On Sunday 16 May 2004 10:48 pm, unixware wrote:
> Thanks Steve for your reply
>
> i think "-tpp6 -xK" i like -march=i686 in gcc
Yes, it is equivalent to -march=i686 in GCC. Likewise, "-tpp7 -xN" is a
replacement for GCC's "-march=pentium4" switch.
> after exporting all these u suggested squid-2.4
> compiles fine :)
>
> i never tried to build rpm but i try if u send me the
> squid.spec file
Well, if you've gotten Squid to build successfully (which it seems you
have from above), it's not really necessary. I was thinking of it more
as a working example.
> do u see any performance difference using icc ?
There was a definite performance improvement using ICC v7.1 vs Red Hat's
GCC v2.96. (In fact I starting using ICC because of the miserable code
generation with GCC as shipped on the Red Hat v7.3 distribution.)
Measured using Polygraph, I got about 15% more transactions/sec with ICC
v7.1 vs GCC v2.96 on a Pentium4 system.
Now I'm using Red Hat v9, which includes GCC v3.2.2. Both GCC and ICC 8.0
are said to feature improved x86 code generation relative to their
previous versions. I haven't benchmarked Squid with both of these newer
compilers, so perhaps I am acting on faith. Nethertheless, I continue to
use ICC, mostly because Intel is serious about showing off the Pentium4,
and ICC is the compiler to beat for x86 code generation. GCC, at least
in v3.2.x, offers little in the way of Pentium4-specific optimization.
One more thing. When building Squid with ICC you'll see a lot of these:
icc: Command line remark: option '-MP' not supported
I can't figure out what is causing these many, many warning messages, but
they seem to have no adverse effect on the build process.
Hope this helps.
Received on Sun May 16 2004 - 23:30:34 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Jun 01 2004 - 12:00:01 MDT