PROGRESS REPORT (8/17)

■ THIS WEEK  (8/11~8/16)

□ Phoenix 2.0

I had a simple test of wordcount on simics.

There is still a degradation after 64 nodes.

But, the speedup increases up to ~250 times (using 64 nodes).

[test_on_gems.pdf]

□ Booksim 2.0

I ran booksim with bimodal uniform traffic. 

using 64 vcs for each destination (8x8 mesh), the throughput is lower than original one.

When buffer size is 1, the decrease of throughput is huge.

When buffer size is getting bigger, the degradation get smaller, but never exceeds the result which does not use destination info.

I not quite sure, but I think it is because of the under-utilization of vcs.

[booksim_bimodal_voq.pdf]

□ sigMP

I presented the paper at the sigMP seminar.

A Novel Cache Architecture with Enhanced Performance and Security (MICRO 2008)

■ NEXT WEEK  (8/17~8/23)

□ Phoenix 

run all applications on gems with different configurations.

□ Booksim 2.0

Test more with different voq options.

Implement the outqueue router and analysis the results.