PROGRESS REPORT (8/3)

■ THIS WEEK  (7/28~8/3)

□ Phoenix 2.0

I analyzed the result of histogram.

average task processing time are increases as # of cores increase.

but, it does not increase dramatically as compared with wordcount.

(new_result_histogram.pdf)

□ Booksim 2.0

I read the book, and drew each traffic patterns.

I'm not sure this is correct, but there are some strange things.

Bit complement (D_i = ~S_i) and Bit reverse (D_i = S_b-i-1) look exactly the same.

for Shuffle(D_i = S_i-1 mod b) and Bit rotation(D_i = S_i+1 mod b), only the traffic direction is different.

and when mesh size is 4x4, transpose and neighbor are the same.

(traffic_patterns.jpg)

we are checking these features with the results.

□ sigMp

no sigMp seminar this week.

■ NEXT WEEK  (8/4~8/10)

□ Phoenix 

run a couple of applications more and compare the result with wordcount.

□ Booksim 2.0

test and compare the result, and analysis the characteristic of each traffic pattern.

□ sigMp

Jaehong asked me to cover for his turn next week (8/14).

so please introduce me a paper for the presentation.