![]() | |||||||||||||
Messages in this thread |
On Tue, 07 Mar 2006, Lee Revell wrote: > If they are doing serious realtime DSP then they should get better > results in userspace anyway, because they get to use the floating point > unit which isn't allowed in the kernel. It's not as though every algorithm needed float just because it said DSP (some of those are actually fixed-point or something like that) at a time. There are lots of algorithms to avoid exactly that, because it costs performance big time. Whenever you can have integer and/or reduce/approximate multiplication by shift+add, people will use it if performance is of paramount importance. And such often is the difference between having a real-time DRM software radio or not, to name just one example I've seen in my vicinity. In a color-space conversion tool (CCITT YUV to RGB) on P-II or P-III (don't recall) emulating fixed point by using integers and then shifting appropriately gave a speedup of well more than three. There must surely be better and reasons than the FPU - licenses had already been mentioned. -- Matthias Andree - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ | ||||||||||||
| Last update: 2006-03-07 23:39 [from the cache] ©2003-2008 | |||||||||||||