Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Talk 11: Nilesh Mahajan #11

Open
rrnewton opened this issue Aug 20, 2015 · 22 comments
Open

Talk 11: Nilesh Mahajan #11

rrnewton opened this issue Aug 20, 2015 · 22 comments

Comments

@rrnewton
Copy link
Contributor

No description provided.

@cgswords
Copy link

Put the communication on the slides!

@ccshan
Copy link

ccshan commented Aug 21, 2015

Watch where you're pointing with the laser pointer and be intentional!

@cgswords
Copy link

Kanor's syntax is hard to read for me. Is it possible to abstract over it maybe, or give pictoral explainations of these (clearly powerful) language constructs?

@samth
Copy link

samth commented Aug 21, 2015

Pictures!

@ccshan
Copy link

ccshan commented Aug 21, 2015

At the "Kanor Equivalent" slide I wonder whether the processing of sender and receiver pairs takes O(n^2) rather than O(n) time using Kanor, where n is the number of processors.

@ccshan
Copy link

ccshan commented Aug 21, 2015

Please enlarge the code snippets (perhaps inserting more line breaks)

@ccshan
Copy link

ccshan commented Aug 21, 2015

Why "Allgatherv" in the slide title?

@cgswords
Copy link

Also, please define the keywords? What does _at_ do? Maybe it does a lot of things?

@samth
Copy link

samth commented Aug 21, 2015

Showing the "formal syntax" is rarely useful -- show examples instead.

@ccshan
Copy link

ccshan commented Aug 21, 2015

I suspect that any point you try to make using a Formal Syntax slide is better made by example without showing any BNF production.

@samth
Copy link

samth commented Aug 21, 2015

@ccshan Jinx.

@cgswords
Copy link

Yes, that. Spend time on smaller examples before diving in, leading us through the language features we need to know before building the big ones.

@cgswords
Copy link

@samth I nearly typed "and the thing Sam is about to post about not showing the formal definition" ;)

@cgswords
Copy link

What guarantees do we have that communcation will always succeed?

@cgswords
Copy link

"Better than what"? You describe these properties and then say we can do better at runtime, but it's hard to tell what you're trying to improve there.

@cgswords
Copy link

Does the programmer define the communication knowledge of their application? Or is this a thing that crops up based on how you write the program, to be used as a general classification technique?

@cgswords
Copy link

For your algorithm, it may be better to gave two computation graphs, one where it holds and one where it doesn't, and talk through the algorithm with them, demonstrating how one fails and one succeeds.

@cgswords
Copy link

What sorts of compiler support are you shooting for? Why is it a problem?

@cgswords
Copy link

Color key for the tests, and maybe use fewer graphs.

@samth
Copy link

samth commented Aug 21, 2015

Have a thank you slide that includes your main points so that it's up during the question period.

@samth
Copy link

samth commented Aug 21, 2015

Feedback

  • Dan: avoid reading your slides
  • Sam: limit yourself to 5 bullets for the whole talk -- use pictures instead.
  • Ken: for each section, write down for yourself what your communication goal is for that section

@ccshan
Copy link

ccshan commented Aug 21, 2015

  • Dan: Not having text on your slides helps you not read them.
  • Cam: Not reading your slides helps you appear competent.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants