PROGRESS REPORT

■ THIS WEEK  (7/21~7/27)

□ Phoenix 2.0

I found that thier scheduler does not work actually.

so they tested using threads which were not fairly distributed to cores.

the threads were mapped core_id = thread_id

therefore, when # of threads <=8, they used only one cpu.

when # of threads=16, they used 2 cpu, when # of threads=32, # of cpu is 4 and so on.

I fixed the code to distribute the threads to cores fairly.

the result is very different and the performance is also very higher than before.

(new results)

□ Booksim 2.0

we had a discussion about the traffic patterns and the expected results.

simple traffic pattern like neighbor is relatively easy to predict but others are not so easy.

we think that we have a misunderstanding of the uniform traffic and need more reseach.

(files of 4 traffic patterns)

■ NEXT WEEK  (7/28~8/3)

□ Phoenix 

re-plot the graph of some other applicatoins

□ Simics & Gems

I think now i should try to run phoenix on Gems.

□ Booksim 2.0

discuss more about the traffic pattern and expeted result.

test more with different config files(using different delay).