search

tonyyxliu / CUHKSZ-CSC4005

Project Materials for CUHK(SZ) Course CSC4005: Parallel Programming
MIT License
79 stars 31 forks source link

Baseline Concern about Project3-Task3 #52

Closed I-am-Future closed 1 year ago

I-am-Future commented 1 year ago

Hi TAs, I found that the given Project3-Task3's baseline for 1 worker is ~37000ms. But I found that I tried many many times, running the pure sequential code is ~42000ms. My odd-even sort implementation for 1 worker is also around ~41000-42000ms. Why the 1 worker baseline could be significantly faster than sequential? No matter there are additional overheads? Can you please give me some improvement hints? Thank you!

MeloYang05 commented 1 year ago

Hello! I think 41000-42000ms is acceptable, don't be too worry. I think when I conduct the experiment, the machine is relatively free, so the performance will be a little bit better.

I-am-Future commented 1 year ago

Yeah, I found that may be the problem. I ran it today 1 a.m., speed is about 39000-40000ms.... But I am not sure whether a in-node communication would be affected significantly? because I think when I am running the program, one node is all assigned to me?

MeloYang05 commented 1 year ago

Yeah, I found that may be the problem. I ran it today 1 a.m., speed is about 39000-40000ms.... But I am not sure whether a in-node communication would be affected significantly? because I think when I am running the program, one node is all assigned to me?

I am not sure, @tonyyxliu may be more familiar with the cluster. In my optinion, I believe there are some load distribution machanism of the cluster.