Re: [Libunwind-devel] [PATCH 0/1] Fast back-trace for x86_64 for only collecting the call stack

[Arun Sharma]

On Thu, May 27, 2010 at 1:03 PM, Lassi Tuura <address@hidden> wrote:

> This is fine in itself. How do you suggest to technically handle a frame
> cache that survives across calls, such that the caller has enough control
> to make the cache big enough to hit almost always after a warm-up period?
>
> I did for example play around with an idea of extra unw_set_foo() calls
> after unw_local_init() but before starting trace itself. I wanted to get
> this patch first out for discussion, so I didn't fully explore that.
>

We already have a caching policy in unw_addr_space.caching_policy
(NONE, GLOBAL, PER_THREAD). src/dwarf implements dwarf_rs_cache based
on this. The implementation was borrowed from ia64.

Perhaps you can reuse the existing policy in unw_addr_space and find a
way of reusing the cache implementation code? One concern is the
amount of code duplication. We'll be going from 2 copies to 3 copies
of very similar code.

 -Arun