On Mar 18, 2005, at 6:57 AM, IOhannes m zmoelnig wrote:
james tittle wrote:
could you please test the attached file ?
...sure! I'll be working on this tonight after work: I didn't get
so did they work then ? i'd rather have these changes committed to the cvs, as they at least have the clamping and are more (i say: "more" :-)) standardized than the last code.
...just tested 'em, and they seem much better so far...maybe not faster in osx(tho I was testing on a 720x480), but then that'll come with the altivec...I was pretty astonished to see that the old rgba conversions actually lightened the picture, and I could only chain together 2 pix_rgba/pix_yuv's before noticeable color screw-ups occurred!
...I'm now migrating the altivec stuff that I've done so far into your new GemPixUtil.cpp that ya sent, so if you wanna commit it, go ahead, or just wait and I can do it when I get this stuff merged...
james
james tittle wrote:
On Mar 18, 2005, at 6:57 AM, IOhannes m zmoelnig wrote:
james tittle wrote:
...just tested 'em, and they seem much better so far...maybe not faster
good to hear
in osx(tho I was testing on a 720x480), but then that'll come with the
well, i didn't promise that they would be faster; i guess they even should be slower due to the clamping; the rest should be pretty much the same (in terms of cycles needed)
altivec...I was pretty astonished to see that the old rgba conversions
we're looking forward to that...
actually lightened the picture, and I could only chain together 2 pix_rgba/pix_yuv's before noticeable color screw-ups occurred!
so what is happening now ? (i mean, i get this greenish thing after >10 to-fro conversions)
...I'm now migrating the altivec stuff that I've done so far into your new GemPixUtil.cpp that ya sent, so if you wanna commit it, go ahead, or just wait and I can do it when I get this stuff merged...
i have submitted it to get it finally in (people kept complaining about some images looking so weird...) altivec (and mmx) (and sse2) will be extra candies, while that change was principially a bug fix.
mfg.a.sdr IOhannes
-
IOhannes m zmoelnig -
james tittle