5 ms·

Haskell Arrays Accelerated with GPUs

24 points by dons 17y ago

ionfish 17y ago
Original PDF link: http://www.cse.unsw.edu.au/~chak/papers/accelerate.pdf http://www.cse.unsw.edu.au/~chak/papers/accelerate.pdf (15MB)
- miloshh 17y ago
  http://hackage.haskell.org/package/accelerate http://hackage.haskell.org/package/accelerate Unfortunately, it seems to be just a reference CPU implemetation, but if the GPU code is ever released, I'd guess it will be here.
masklinn 17y ago
Dons, could you please stop linking to scribd.com? It's slow, it's bloated and it's overall an unusable pain whatever the platform you're trying to run it on.
- gjm11 17y ago
  You say that as if he's doing it habitually. So far as I can tell from the HN submissions list, he's done it exactly once to date. (And, on one other occasion, submitted a PDF link directly -- which of course HN adds a scribd link to in case, er, someone feels that they haven't visited scribd often enough lately. At least I assume that's why people use scribd links; I can't think of any other reason.)
  - masklinn 17y ago
    He's also done it at least twice on Reddit in the past 3 days. Though you're right that my wording was pretty awful. I'd edit and rephrase, but apparently I can't edit.
- dons 17y ago
  The author's now posted a blog post on this talk. Sorry for breaking the news early. http://justtesting.posterous.com/haskell-arrays-accelerated-0 http://justtesting.posterous.com/haskell-arrays-accelerated-...
esessoms 17y ago
I'd be grateful if anyone could point me to the actual code for this. I'd like to give it a try for myself, but all I seem to be able to dig up is research papers.
- jerf 17y ago
  The Haskell community is fond of mistaking research papers (in PDF!) for documentation, it's quite likely you've found all there is. Just as they are also fond of mistaking a list of function names and type signatures for documentation. (On good days, they may deign to add a sentence about what the function does.) (Haskell guys, I love ya, but you have some of the oddest blind spots sometimes.)
  - DannoHung 17y ago
    I think it's because of overexposure to category theory. It gets to the point where they see a name, the see a type signature, they assume the developer isn't pathological, and figure out what they need.
  - esessoms 17y ago
    It's too easy to tease, but I will try to resist. But, seriously, they've been talking about this GPU work for upwards of a 1.5 years now. Usually we get something to play with, even if it's pretty badly broken (I'm thinking NDP here), so the lack of code here is pretty atypical. The accelerate link above is news to me, though, so that's helpful. Hope to see more soon! * EDIT (to back up my 1.5 years claim): earliest mention, May 15, 2008: http://www.cs.chalmers.se/Cs/Grundutb/Kurser/svh/Slides/may-15-2008-Obsidian-shortp2.pdf http://www.cs.chalmers.se/Cs/Grundutb/Kurser/svh/Slides/may-... Chak's own work, Jan 10, 2009: http://www.citeulike.org/user/iff/article/3872848 http://www.citeulike.org/user/iff/article/3872848 http://www.cse.unsw.edu.au/~chak/papers/LCGK09.html http://www.cse.unsw.edu.au/~chak/papers/LCGK09.html and, of course, accelerate ("reference implementation") was only released Aug 17, 2009: http://hackage.haskell.org/package/accelerate http://hackage.haskell.org/package/accelerate
- dons 17y ago
  It's on Hackage, as always, http://hackage.haskell.org/package/accelerate http://hackage.haskell.org/package/accelerate
tsuraan 17y ago
Stunning. I had always thought scribd was the most annoying thing ever, but they revamped their UI and completely proved me wrong. Now, to actually get to the pdf (I don't run flash), I have to go to "classic view", click the download as pdf option, login, walk through about three pages of demands that I "give back" to scribd, and then I can actually get my pdf. It's always impressive when bad gets worse, I guess. Why not just link to the actual content?

24 points by dons 17y ago

Original PDF link: http://www.cse.unsw.edu.au/~chak/papers/accelerate.pdf http://www.cse.unsw.edu.au/~chak/papers/accelerate.pdf (15MB)

miloshh 17y ago
http://hackage.haskell.org/package/accelerate http://hackage.haskell.org/package/accelerate Unfortunately, it seems to be just a reference CPU implemetation, but if the GPU code is ever released, I'd guess it will be here.

masklinn 17y ago

Dons, could you please stop linking to scribd.com? It's slow, it's bloated and it's overall an unusable pain whatever the platform you're trying to run it on.

gjm11 17y ago
You say that as if he's doing it habitually. So far as I can tell from the HN submissions list, he's done it exactly once to date. (And, on one other occasion, submitted a PDF link directly -- which of course HN adds a scribd link to in case, er, someone feels that they haven't visited scribd often enough lately. At least I assume that's why people use scribd links; I can't think of any other reason.)
- masklinn 17y ago
  He's also done it at least twice on Reddit in the past 3 days. Though you're right that my wording was pretty awful. I'd edit and rephrase, but apparently I can't edit.
dons 17y ago
The author's now posted a blog post on this talk. Sorry for breaking the news early. http://justtesting.posterous.com/haskell-arrays-accelerated-0 http://justtesting.posterous.com/haskell-arrays-accelerated-...

esessoms 17y ago

I'd be grateful if anyone could point me to the actual code for this. I'd like to give it a try for myself, but all I seem to be able to dig up is research papers.

jerf 17y ago
The Haskell community is fond of mistaking research papers (in PDF!) for documentation, it's quite likely you've found all there is. Just as they are also fond of mistaking a list of function names and type signatures for documentation. (On good days, they may deign to add a sentence about what the function does.) (Haskell guys, I love ya, but you have some of the oddest blind spots sometimes.)
- DannoHung 17y ago
  I think it's because of overexposure to category theory. It gets to the point where they see a name, the see a type signature, they assume the developer isn't pathological, and figure out what they need.
- esessoms 17y ago
  It's too easy to tease, but I will try to resist. But, seriously, they've been talking about this GPU work for upwards of a 1.5 years now. Usually we get something to play with, even if it's pretty badly broken (I'm thinking NDP here), so the lack of code here is pretty atypical. The accelerate link above is news to me, though, so that's helpful. Hope to see more soon! * EDIT (to back up my 1.5 years claim): earliest mention, May 15, 2008: http://www.cs.chalmers.se/Cs/Grundutb/Kurser/svh/Slides/may-15-2008-Obsidian-shortp2.pdf http://www.cs.chalmers.se/Cs/Grundutb/Kurser/svh/Slides/may-... Chak's own work, Jan 10, 2009: http://www.citeulike.org/user/iff/article/3872848 http://www.citeulike.org/user/iff/article/3872848 http://www.cse.unsw.edu.au/~chak/papers/LCGK09.html http://www.cse.unsw.edu.au/~chak/papers/LCGK09.html and, of course, accelerate ("reference implementation") was only released Aug 17, 2009: http://hackage.haskell.org/package/accelerate http://hackage.haskell.org/package/accelerate
dons 17y ago
It's on Hackage, as always, http://hackage.haskell.org/package/accelerate http://hackage.haskell.org/package/accelerate

tsuraan 17y ago

Stunning. I had always thought scribd was the most annoying thing ever, but they revamped their UI and completely proved me wrong. Now, to actually get to the pdf (I don't run flash), I have to go to "classic view", click the download as pdf option, login, walk through about three pages of demands that I "give back" to scribd, and then I can actually get my pdf. It's always impressive when bad gets worse, I guess. Why not just link to the actual content?