Don’t forget the reading group tomorrow. Two quick questions about the papers being presented:
(1) What is the differnce between for-loops and while-loops in the roccc system? If no difference, are for loops converted into while loops or the other way around? If there is a difference, what is the major difference?
(2) Compared to ROCCC, what major philosophical difference is highlighted by the CHIMPs effort?
Ron
Andy’s Workload:
1. RAW Paper – Help Will with Software and Paper writing
2. IPDPS Paper – Test Multiple Disk Nodes and then Multiple Blast Nodes
That will take me all week I am sure.
Last week
Rewrote the bcast core, this one used less fifos and should have less bugs
This week
Synthesize bcast_core with updated router
Start reduction (integer add)
Last week:
Read up on conjugate gradient and sparse matrix multiplication
This week:
Working on the code for conjugate gradient
Last week:
Developed the network device driver over our AIREN network;
The driver can now support “ping” between two nodes
This week:
Fully test this driver with “telnet” or other TCP/IP-based program
Benchmark our AIREN network under Linux
Gig ethernet support for the cluster
Last week: Worked on HW and Literature survey for MD
This week:
1. More of Literature survey and,
2. MD Software Profiling and Code Flow Analysis
Last week:
Completed the design and tested functionality of the force compute core. Ran a few preliminary tests, speedup looks very possible.
Completed a large portion of the background chapter.
This Week:
Run the Force compute core as a part of NAMD and test for correctness.
Integrate the Distance and force Compute Core.
Optimize software and hardware to improve speed of operation.
Complete the ‘background’ chapter.