April 4

We have a lot of code, but we don't have a complete running system yet. We intend to have all major parts functioning by our Thursday meeting.

Tim: Worked on the SGI hardware performance measures. The code isn't quite finished yet. It needs to be finished, then incorporated into the library.

I did a systematic renaming of variables throughout the library. The client wants the library to be portable, so the code needs to be legible to someone in the future. I recorded my name changes in a file called "rename" to help Xiao know what's up.

Rob: This week has been spent largely trying to finish the implementation of the GistView application. My portion of the work this week has been thwarted by a number of technical problems, some of which remain unresolved as I write this. We've opted to use a more sophisticated interface which has caused me a number of problems. The classes we're using are either buggy, in some form, or I'm running in to some sort of process or memory limit on eagle and capefear which prevent the application from running. It starts, then just hangs without ever displaying anything. I've thus far wasted at least 6 hours trying to track down the problem. *grumble*

The problem I was having does appear to be related to eagle/capefear as I can run it on a Sun at work and PC at home without problems. I consider it resolved. I just won't run the app on eagle/capefear for now. If I have time it might be worthwhile to find out exactly why it won't work on eagle/capefear, but it isn't critical for now.

I've also written and debugged the header and footer parser routines, which now work at least if certain assumptions about the input are made which aren't part of the file spec. I've modified the GistView class so that the loadFile function actually does load the file, where previously we used a demo loader function.

The preference object does attempt to load itself and sets up defaults if the file isn't present. Nothing in the application really uses them yet, but we can begin to do so. One snag in this is that I'd like to have the option to save them on exit, but this appears not to happen if the user closes the window rather than choosing File:Quit.

I hope to verify, test, and debug the event parser today. I also need to get the legend panel working, but have been deferring that in favor of work which has a greater impact on the functionality of the application.

Tiger: Continue working on java GUI design and coding. Got profile graph showing fake data. Got the horizontal and vertical scrolling of event window working correctly with GistRule and index. A lot of code rewriting and modification. Trying to get printing working, and now the profile graph printing is kind of working, but the event window still doesn't work.

Liusong: I am writing a technical manaul.

Xiao: This week we continued to work on the Gist Library part. Based on the sequential part, I went further to deal with the real parallel program. Actually this program was used our "parallel and distributed computing" class; therefore, we could it is a "real one". while dealing with the real parallel program. i found some problem i did not think of before,like dumping the memory content into the log file. Originally, what i am doing is, which processor's buffer is full then it will dump all the processor's content into the log file because we want to keep the log file in the time order. But now it is parallel program and maybe many processors' buffers are full at the same time. therefore, the log file is very messy. So i added some syncroniuzation mechanism to gurantee that each time only one processor is accessing the log file. After sloving that problem, i began to improve the performance of this program. as you know, when we use more processors to solve one problem, we should need less time. However, because of the embedded Gist calls, once when increase the number of the processors, we need syncronize more processors' work and more processors need to access the log file. So it requires more time but deos not need much more. So later on i modified the code to access the log file and made the time much less.