John Biddiscombe
John Biddiscombe
Issues #3348 and #3553 should not be closed as they are under consideration at the time of writing.
Agreed. I even floated the idea some time back of switching out the entire scheduling loop so that we could do away with the whole task creation and queuing using...
just to bump this. I ran a full set of tests on my laptop a few days ago and noticed this was one of the few that failed.
Not a timeout, just a regular fail. I'll look into it when I have a moment ``` test 880 Start 880: tests.unit.component.migrate_component 880: Test command: /bin/python "/home/biddisco/build/hpx/bin/hpxrun.py" "/home/biddisco/build/hpx/bin/migrate_component_test" "-e" "0"...
ooh. I have already a patch for that. I will enable it on this branch and report back
This is also connected to the idea of N-ary tasks. In that case, promoting the callable object (as suggested in #3348) to be created earlier and storing things like counters...
Note that in the examples I pasted in above, it would actually be nice to have this functionality with `hpx::apply` since we don't need to return the futures and wait...
Internally `for_loop` is creating tasks and then waiting on the futures. I could use the `for_loop` in my example (and I will for the time being), but I wanted to...
Yes. I am aware of the existence of the latch - this is what I want to use. My question is about how we could nicely fire off tasks and...
As I wrote above. I would like to mix tasks that are 'important' and should be counted in the latch, with tasks that are not important and can just run...