On 31/10/2024 08:23, Alexandre Oliva wrote:
On Oct 25, 2024, "Andre Vieira (lists)" <andre.simoesdiasvie...@arm.com> wrote:
I have to admit I am not super familiar with long doubles, either than
knowing they are 128-bit FP representations... but bisect has pointed
me to this patch when investigating a regression on
aarch64_be-none-elf for the libstdc++ testcase:
26_numerics/complex/13450.cc
Can you just confirm this is 'wrong' and that the problem is likely to
be with the big-endian implementation of sqrtl in I'm guessing newlib?
Yeah, this looks like a bug in newlib. sqrtl is always defined there,
it manipulates long doubles' internal representation, but unlike double
and float, its ieeefp.h doesn't distinguish between big and little
endian. That should be easy to fix by defining another version of
struct ieee_ext on big endian systems.
Thanks, reported it in https://sourceware.org/bugzilla/show_bug.cgi?id=32326
