Windows 64bit XBMC

Do you plan to develop 64bit XBMC for windows?

You probably should provide some reasons as to why...
One could think about using >2GB RAM, but well - is that necessary for XBMC? Never had it hit around 1GB while running, so... what is your reason for wanting a x64 version?

And don't bring the "speedup" reason. There is no speedup (tested on OSX where we have 64bit builds).

64 is a higher number than 32, so therefore it must be better.

Bigger is always better, at least until it is too big.

You have speedup only if you optimize your code to use the extra standard/SSE registers.
It's possible almost only on ASM code, and if the algorithm can realy take a great advantage of using more registers.
On some video processing filters/jobs, you can sometimes achieve a speedup of around 40%. That's what i have with x264, when i'm using it to encode video, x64 version is 40% faster than x86 version. If it's 'simple' C code you compile with x86 or x64 compiler, you'll probably not see noticeable speedup.

Come on...
Encoding is of course one of the things that benefits from the x64 - because this is one of the tasks that highly benefits from calling more than 2GB.
Decoding - and this is what XBMC is about, since XBMC does not encode video - is not. You will have no benefits from using 64bit in this case, even if you use assembler.

Also - your statement makes not much sense to begin with, because we want the GPU to decode our videos. It is built for that, isn't it?

GPU decode only some specific format, corresponding often to more than 90% of the video format you'll find in non anime world. Enter anime world, and you're lucky if GPU decode 50% of your files, without even talking of 10bits h264. I personnaly don't use GPU for video decoding (dxva is disabled), but in case of GPU decoding, it's true 64bits will do nothing. In case of CPU decoding, i've never coded codec, so, i don't know how much you can speed up algorithm using the 64bits specific features. Another information so : The resizes feature (spline36 at least i think) in avisynth were speed up to 30% on x64 version against x86 version (optimizing code using extra registers avaibles).
And my encodes uses less than 2GB, so gain is not here, but algorithms greatly benefits of the x64 specific features.
So, it's possible that optimizing for x64 may speed up some features, impossible to realy tell wich ones and how much, and the so the huge time/work necessary will probably be not worth it.

Well - okay. Never thought about anime
Expanding code to the 64bit registers may make sense if it's about specialized features, there is with no doubt some stuff which _will_ benefit from using x64.

Yeah, you said it - it _may_ speed up some features and there is no way to tell which features at this moment, so spending lots and lots of time on coding a x64-version probably really isn't worth it.

From the above discusions I have learned that GPU decoding benefits nothing from X64. Thanks