On 25 Feb 2009, at 16:21, Alexander.Gorshenev at Sun.COM wrote:

> On Wed, 25 Feb 2009, Chris Ridd wrote:
>
>>
>> On 24 Feb 2009, at 20:23, C. Bergstr?m wrote:
>>
>>> Since the xorg driver size issue I've been looking at the objdump  
>>> of some binaries and the code generation between sun cc, gcc and  
>>> open64..
>>> CSiBE gave some interesting rough size comparison numbers, but was  
>>> only able to compare open64/gcc.. Besides things like  
>>> SPEC2000int.. what else if any are people using to check for  
>>> performance/size regressions in the sun toolchain?  I know there's  
>>> full time employees making sure solaris stays top notch, but can  
>>> any of these details/sources be published?
>>
>> I noticed that Phoronix recently did some comparisons between gcc  
>> and SS 12.
>

[link deleted]

> Oh, not again! There are serious apples vs. oranges issues in the
> comparison to the level making it completely irrelevant.  They compare
> optimized compilations of one compiler with no-opt compilations of
> another compiler.  They compare openmp builds of one compiler with
> non-openmp builds of another.

They did only test default options for each compiler... which is a bit  
odd, and I agree it makes the tests fairly meaningless.

> The build and run logs or at least details are not provided, so people
> around trying to reproduce the result get completely different  
> figures.
>
> See for example the comments to the article itself.
> http://www.phoronix.com/forums/showthread.php?t=15534&page=2

They did note that they were talking to someone on the SS team about  
the results. Any ideas who, and what their comments were?

Cheers,

Chris

Reply via email to