To show a bit how deblur is working (lo res source):
3 Likes
New release version (13.12.2020).
Notable changes:
- Deblur final color clamping much improved, now it has dynamic range and allows stronger applications of deblur, or weaker.
- Some speedup, code cleaning
- 2 pass version has joint parameters now for easier use. If you change something with a parameter in one shader with an text editor, the other shader must be corrected too in an exact manner or it won’t compile.
Download link:
https://mega.nz/file/Jt5RCaLJ#W7G-d2pHMS760peIId0XTm5D91gHtrHUJzE-dEYEuqY
Example of ‘extreme’ shader usage:
Didn’t expect such funny image. 
3 Likes
BTW, That deblur.fx worked on PS3 full speed. Thanks!
2 Likes
Meanwhile i took some time to think about the new shaders/options and Deblur had receive a decent upgrade for new looks.
Will post the ‘shader pack’ soon.
Not to forget, many tweaking options… 
4 Likes
The evolution is very apparent! It’s looking great!
2 Likes
It was a bit against my expectations, but deblur works better with plain gaussian blur compared with the resampler. Was a bit stubborn first, but after giving it a thought deblur increases contrast on it’s own and since gaussian blur decreases contrast it’s a nicer combo.
New release version (15.12.2020).
Notable changes:
- some relevant presets added
- deblur got many nice tweaking options
- hosted on github now
Download link:
3 Likes
Do you know a way or app to profile these slang shaders?
I only profile cg shaders using nvidia nvshaderperf. It work very well.
About deblur, have you thought/tried to deblur just using lum? If it works, it should speed things up.
1 Like
The shaders itself have dynamic range, quite hard to profile, but i like to use the Imageviewer core in Retroarch with a test image. It’s a bit more realistic as RL performance is measured. I also used GPU ShaderAnalyzer, but it didn’t distinct between normal and special functions.
Color calcs gave me decent results with deblur, although i had a luma-deblur running quite a while ago. I think it’s comes into consideration for a performance version indeed.
Otherwise the new deblur shader is quite fast, if you use small loops. It could be optimized, as well as presets, but i try to avoid too much version entropy.
2 Likes
I got access to the shader repos again, so I can finally merge in you guys’ stuff, I just haven’t gotten around to actually doing anything with it yet. I’ll merge in the outstanding PRs in the next day or two and then we can see about getting you and @Hyllian’s stuff up to the latest versions.
7 Likes
Thanks @hunterk. The crt-guest-dr-venom2 is finished i guess, also considering the feedback.
I also want to calm the spirits a bit with the new shader presets posted in this topic. 
I posted more ‘extreme’ smoothing setups, but it goes the other way also, regarding scalefx and in general.
4 Likes
Been working on a very sharp ScaleFX upscaler recently and i’m also surprised with the results. It can upscale to any resolution quite well, but did only try 4k in virtual mode above 1080p. It’s also capable of some AA without extra pass, which can be nice.
Will post it on Github soon.
3 Likes
Looking good. Are you using deblur and some aa to get this? (The vanilla scalefx is noblend)
On a side note, xbr uses analytic line functions to be resolution independent. The only drawback is that it gets too complex to create a lvl6 shader as scalefx. Scalefx is fixed at 3x res, though it gives simplicity to achieve lvl6. Anyway, it gives me some idea I could work on xbr-noblend to achieve lvl6 by fixing resolution at 3x too…
2 Likes
Yes, i modified deblur somewhat, but didn’t expect consistent results. I guess it’s an experimental method, but is working quite ok. 
It’s on github now. For smother results last pass can do some over scaling also.
2 Likes
Hey mate , thanks for your efforts.
the preset you just posted fail to load for some reason , I put them in the appropriate folder and what not , would you please upload the recompiled version on any hosting site? thanks
These might be precision issues related to high deblur or sharpness parameters. I pushed them a bit regarding the range, but that’s the limits of my adapter, other might have different ceils.
Could you please report the parameter range in which it’s still without glitches? Also changed the file on github…might work better now.
I only used the default parameters values. That being said ,I have tried multiple adjusts of the parameters but even on lowest values, the glitch persists
1 Like
Should be fixed now. The bad was my adapter/drivers compensate too many exceptions in by covering it up.
Found another “invisible” bug, added two more shader modes, for smoother or sharp definition.
1 Like
can confirm , it’s working now. level of detail is kinda on low side tho espcially compared to NEDI3 , but I do like the crisp look.
1 Like
I’m continuing my work on the same topic (resamplers), and i guess these are good results for now with downsampling too…
A custom width custom lanczos kernel is used this time.
Example:
ordinary filter size:
Increased filter size:
Need to do some stuff, i guess i’ll release the shaders soon.
7 Likes