Naoki Shibata

Is there any documentation on how to do it?

If it is possible to improve the computation speed in that way, someone would have already done it. You know, scalar computation of trigonometric functions is an older problem than...

I am now considering adding FMA support to aarch32. AArch32 processors with vfpv4 support have FMA. Such processors include cortex-a5, a7, a15 and a17. So, this will benefit almost all...

Yes, that would be very nice. Is bumpversion.sh the following one? https://gist.github.com/pete-otaqui/4188238

For most of them, it is basically possible. Because it involves randomized optimization, the generated coefficients are not bitwise identical.

Please tell me which functions you want to regenerate coefficients for, if you have difficulty in doing so.

The settings are included in dp.h and sp.h.

For some functions, you need multiple trials.

Yes, it’s possible. I didn’t do that because I didn’t see a demand by users for generating bitwise identical results.

erfc in sleef is slow. I didn't think that people would care about erf or gamma.