I've been working with ImageMagick on the Raspberry Pi (model B) and have noticed a bottle neck in the area of image conversions.
When converting between bmp, jpg, png, etc. the speeds are acceptable but I also have a need to convert RAW formats.
In particular, I have been converting NEF to jpg at 70% for benchmark tests.
The Raspberry Pi takes just over 5 minutes to convert a single 12.7 MB image compared with 14 seconds on my Ubuntu Laptop.
Of course, I understand the limitations of the Pi's hardware but I am very interested in how I can really get the most from the performance.
I already know about overclocking and intend to implement this tomorrow when I work on it next.
Information:
- Raspberry Pi uses heavily modified version of Raspbian (lightweight, no X, unnecessary software stripped out, running Apache, running own software)
- A class 10 4GB SD card is used
- Benchmarks taken without overclocking
- Totally headless
Some things I am considering:
Overclocking
Finding another piece of software which may be able to perform conversions on raw faster
Trying to find a way to utilise the GPU as well as the CPU (possibly forking the necessary code for conversions and attempting modifications to use GPU)
Allocating all possible RAM to CPU
Any help is greatly appreciated, whether it be advice on optimisations or another piece of software that could be used.
I don't expect it to be as fast as a laptop but ideally the speed would increase to around 1 - 2 minutes per image.