From: Paul H. Hargrove (PHHargrove_at_lbl_dot_gov)
Date: Thu Apr 16 2009 - 12:37:50 PDT
Nikita, More answers below. -Paul Nikita Andreev wrote: > Hello again. > > I want to ask some general questions. > > 1. There is an implicit barrier at the end of the program. Is there an > implicit barrier at the start? There is no *specification* that says there is an implicit barrier at the start of the code. Therefore it is possible that the first statement in on thread could execute much earlier than the first statement in another, for instance if startup times are not uniform. However, the Berkeley UPC runtime does contain a barrier very shortly before the beginning of main() due to its own initialization requirements. > 2. Does UPC compiler generate any other implicit barriers somewhere in > the code (i.e. after upc_forall)? There are no implicit barriers in the UPC specification, but any upc_all_*() function is going to involve some sort of collective communication and therefore the implementation *might* include a barrier, but is not required to. There is NOT an implicit barrier before or after a upc_forall() in the Berkeley UPC implementation. > 3. Are you going to add UPC-IO events to GASP and when? I don't know the answer to this one. > > Thank you for the answers. > > With all regards, > Nikita. -- Paul H. Hargrove PHHargrove_at_lbl_dot_gov Future Technologies Group HPC Research Department Tel: +1-510-495-2352 Lawrence Berkeley National Laboratory Fax: +1-510-486-6900