pow(3): warn against ineffecient uses?

pow(3): warn against ineffecient uses?

[Joachim Wuttke]

[-- Attachment #1: Type: text/plain, Size: 560 bytes --]

Maybe it would be appropriate to warn against inefficient
uses of the function pow. To make a very first proposal
for a paragraph to be added to pow(3):

In time-critical code, pow with small integer or simple
fractional exponents should be avoided in favor of
straightforward multiplication, division or square or
cubic roots: prefer x*x over pow(x,2), 1/x over pow(x,-1),
sqrt(x) over pow(x,0.5), or cbrt(x*x) over pow(x,2.0/3).

My motivation for this proposal comes from student code
in which I have seen exactly these mistakes.

- Joachim


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 5110 bytes --]

Re: pow(3): warn against ineffecient uses?

[Florian Weimer]

* Joachim Wuttke:

> Maybe it would be appropriate to warn against inefficient
> uses of the function pow. To make a very first proposal
> for a paragraph to be added to pow(3):
>
> In time-critical code, pow with small integer or simple
> fractional exponents should be avoided in favor of
> straightforward multiplication, division or square or
> cubic roots: prefer x*x over pow(x,2), 1/x over pow(x,-1),
> sqrt(x) over pow(x,0.5), or cbrt(x*x) over pow(x,2.0/3).

I don't think transforming pow (x, 2.0 / 3) to cbrt (x * x) is correct
because the result is not correctly rounded.  I'm not sure if it is a
good idea to add those examples to the manual page without discussing
the cases where the transformations might not be valid.

If GCC does not handle some of the common cases from real-world code,
that could be fixed in GCC (and GCC can do more transformations under
-ffast-math, too).
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: pow(3): warn against ineffecient uses?

[Michael Kerrisk (man-pages)]

On 03/31/2017 01:17 PM, Florian Weimer wrote:
> * Joachim Wuttke:
> 
>> Maybe it would be appropriate to warn against inefficient
>> uses of the function pow. To make a very first proposal
>> for a paragraph to be added to pow(3):
>>
>> In time-critical code, pow with small integer or simple
>> fractional exponents should be avoided in favor of
>> straightforward multiplication, division or square or
>> cubic roots: prefer x*x over pow(x,2), 1/x over pow(x,-1),
>> sqrt(x) over pow(x,0.5), or cbrt(x*x) over pow(x,2.0/3).
> 
> I don't think transforming pow (x, 2.0 / 3) to cbrt (x * x) is correct
> because the result is not correctly rounded.  I'm not sure if it is a
> good idea to add those examples to the manual page without discussing
> the cases where the transformations might not be valid.
> 
> If GCC does not handle some of the common cases from real-world code,
> that could be fixed in GCC (and GCC can do more transformations under
> -ffast-math, too).

Thanks, Florian, for jumping in on this. I don't have enough
knowledge of these details, so your comments help a lot. Lacking
any further refinement, I'll drop this proposed change for now.

Cheers,

Michael

-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html