ZZogl -- Zero GS KOSMOS fork

[Grimspiller]

I only tested FF12,and it's good.You have improved this plugin a lot.I remember when it ran 5 FPS

[wingnux]

ZZogl is almost fast as gsdx,I tried it in some games.Have you had a chance to compare results?

Not at all! I was able to finish GoW1 and Mortal Kombat: Deadly Alliance on windows but both are almost unplayable using zzogl =/ FF12 is still pretty slow and have lots of glitches too.

The bright side is that the development of zzogl is going pretty strong so I trust it won't be long until it's as fast as gsdx (dx10/11).

[fares1131]

plaese send zzogl 0.19.187 to me

[laharl2k]

ZZogl is almost fast as gsdx,I tried it in some games.Have you had a chance to compare results?

Not at all! I was able to finish GoW1 and Mortal Kombat: Deadly Alliance on windows but both are almost unplayable using zzogl =/ FF12 is still pretty slow and have lots of glitches too.
...

maybe it's because of the graphic drivers. At leats with fglrx it's much slower on linux than on windows. And if u use the OSS drivers it's even more.
I've no experience with nvidia, so no idea in your case, but be sure to check that it's not a driver side problem.

[wingnux]

Unfortunately it's not a driver issue since I have the latest nvidia binary drivers installed =/

[Grimspiller]

I use OS:windows xp and ATI drivers.I said that BIG progress is made since last year.

[gregory]

In my humble opinion there is clearly a driver/opengl impact. To sell gaming card, vendors need to have the best directx driver. I thinks professionals (aka opengl users) use some different code path.

Just for information the target of the amd OSS driver is 50-60% of the proprietary version so do not expect too much.

Edit I have a question on texture parameter.
S & T are rounded down in GIFPackedRegHandlerUV. But Q seem not to be rounded, why ?

[gregory]

I give a quick glance at the new local local transfer code. I think it would be possible to do block copy to improve local -> local transfers.

As I understand, the src & dst buffer are aligned on block boundaries. So when color format have same block layout, a batch copy is possible (maybe need some mask for 24,8h,4hh,4hl).

I read that pixel depth must be identical between src & dst. So the special case conversion are only 8 <-> 8H and 4 <-> 4HL <-> 4HH, maybe it is possible to do somethings with sse2.

Any comment ?

[Zeydlitz]

Of course, you right. Look at my code. Whole transfer code could and should be redone,

[aeridus]

How do you compile this code in Windows? I tried using the vcproj file to compile it in Microsoft Visual Studio 2008 it came up with a ton of errors. I put the GL header files and zlib files in the appropriate places...