Commits on Aug 18, 2015

1. Fix wrong math in dynamic_vector allocation. …

   There were two problems with this expression. First, the order of
   operations was wrong (missing parentheses). Second, the size of a
   strided subsection of an array is the size over the stride rounded up,
   rather than down, as in typical integer division.
   
   This bug showed up in unit tests with some compiler flag modifications,
   but the affected code is not exercised in CESM (or probably anywhere
   else, other than unit tests). The reason that unit tests did not catch
   this issue earlier is that the bug has no effect under Fortran 2003
   automatic allocation rules. If the array is too small, automatic
   re-allocation prevents a segfault. If the array is too big, this is not
   a problem, since the vector type is deliberately designed to reserve
   extra space.
   
   However, we currently use some compilers without Fortran 2003 allocation
   rules enabled. This seems true of PGI and Cray in particular. On these
   compilers, it seems to be worth fixing the issue in case the strided
   access feature of the dynamic vector is used in the future.
   
   Testing:
   
       Unit tested only (this code is not currently exercised).
       Status: bit for bit

   

   quantheory committed Aug 18, 2015
   Configuration menu

   • View commit details
   • Copy full SHA for 8562aab
   • Browse repository at this point

   Copy the full SHA

   8562aab View commit details

   Browse the repository at this point in the history