Tim Blechmann
Tim Blechmann
> Locking/unlocking is really just a single atomic exchange or compare-and-swap, so I am pretty optimistic that the performance overhead would be negligible. But yes, we should still measure it!...
> We would definitely need to benchmark. the difference may be more profound on some platform than on others. from my gut feeling i'd expect a difference when running on...
could you try to changes this? ``` static inline float32x4_t gen_sign_mask(void) { static const int sign_mask = 0x80000000; float * casted = (float*)(&sign_mask); return vdupq_n_f32(*casted); } ``` to ``` static...
could you create a PR? i cannot test it, but if you're running on a ppc hardware, it might be a good idea to execute the unit tests
i'm not sure if the ppc/altivec code was ever used in production, i merely wrote the altivec version as a proof of concept with a gcc of 10 years ago...
nova-simd doesn't have a stable API, so sc3-plugins and supercollider should use the version that is bundled (or referenced as git submodule). i know that i'm disagreeing with the debian...
they might use the same version and API changes might be rare ... but the point is: i don't want to give any guarantees on it, that's why i'd suggest...
actually, i wonder if all of `static_vector` could potentially be `constexpr`
any update on this one?
* only a single system call is required (`open` or `create_file` take care of the rest), so it should be slightly faster * it reduces the risk of race conditions...