|Operation Latencies email@example.com (2004-02-08)|
|Re: Operation Latencies firstname.lastname@example.org (TOUATI Sid) (2004-02-12)|
|Re: Operation Latencies email@example.com (2004-02-12)|
|Re: Operation Latencies firstname.lastname@example.org (Tommy Hoffner) (2004-02-13)|
|Date:||8 Feb 2004 22:07:39 -0500|
|Keywords:||optimize, architecture, question|
|Posted-Date:||08 Feb 2004 22:07:39 EST|
Many operations have variable latencies...so I was wondering how
compilers do scheduling for VLIW and how they calculate the critical
path of a data dependence graph when the latencies of operations are
not known at compile time.. Some scheduling like s/w pipelining
requires calculation of RecMII which needs to know the critical path
in the ddg... How can you be sure you have the critical path when the
latencies can vary?
What are the usual approaches ?
Return to the
Search the comp.compilers archives again.