YukunXue/CNN_accelerator

Convolutional parallel accelerators deployed in FPGA（面向硬件的卷积加速器）

CNN acclerate

1. input && conv && output

input (uint8): 48 * 48 *3

conv (uint8): 3 * 3 * 3

output : 46 * 46 *3

2. File tree discription

1. build :

   run build.bat

   run build.bat to bulid the project and to get the vcd file checking the wave

2. c :

   c project, used to check the correct of the element in array(Fi or Fo) by changing the array Fi, conv

3. data :

   store the data file (like input signals or result data and so on)

4. py :

   by run python gen.py

   get the Fi data, or conv data of coe file used by rom.

   by diff_test to check the result

5. rtl :

   Here are two files:

   1. pe.v
   2. top.v

   pe module used to simulate the conv op and just execute mulitiply operation.

   top module instant 46*3 pe modules

6. sim:

   tb_top.sv arrange the data to flow in

3. detail and process

while the top module is very simple:

while the parallel