I should clarify that the purpose for the patch was that the interface, as I originally implemented it in libjpeg/SIMD, did not do anything useful. It was always choosing the SSE2 code path. This patch was to make the interface work the way it was intended.
It would be nice to have a way to do this at a programmatic level without exposing the environment variables, but I don't know of a way to do that. DRC wrote: > That's exactly the point. This is an undocumented interface used for > forcing the MMX and SSE code paths on an SSE2-capable system so I can > verify the performance and stability of those code paths. It is not > ever going to be invoked by an end user. > > This interface was implemented as a series of environment variables > because we didn't want to add functions to the libjpeg API, and there is > no extensible configuration feature in that API that we could use to > pass a "force MMX" or "force SSE" flag. > > Pierre Ossman wrote: >> I'm not sure I like this. Although it does take the word "force" more >> literally here, the previous code was safe from foot shooting. With >> this you can enable e.g. SSE2 on a cpu that doesn't support it and end >> up with SIGILL. >> >> Rgds ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ Tigervnc-devel mailing list Tigervnc-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tigervnc-devel
