Course Announcements

23 January 2006

Dear ENEE 759A Student,

By receiving this email, you are receiving confirmation that you have been added to the mailing list for ENEE 759A. I look forward to seeing you in class on January 26th, and getting to know you better throughout the Fall semester.

regards,
Don.

1 February 2006

Tomorrow, I need to hold my office hours 1/2 hour earlier, from 2:30-3:30. Sorry for the inconvenience.

6 February 2006

Last Thursday, I handed out Homework #1. This homework is due in class February 16. I also mentioned there is a problem with the compiler, "alcc," which is necessary to complete the homework. To fix this problem, I have given all of you access to a Glue machine where the compiler is properly installed. That Glue machine is called "ruggieri.eng.umd.edu" (the alias for this machine is "rug.eng.umd.edu"). By late this afternoon, you should be able to log into this machine using your normal Glue username and password. If by tomorrow you can't access this machine, please let me know. On this machine, include "/projects/alewife/tools/sun4" into your path environment variable (instead of "/software/stradivari4/class/759a/tools/sun4," as instructed by the lab handout). You should now be able to execute "alcc" correctly.

Please only use "ruggieri" for compilation (i.e., for running alcc). Homework #1--and other homeworks--require you to run the Alewife multiprocessor simulator, nwo, as well. This software is very compute-intensive, but should work from any Glue machine. Since there's only 1 machine for compilation, it's important that none of you run the nwo simulator on ruggieri; otherwise, the load will become too high and students will not make sufficient progress.

7 February 2006

Apparently, most of you do not have access to ruggieri at this time. The ECE Help Desk is looking into this, and I'll let you know when it gets resolved (unfortunately, I don't have direct control over this).

8 February 2006

At this point, your account on ruggieri should be active. Please try your login ASAP, and let me know if you still can't get in. If you can login, then you can follow the instructions I gave before to get access to the alcc compiler. Remember, ruggieri is only for running the compiler; you should use any other Glue machine to run the NWO simulator.

Due to the delay in getting things up, I am extending the deadline for Homework #1 to Tuesday, February 21.

9 February 2006

Today, I put a new paper on the web reading list--this paper describes the NYU Ultra shared memory machine. When you download this paper, you'll actually get a tarball of two papers. You should print and read the paper in the "main.ps" file. The other file is a longer tech report (which you're welcome to read, but you can get the gist of the idea from the shorter paper).

9 February 2006

I have a conflict on Thursdays 3-4, so I'd like to change to Thursdays 2-3. I'll keep my hours on Tuesdays from 3-4. Sorry for any inconvenience.

28 March 2006

Here's the order of presentations for the term projects resulting from the drawing we had in class today:

1. Matt
2. Inseok
3. Bhuvan and Amit
4. Katie
5. Chenjie
6. Wanli and Meng-Ju
7. Nghi
8. Hojin
9. Xu
10. Joe
11. Rakesh and Alessandro

We will have roughly 3 presentations per day, beginning May 2nd. You should plan 25 minutes for your presentation, including questions. If you haven't already, I would highly recommend starting on the project.

3 April 2006

Some of you will be doing projects using NWO to gather multiprocessor traces. (For example, you might build a cache coherence simulator, and then feed the NWO traces into your simulator to study different coherence protocols). To help those of you doing such projects, I have provided a tracing facility in NWO. To get access to this facility, simply start the NWO simulator just like you would for your homeworks, and then load the memory trace functions into the simulator:

(load "/software/stradivari4/class/759a/nwo-trace/trace-mp-data.so")

With these functions loaded, you can type "(trace-processors)" and this will turn on tracing. Then, when you run any program in the simulator, the simulator will generate a trace file during the run (beware, the trace files can be *very* large). These trace files can then be used for your projects.

It's up to you to decide which benchmarks to trace and study. I have also prepared some SPLASH benchmarks in the course directory under "pgms". Here, you will find 5 benchmarks from SPLASH:

barnes: This is a hierarchical N-Body simulation
blu: Blocked LU factorization
gauss: Gaussian elimination
mp3d: rarified particle flow simulation
water: n-squared particle-particle simulation

In each sub-directory, you will find source code, and sample inputs and outputs. If you are interested in using some of these benchmarks, take a look at the directories, and if you need help on how to run the benchmarks, let me know.

Hopefully, this will help some of you get started on your projects. Anyone who needs help getting started should meet with me.

5 April 2006
There was some temporary outage on ruggieri today, but it should be fixed now. Let me know if any of you continue to have problems getting access to the alcc compiler. As a reminder, please don't run anything on ruggieri except for the compiler. For example, you should run NWO on another Glue machine.

12 April 2006
During tomorrow's lecture, we will complete the interconnection network topic. As I've mentioned before, the remaining lectures before project presentations begin will be structured around research papers. For the second half of tomorrow's lecture, we'll start looking at Thread Level Speculation (TLS) covered by papers [17] and [18] on the reading list web page. These papers are available on-line, and hyperlinks to the papers appear on the web page. It's important that you read these papers; at the least, you need to bring copies of these papers to class. Otherwise, the lecture won't make much sense.

18 April 2006
As mentioned in previous emails and in class, you will make a 25-minute presentation (including questions) on your term project; the presentations will follow the ordering of numbers drawn in class. In addition, you will hand in a writeup of your project. I will accept project writeups at any time up until Monday May 15. Since that date is after our last meeting, you can hand the report to me in person at my office, or place it in my mail box in the faculty mailroom. Each group should hand in 1 report, and the report must be a hardcopy.

15 May 2006
Just a reminder that your final project reports are due today. You can hand them to me in my office or place them in my mail box. (You must hand in a hardcopy). Best of luck on finals, and enjoy your summer!