super4xsai?

bantza · Post by **bantza** » Mon Nov 21, 2005 7:44 pm

I was just speculating... Would it be possible to include larger scaling on 320x240 games than 2x. Like applying for example super2xsai twice (if it doesn't slow down the machine too much) to get 1280x960 resolution.

Post by **sev** » Mon Nov 21, 2005 8:16 pm

No, we decided not to support any 4x scalers. Maybe that will get change in the future, who knows.

Eugene

JamesWoodcock · Post by **JamesWoodcock** » Tue Nov 22, 2005 9:28 am

For LCD displays, having a 4x scaler would be great, going full screen causes the image to lose quality.

Dark-Star · Post by **Dark-Star** » Tue Nov 22, 2005 3:30 pm

ScummVM could support a perfect aspect ratio correction with a 5x/6x scaler which scales up to 1600x1200 (for really high-res LCDs or 19'' CRTs):

a 320x200 game scaled x5 in x-direction and x6 in y-direction gives 1600x1200, without any scaling artifacts.

I'm thinking about getting such an LCD so such a scaler would be ideal

--Darkstar

GUN · Post by **GUN** » Tue Nov 22, 2005 6:56 pm

Dark-Star wrote:ScummVM could support a perfect aspect ratio correction with a 5x/6x scaler which scales up to 1600x1200 (for really high-res LCDs or 19'' CRTs):

a 320x200 game scaled x5 in x-direction and x6 in y-direction gives 1600x1200, without any scaling artifacts.

I'm thinking about getting such an LCD so such a scaler would be ideal

--Darkstar

butg the needed CPU Powerr is slightly much higher :S

Dark-Star · Post by **Dark-Star** » Thu Nov 24, 2005 12:06 am

Well I don't think that an integer x5/x6 scaler would require more CPU power than one of the more advanced 2x scalers. I mean it's just a matter of duplicating pixels (horizontally) or even whole scanlines (vertically), no fancy math involved.

However I didn't benchmark it so I can't be 100% sure. Feel free to prove me wrong

--Darkstar

LogicDeLuxe · Post by **LogicDeLuxe** » Thu Nov 24, 2005 1:21 am

Dark-Star wrote:I mean it's just a matter of duplicating pixels (horizontally) or even whole scanlines (vertically), no fancy math involved.

But then, it wouldn't be a super4xsai nor super5/6xsai or whatever. If a non enhancing scaler is what you want, a nearest neighbour resize code would make sense here. With that the window could be resized to any dimension, including 5x/6x full screen.

Dark-Star · Post by **Dark-Star** » Thu Nov 24, 2005 9:08 am

gah, of course you're right. First I was thinking only about point upscaling (which, combined with some scanlines, would make an almost perfect aspect-ratio-corrected display on an 1600x1200 CRT) but if you consider SuperSAI then of course the CPU requirements would be much higher.

However, pointsampling would probably be enough at this resolution...

--Darkstar

zorbid · Post by **zorbid** » Sun Nov 27, 2005 6:46 pm

If your graphic card supports pixel shaders (v2.0, for sure, maybe even 1.4), you can try Moe's OpenGL HQ SDL build, that implements an hardware accellerated HQ-any-X scaler. The scale factor can be any number, integer or not, and the horizontal and vertical scale factors aren't bound.

LogicDeLuxe · Post by **LogicDeLuxe** » Sun Nov 27, 2005 8:58 pm

zorbid wrote:If your graphic card supports pixel shaders (v2.0, for sure, maybe even 1.4)

v2.0 what?

Jimbob · Post by **Jimbob** » Sun Nov 27, 2005 9:04 pm

LogicDeLuxe wrote:
zorbid wrote:If your graphic card supports pixel shaders (v2.0, for sure, maybe even 1.4)
v2.0 what?

OpenGL or pixel shaders on the gfx card, I assume.

Jimbob

zorbid · Post by **zorbid** » Sun Nov 27, 2005 9:53 pm

Yep, Pixel Shaders v2.0 or 1.4 (I thought it was obvious... didn't think of OpenGL versions.)