You are viewing paulmck

Previous Entry | Next Entry

SequentialCaveman
I presented “Is Parallel Programming Hard, And, If So, Why?” at the Multicore and Parallel Computing Miniconf, and had the good fortune of having Linus in the audience. As you will quickly learn if you post to LKML, Linus often gives very insightful (and sometimes quite pointed) feedback. My presentation was no exception.

Linus noted that many parallel programming experts are unwilling to admit that there are algorithms that do not parallelize nicely.

As you might expect, I responded by saying that parallelism is an optimization, and like other optimizations has places where it works well and places where it does not.

What else could I have said?

Comments

( 2 comments )
dvyukov
Jan. 26th, 2011 08:15 am (UTC)
But how would you have responded?
Linus, you statement makes no sense. Why people must will to admit irrelevant statement? It's like say that many experts are unwilling to admit that Bloom filters can't store information about transactions nicely. It's not an algorithm that you must parallelize. It's a good solution to a problem that you must find. If good means fast in your context, and fast means parallelized, then you need to find an algorithm that can be parallized nicely as part of the solution. For some problems it's very difficult or impossible, it's Ok, it's like there is no data structure that provides O(1) inserts, O(1) deletions, (1) lookups and O(1) ordered iteration - it's not considered as a fundamental problem in theory of data structures, right?

Is it enough mutually pointed? :)
(Anonymous)
Jan. 26th, 2011 11:17 am (UTC)
claim insurance
Nice site, I envy it, but it seems a little weird when under mozzila browser, keep it moving:)
( 2 comments )