A recent patch decreased GCC's memory consumption.

Andrew Pinski pinskia@gmail.com
Thu Nov 2 07:52:00 GMT 2006


On Thu, 2006-11-02 at 07:49 +0000, gcctest@suse.de wrote:
> Hi,
> 
> I am a friendly script caring about memory consumption in GCC.  Please
> contact jh@suse.cz if something is going wrong.

> comparing insn-attrtab.c compilation at -O2 level:
>   Ovarall memory allocated via mmap and sbrk decreased from 134058k to 129338k, overall -3.65%
>     Overall memory needed: 134058k -> 129338k
>     Peak memory use before GGC: 92593k
>     Peak memory use after GGC: 84705k
>     Maximum of released memory in single GGC run: 30380k
>     Garbage: 319045k
>     Leak: 9345k
>     Overhead: 36716k
>     GGC runs: 247
> Head of the ChangeLog is:
> 
> --- /usr/src/SpecTests/sandbox-britten-memory/x86_64/mem-result/ChangeLog	2006-11-02 02:40:15.000000000 +0000
> +++ /usr/src/SpecTests/sandbox-britten-memory/gcc/gcc/ChangeLog	2006-11-02 06:19:57.000000000 +0000
> @@ -1,3 +1,9 @@
> +2006-11-01  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
> +
> +	PR middle-end/29335
> +	* builtins.c (do_mpfr_sincos): New.
> +	(fold_builtin_1): Use it to fold builtin sincos.
> +


This change does make sense for the decrease at all because we should
not have any fp in insn-attrtab.c.

-- Pinski



More information about the Gcc-regression mailing list