| Re: [eigen] Remaining beta3 blockers |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Remaining beta3 blockers
- From: Gael Guennebaud <gael.guennebaud@xxxxxxxxx>
- Date: Tue, 23 Nov 2010 20:27:40 +0100
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:mime-version:received:in-reply-to :references:from:date:message-id:subject:to:content-type :content-transfer-encoding; bh=Ab2GtfJ+r6syLCALOjQpMVUZMc/ubgxBFJNsNeNhaMY=; b=Hl0FVuYlP6yJj7ZoSZrAhXreymcYI6UxYO7Bm0OK4o+SCN69XVxCzMEXN98UX7KTSF F6V7APoqF389rZhPT+Rp7LOUR67tj0s7ERg5d35LNDTERj2Kc6M4RZNu4HlNXP2hzakx wNhEM4KRAXnhgSjatEc1d3CUdMOBdL5Xx2pfs=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:from:date:message-id:subject:to :content-type:content-transfer-encoding; b=T8fECYGLcFfS7e3nl18M3WaGKBbP1oaokH8qSipJveuNAzzRguf4n+Fv6oMkBLqx1P hUi/CAGgcu22czXQQhOLiPpj3HMy3a+FKV2JfL9NaFYjyJ0ctPLL9qcQmwgrHbfiWzkP cJQT+h2HuEuJ/SqkX9JziIFYhFzK6Jpi7Q6s8=
> 51: no malloc in fixed-size products
>
> Gael, are you looking at it? If needed I'm OK to have it block only 3.0 final.
This one is a quite big challenge because it is tightly coupled with
multi-threading. Basically, fixing it before having find a good and
scalable solution to parallelize all products with less as possible
code redundancy means doing the work twice.... :(
Currently, not that general matrix-matrix and matrix-vector products
are already ok with respect to no malloc for fixed-sizes. Only special
products have yet to be fixed.
> 53: Sparse - API stability for basic features
>
> Gael?
I'm think it is more reasonable to postponed it for 3.1 since I'm
still not really convinced by my current choices.
> 54: Const-correctness
>
> The biggie. Discussed with Gael on IRC. The way to go is to make
> Map<const T> not have the Lvalue bit, so that absence of LvalueBit
> will propagate to other expression automatically, the tricky part is
> that this kills the assumption that DirectAccessBit => LvalueBit.
>
> I'll have a go at it unless Gael does it first...
yeah, we'll see who's first ;)
> 57: Uniform external library support
>
> Gael?
this won't be needed for any stable 3.0 module, so we can postponed it.
> 73: 3.0-beta2: Intel incomplete type error breaks several unit tests
>
> BEEP BEEP BEEP! This one is not yet assigned to anybody. Any volunteer?
I'll check it.
> 89: householder_1 and _2 are continuously failing with segfaults
>
> BEEP BEEP BEEP! This one is not yet assigned to anybody. Any volunteer?
>
> 91: scalar_product_traits is missing some specializations
>
> BEEP BEEP BEEP! We're still deeply stuck in the sauerkraut with this
> one. This is a terrible MSVC 2010 bug that we hit since we use a
> nested 'internal' namespace and we don't know what to do about it.
>
> 100: Finalize the array-as-scalar feature
>
> Gael ?
This one would make a good "junior" jog, so if someone want to step
by? Otherwise I'll do it, np.
gael