Has anybody else seem significant performance hits when setting the initial values of an array (to be used in a summation of some form) to zero using:
when using the Fortran accelerator compiler
If so, how did you work around this? Obvious solutions I am currently using are looping over the array in an initial separate data region or copying in an initally zeroed array.
Is there perhaps an optimised intrinsic function for this that i am unaware of?