Bill Long longb at cray.com
Fri Jan 29 16:12:10 UTC 2021

> On Jan 29, 2021, at 9:59 AM, John Reid via J3 <j3 at mailman.j3-fortran.org> wrote:
> Bill, Rich, Steve Lionel
> If  COUNT_RATE is real, its kind determines the accuracy of its value. For a 64-bit clock, the user may be perfectly happy to compute all actual times to single-precison accuracy. Each time is an integer difference in COUNT values times COUNT_RATE. Therefore, I see its kind as having nothing to do with which clock is involved.

If the COUNT_RATE is real, its KIND determines the value (not the accuracy of the value). 

The COUNT_RATE is a different numerical value for the 32 and 64-bit “clocks”.  Making sure that the 64-bit clock rate goes with the 64-bit counts is essential to making sure we don’t invalidate existing codes.  We don’t have to say “64-bit”. We can leverage the “numeric storage units” language, since it applies to both integer and real types. 


Bill Long                                                                       longb at hpe.com
Engineer/Master , Fortran Technical Support &   voice:  651-605-9024
Bioinformatics Software Development                      fax:  651-605-9143
Hewlett Packard Enterprise/ 2131 Lindau Lane/  Suite 1000/  Bloomington, MN  55425

More information about the J3 mailing list