search

yuyuranium / FPGA-Project-2022-simple-tpu

Systolic array based simple TPU for CNN on PYNQ-Z2
18 stars 2 forks source link

Add MM accelerator #2

Closed yuyuranium closed 2 years ago

yuyuranium commented 2 years ago

I've been working on the verilog code for the accelerator. In this draft PR you can see my progress and give me some reviews so that I can improve my codes.

japoka410666 commented 2 years ago

I have tested global_buffer.v but failed at implementation. Here is the most primary error messeage in my opinion.

[Place 30-415] IO Placement failed due to overutilization. This design contains 267 I/O ports
 while the target  device: 7z020 package: clg400, contains only 255 available user I/O. The target device has 255 usable I/O pins of which 0 are already occupied by user-locked I/Os.
 To rectify this issue:
 1. Ensure you are targeting the correct device and package.  Select a larger device or different package if necessary.
 2. Check the top-level ports of the design to ensure the correct number of ports are specified.
 3. Consider design changes to reduce the number of I/Os necessary.
yuyuranium commented 2 years ago

Thanks for testing! I think you are right. The error occurs because Vivado wants to connect all the port, say 267 pins, to the I/O ports, which I think means the 255 pins on the PYNQ board.

How about the utilization report? Does Vivado synthesize the global buffer successfully? Does Vivado utilize BRAM's?

japoka410666 commented 2 years ago

The following shows the utilization after changing GBUFF_ADDR_BEGIN to 12'h000. The I/O BUFs can be neglected.

image

yuyuranium commented 2 years ago

Looks great! It's a great relief to know that our memory model is synthesizable. :smiling_face_with_tear:

yuyuranium commented 2 years ago

I want to change the project name into simple-tpu instead of mm-acc, since the former sounds more awesome. XD

Therefore, you can see that in the commit message of edbee34 I used Add tpu rather than Add mm acc. (i.e., tpu is equivalent to mm acc)

yuyuranium commented 2 years ago

I think I've finished the coding for the TPU. However, I didn't do any tests yet. Let's merge this branch into main first and I will open another branch for testing, simulating or modifying the code. Please have a look and feel free to give me comments. Thanks!

japoka410666 commented 2 years ago

Awesome work! Thanks for your contributions! 💯

WillyChennnn commented 2 years ago

Thanks for work. You are my god...