Speculation in a Shared Cache

Andrew Faulring faulring@cs.cmu.edu	Yan Karklin yan@cs.cmu.edu

CS 15740 Project, Fall 2000

Project Web Page

Milestone Report: http://www.cs.cmu.edu/afs/cs/user/faulring/15740/project/milestone.html
Project home page: http://www.cs.cmu.edu/afs/cs/user/faulring/15740/project/index.html

Summary

Overall, the project has proceeded smoothly, and we have made good progress. However, we underestimated the time it would take for us to understand and modify the existing cello simulator code.

Major Changes

No major changes, except that our suite of benchmarks has changed. Greg Steffan has made available an alternate set of much more simple benchmarks. We would still like to run the standard benchmarks if time permits.

Accomplishments So Far

We have updated the simulator to support a single TLS shared cached. This involved modifying the cache and memory system objects. At this point we can run the simple benchmarks that Greg has provided to us.

Meeting Our Milestone

As mentioned before we have changed the type of benchmark that we will run. We are able to successfully run the benchmarks provided by Greg, but we need to meet with him to help us interpret the results.

Surprises

In general, no surprises. Though it took longer to modify the code than we had expected.

Revised Schedule

Week	Task
19 Nov	Meet with Greg to discuss how to interpret benchmarks (Andrew & Yan) Generate a set of test cases to ensure that our modifications are working properly. Greg has offered to show us how to run the simulator in a special mode so that we can thoroughly test our modifications. (Yan) Run new benchmarks and figure out how to interpret results. (Andrew)
26 Nov	Finish running all tests. (Andrew) Write report. (Yan & Andrew)
3 Dec	Turn in Report on Monday, 4 December

Resources Needed

We still need to meet with Greg to discuss how to interpret the raw benchmark output.