Back
< next | previous >
19/07/2010

For the BioMed paper I need to run extra sims for 2 wide systems, this is due to V having the execution data of a 2 wide system. The reason the Sims failed originally was due to me forgetting to include the machine model files so it was being compiled as a 4 wide and then causing seg faults with the counters.

Sims are currently re-running on the computers in the lab.

Scott helped with the smallocs for h.264 and it now runs under gcc, I need to get this working through the toolchain.
13/07/2010

mpeg2enc has now passed and been verified through the Simulator. Also have run aha and distray benchmarks. Next job will be h.264 decoder (available in new VEX testbench). Initial run shows that smalloc doesn't seem to work for it, need to get this working.

Presented paper at ISVLSI last week, main questions/comments were regarding to the fact the the programmer would have to generate the threads rather than then being spawned automatically. Need to think of possible extensions and if this would be possible?

I have started a benchmark code section on the trac server, I will go back and run all old code through the next tools and check they still work.
25/06/2010

Currently using the mpeg2 encoder as a test bench. This has required implementing fileIO to read in the files and print out.

Multiple operations have been implemented in the Simulator. I have just got arguments from the command line to be passed to the FUNC_main of the Simulator. it is currently compiling and will take a few hours before I can test it.

It compiled previously and ran without segfauling, however, it exited out due to there being no inputs from command line. After mpeg2 has run V would like to run h.264 (there is a version with the new vex 4.32)
04/06/2010

Currently have a working version on malloc, implemented by Scott. I have tested it on a sample malloc program and it worked. Need to test it on a bigger program now.

I have downloaded some papers from IEEE Xplore and will read them over the weekend?

Need to look into the _impure_ptr problems from before as that might be the difference between having lots of things working and not. From looking at it again it seems that it does load from memory (_impure_ptr+0)[$r0.0], its just that I don't know where that is, need to do some investigations with VEX.
03/06/2010

We have decided to use the coremark benchmark as it is the only embedded and threaded piece of C code I can find (even though there are problems with memory operations due to the way VEX defines out of memory operations).

Still need to:
- read papers/magazines
- implement malloc (got a working version on Mac which seg faults on Linux :S)

Paid the money for ISVLSI but can't get any funding for travel. No surprises there then.