On 10/21/2014 03:32 PM, Vesa wrote: >> The "Disabling denormal floats at the code level" is specially useful >> for LMMS. >> >> AFAIK the usual procedure to avoid denormals is as follows: >> >> On the engine code, include the headers needed for _mm_getcsr/_mm_setcsr >> >> #include <xmmintrin.h> >> >> >> When the engine thread is initializing call this: (on all the threads >> that do audio-processing) >> >> _mm_setcsr( _mm_getcsr() | 0x8040 ); >> >> >> The plugins and LMMS engine/audio code will need to be built with >> fastmath and SSE2, the gcc arguments are: >> >> -ffast-math -mtune=generic -msse -msse2 -mfpmath=sse >> >> >> Note that this requires SSE2. >> But if LMMS does this, the issue with denormals will very likely go away. >> >> >> PS: I tried building the full LMMS code with "-ffast-math", but that >> breaks the timeline. >> So if this is going to be introduced it will need to be carefully >> selected (only audio engine and plugins). > Thanks for the info. > > We've considered denormal-stripping and SSE2 before, but nothing came of > those discussions - IIRC, it was thought that there wasn't enough > benefit in doing so... but maybe that could be re-evaluated now, if it > could prevent issues with fx plugins.
Maybe we can evaluate this again... if we can prevent problems with DSP and fix long-standing issues, isn't that enough of a benefit to consider this? Is anyone really running LMMS on pre-SSE2 hardware anymore? ------------------------------------------------------------------------------ _______________________________________________ LMMS-devel mailing list LMMS-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lmms-devel
