Luca Barbato <lu_z...@gentoo.org> writes: > On 11/9/12 6:13 PM, Mike Melanson wrote: >> When Libav performs CPU feature detection (i.e., does this chip have VFP >> or NEON?) on ARM, the options are determined at compile time. This is in >> contrast to x86 where it performs feature detection at run time. >> >> I just wanted to learn about the reasoning behind this. Is it not >> possible to do run time feature detection on ARM? Or was it reasoned >> that run time feature detection isn't as useful for a primarily embedded >> platform (where software will be more static)? > > IIRC there isn't a safe way to detect features that isn't system > specific (e.g. parsing the elf aux header/vector) or dangerous on > libraries (sigill handler tests).
That is part of the reason. The other part is that it is beneficial to build _all_ the code for the specific target, not just the functions that are selected at runtime. -- Måns Rullgård m...@mansr.com _______________________________________________ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel