Week 3 Report
June 14 - 18, 2010
During Week
3, I started to formulate a way to assign multiple
matrixes to a vector onto a FPGA board using the Verilog
language. The main problem was
that the FPGA chip cannot hold a very large matrix. For
example, the matrix I am working on is 1024x1024, so my
advisor and I came up with a solution that we should
treat it as a vector by vector multiplication. And then stream
the data into the chip. The chip will have several MAC
(multiplier accumulators) to multiply the data very fast.
Also, my advisor suggested CoreGen which is part of the Xilinx ISE WebPack
(A Verilog
synthesizer), which will automatically code the multiplication
and the addition hardware block. All I need to do now is to
adjust the scheduling or any other adjustments to reach
optimal performance.
You can
reach me at fernandez.andres@sbcglobal.net
|