Aw: Re: [Ping][Patch] Speedup Riemann Zeta

[Ed Smith-Rowland]

On 09/01/2014 08:39 PM, Florian Goth wrote:
> Hi all!
> I've looked into the draft for 29124 and as far as I understand it, it involves
> an extension of <cmath> like
> namespace std {
> float cos(float x) {}
> float sin(float x) {}
> .
> etc...
> .
> #ifdef __STDCPP_WANT_MATH_SPEC_FUNC__
> #define __STDCPP_MATH_SPEC_FUNCS__
> ...
> float riemann_zeta(float s) {}
> ...
> #endif
> }
>
> The implementation of all the special functions is already present due to Edwards work in the TR1 folder of gcc's current implementation.
> So we would in theory just move over everything we already have in the TR1 folder?
> I suppose the polylog will go somewhere there too into ::__detail .
> So how do you guys start with this? Somebody sets up a repository on github?
>
> On a sidenote how long do I typically have to wait for a confirmation of the paperwork...?
>
> Florian.
>
>> Gesendet: Dienstag, 19. August 2014 um 17:18 Uhr
>> Von: "Paolo Carlini" <paolo.carlini@oracle.com>
>> An: "Florian Goth" <CaptainSifff@gmx.de>, "libstdc++@gcc.gnu.org" <libstdc++@gcc.gnu.org>, 3dw4rd@verizon.net
>> Betreff: Re: [Ping][Patch] Speedup Riemann Zeta
>>
>> Hi,
>>
>> On 08/19/2014 04:54 PM, Florian Goth wrote:
>>> Anything wrong with it?
>>> I would like to learn which guidelines I have to respect,
>>> before I try to get the larger patches for the Polylog function into libstdc++.
>> I suppose we have to wait for your paperwork anyway, right? Other than
>> that, in case nobody mentioned it already, I wanted to clarify that
>> since the TR1 version of these facilities is in deep freeze mode, all
>> the new code belongs to a new, separate, directory for 29124:2010, which
>> I think Ed already mentioned. I suppose we want first to copy to the new
>> directory the TR1 code, do all the mechanical changes, eg, namespaces,
>> deletions, and then start committing the new contributions. Maybe Ed is
>> willing to help with that...
>>
>> Paolo.
>>
You should be getting your paperwork pretty soon I should think.
Maybe wait another week and ping somebody...

I had started a patch a long time ago to move everything over from tr1 
to bits.
I think now I'd rather *copy* everything.
This way we can start hacking the new 29124 functions without touching tr1.
At first this will be annoying duplicate but later

I'll fix up a new patch and post it (that's how we roll here unless a 
patch is really large and takes most of a cycle to research).
The cmath 29124 initial migration should be mostly mechanical.

When that's in we'll have a pace for you new work.
Extensions such as airy functions and the hurwitz_zeta function should 
go under a namespace __gnu_cxx for gnu extensions.
Implementation files under __detail.

Ed