Re: [Valgrind-users] Expected slowdown

[Julian S. <js...@ac...>]

> It looks like your program is causing a lot of very similar and/or
> identical stack traces to be created, and the comparisons against the
> existing ones are taking up a lot of time. 

That was my guess too.  But later it occurred to me, Christoph
said the program first has a nutshell (container, I guess?) 
which allocates a lot of memory - maybe 1 M blocks, and then the
main algorithm runs.  Suppose the nutshell fills the table 
execontext table up a lot.  Then when the algorithm runs and 
does malloc/free, these will be expensive because each one 
requires searching through long hash chains.

If that's true one simple fix would be the usual hack of moving
an ExeContext in a chain one step forwards for each successful
search.  Various other tables already do that.

I suspect this is easy to fix if we have a way to reproduce 
the problem.

J