FEniCS/PDESoft Abstracts
Hi all, Here are my drafts of potential talks to submit: FEniCS'14: https://gist.github.com/kynan/9c9c048b0c05fa8d0188 "Firedrake: Composing Domain-specific Abstractions for Productive and Efficient Computational Science" PDESoft 14: https://gist.github.com/kynan/025857adaa21bfabdd59 "PyOP2: a performance-portable framework for parallel computations on unstructured meshes" Cheers, Florian
I have updated my FEniCS abstract: "Firedrake: Re-imagining FEniCS by Composing Domain-specific Abstractions" https://gist.github.com/9c9c048b0c05fa8d0188 I'll also submit an abstract for a 20min talk at EuroSciPy 2014 Cambridge, UK 27-30 August 2014: https://www.euroscipy.org/2014/ "Firedrake: a performance-portable finite-element computation framework" https://gist.github.com/2c05a630c15c428e8da0 The deadline for EuroSciPy is tonight, sorry for the short notice. Many thanks, Florian On 27/03/14 14:21, Florian Rathgeber wrote:
Hi all,
Here are my drafts of potential talks to submit:
FEniCS'14: https://gist.github.com/kynan/9c9c048b0c05fa8d0188 "Firedrake: Composing Domain-specific Abstractions for Productive and Efficient Computational Science"
PDESoft 14: https://gist.github.com/kynan/025857adaa21bfabdd59 "PyOP2: a performance-portable framework for parallel computations on unstructured meshes"
Cheers, Florian
On 14/04/14 20:33, Florian Rathgeber wrote:
I have updated my FEniCS abstract: "Firedrake: Re-imagining FEniCS by Composing Domain-specific Abstractions" https://gist.github.com/9c9c048b0c05fa8d0188
Hi Florian, I don't think it really tells anybody what the advantage of firedrake is --- """ optimised implementations for different hardware architectures can be automatically generated without any changes to a single high-level source. Performance matches or exceeds what is realistically attainable by hand-written code. """ That last sentence is already true for FEniCS. The first sentence implies you'll show GPU results. Is that true? If so, say so; if not, remove it? To be specific, what concrete advantages does firedrake offer? I think you should address this in your abstract. Hope this helps, Patrick
On 14/04/14 20:40, Patrick Farrell wrote:
On 14/04/14 20:33, Florian Rathgeber wrote:
I have updated my FEniCS abstract: "Firedrake: Re-imagining FEniCS by Composing Domain-specific Abstractions" https://gist.github.com/9c9c048b0c05fa8d0188
Hi Florian,
I don't think it really tells anybody what the advantage of firedrake is ---
""" optimised implementations for different hardware architectures can be automatically generated without any changes to a single high-level source. Performance matches or exceeds what is realistically attainable by hand-written code. """
That last sentence is already true for FEniCS. The first sentence implies you'll show GPU results. Is that true? If so, say so; if not, remove it?
To be specific, what concrete advantages does firedrake offer? I think you should address this in your abstract.
Thanks, Patrick. I've modified the last paragraph to more clearly point out the advantage of the composition of abstractions and removed the sentence on performance entirely. The first sentence is intentionally vague: it would be great to have GPU results to show, but we'll have to see how far we have got by then. I'm not specifically claiming performance portability, because we can't quite live up to that at the moment. Florian
Hope this helps,
Patrick
I think it is reasonable to say that Firedrake is targetting performance portability, and these are the first steps towards that. --cjc ________________________________________ From: firedrake-bounces@imperial.ac.uk [firedrake-bounces@imperial.ac.uk] on behalf of Florian Rathgeber [f.rathgeber10@imperial.ac.uk] Sent: 14 April 2014 21:06 To: firedrake Subject: Re: [firedrake] FEniCS/PDESoft Abstracts On 14/04/14 20:40, Patrick Farrell wrote:
On 14/04/14 20:33, Florian Rathgeber wrote:
I have updated my FEniCS abstract: "Firedrake: Re-imagining FEniCS by Composing Domain-specific Abstractions" https://gist.github.com/9c9c048b0c05fa8d0188
Hi Florian,
I don't think it really tells anybody what the advantage of firedrake is ---
""" optimised implementations for different hardware architectures can be automatically generated without any changes to a single high-level source. Performance matches or exceeds what is realistically attainable by hand-written code. """
That last sentence is already true for FEniCS. The first sentence implies you'll show GPU results. Is that true? If so, say so; if not, remove it?
To be specific, what concrete advantages does firedrake offer? I think you should address this in your abstract.
Thanks, Patrick. I've modified the last paragraph to more clearly point out the advantage of the composition of abstractions and removed the sentence on performance entirely. The first sentence is intentionally vague: it would be great to have GPU results to show, but we'll have to see how far we have got by then. I'm not specifically claiming performance portability, because we can't quite live up to that at the moment. Florian
Hope this helps,
Patrick
participants (3)
-
Cotter, Colin J
-
Florian Rathgeber
-
Patrick Farrell