Hello the cuda community, I have learned lot of things here
After a phd thesis in applied mathematics (inverse problems in electromagnetism), I found a developper job. It’s just pure coding (gui…) which is not very thrilling. So I decided to continue some research work by my own. Would maybe be interesting in my cv, maybe participating in some open source project or opening a blog, I don’t know yet. I learned some basics on cuda and open cl which seems to be very promising tools. Searching for connex topics to my research one, I studying some image processing methods, medical image processing, would be great but real datas seems to be hard to find.
By the way, I was thinking of using a video decoder (xvid for instance) to write a real time image processing tool. The obvious processing would be for each frame decompressing the actual frame applying some process and display it using open gl. I was wondering if it would’t be more elegant to modify my decoder to decompress directly into a cuda array or cuArray with texture fetches that would be directly processed using cuda. What do you think about my second solution, would it be better usig the line by line copy xvid -> open gl texture -> processing -> display, than bitmap -> open gl texture -> processing -> display ?
In the second case I should create an open GL texture rebind it to a cuArray and copy my frame into it line by line. Is there a method to create directly a pbo starting with a cuArray or an array ?
ps : excuse me for my english. You can even tell me the bugs faults by private message. I would help improving my english