Paper permissions on cse.unsw.edu.au

[Alex Mason]

I'm trying to access the paper linked from the repo site at ​Data Flow Fusion with Series Expressions in Haskell, but it doesn't appear to be available. Is there a version available elsewhere, and could that link be updated?

Cheers,

Alex

[Amos Robinson]

There's another copy here: http://www.cse.unsw.edu.au/~amosr/papers/lippmeier2013flowfusion.pdf

I'll let Ben deal with either updating the link or fighting with CSE

--
You received this message because you are subscribed to the Google Groups "Haskell Repa" group.
To unsubscribe from this group and stop receiving emails from it, send an email to haskell-repa...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Alex Mason]

Thanks Amos! I’d always wondered what happened to the university hosting when you leave the university…

Really looking forward to giving this a read, we might have a project where this would be useful at work (along with all the classical Repa array fusion stuff from Repa 3), used in conjunction with distributed processing.

Cheers,
Alex

[Amos Robinson]

The problem with data flow fusion is it only supports a few combinators and it's hard to add new ones. My new system, machine fusion, solves this but we are still working out the best way to present it. But I haven't thought about any sort of parallelisation for it yet. The secret sauce is to treat each combinator as its own sequential process, and then all processes are run concurrently. Then you "just" choose a sequentialisation of the concurrent network, ending up with a single sequential process.

Depending on your use case, if you don't use inputs or intermediate results multiple times, and only have one output array, the staged computation paper "Stream fusion, to completeness" is worth reading. They claim to fuse "any combination of combinators" but that's a bit of a fib since using values multiple times duplicates work.

[Alex Mason]

Sounds interesting, I’ll take a look. We may be in a situation where repeated work isn’t a problem (or where caching results isn’t a problem) so we’ve got a bit of freedom to use the best results.

We should catch up next time I’m in Sydney, I’d love to hear your thoughts this project.

Cheers,
Alex

[Ben Lippmeier]

On 9 Feb 2017, at 12:21 AM, Alex Mason <axm...@gmail.com> wrote:

I'm trying to access the paper linked from the repo site at ​Data Flow Fusion with Series Expressions in Haskell, but it doesn't appear to be available. Is there a version available elsewhere, and could that link be updated?

Where did that link come from? I tried googling for the paper title and got Manuel’s version, which works. The version accessible from my home page is hosted on my own VPS, i.e. http://benl.ouroborus.net/papers/2013-series/flow-Haskell2013-rev1.pdf

Ben.

[Amos Robinson]

I think it's from here:

 http://repa.ouroborus.net/#DataFlowFusion

[Ben Lippmeier]

Thanks, I’ve fixed this, and the others.