This is the mail archive of the java@gcc.gnu.org mailing list for the Java 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: Anyone else sees a bootstrap failure (Linux-x86)?


Andrew Haley wrote:
>  > "-pipe" is useful and eliminates a lot of temporary files creation,
>  > somewhat speeding up the build.
> 
> Not on any sane platform.  Really, have you ever seen a significant
> speedup?  Or is file creation so horribly slow on Windows that this is
> a real issue?

Ok, I did some measurements on Win2K and Solaris 8 with
a GCC 3.4 snapshot and 3.3.2 respectively.

On Windows 2000, I could not measure any difference
between "-pipe" and "normal" compilations (using the
'time' command as provided by MSYS's bash).

On Solaris 8, I did see an improvement with "-pipe",
but it was always a ~5% improvement. A representative
run would be:
-------------------------- 8< -----------------------
    ~/tmp > time gcc -pipe foo.c

    real    0m0.550s
    user    0m0.350s
    sys     0m0.100s
    ~/tmp > time gcc foo.c

    real    0m0.579s
    user    0m0.380s
    sys     0m0.050s
-------------------------- 8< -----------------------

"foo.c" is a generated file with 51 simple
functions - not terribly representative of "real" code,
I agree. The same file was used on both Win2K and Sol8.

I guess with more complicated code, more time
would be spent by GCC in optimisations rather than
I/O (and this trend is exacerbated in 3.4!) so
this represents hardly 5 minutes saved in a two
hour bootstrap. :-/

But I *still* maintain that GCC should not ICE
if I specify "-pipe" in bootstrap flags.

Ranjit.

-- 
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/


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