This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH v3 2/9] Add new log implementation
- From: Joseph Myers <joseph at codesourcery dot com>
- To: Szabolcs Nagy <szabolcs dot nagy at arm dot com>
- Cc: GNU C Library <libc-alpha at sourceware dot org>, <nd at arm dot com>
- Date: Mon, 2 Jul 2018 21:54:59 +0000
- Subject: Re: [PATCH v3 2/9] Add new log implementation
- References: <7a77bac7-7e68-aa63-55c5-c11ca47a8692@arm.com>
Similar comments apply as to the exp patch about there being various cases
in the code for e.g. LOG_POLY_ORDER with no obvious reason why it would be
appropriate for some architectures to use different values from others, or
how the different values have been tested. If such values can sensibly be
changed for some architectures I'd expect detailed comments in
math_config.h discussing possible valid values and why different values
might be appropriate on some architectures, and information in the patch
submission about how the different cases were tested. Otherwise, I'd
expect to have at most #errors to document where the code depends on the
value of such a macro.
Similarly for log2 to a lesser extent (it looks like the code there
doesn't actually handle the other cases in general, but the patch
submission should still discuss e.g. how both fma and non-fma cases were
tested). Likewise for pow.
--
Joseph S. Myers
joseph@codesourcery.com