Page MenuHomePhabricator

native-arch-optimization not working correctly?
Open, Incoming QueuePublic


If I understand it right, then the native-arch-optimization feature is about runtime detection of various CPU features and using them only when available. For example NEON on ARM. But the current buildsystem injects -mfpu=neon into the global compiler flags (using add_global_arguments()), so all files are built that way and not only the various optimized *_neon.c ones. On the other hand SSE3 is handled via native_arch_opt_c_args, which looks correct to me. Altivec on PPC is handled simillarly to NEON, enabled for all source files. Do I misunderstand the native-arch-optimization feature or is there a bug? IIRC it worked correctly with the autoconf based build system.

sharkcz created this task.Oct 17 2019, 2:47 AM
ProhtMeyhet added a project: Restricted Project.Nov 16 2019, 6:42 PM
ProhtMeyhet added subscribers: bu5hm4n, ProhtMeyhet.

I assume you are talking about efl?