This is really shaped up into an awesome shader, great supplement to the XBR shader set.
However…
The newest version of this (i.e. version 2.6 and version 2.6a do not work)
Version 2.5 (the new 5 pass) works great. Tried with retroarch PS3 and is much faster than the previous version.
However, 2.6 and 2.6a don’t load. I tried both the links on this forum provided by Spookyfox, and the github links. Those files are corrupt or not working in some way. The current version of PS3 retroarch, unfortunately, has the ‘non-working’ version of this great shader (i.e. 2.6a).
the 7zip archive in the first post is ok, you just have to use the recent 7zip version to extract it correctly. but those are the same files as the ones on the common shader repository. so it doesn’t matter where you get them from.
and your issue is probably a platform specific one. since I don’t have a PS3 I’d need at least an error log to do anything about it. I can only tell you that mdapt v2.6a is not broken in general. I just tested it again with the newest retroarch version for windows and it works just fine. I also know for a fact that it works for other users as well on pc.
Wish I knew how to generate an error log. :-/ I’m a noob/dummy when it comes to these things.
I will try one other thing, to see if I can get it to work. But I can say for sure, at least for ps3 users, that since the change from 2.5 to 2.6 and beyond the shaders no longer work.
I will try one other thing just to make sure it’s not something on my end but it appears to be a change in the shaders itself, unless the ones you posted here are corrupt in some way. Let you know what I find.
Well, I tried everything in the book. Still doesn’t work on PS3. :-/
Anyone here using PS3 retroarch know how to run an error log for Spookyfox? Don’t know how to do that.
Basically, the newest version of your mdapt filter doesn’t work on PS3. The most recent version that does is 2.5. And it works very well, especially with the lvl 3 filters. The slow down is negligible and accompanies that xBR filter well.
Of course I’m still excited about the optimizations in 2.6 and 2.6a since any optimizations will make these multipasses work faster so Im hoping they can be fixed to work on PS3.
I understand that you are saying they work elsewhere in Windows. Whatever changed from version 2.5 to 2.6 is now broken with PS3 if we rule out the integrity of the files themself (which you are saying are good on here and at github when used with other systems?). Wish I knew how to generate an error log. It’s not a question of buggy sideeffects or anything like that.
They basically don’t work at all. Even the preset that comes with 2.6 and 2.6a refuses to apply the filter(s) to the game’s texture. When selected, the preset kicks out and nothing shows up in the passes section. If I apply the passes individually, they are there now but do nothing.
Guess if someone here knows something about PS3 Retroarch and wants to take a crack at it, it’s all you since I don’t know how to code and such.
For the record, these are some of the best filters around. Without them, many games (especially genesis games and older arcade games) look really dated. This combined with xBR creates a true HD finish that breathes new life into older retro games. They are a must have and a miracle of ingeniuity. So clearly I’m biased.
In any case, thanks boys for looking into it. Hoping ‘future’ revisions of this filter will be PS3 compatible. In the meanwhile, while limited to version 2.5, I’m still very satisfied and recommend everyone use it for their games!
nice to hear that you like it so much. but I didn’t introduced new optimizations with 2.6(a). there is only a small change in the algorithm behavior and the image output. if anything then 2.6a is slightly slower than 2.5. maybe other shader developers see further ways to speed it up but I don’t know how. I already made huge optimizations in contrast to the early versions of mdapt. so if it doesn’t run full speed on PS3 try to combine it with a less demanding xbr variant (my personal favorite is 2xBR-Hybrid-v5-gamma.cg).
regarding the error log, check out the thread for the ps3 version of retroarch. I bet you can get help there:
RetroArch PS3 releases
ps: I see your posts in my thread. no need for sending them as pms to me as well
Hey sweetie, is the same true that 2.6 is a bit slower than 2.5? If so, problem solved on my end at least. Would rather use the quicker of the two. 2.5 more than gets the job done and looks great as far as I can see. I remember when this first came out it was kind of glitchy. But now it’s practically flawless!
But I was still curious: which is faster, 2.5 or 2.6? (leaving out version 2.6a and going back to regular 2.6)
according to my update 12/06/13 log (which was 2.6) I introduced a new color metric with this version, which is slower than the standard euclidean distance for sure. so yeah 2.5 is faster than 2.6(a), but the speed difference should be negligible anyway.
For us PS3 users using the antiquated hardware we are limited to, ANY speed bump (even small) is a reason to celebrate given the strain the multi pass shaders put on game performance. haha.
So, problem solved. v2.5 is the best one for us PS3 users looking to maintain the best possible, speed/performance.
In any case, thanks for your hard work. You rock babe.
yeah, I don’t really know what to do with these specific hardware problems. I hope some of the more technical shader devs will figure it out in time.
anyways, I updated mdapt with an idea which crossed my mind for a while. since in pass2 the shader counts the dithering pixels outgoing from the one in question, the pixels which are at the border of dithering areas are often ignored. I implemented some rules to add them again afterwards. the effect is subtle so I won’t post new screenshots.
I was helping out over at the brazilian outerspace forum with the shader usage. Since the Sega Mega Drive / Genesis is pretty popular there I saw the need of a more aggressive treatment, cuz many games for this console are full of dithering and mdapt acts too cautiously. So I went back to the old 2pass-version and mixed it with new algorithm techniques of the recent one. The result is “gdapt”:
Thank you for this! Just what I was looking for. Can’t wait to try it out.
Question: any practical differences from your 5 pass? Is this tailored only FOR Sega genesis? Is this faster (better performance) than the 5 pass? Would I lose something if I used this in place with OTHER emulators?
In other words, is this essentially a more aggressive version of your 5 pass shaders in a two pass form? Any reason I shouldn’t use it in place of the 5 pass with other games?
you can use it for any game with dithering, specific games for the genesis merely provided the idea for that. the difference is that gdapt only uses 2 passes instead of 5 (therefore it’s faster), the dithering treatment is more aggressive but it also makes more errors and it only knows of horizontal dither blending where mdapt handles checkerboard patterns in another special way.
and nope, gdapt is more like going back to the past in the version history. you could say that it’s an enhanced variant of the old mdapt v1.2 with the new mathematical approaches of v2.6.
updated mdapt to v2.7. there was some major issue with the sigmoid function in pass2. I kinda randomly noticed that with very big absolute values my approximation of tanh produces wrong results. funny enough it also happens with the official implementation. I fixed it by clamping the input. apart from that there are some minor improvements. the full changelog is in the first post.
alongside a readme you can find my favorite cgps I like to use on the shader repo.
did you change any parameters and are you using the cgp file? try using opengl and see if it works.
if this doesn’t help, plz post a screenshot and the name of the game.
No, I didn’t change anything, I’ using Win7x64 with latest drivers, and the effect is indistinctly of system or game. Always using Gl, but also tried d3d, there are other shaders that don’t work like the motion blur shaders etc. But if you tell me this shader is supported then something must be wrong on shader or maybe nvidia settings (haven’t touched those though). For example using mdapt+xbr-hybrid+aa.cgp I get this on Sonic Advance:
seems to be a hardware specific thing again. hard to tell what exactily is causing this. is gdapt giving you the same problem? you could try to find out which pass is the one with the issue by decreasing the pass number in the shader options. here is an example how every pass is supposed to look like:
