These are chat archives for uwhpsc-2016/uwhpsc-2016

10th
May 2016
djg25
@djg25
May 10 2016 00:36
Hi Chris, for some reason when I run "make test", it only executes the timing tests and not the tests I wrote to compare my computed values of the integral with the values returned by numpy.trapz and numpy.simps
Chris Swierczewski
@cswiercz
May 10 2016 19:07

Office Hours - Start

(Sorry for getting here late.)
@djg25 Did you write your tests in a TestX class? If so, maybe the timing functions returned an error and caused the test script to quit before running the tests?
Nathan Schiele
@nschiele
May 10 2016 19:10
blob
I'm getting some pretty wild variations in my plots of timed_simps_parallel_chunked
Am I doing something wrong?
Chris Swierczewski
@cswiercz
May 10 2016 19:12
No. I mentioned in class today that because there are hundreds of other people using SMC as well that, periodically, a thread may have to wait for a CPU to become available.
If you run your timing code several times you should see some characteristic behaviors, though.
Sometimes you just have to sift through the noise.
Nice, plot!
Nathan Schiele
@nschiele
May 10 2016 19:14
Yeah - I remember you saying that. Can it even effect me running though a bunch of tests?
Chris Swierczewski
@cswiercz
May 10 2016 19:15
What do you mean by "a bunch of tests"?
Repeated generations of the plot?
Increasing int repeat to something larger?
Or running the test suite?
Nathan Schiele
@nschiele
May 10 2016 19:17
I mean, when I run timed_simps_parallel_chunked in a for loop. Those iterations are running fairly quickly sequentially. Will the effect from other students show up on different iterations of the for loop? (my guess is the answer is yes, just want to confirm).
Hugh Krogh-Freeman
@hughkf
May 10 2016 19:18
@cswiercz Just to confirm, AMATH 583 students don't need a proctor for the final because it's online?
Chris Swierczewski
@cswiercz
May 10 2016 19:19
@nschiele In short, most likely. It's difficult for a thread to stake an exclusive claim on a core. You should see these discrepancies even if you ran the code on your personal machine. Your OS has a bunch of background processes running. On top of that, the hardware scheduler doesn't necessarily assign instructions in the same order with each run of the code.
@hughkf No proctor needed.
Hugh Krogh-Freeman
@hughkf
May 10 2016 19:23
@cswiercz Awesome
Chris Swierczewski
@cswiercz
May 10 2016 19:28
A word of warning: uwhpsc-2016/homework3#11
Nathan Schiele
@nschiele
May 10 2016 19:33
Here's a nitpicky question - should I include a legend for which colors correspond to which num_thread argument, or is just having the correct colors sufficient?
Chris Swierczewski
@cswiercz
May 10 2016 19:36
Well, having a legend without identifying the meaning of each color is pretty meaningless. :)
The legend should identify which color corresponds to which number of threads.
Hugh Krogh-Freeman
@hughkf
May 10 2016 19:53
@cswiercz My grade for homework #2 is "-". Is that simply because it's currently being graded? :worried:
Chris Swierczewski
@cswiercz
May 10 2016 19:54
Yes. It takes a while to grade homework.
But it should be ready by Thursday.
Hugh Krogh-Freeman
@hughkf
May 10 2016 19:55
Whew! That's a relief
And completely understandable
Chris Swierczewski
@cswiercz
May 10 2016 20:02
No worries. I'll make an annoucement when grades are ready.

Office Hours - End