Bottlenecks?

Bottlenecks?

Has anyone tried to parallelize the problem with explicit inter-statment dependencies worked into the parallelism? I've pondered and pondered, but I can't seem to find a model that is simple enough to produce any speedup except in very specialized cases (like a whole slew of output() statements).

4 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

Don't be too dismayed! This is after all an area of active research - parallel execution of sequential programs. While we don't have conditional execution to contend with, there is plenty to work with in simply dividing simple dependencies amongst multiple threads.

I have a scheme layed out to do this but I haven't completed my expression parser. Until that is done I cannot run tuning tests (what size buffers etc...) If it succeeds you will find out after the competition. It is too early for me to say if this will introduce more redundant work than payback.

Jim Dempsey

www.quickthreadprogramming.com

@Jim: Good luck!FWIW I really like this problem because it illustrates very clearly the need to separate the parallel and serial parts as much as possible to scale, altough ultimately I think it will again be I/O bound.

Leave a Comment

Please sign in to add a comment. Not a member? Join today