[racket-dev] gc vs assignment

On Tue, Aug 24, 2010 at 7:05 AM, Matthias Felleisen
<matthias at ccs.neu.edu> wrote:
>
> Catching up with some mail.
>
> Neil wrote:
>
>> Avoiding allocation reduces GC collects, which reduces stutters and hitches.
>
> My (possibly old) understanding of GC and mutation tell me that this is one
> of those prejudices that programmers should get rid of.

In my experience, GC can easily be tuned to take less than 5% or so
of the runtime (on average).  Optimizing beyond that is most definitely
not worth it.

A decent generational collector takes almost zero time to deallocate garbage,
so allocation of very temporary structure has little penalty.
However, allocation
of long-lived structure will have a bit more overhead.  If that
overhead is really
a problem (that is, if it takes more than about 10% of your runtime), then the
solution is to pre-allocate, or instruct the allocator to put the object in
`permanent' storage from the get-go.

I'm surprised that racket3m uses page protection.  Taking a hardware trap
can often be thousands of times slower than taking an inline conditional
branch.


-- 
~jrm