Updated GSOC Ideas (markdown)

Alex Clark ☺ 2014-02-14 01:59:30 -08:00
parent 3eecfde70f
commit 475a735824

@ -8,12 +8,12 @@ _From Prashant Shrivastava, a potential GSOC participant who contacted Alex Clar
## Performance ## Performance
_Suggested by Eric Soroos, Mentor _Suggested by Eric Soroos, Mentor_
Investigate speeding up core operations by vectorizing core functionality. This could be a combination of autovectorization in the compiler, manual recoding of tight loops in vector form, or writing kernels for OpenCL. Investigate speeding up core operations by vectorizing core functionality. This could be a combination of autovectorization in the compiler, manual recoding of tight loops in vector form, or writing kernels for OpenCL.
## Stability ## Stability
_Suggested by Eric Soroos, Mentor _Suggested by Eric Soroos, Mentor_
Pillow decodes a lot of image formats, some of which are safer and better specified than others. Experience with [other projects](http://googleonlinesecurity.blogspot.com/2014/01/ffmpeg-and-thousand-fixes.html) has shown that there are many bugs lurking in otherwise functional code. Running a large corpus of valid, invalid, or otherwise fuzzed images while recording and investigating errors and crashes should improve the robustness of Pillow. Pillow decodes a lot of image formats, some of which are safer and better specified than others. Experience with [other projects](http://googleonlinesecurity.blogspot.com/2014/01/ffmpeg-and-thousand-fixes.html) has shown that there are many bugs lurking in otherwise functional code. Running a large corpus of valid, invalid, or otherwise fuzzed images while recording and investigating errors and crashes should improve the robustness of Pillow.