In Homework 2 function __builtin_assume_aligned is to be used in
kernel code. That requires CUDA 11.2 or later to compile, I recently
learned. Until a few minutes ago the lab machines all had CUDA 11.0.
I've updated some of the machines to 11.3, including those with CC 7.5
16 April 2021, 14:54:25 CDT
For next week (19 April) read the two papers as assigned
in Homework 3. Focus on the two papers
mentioned at the top, additional papers are listed as background.
13 April 2021, 14:23:49 CDT
For the homework assignment try to start with the machines with with
CC 7.5. These machines will collect the performance counter needed for
the assignments. Most of the other machines are capable of doing so
but have not been set to do so yet. Running code on these machines
will result in a unhelpful error message. (That itself is an error,
there should be a helpful error message.)