Re: New to PostgreSQL, performance considerations

From: "Daniel van Ham Colchete" <daniel(dot)colchete(at)gmail(dot)com>
To: "Florian Weimer" <fweimer(at)bfk(dot)de>
Cc: "Cosimo Streppone" <cosimo(at)streppone(dot)it>, "Postgresql Performance list" <pgsql-performance(at)postgresql(dot)org>
Subject: Re: New to PostgreSQL, performance considerations
Date: 2006-12-12 15:57:20
Message-ID: 8a0c7af10612120757t3e2add5ckb512af8f794a2235@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On 12/12/06, Florian Weimer <fweimer(at)bfk(dot)de> wrote:
> * Cosimo Streppone:
>
> > "-O0" ~ 957 tps
> > "-O1 -mcpu=pentium4 -mtune=pentium4" ~ 1186 tps
> > "-O2 -mcpu=pentium4 -mtune=pentium4" ~ 1229 tps
> > "-O3 -mcpu=pentium4 -mtune=pentium4" ~ 1257 tps
> > "-O6 -mcpu=pentium4 -mtune=pentium4" ~ 1254 tps
>
> -mcpu and -mtune are synonymous. You really should -march here (but
> the result is non-generic code). Keep in mind that GCC does not
> contain an instruction scheduler for the Pentium 4s. I also believe
> that the GCC switches are not fine-grained enough to cover the various
> Pentium 4 variants. For instance, some chips don't like the CMOV
> instruction at all, but others can process it with decent speed.

You can use -march=pentium4, -march=prescott and -march=nocona to the
different Pentium4 processors. But you have to use -march (and not
-mcpu or -mtune) because without it you are still using only i386
instructions.

Daniel

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message David Boreham 2006-12-12 15:58:21 Re: New to PostgreSQL, performance considerations
Previous Message Tom Lane 2006-12-12 15:47:42 Re: New to PostgreSQL, performance considerations