The project was developed for course Computer Systems.
Vivado HLS is used to optimize GEMM operation. In these seminar we explore the use
of pipeline directive and array_reshape directive. We analyse performance
and hardware resources utilization using matricies of 3 different sizes:
16 x 16, 32 x 32, 64 x 64.
Report of the work done is available report.pdf. In analysis/ folder all of
the synthesis results are also avaiable.
In order to use the source code you should include gemm.cpp and gemm.h in your Vivado project.
Select gemm as the top function. You can also include the test bench file: gemm_tb.cpp.
You can generate random test matrices and expected result of GEMM operation
using the python script as follows: python gemm.py <MATRIX_SIZE> <VALUES_RANGE> <ALPHA> <BETA>.
The script prints the matricies in C format in the terminal, you can simply copy and
paste the terminal output into test bench file gemm_tb.cpp.