native floating point sinint

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

native floating point sinint

Colin Macdonald-2
Hi Joel,

I wonder if you could take a look at this old pull request:

https://github.com/cbm755/octsympy/pull/425

My feeling is drop this but I thought you might offer a perspective
after your summer project.

Background: the symbolic package inherited a bunch of code from the
specfun pkg.  This included "sinint" and "cosint" implementations.  They
were not very good so I replaced them [1] with rather slow,
nonvectorized calls to the SymPy (ultimately the arbitrary precision
python library "mpmath").  The PR is the original specfun code, with a
few failing tests added by me.

[1] https://github.com/cbm755/octsympy/blob/master/inst/%40double/sinint.m

Reply | Threaded
Open this post in threaded view
|

Re: native floating point sinint

Colin Macdonald-2
On 2017-08-16 02:22 PM, Colin Macdonald wrote:

> I wonder if you could take a look at this old pull request:
>
> https://github.com/cbm755/octsympy/pull/425
>
> My feeling is drop this but I thought you might offer a perspective
> after your summer project.
>
> Background: the symbolic package inherited a bunch of code from the
> specfun pkg.  This included "sinint" and "cosint" implementations.  They
> were not very good so I replaced them [1] with rather slow,
> nonvectorized calls to the SymPy (ultimately the arbitrary precision
> python library "mpmath").  The PR is the original specfun code, with a
> few failing tests added by me.
>
> [1] https://github.com/cbm755/octsympy/blob/master/inst/%40double/sinint.m

I'm sorry, wrong person!

*Michele*, I meant to refer the above to you!

best,
Colin

Reply | Threaded
Open this post in threaded view
|

Re: native floating point sinint

Michele Ginesi
On 08/16/2017 11:24 PM, Colin Macdonald wrote:

> On 2017-08-16 02:22 PM, Colin Macdonald wrote:
>> I wonder if you could take a look at this old pull request:
>>
>> https://github.com/cbm755/octsympy/pull/425
>>
>> My feeling is drop this but I thought you might offer a perspective
>> after your summer project.
>>
>> Background: the symbolic package inherited a bunch of code from the
>> specfun pkg.  This included "sinint" and "cosint" implementations.  
>> They were not very good so I replaced them [1] with rather slow,
>> nonvectorized calls to the SymPy (ultimately the arbitrary precision
>> python library "mpmath").  The PR is the original specfun code, with
>> a few failing tests added by me.
>>
>> [1]
>> https://github.com/cbm755/octsympy/blob/master/inst/%40double/sinint.m
>
> I'm sorry, wrong person!
>
> *Michele*, I meant to refer the above to you!
>
> best,
> Colin
Dear Colin,
I'm giving a look to the functions. Unfortunately seems that there is
only a series expansion (no continued fraction or other strategies). But
it is possible to use some relations in order to use expint to compute them.
I took few minutes to try to implement sinint using some of these
relations (Abramowitz 5.2.19 and 5.2.21).
The result far to be perfect but I think it could be a good starting point.

--
Michele Ginesi

sinint.m (2K) Download Attachment