Richard Guenther wrote:
>> I wrote:
The speed up is around 5 %.
Is this because of cbrt or a combined effect? Can you measure the cbrt
effect in isolation?
This is a combined effect (assuming that things like x**1.5 didn't
expand to x * sqrt (x) before) - however, the number of opportunities of
converting x**N.5 -> x**N * sqrt(x) is very limited in our code.
I artificially up'd the number of opportunities for the x**(N+{1,2}./3.)
-> x**N * cbrt(x)[**2] by changing about two dozen places where
x**0.333333 or x**0.3333 or x**0.6666666 [sic] was used instead of
x**(1./3.) or x**(2./3.). In fact, I'd like to use the carrot of
potential optimizations to get people to use the clearer x**(1./3.) (we
also have a legitimate use of x**0.3, where the 0.3 really is 3./10.)
I can measure the contribution of the cbrt effect in isolation, though
(given the above change of HIRLAM source code).
Cheers,
--
Toon Moene - e-mail: [EMAIL PROTECTED] - phone: +31 346 214290
Saturnushof 14, 3738 XG Maartensdijk, The Netherlands
A maintainer of GNU Fortran: http://gcc.gnu.org/fortran/
Who's working on GNU Fortran:
http://gcc.gnu.org/ml/gcc/2006-01/msg00000.html
