On Nov 14, 2006, at 5:58 AM, IOhannes m zmölnig wrote:
Modified Files: GemPixConvert.h GemPixConvertAltivec.cpp Log Message: disable YUV422_to_YV12_altivec() for gcc<4.0 via the NO_VECTORINT_TO_VECTORUNSIGNEDINT define (the fix yesterday coukdn't have worked); btw YUV422_to_YV12_altivec() is really _unused_ in the current sources!!
...um, yeh, unused in the current GEM sources, but used in the source for [pix_2pdp] ;-) Part of my attempt to make "roundtrip" color conversions fully altivec'ed...
jamie
james tittle wrote:
On Nov 14, 2006, at 5:58 AM, IOhannes m zmölnig wrote:
Modified Files: GemPixConvert.h GemPixConvertAltivec.cpp Log Message: disable YUV422_to_YV12_altivec() for gcc<4.0 via the NO_VECTORINT_TO_VECTORUNSIGNEDINT define (the fix yesterday coukdn't have worked); btw YUV422_to_YV12_altivec() is really _unused_ in the current sources!!
...um, yeh, unused in the current GEM sources, but used in the source for [pix_2pdp] ;-) Part of my attempt to make "roundtrip" color conversions fully altivec'ed...
ah i see, so we should either make gem2pdp respect the "NO_VECTORINT_TO_VECTORUNSIGNEDINT" or fix the wrong types in YUV422_to_YV12_altivec().
would casting help to solve the problem? proabably gcc4 does the casting from signed to unsigned implicitely. since no type conversion is used, the cast should be fast as light (but i might be very wrong here...)
mfga.sdr IOhannes
jamie _______________________________________________ GEM-dev mailing list GEM-dev@iem.at http://lists.puredata.info/listinfo/gem-dev
-
IOhannes m zmoelnig -
james tittle