Code Review #3
Description
Hi Matthew,
Nice job on your project. I'm impressed on how much functionality you have put together.
I have a few suggestions on how you could consider improving the usability of your code.
-
Your example use cases that you included in the response to Stephen’s post were very helpful for me. I recommend including those samples/usage notes in your repository and README so a user can verify their installation of your code.
-
Additional installation instructions in your README would also be helpful, such as the commands you used to build with CMAKE, and the CMAKE version.
-
Are there restrictions on the allowed input/output types to the pixelCompressor.py or pixelExtractor.py? I used png, which worked fine, and PIL Image.open() has a lot of support. Perhaps you could mention which ones you have tested with or any that you have found that aren't supported.
-
For pixelCompressor.py, if an invalid number of arguments is specified, then the program errors. Perhaps add input handling and print usage notes for cleaner execution.
compressPixelData(sys.argv[1], sys.argv[2])
IndexError: list index out of range
- When I tried to execute the first time, I was missing the ‘numpy’ library. To tidy things up, and for reproducibility perhaps you could include a requirements.txt file with the required Python libraries and versions you used.
Lastly, one of the example images I used produced a "segmentationFault". I ran it with various parameters, including some of the same as you had on your /home/sunnyonyxtom/Documents/GPU/imageProcessing/cat.pngexample runs, with the same result. I traced it to line 191 in imageProcessing.cu:
cudaMemcpy(outputPixelArrays[i], inputPixelArrays[i], width * height * sizeof(double),
cudaMemcpyHostToHost);
but couldn't get to the root of the issue. Perhaps it's related to the buffer issue mentioned by Stephen, so I am attaching the png file in case you'd like to run against it as a test case.
