[Developers] MoL Scheduling

Tom Goodale goodale at cct.lsu.edu
Fri Sep 7 09:24:52 CDT 2007


On Fri, 7 Sep 2007, I.Hawke wrote:

> Yosef Zlochower wrote:
>>  I think that two MoL routines ( MoL_RestoreSandR,MoL_FinishLoop)
>> are schedules incorrectly.
>> These routines are scheduled
>> "IN MoL_Evolution AFTER MoL_PostStep",
>> but MoL_PostStep is not scheduled in MoL_Evolution,
>> it is in the subgroup MoL_Step. Should this rather be
>> "IN MoL_Evolution AFTER MoL_Step"
>>
>> I tried to schedule one of my routines after MoL_FinishLoop,
>> and the result was that MoL_FinishLoop got scheduled before MoL_Step.
>>
>
> I had always assumed that if B is scheduled in A and C is scheduled
> after B, then C would implicitly be scheduled after A (in the case where
> it is not explicitly scheduled in A). I cannot remember if the precise
> statement is meant to explicitly give the reason for the schedule
> dependency or is just historical. Tom, does the sorter not recognize
> this case?

Scheduling statements are only considered relative to things within the 
same schedule group/bin, i.e. in your example above you have to explicitly 
schedule C in A --- without that the scheduler just looks in the current 
group, sees there is no B there and ignores the AFTER.

Cheers,

Tom


More information about the Developers mailing list