Updated the Final Exam with more information
on branch divergence needed for Problem 2. Also fixed a formatting
problem in Problem 3 which resulted in a part (b) that didn't
seem to ask anything.
Grades for the last two homeworks will be sent out
by early next week, certainly before the final
exam is due.
28 April 2023, 11:17:20 CDT
The next homework has been assigned,
due 1 May 2023, late submssions accepted until Wednesday.
This involves reading papers. Questions in the
take-home final will also be about those papers.
26 April 2023, 13:42:07 CDT
The homework will be copied on Friday morning.
For examples on how to perform reductions within
a block and within a warp look at the code in
cuda/intro-vtx-transform/vtx-xform-sum.cu. Probably
the best one to use is sum_wp. Note that the
second argument is the lg of the number of values. Setting
n_vals_lg=2, for example, will perform the sum four
groups of consecutive threads. So within a warp eight different
sums is computed.
19 April 2023, 18:44:05 CDT
The next homework handout has been updated
with information on how to run and interpret the microbenchmarks.
19 April 2023, 11:27:59 CDT
The next homework has been assigned,
it is due 26 April 2023.
14 April 2023, 15:53:09 CDT
The updated classroom demos have been pushed to the git repo.
For those who rather not git pull, here
small matrix code presented
in the last few lectures. Its path in the repo
is ./cuda/intro-vtx-transform/vtx-xform-size.cu.
21 March 2023, 17:18:56 CDT
Homework 1 will be collected Wednesday evening, 22 March 2023.