It's implicitly on for x86_64 (all x86_64 chips support a minimum of SSE2), unless you use USE_SIMD=0 to force it to avoid that code path and fall back on scalar code.
And that's why I so often don't notice flaws in the fallback -- I only use 64 bit machines. The problem is mostly 32 bit machines, which will only use SSE paths if you specifically tell it to. (Though in this one case I made a bigger mistake -- I did something that would require SSE 4.2, and didn't have proper fallback for SSE2.) I'll be better about this. Before tagged releases, I'll do builds with USE_SSE=0 and a USE_SSE=sse2, in addition to my usual defaults, to confirm that all configurations compile cleanly. On Mar 12, 2015, at 9:07 AM, Richard Shaw <[email protected]> wrote: > On Thu, Mar 12, 2015 at 8:03 AM, Larry Gritz <[email protected]> wrote: > Pushed this fix (and one other, though it was harmless), tested (with > USE_SIMD=0, as I should have originally), and re-tagged. Sorry for the > trouble. > > Thanks for the quick response! > > So USE_SIMD defaults to off? Is there any detection mechanism to turn it on > automatically or can I assume it's OK to enable for certain platforms, i.e. > x86_64? > > Thanks, > Richard > _______________________________________________ > Oiio-dev mailing list > [email protected] > http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org -- Larry Gritz [email protected]
_______________________________________________ Oiio-dev mailing list [email protected] http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
