[eigen] Re: [Bf-committers] How about adding boost uBLAS library to blender/extern? |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: bf-blender developers <bf-committers@xxxxxxxxxxx>
- Subject: [eigen] Re: [Bf-committers] How about adding boost uBLAS library to blender/extern?
- From: Keir Mierle <mierle@xxxxxxxxx>
- Date: Tue, 27 Jan 2009 10:09:01 -0800
- Cc: eigen <eigen@xxxxxxxxxxxxxxxxxxx>
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:in-reply-to:references :date:message-id:subject:from:to:cc:content-type :content-transfer-encoding; bh=kaBvgTuRZ4GTSPjWm9E22Y6ga7ud4KxHPn+JsAPMoso=; b=P9sHdj+y55O/ZDwjveO8S1KsvcU1FPqvFx641nLNqaboYj0TxmfnbigNo0o1aMfkqw Q0PZWf/T/+J8ctKX/Bx3dXZD+bij/dCMKjyNXICzfostH90JZeNhxZzslR/DFxJWz80F Hy3SSlxLjiIkFDwDlcUDIOcp+2DWVZdW0QLdw=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type:content-transfer-encoding; b=cMUnjLQvK+pPAa7gsk2EvgD5lOcNpCabvxJOpkzMJxaalg0P7v8EjzPdBjOfozaBUE ZiIGFjWf8V895I6+SIGgV2ckTdSgPofWR7AXds9SHAqgUZmIBCw1PChEGdEMnP2isdny 0lV3RWaFjYaP9And8mbN0sWRdD09nJ85fsHIE=
On Tue, Jan 27, 2009 at 7:02 AM, Shaul Kedem <shaul.kedem@xxxxxxxxx> wrote:
> Thank you for the info benoit.
>
> Can you please ask him how he debug the template on each platform?
Usually GDB. However, as a user of Eigen for libmv, I find the
debugger / eigen issue is irrelevant because the bugs are all in my
code, not Eigen. If anything, your code is more likely to work with
Eigen after compiling (compared to C matrix libraries) because the
types have much more information about the computations and will
refuse to compile nonsensical expressions.
Switching to Eigen decreased the rate at which I add bugs to libmv.
> what is his main dev platform?
Linux.
Keir
>
>
>
> On Tue, Jan 27, 2009 at 4:09 AM, Benoit Bolsee <benoit.bolsee@xxxxxxxxx> wrote:
>> Regarding Eigen2, I received additional explanations from the main
>> developper Benoit Jacob. It shows a nice commitment to future
>> improvements:
>>
>>> it has automatic SIMD and cache friendly optimization
>>
>> Yup indeed :)
>>
>>> it provides out of the box SVD decomposition
>>
>> Indeed. The current SVD code, in Eigen 2.0, is taken from the public
>> domain JAMA. It's decent but not a killer. My top priority right now is
>> to get a much more SVD, written from scratch, in Eigen 2.1. In
>> particular it will be extremely stable numerically, it should have
>> efficient fixed-size specializations, and it should of course take full
>> advantage of the Eigen framework. My motivation is that this is one of
>> the most asked-for features.
>>
>> Eigen 2.0 will be released next week. Eigen 2.1 will be used heavily by
>> KDE 4.3 so has to be released around June or July.
>>
>>>MSVC 2005 and up (MSVC 2008 for SSE
>>>optimization), gcc 3.3 and up (gcc 4 for SIMD optimization), recent ICC
>>
>>>(9? 10?, 11?).
>>
>> GCC 4.2 for SIMD. For ICC, I don't know exactly because it's not me who
>> tested it. There were at least successful reports with ICC 10.
>>
>>> However, current version is beta6,
>>
>> Yep, final release next week.
>>
>>>API is stable for dense matrices
>>
>> Mostly. In Eigen 2.0 we marked many parts as experimental to retain some
>> freedom, although we will ensure 99.9% compatibility for our important
>> users. If Blender uses Eigen we will work with you to minimize trouble.
>> Our goal is to shed the experimentalness and extend the API stability
>> guarantee to much more of Eigen for the 2.1 release around June.
>>
>>>it is already used by many projects
>>
>> Indeed new projects switch every week.
>>
>>> and available as a package in various
>>>Linux distributions.
>>
>> Indeed, mostly thanks to KDE and KOffice depending on it.
>>
>>
>> _______________________________________________
>> Bf-committers mailing list
>> Bf-committers@xxxxxxxxxxx
>> http://lists.blender.org/mailman/listinfo/bf-committers
>>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers@xxxxxxxxxxx
> http://lists.blender.org/mailman/listinfo/bf-committers
>