Workshop Feedback Summary
Workshop on Mixing Logic and DRAM:
Chips that Compute and Remember

Sunday, June 1st, 1997, 8:30am-5:30pm
Denver, Colorado

Organized as part of the 24th Annual International Symposium on Computer Architecture

The following feedback was collated by Liz Pennell. Please send corrections or comments to pennell@eecs.harvard.edu

Overall summaries:

Organize another workshop?

• Yes, but more focused would be better.

• Yes, but only if some more work occurs and the scope of the workshop is narrowed.

• Yes, but only if there is new research.

• Perhaps, if there are enough ideas to fill another workshop.

• Maybe two years from now.

Comments from attendees:

• Several people expressed satisfaction with the "open mike" session. Here's a sampling of the comments: "I think the last session (free mike) is excellent." "Open mike session was good." "Open mike very good."

• It seems that the inclusion of some papers in the roundtables without a formal presentation didn't work too well: "Not good that some papers were not presented. Either reject or make them present." "Some papers were not presented but were part of round tables anyway. That didn't really work--they didn't come into the discussion very much." Perhaps we should include these papers as posters in a separate session next time.

• There was some feeling that the lack of a formal set of proceedings would hurt the future impact of this workshop. Here are two specific comments:
  • "Publish the papers in some kind of proceedings for the workshop: without a proceedings that includes all papers, I'm afraid the authors will not be recognized."

  • "Without a proceedings, the workshop will be forgotten in history. The Web is a great way to prepublish papers, but for now, the above is unfortunately true."

• Several people complained about the cost of the workshop. "Should be cheaper." "Registration fee was very high." "Too expensive."

• "This workshop satisfied two issues. 1. An analysis of the technology (1st paper) and 2. Opened the discussion to new, interesting architectures. I would like to have seen, in addition, an overview of the issues, technology, etc. to set the stage for the later presentations."

• "Patterson and Smith did a nice job of running the workshop, as expected. I was a bit disappointed in some of the presentations--especially the SIMD/DIM stuff. I would have liked to see us get down to real issues: cost tradeoffs, process differences, architectural implications, and have some more tutorial-ish stuff followed by debates. But it was fun."

• "Very good workshop overall. Not that controversial. Last session was very good though."

• "Overall the workshop was good. However, some papers didn't clearly belong to this workshop."

• "Good workshop. Bravo!"

• "Learned a lot, good use of time. Liked the long breaks, speaker panels at end of sessions, open mike."

• "The room was small, we could have a room in which we have elevated back seats (so as you can guess, I was sitting in back!)"

• "Logic and DRAM integration is an emerging research area, with researchers coming from many perspectives (EE background, computer architecture, software). It is very important to encourage cross-pollination between these groups, and between industry and academia. I recommend strongly that the workshop should be offered again."

• "If a paper arouses a lot of questions/discussions we should allow more time for this. Presentation time could be reduced."

• "I loved the format of 3 talks, panel discussion, and breaks."

• "A more focused topic might be better in the future."

• "Surprisingly little novel content...maybe IRAM is a packaging 'integration' concept more than it is an architectural concept."

• "Segment talks into technology, uprocs, mp's, appl. specific. We need to understand tech & manufacturing issues better to design a better arch."

• "It would have been useful to divide it into general-purpose processing (how DRAM affects memory hierarchies) and special purpose processing (how DRAM and logic combined can speed up specific, special tasks). Both have considerably different fields of interest."

"Future workshop topics: limit scope to focus work, to limit attendance, to focus the expertise of the attendees.
• Microprocessor w/ integrated DRAM: how should commodity processors leverage DRAM.
• Single chip systems with integrated DRAM: organization of single memory chip systems.
• High-performance DRAM: macro organization optimizing for something other than identity (likely this is a circuit-technology workshop that does not compliment ISCA very well).
• SIMD/Vector processing w/ integrated DRAM: keep the SIMD people away from the rest of the world. (Has anything worthwhile occurred when SIMD people were in a workshop with general purpose microprocessor people?)

In another year, enough new work will occur in the area of IRAM that it will support true workshops in all of these topics."

Return to workshop web page