tech-userlevel archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: float128 in libstdc++



On 25.06.2017 12:30, Martin Husemann wrote:
> On Sat, Jun 24, 2017 at 04:03:49PM +0200, Kamil Rytarowski wrote:
>> arch/sparc64/c++config.h:/* #undef _GLIBCXX_USE_FLOAT128 */
> 
> Something is definitively strange here (as Joerg noted). We use 128 bit
> long doubles via softfloat on sparc64. We should actually offer the
> hardfloat version too and fix all fallout from the emulator (if there
> are any left), but last tests showed this to be slower.
> 

__float128 is not listed for sparc64 hardware in the GCC documentation:
https://gcc.gnu.org/onlinedocs/gcc/Floating-Types.html

>> SYSV ABI AMD64 notes float128 optionally and software emulated for at
>> least order hardware.
> 
> What amd64 cpus support it in hardware (or which of the strange feature
> bit acronyms matches it)? Of course we can not make the library features
> depend on runtime cpu support detection.
> 

Right now I don't know any x86_64 cpus that can do __float128 in
hardware, at least few years ago there were none.

> Martin
> 



Attachment: signature.asc
Description: OpenPGP digital signature



Home | Main Index | Thread Index | Old Index