Handouts: Syllabus and massively-slugfull computing.
Topics: Overview and an introduction to the MasPar.
Comments: Documentation can be found on ganesha in /usr/maspar/doc, or you can read an excellent tutorial (please see me if you'd like a hardcopy to xerox). You can also find a few things in my ~rph/220 directory, and MasPar manuals in the technical library (please do not remove).
Homework: For Tuesday: Chapter 1 problems 1-4, 6, 8. For Thursday: On the Maspar, program either matrix multiplication (for 64 by 64 and 1024 by 1024 arrays) or prefix sum (4096 and 1M elements). Use doubles for both projects, and time your programs.
Topics: Finished up MPL, shared memory, data parallel, list doubling, PRAM models.
Comments: You can find example mpl code with a makefile and timer routine in ~rph/220/sort directory. The linear commication routines are in /projects/compbio/ultimate/src/maspar, and documentation is on the WWW . I also have a copy of the draft document for MasPar C (written before the compiler group was eliminated), and a technical report on the Blitzen processor, a successor to MPP from MCNC and NCSU (that may not have gotten beyond chips), and a link to the MGAP page at Penn State.
Homework: Occam fine-grain matrix multiply for Thursday.
Topics: Data parallel languages, *Lisp, C*, basic algorithms, CSP & Occam.
Comments: Here are the MPL environment settings I use.
Handouts: SDSC Cray and Paragon accounts
Homework: Individually assigned: Chapter 2: 1-3, 12, 14, 17. Everyone: Chapter 3: 1, 2, 5.
Topics: MasPar performance on matrix multiply and prefix sum. Problem decomposition.
Homework: For 4/18: 3.19, 3.20, 4.7, 4.8. For 4/22: Communication timing on various machines (one per person).
Topics: Finish chapter 3. Floorplanning: branch and bound searching.
Comments: You now all have non-zero allocations on both the Cray and the Paragon. There are users guides on the SDSC WWW pages. Because of a conference, final project will be due at noon on June 10th, not June 12th as originally scheduled.
Homework: Christian volunteered to do connected components on the MasPar.
Topics: Chapter 4 (performance, analytical models), preliminaries for FFT.
Homework: Use DFT to multiply 2xx+3x-7 and 5x+2. Do exercise 4.6.
Topics: Results of communication experiments. FFT.
Comments: New communication results for Thursday.
Handouts: MPI Tutorial selected transparencies.
Homework: MPI programming for Thursday (see /projects/kestrel/programs/mpich --- I've compiled the sun4 and IRIX versions of the code).
Topics: Finish up FFT, MPI.
Handouts: FFT
Homework: Paper Linda program
Topics: Linda, CC++.
Comments: Project proposals due. Here's the Linda group home page
Homework: Paper CC++ program.
Topics: MPI program discussion. CC++. Start FM.
Homework: For Thursday: Paper CC++ and (FM or HPF) programs. Problems 7.8 and 7.9.
Topics: FM, HPF.
Comments: I've installed another MPI implementation called LAM , specifically designed for networked workstations, that includes an X-based debugger. To run xmpi, you will need to add /projects/kestrel/programs/lam/bin to your path. You may wish to merge the application defaults into your xrdb. Start thinking about what language you want to go find a little information on. Possibilities include PVM, pC++, Mentat, Strand, PCN, and Concurrent C, and there are many many others.
Topics: Program discussion.
Comments: C90 vs T3D. Be sure that you're actually running your programs on the T3D, not on the C90: C90 time is much more expensive, and it looks like nearly all of our 200 service units have gone to C90 time rather than T3D time.
Handouts: Moldovan article.
Homework: Short paper on your project due today.
Topics: SIMD Algorithm Mapping
Handouts: Kestrel information.
Homework: For the 23rd, two matrix algorithm mappings of sequence comparison different from the one given in class. Even better, do several more including a few incorrect mappings, giving reasons as to what is incorrect.
Topics: SIMD Mapping, Kestrel.
Homework: Programming language overviews this week. For Thursday the 30th, a Kestrel program for poly mult, convolution, weighted average, poly eval, sequence comparison, or something else.
Topics: Matrix multiplication mapping, Strand, Sisal, C-Linda.
Topics: PCN, Concurrent C, pC++.
Homework: Kestrel program due.
Topics: Project presentations and various topics. Be sure to have a xeroxed 1-page handout for everybody on your project. Weikai and Kwansik.
Topics: Project presentations and various topics. Christian, Joe, and Marcelo M.
Homework: Project reports due at 2pm.
Comments: Fifth Annual Parallel Processing BBQ at Natural Bridges State Park at 3pm. Here are descriptions of the Park's geology.