Skip to content

Latest commit

 

History

History
107 lines (81 loc) · 4.19 KB

README.md

File metadata and controls

107 lines (81 loc) · 4.19 KB
Raw

CPA-SIM

A (very) simple SIMD Cellular Processor Array Simulator (also known as Focal-Plane Processors, Vision-Chips etc.) based on OpenCV.

  • Customizable -> You can hack in your own CPA
  • Based on OpenCV with OpenCL modules. Runs on CPU and GPU

SCAMP Module

The only CPA currently implemented is loosely based on the SCAMP 5 chip. There are some notable differences to the physical chip:

  • No noise
  • Values are hard-capped
  • The NEWS register does not update on neighbor transfers
  • The F register does not update on arithmetic operations
  • Some reset-on-write logic (R12?) is not implemented
  • All operations can work in-place, where there are some limitations to this on hardware
  • No async operations
  • Higher level instructions

The following instructions are supported:

Instruction Explanation
rpix Reads in value from camera
add Adds two values
sub Subtracts two values
north Takes value from north cell
east Takes value from east cell
south Takes value from south cell
west Takes value from west cell
mov Copies value from one register to other
in Inputs a value to a register
where Set FLAG on pixels where value > 0
all Set FLAG on all pixels
div2 Divides value by 2
d_nor NOR operation (or NOT)
d_mov Moves digital values from one register to other
d_north Get digital value from north cell
d_east Get digital value from east cell
d_south Get digital value from south cell
d_west Get digital value from west cell
d_where Set FLAG on pixels where digital register 1
d_select_pattern Select pattern with addresses/don't cares
d_load Loads a digital value

Getting Started

Compiling

mkdir build
cd build
cmake ..
make
./cpa-sim

Setting the OpenCL device

One can set the device to run the code on by setting th OPENCV_OPENCL_DEVICE Environment variable

For example, to run on the first GPU of the system you could set the variable to OPENCV_OPENCL_DEVICE=:GPU:

Results

I could only run this on a pretty bad Macbook GPUs (GeForce 750m and Radeon Pro 560), which provide limited speedup compared to CPU. I can imagine that a truly powerful GPU could provide larger speedups.

Debugging

ALl the computations happen on a remote device, such as a GPU. Therefore, in normal operation, the images are only downloaded for the UI when update_ui() is called on the simulator instance.

This speeds up things considerably. However, one can no longer step through a program with a debugger and see the effect of every instruction.

To do this, there is a super-debug mode, that downloads the image and updates the UI after every instruction. This is incredibly slow and only meant to manually step through the program!

Enable super-debug mode:

cmake -DSUPER_DEBUG=ON ..

(Or create a launch configuration for this in your IDE)

License

MIT License

Copyright (c) 2018 Thomas Debrunner

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.