Wish my programming coworkers knew something about work stealing. But they only know about work giving.... i'm drowning in work ... please steal something from me.
this talk and the 3d imaging thing were great. i don't necessarily agree with all the points in either, but the format was amazing compared to other talks. some good things in these two talks: informative concise articulate coherent/sequential (says a thing, builds on it, defines an aditional thing, either joins them or elaborates) the only issue in this one at a presentation level was not repeating questions during Q/A.
that continuation stealing thing is exactly what I need! The work should run just like a big for loop on single core, amazing stuff! Is there any implementation of this by chance? Need to run it on 32bit RISC-V embedded processors.
In the recursive fib example, am I correct that it is doing the full (non-optimal, expensive, duplicate-work) version of fib calculation as an example? Does tbb provide a way to know that it has already calculated fib(2) rather than re-evaluating it?
Wish my programming coworkers knew something about work stealing.
But they only know about work giving.... i'm drowning in work ... please steal something from me.
this talk and the 3d imaging thing were great. i don't necessarily agree with all the points in either, but the format was amazing compared to other talks.
some good things in these two talks:
informative
concise
articulate
coherent/sequential (says a thing, builds on it, defines an aditional thing, either joins them or elaborates)
the only issue in this one at a presentation level was not repeating questions during Q/A.
Great video for understanding work-stealing. Very neat explanation which does not really require any c++ knowledge.
Amazing talk!
that continuation stealing thing is exactly what I need! The work should run just like a big for loop on single core, amazing stuff! Is there any implementation of this by chance? Need to run it on 32bit RISC-V embedded processors.
In the recursive fib example, am I correct that it is doing the full (non-optimal, expensive, duplicate-work) version of fib calculation as an example? Does tbb provide a way to know that it has already calculated fib(2) rather than re-evaluating it?
no