Team Status Report
RISKS
Currently, our greatest risk is not completing all of the features we anticipated. John is still working on transferring data from the external computer to the SoC over Ethernet. This will soon be the bottleneck if it is not completed, because we need a video on the board to test the system as a whole. Our contingency plan is to transfer the video to the board via SD card if we can’t create a working wired connection.
DESIGN CHANGES
There hasn’t been any major design changes. The only potential change is that we might change the way of transferring the video data from using ethernet to uart because of a lack of established documentation/tutorial
UPDATED SCHEDULE
There is no schedule updates.
Maxwell Johnson
Personal work and Progress:
This week, I focused on the RTL and interface for our accelerator. We already had a connection to the PS via BRAM from last week. I completed an accelerator that reads pixel data from BRAM, computes the energy, and writes the energies to another region of BRAM. We use a multi-buffering scheme so that the PS can write pixel values and read energies while the PL computes energies. A region of the BRAM is dedicated to synchronization: when the PS has filled in pixel values for a frame, it writes a 1 to a specific address (as well as some frame meta-data); when the PL has calculated the energy map of a frame and written it to memory, it writes 0 back to that address. I tested the accelerator on the board with a small set of test frames to confirm functionality. The RTL for the accelerator is here, and the C code for the processor’s interface is here.
Deliverables:
This week, I’m going to help John and Riki with their portions of the project for the demo, since the accelerator is working. After the demo, I’m going to focus on optimization. The bottleneck is currently the PS transferring data to the BRAM. I’ll attempt to connect the PL directly to the PS’s RAM, which will decrease the amount of data movement required and should relieve the PS of some of its workload. I will also begin working on the final presentation and report.
John Zhang
Personal Work
This week I have been focusing on achieving the data transfer between the PC and the DDR3. I was able to get an echo server up on the board but no further progress. I tried to set up a tftp server on the board per a tutorial online, but the tutorial is outdated and it is for a different board. I’ve just started to look into data transferring through UART.
Progress
Echo server on the board. Researched on different alternatives.
Deliverables
I will get the data transfer working in the next few days.
Riki Khorana
Personal Work and Progress
Right before the start of this week, I finally debugged the software algorithm and had a final product. While the outputted video was not something that we expected (too coarse), we managed to achieve our first goal of implementing a seam carving algorithm to alter video playback speed.
This week I focused on figuring out the interfacing of the software application with the hardware component. I mainly worked with Maxwell to figure out a viable bit stream format, and memory access conventions.
Deliverables
I am expected to continue and complete integration by the demo this week. I am planning to spend a whole Sunday dedicated to this task.
I am also expected to research for more videos throughout the next few days in order to find which type of videos work well with our algorithm.