I wanted to do pdf to html conversions in GPU.

In case of integrated graphics cards, it is not having its own RAM. So in those cases, transfer overhead will never be there. So in case of pdf to html conversion, parallel processing and size of file are considerations. Size of file will never be too large. Even though transfer overhead may be a problem if we use dedicated cards. If we use integrated cards, this problem wont be there.

Considering Parallel processing, I think, parallelization could be achieved page-wise. Will it work?

Is there any parallel or GPU implementations for pdf to html conversions? (Till now I could not find any one)

I have already posted in https://stackoverflow.com/questions/36199864/how-to-parallelize-pdf-to-html-conversion-on-gpu

The post depicts, it is not feasible. But I am not clear why is it not feasible. We can parallelize page-wise. Why cant we do this?

Why cant we do this conversion in GPU?

Is there any white papers published by NVIDIA regarding this? (I did not find even one)

Any ideas at this time will be very useful

Thanks in advance

  • Even if no programming language is involved, it's still a programming question, I don't think you'll get better answer here, then you received on stackoverflow. Aug 9, 2016 at 6:03
  • Thank you. May be. But I am need of reason in detail. Is that impossible or possible. The fact is about possibility or feasibility?
    – Vanns
    Aug 9, 2016 at 6:42
  • Due to the requirement of complex logical processing I'd suspect that this task is entirely unsuitable for GPU conversion. If all you want is to pull images out of a PDF and poke them into a webpage then there is nothing complex for a GPU to do. If you want to take the text in a PDF and put it into html then all there is to do is to wrap that text in html blocks, again nothing for a GPU to do. OCR might be workable on a GPU but that is a highly complex task and not the kind of work you'd put in a free PDF to html converter.
    – Mokubai
    Aug 9, 2016 at 7:04
  • Just saying "do it in the GPU" isn't magically going to make tasks faster because GPUs are not complex branching processors in the way the main CPU is. A lot of tasks that are "easy" on a CPU have to be very painstakingly mapped out for conversion to GPU land and most of the time the return on investment simply isn't worth it. There are also so many different performance levels in GPUs that a task that could perform 10x faster than a CPU on a high end card could perform 10x slower than the CPU on a low end GPU. Use the right tools for the job you have. don't mangle the tools to fit your job.
    – Mokubai
    Aug 9, 2016 at 7:10

1 Answer 1


OCR can be done in a GPU but other components would not necessarily be any faster using a GPU. Your GPU is a Single Instruction Multiple Data processor (SIMD), this means that it can do the same operation on tons of data all at once. Your CPU (assuming it's multicore) is a Multiple Instruction Multiple Data processor (MIMD), which means it is capable of performing many different operations on different data at the same time.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.