[J3] floating-point status

Bill Long longb at cray.com
Mon Oct 22 09:32:56 EDT 2018


Thanks, Bob.  That’s reassuring. 

But, focusing more closely, a typical image runs on a processor core or node that has many threads.  How does a programmer deal with flags and traps that occur on a specific thread?  GPUs are a notable instance of this issue, but even hyperthreads or different cores on the same chip are problems. 

Does the IEEE spec take into account threads? Or does it assume that every entity in the processor that can perform floating point operations has its own set of flags and trap modes (and rounding modes)?  

I see a possible correspondence with POSIX IO that was never designed to deal with parallel I/O systems.  As a result, that standard is essentially dead in the sense of being irrelevant and ignored. It would be unfortunate if IEEE went down the same path.  

Cheers,
Bill


> On Oct 19, 2018, at 4:03 PM, Robert Corbett via J3 <j3 at mailman.j3-fortran.org> wrote:
> 
> Bill asked if there is a set of
> floating flags and modes for each
> image.  The answer is given in
> Note 1 in Clause 17.7.  The note
> cites the normative text that
> supports the answer.  The answer
> is yes.
> 
> Bob Corbett

Bill Long                                                                       longb at cray.com
Principal Engineer, Fortran Technical Support &   voice:  651-605-9024
Bioinformatics Software Development                      fax:  651-605-9143
Cray Inc./ 2131 Lindau Lane/  Suite 1000/  Bloomington, MN  55425




More information about the J3 mailing list