This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Faster compilation speed


>   From: 
>        "David S. Miller" <davem@redhat.com>
> 
>    From: Richard Henderson <rth@redhat.com>
>    Date: Mon, 19 Aug 2002 10:29:09 -0700
> 
>    Well, no, since SET, MEM, REG, PLUS all have two arguments.
>    And thus are all allocated from the same page.
> 
> Ok, how about walking from INSN down to the SET?  The problem
> does indeed exist there.
> 
> Next, we have the fragmentation issue.  Look at the RTL you
> have right before reload runs on any non-trivial compilation,
> and see where the pointers are.
> 
> So the problem is there.

This is consistent with my tests; I found that a simplistic allocation which
put everything on the same page, but which never freed anything, actually
bootstrapped GCC faster than the standard GC.

The GC was never supposed to make GCC faster, it was supposed to reduce
workload by getting rid of memory problems. But I doubt it achieves that
objective. Certainly, keeping track of all the attempts to 'fix' GC has burned
a lot of my time.

Tim Josling


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]