boostorg / boostorg/math

arm64_v8a: ibeta_inv exceeds 200 iterations for some arguments

Open
#442 8 comments 0 reactions 0 assignees View on GitHub
Dominant language
C++
Stars
374
Forks
264
Avg merge
1d 23h
Merged PRs (30d)
17

Description

Hi, when compiling for 64-bit ARM (arm64_v8a), ibeta_inv(alpha, beta, x) produces

boost::exception_detail::clone_impl >: Error in function boost::math::ibeta(long double, long double, long double): Root finding evaluation exceeded 200 iterations, giving up now.

when either alpha or beta are smaller than about 3e-3. For other target processors I've tried (arm_v7a, x86, x86_64) this works fine.

I've now hacked up my own custom root finder for these cases which converges in just a few iterations. My code's certainly not up to boost standards, but I'm happy to share it as a starting point if someone wants to use it to improve the ibeta_inv root finder in this regard.

Though I suspect that something else is going on here - the bug may be indicative of some deeper numerical issues on arm64_v8a?

Contributor guide

No contributing guide indexed for this repository

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.