Re: [Lse-tech] gcc 2.95 vs 3.21 performance

Re: [Lse-tech] gcc 2.95 vs 3.21 performance

[Andi Kleen]

On Mon, Feb 03, 2003 at 03:05:06PM -0800, Martin J. Bligh wrote:
> The results below leaves me distinctly unconvinced by the supposed 
> merits of modern gcc's. Not really better or worse, within experimental
> error. But much slower to compile things with.

Curious - could you compare it with a gcc 3.3 snapshot too?

It should be even slower at compiling, but generate better code.

-Andi

Re: [Lse-tech] gcc 2.95 vs 3.21 performance

[Dave Jones]

On Mon, Feb 03, 2003 at 03:05:06PM -0800, Martin J. Bligh wrote:
 > People keep extolling the virtues of gcc 3.2 to me, which I'm
 > reluctant to switch to, since it compiles so much slower. But
 > it supposedly generates better code, so I thought I'd compile
 > the kernel with both and compare the results. This is gcc 2.95
 > and 3.2.1 from debian unstable on a 16-way NUMA-Q. The kernbench
 > tests still use 2.95 for the compile-time stuff.
 > 
 > The results below leaves me distinctly unconvinced by the supposed 
 > merits of modern gcc's. Not really better or worse, within experimental
 > error. But much slower to compile things with.

What kernel was kernbench compiling ? The reason I'm asking is that
2.5s (and more recent 2.4.21pre's) will use -march flags for more
aggressive optimisation on newer gcc's.
If you want to compare apples to apples, make sure you choose
something like i386 in the processor menu, and then it'll always
use -march=i386 instead of getting fancy with things like -march=pentium4

        Dave

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: [Lse-tech] gcc 2.95 vs 3.21 performance

[Martin J. Bligh]

>  > People keep extolling the virtues of gcc 3.2 to me, which I'm
>  > reluctant to switch to, since it compiles so much slower. But
>  > it supposedly generates better code, so I thought I'd compile
>  > the kernel with both and compare the results. This is gcc 2.95
>  > and 3.2.1 from debian unstable on a 16-way NUMA-Q. The kernbench
>  > tests still use 2.95 for the compile-time stuff.
>  > 
>  > The results below leaves me distinctly unconvinced by the supposed 
>  > merits of modern gcc's. Not really better or worse, within experimental
>  > error. But much slower to compile things with.
> 
> What kernel was kernbench compiling ? The reason I'm asking is that
> 2.5s (and more recent 2.4.21pre's) will use -march flags for more
> aggressive optimisation on newer gcc's.
> If you want to compare apples to apples, make sure you choose
> something like i386 in the processor menu, and then it'll always
> use -march=i386 instead of getting fancy with things like -march=pentium4

Kernbench compiles 2.4.17, because I'm old, slow and lazy, and that
was what was around when I started doing this test ;-)

But the point is still the same ... even if it is doing more agressive
optimisation, it's not actually buying us anything (at least for the kernel)

M.

Re: [Lse-tech] gcc 2.95 vs 3.21 performance

[Momchil Velikov]

>>>>> "Martin" == Martin J Bligh <mbligh@aracnet.com> writes:

    Martin> But the point is still the same ... even if it is doing
    Martin> more agressive optimisation, it's not actually buying us
    Martin> anything (at least for the kernel)

which might be due in part to ``-fno-strict-aliasing'' used to compile
the Linux kernel.

~velco

Re: [Lse-tech] gcc 2.95 vs 3.21 performance

[Albert Cahalan]

Momchil Velikov writes:
> "Martin" == Martin J Bligh <mbligh@aracnet.com> writes:

>> But the point is still the same ... even if it is doing
>> more agressive optimisation, it's not actually buying us
>> anything (at least for the kernel)
>
> which might be due in part to ``-fno-strict-aliasing''
> used to compile the Linux kernel.

This is fixable for any gcc implementing __may_alias__.

Linux uses -fno-strict-aliasing because people like
to cast a (foo*) to an (int*) instead of using a
union or (char*) as required by the C language.
When -fno-strict-aliasing was added to the command
line, gcc did not offer the __may_alias__ attribute.

BTW, in case any gcc hacker is paying attention,
the documentation fails to mention the gcc version
required for this or any other attribute. Also it
would be nice to have an option to ditch the (char*)
exception; it's junk when you have __may_alias__.

Re: [Lse-tech] gcc 2.95 vs 3.21 performance

[Falk Hueffner]

Albert Cahalan <albert@users.sourceforge.net> writes:

> BTW, in case any gcc hacker is paying attention, the documentation
> fails to mention the gcc version required for this or any other
> attribute. Also it would be nice to have an option to ditch the
> (char*) exception; it's junk when you have __may_alias__.

I don't think a switch that makes standard compliant source break in
potentially very subtle ways is a good idea. Just use "restrict".

-- 
	Falk