This is the mail archive of the gcc-help@gcc.gnu.org mailing list for the GCC 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: Fwd: GCC Optimization for expressions


Hello Tim,

I am grateful to your comments. I had started this thread to know
about a concept regarding difference in execution timings in two
scenarios i had mentioned. I guess, I still havn't got answer to that.
I need to know among those scenarios, in which case GCC would give
better performance? I am not clear about this concept, and am seeking
clarification. I gave details of my work just to give a feel,
otherwise I need to know about the concept, irrespective of the work
being done.

Regards,
Nirav Shah.



On Mon, Dec 1, 2008 at 3:52 AM, Tim Prince <TimothyPrince@sbcglobal.net> wrote:
> Nirav Shah wrote:
>
>>
>> I am working on study of data sharing characteristic of Blackscholes
>> application of Parsec benchmark. I need to come up with better data
>> sharing foot print of execution.
>
>> No expression is evaluated in loop nor any particular intermediate
>> result is used more than once. Given this, can I get benefit in terms
>> of lesser clock cycle to execute the computation if done in a single
>> statement as compared to done over multiple statements with local
>> variables?
>>
>> Regards,
>> Nirav Shah.
> Black-Scholes benchmarking has consumed a great deal of technical
> marketing time, without in most cases divulging source code.   Referring
> to a public example, http://www.espenhaug.com/black_scholes.html which
> doesn't seem to bear much relationship to your comments.  That example
> shows an apparent total ignorance of Horner's rule, along with the
> unexplained sequence (1 - (1 - .....), protected against simplification by
> -ffast-math.  Maximum delay and cancellation of precision for minimum effort.
> The Fortran-ish version (not compliant with any Fortran standard, but may
> "work") tries to prevent optimization by using floating point exponents,
> but that's not as much of a sin as evaluating a polynomial with multiple
> pow() evaluations.
>
> The whole thing is complicated enough that you would want to use at least
> gprof to determine where you could work usefully to improve performance.
> If you were set this exercise as a university project, was it really the
> intent to solicit uniformed opinions about how to proceed?
>
> I'm amazed at the Parsec site claim to support world-wide downloads, it's
> so slow.
>
>


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