Dec 07, 2013 6:28 pm

I get the same result with or without reset, cold or hot, sorry.

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 6:36 pm

Xerus wrote:I get the same result with or without reset, cold or hot, sorry.


Ok thanks, that's helpful (even if it doesn't seem like it) :)

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 7:36 pm

Ok I quickly migrated the CC pattern into a second overscan configuration and I'm getting the same result in Hatari and my STE.

However the right edge is cropped in STeem and the timings appear to be wrong. Looks like this mode isn't emulated in STeem.

I'm mainly interested in the real HW at the moment so I'll upload a test for that tonight to see how it compares.

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 7:59 pm

Here's a test binary for overscan config #2. The purpose is just to test the left/right border combined with the palette timings.

It will look a bit ugly because the migration process isn't complete - the image is grainy with glitches/streaks and the low border isn't open yet. To be sure I attached a reference image - it should look like that.

The risk with this one I think is the top border but it's easy enough to adjust if not right. Hopefully the rest is ok. If the left/right border is broken there is some room in this config to change the timings (the previous one was more difficult).

If it works on both your machines I'll finish the process and re-convert the original image properly.
You do not have the required permissions to view the files attached to this post.

User avatar
Xerus
Moderator
Moderator
Posts: 1233
Joined: Fri Dec 13, 2002 9:31 pm
Location: France

Re: PhotoChrome v6

Postby Xerus » Sat Dec 07, 2013 9:03 pm

Here a capture (identical on the two ST) :
You do not have the required permissions to view the files attached to this post.

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 9:07 pm

Great, thanks for helping!

I should have a new, more complete test in a day or so.

User avatar
Xerus
Moderator
Moderator
Posts: 1233
Joined: Fri Dec 13, 2002 9:31 pm
Location: France

Re: PhotoChrome v6

Postby Xerus » Sat Dec 07, 2013 9:09 pm

:cheers:

User avatar
Steven Seagal
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2018
Joined: Sun Dec 04, 2005 9:12 am
Location: Undisclosed
Contact:

Re: PhotoChrome v6

Postby Steven Seagal » Sat Dec 07, 2013 10:00 pm

dml wrote:However the right edge is cropped in STeem and the timings appear to be wrong. Looks like this mode isn't emulated in STeem.


Yo
When you see a problem in Steem, could you please:
- specify the version
- also report in steem/development section if it's the last version (3.5.4 these days)
It would help me.
Also if it works on hardware please don't try to make it compatible with emulators: it's our job to make the emulators conform to the hardware.
In the CIA we learned that ST ruled
Steem SSE: http://sourceforge.net/projects/steemsse

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 10:52 pm

Steven Seagal wrote:Yo
When you see a problem in Steem, could you please:
- specify the version
- also report in steem/development section if it's the last version (3.5.4 these days)
It would help me.
Also if it works on hardware please don't try to make it compatible with emulators: it's our job to make the emulators conform to the hardware.


The version I am testing with is 3.4.1.

I won't try to do anything to make it work specially in STeem - but I may explore a few cycles +/- in case it suddenly registers as a mode which is both valid/emulated and very near the one I'm using.

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sat Dec 07, 2013 11:02 pm

So I completed migrating the rules for the new overscan routine, and implemented the bottom border removal.

In fact I thought the bottom border was a big problem because the second overscan routine steals cycles from the main blit block (90 nops -> 87) and the blit fits exactly in that block. Fortunately the leftover nops from the previous line were just enough to partly set the blitter up again and allow it to fit. This required 2 'special' lines instead of 1 for the bottom border and was a very tight fit.

However it did mean there are no special scanline rules now - the whole image can be converted with a single rule for 272 lines (unlike the previous version, which loaded 8 colours at the end of the line instead of 10 - the rule file had 3 separate sections: 227+1+44).

It's happy in Hatari and my STE with wakeup tests (so far) but unhappy in STeem. I'll look at that more closely another day.

Will post another test soon.

User avatar
Steven Seagal
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2018
Joined: Sun Dec 04, 2005 9:12 am
Location: Undisclosed
Contact:

Re: PhotoChrome v6

Postby Steven Seagal » Sun Dec 08, 2013 9:06 am

dml wrote:
The version I am testing with is 3.4.1.



I ask this because you won't get the same problems with v3.5.4.
In the CIA we learned that ST ruled
Steem SSE: http://sourceforge.net/projects/steemsse

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sun Dec 08, 2013 10:33 am

Steven Seagal wrote:I ask this because you won't get the same problems with v3.5.4.


I normally don't change versions of programming tools very often - changing code and tools together regularly can lead to some strange problems. So I tend to stick with a version of something much longer than most people would (it's not all lazyness on my part ;-) ).

However I'll upgrade and retest with this to see what happens. If it looks correct I'll need to stick with the newer version.

Thanks!

User avatar
dml
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 3474
Joined: Sat Jun 30, 2012 9:33 am

Re: PhotoChrome v6

Postby dml » Sun Dec 08, 2013 12:25 pm

I have attached timing diagrams for the old and new overscan + multipalette timing patterns. They are nearly identical, with the same colour change count & density.

The pattern is nearly uniform across the display, except for slight skewing on the right side near the border toggle.

old (unstable)

pcsv6timingd.png


new (fixed)

pcsv62timingd.png
You do not have the required permissions to view the files attached to this post.


Social Media

     

Return to “Coding”

Who is online

Users browsing this forum: No registered users and 5 guests