|
|
Hi,
> BUT, it has the disadvantage that I don't think it is possible to avoid a
> wrapper level of indirection to massage the arguments. Personally, I don't
> see that as a design goal: MPI internally surely will call dozens of wrappers
> until the actual work gets done?
>
>
This is not true for our MPI versions; for example the MPI send calls
for SX and Earth
Simulator don't have any wrapper up to the assembler routine which is
putting the data
on the interconnect. And this is true also for the Fortran versions of
send routines.
And I would expect that the number of wrappers could become crucial
if the clock frequency is reduced in future systems (in order to save
energy).
Hubert
|
|
|
![[Flag]](../../../../../images/flag-us.gif){kind=small}