Hi,
Just in case some of you haven't seen this already: http://blog.chromium.org/2010/12/new-crankshaft-for-v8.html
At a high level, they have ideas similar to ours, but then, the devil is usually in the details for this kind of work.
Bruno
Afficher les réponses par date
It does sound *extremely* similar to the ideas I've put forward... And they have a working build, apparently.
I'm not sure if it is exactly what I wanted to do, because they don't have any mention of things like type inference. They could essentially be describing what the Java HotSpot VM is doing. We would have to checkout their bleeding edge branch to have some idea.
Marc, do you know these two people: Kevin Millikin and Florian Schneider?
- Maxime
Hi,
Just in case some of you haven't seen this already: http://blog.chromium.org/2010/12/new-crankshaft-for-v8.html
At a high level, they have ideas similar to ours, but then, the devil is usually in the details for this kind of work.
Bruno_______________________________________________ Tachyon-list mailing list Tachyon-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/tachyon-list
On 2010-12-07, at 4:19 PM, chevalma@iro.umontreal.ca wrote:
It does sound *extremely* similar to the ideas I've put forward... And they have a working build, apparently.
I'm not sure if it is exactly what I wanted to do, because they don't have any mention of things like type inference.
"3. An optimizing compiler which recompiles and optimizes hot code identified by the runtime profiler. It uses static single assignment form to perform optimizations such as loop-invariant code motion, linear-scan register allocation and inlining. The optimization decisions are based on type information collected while running the code produced by the base compiler."
They could essentially be describing what the Java HotSpot VM is doing. We would have to checkout their bleeding edge branch to have some idea.
I think we should cancel the Tachyon project... I'll give a call to the Mozilla people to let them know there's no point in coming to meet with us. ;-)
Marc, do you know these two people: Kevin Millikin and Florian Schneider?
- Maxime
I've corresponded with Kevin. I think you can contact him directly if you have questions about V8.
Marc
I think we should cancel the Tachyon project... I'll give a call to
the Mozilla people to let them know there's no point in coming to meet with us. ;-)
Sarcasm carries very poorly by e-mail.
Since I'm being asked to come up with something major and novel by the end of my Ph.D., perhaps you should understand why these announcements make me uncomfortable. The novelty factor kind of goes away if somebody beats me to the punch by a couple of months/years.
I've corresponded with Kevin. I think you can contact him directly
if you have questions about V8.
Should I talk to him directly or to their HR department if I want to land a job at Google?
- Maxime
On 2010-12-07, at 17:13 , Maxime Chevalier-Boisvert wrote:
I think we should cancel the Tachyon project... I'll give a call to
the Mozilla people to let them know there's no point in coming to meet with us. ;-)
Sarcasm carries very poorly by e-mail.
Since I'm being asked to come up with something major and novel by the end of my Ph.D., perhaps you should understand why these announcements make me uncomfortable. The novelty factor kind of goes away if somebody beats me to the punch by a couple of months/years.
I've said this before, but optimistic optimizations are not really a novelty. Most JITs will do that to a certain extent. I think the specific (optimistic) optimizations you'll come up with for JS will be your contribution. We're not building a product, we're exploring research ideas to see how far we can push the envelope. There are plenty of ideas to try, I'm not really worried about V8 at the moment.
Bruno
I've said this before, but optimistic optimizations are not really a
novelty. Most JITs will do that to a certain extent. I think the specific (optimistic) optimizations you'll come up with for JS will be your contribution. We're not building a product, we're exploring research ideas to see how far we can push the envelope. There are plenty of ideas to try, I'm not really worried about V8 at the moment.
Your comment does reassure me. I suppose that there are indeed plenty of ideas to try, and that I probably shouldn't be overly attached to that very specific one, and that my Ph.D. doesn't have to fail because someone implemented something similar to what I envisioned doing my thesis on. Once Tachyon is bootstrapped, it should be possible to extend JavaScript in a quasi-infinite number of ways, making the exploration of even more ideas possible... Parallelism, numeric computations, distributed computing, serialization, and then all the rest of computer science.
Anyways. On the topic of ideas, and collaborations. So far, we've actually been mostly private about the project. Apart from our talk at CASCON, and the correspondences Marc has had with Google people, our project has little public presence. Do you think that perhaps it might soon be time for us to have a blog, like Mozilla and Google do? Going more public with the project might help me alleviate my own fears of idea theft (if there is such a thing), and might generate more interest in the project (e.g.: from talented students, among others...).
I think that basically, most of what we need for a web-presence is a Wordpress blog and a public GIT repository. Both of these shouldn't be a huge pain to set-up. Perhaps we could look into those some time next semester.
- Maxime
On 2010-12-07, at 17:58 , Maxime Chevalier-Boisvert wrote:
I've said this before, but optimistic optimizations are not really a
novelty. Most JITs will do that to a certain extent. I think the specific (optimistic) optimizations you'll come up with for JS will be your contribution. We're not building a product, we're exploring research ideas to see how far we can push the envelope. There are plenty of ideas to try, I'm not really worried about V8 at the moment.
Your comment does reassure me. I suppose that there are indeed plenty of ideas to try, and that I probably shouldn't be overly attached to that very specific one, and that my Ph.D. doesn't have to fail because someone implemented something similar to what I envisioned doing my thesis on.
Again, the idea of optimistic optimizations is not new. It's what you do with it that will be.
Once Tachyon is bootstrapped, it should be possible to extend JavaScript in a quasi-infinite number of ways, making the exploration of even more ideas possible... Parallelism, numeric computations, distributed computing, serialization, and then all the rest of computer science.
I meant there are plenty of ideas to try for compiling programs written for the *current* JS spec, but these ideas are fun too :)
Anyways. On the topic of ideas, and collaborations. So far, we've actually been mostly private about the project. Apart from our talk at CASCON, and the correspondences Marc has had with Google people, our project has little public presence. Do you think that perhaps it might soon be time for us to have a blog, like Mozilla and Google do? Going more public with the project might help me alleviate my own fears of idea theft (if there is such a thing), and might generate more interest in the project (e.g.: from talented students, among others...).
I think that basically, most of what we need for a web-presence is a Wordpress blog and a public GIT repository. Both of these shouldn't be a huge pain to set-up. Perhaps we could look into those some time next semester.
1 - If we make things public too early, people who stumble upon the project have a tendency to dismiss it as a toy and look elsewhere. I think we'll reach the right time for going public soon, but that we're not quite there yet. 2 - Setting up the resources takes time, and it seems that the time we have now is better spent on something else.
Just my opinion of course.
Bruno
-
Bruno Dufour -
chevalma＠iro.umontreal.ca -
Marc Feeley -
Maxime Chevalier-Boisvert