PDF.js also seems to draw the page by repeatedly calling ctx.fillText one character at a time, which I assume isn't too good from a hardware acceleration point of view?
Thanks Liam Wilson On Wednesday, May 28, 2014 7:42:43 PM UTC+1, [email protected] wrote: > > I just found this email by googling "emscripten poppler", as it is a > mystery to me why such a solution hasn't obsoleted pdf.js yet. > > The "hardware acceleration" argument doesn't hold water. GPUs are no > silver bullet here. Typical text rendering is hard to make to utilize GPUs > in a meaningful way, so benefits here should be considered speculative. On > the other hand, there is no question that a plain software renderer could > be an magnitude faster than PDF.js, as PDF.js is slower on my core i7 than > XPDF was on my Athlon 12 years ago. Emscripten'ing and using WebGL to > present the resulting surface (or, when WebGL is not available, Canvas 2D) > should thus handily beat PDF.js. > > Benoit > > > > On Monday, November 25, 2013 1:31:29 PM UTC-5, azakai wrote: >> >> I did discuss this with someone working on pdf.js once. I think overall, >> pdf.js will potentially be faster since it can render using the browser's >> hardware acceleration, while compiled poppler will use software rendering. >> But poppler will be more precise in rendering in some cases. Definitely the >> comparison would be interesting to do. >> >> - Alon >> >> >> >> On Sun, Nov 24, 2013 at 8:56 PM, 王璐 <[email protected]> wrote: >> >>> Hi all, >>> >>> Recent I updated poppler in the emscripten repo, and managed to make >>> it works: https://github.com/kripken/emscripten/pull/1854 >>> >>> The old demo is down, as well as many demos on the same site, I >>> wonder if they are still maintained. Anyway I'm interested in creating a >>> new one myself on Github, and I'll post back once I make it running. >>> >>> I wonder if you have ever compared the old poppler demo with pdf.js: >>> >>> - Size: The old one was pretty big (12M), but now it has become much >>> smaller, around 5M (probably thanks to closure compiler?). PDF.js seems to >>> be around 3M without closure (but it does not support closure so far). >>> - Features: I'd bet that poppler supports much more PDF features due >>> to its long history and reception, but also I rarely see the messages >>> showing that some PDF is not supported by PDF.js yet, so PDF.js should have >>> also supported most useful features. >>> - Speed: Might be an interesting comparison, but I cannot predict the >>> result at all. >>> >>> Since I did not find anything with Google, I wonder if you guys have >>> any discussion with PDF.js guys. I guess that it might be interesting to >>> replace the PDF parsing code of PDF.js with a compiled poppler. >>> >>> What do you think? >>> >>> >>> regards, >>> - Lu Wang >>> >> -- You received this message because you are subscribed to the Google Groups "emscripten-discuss" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
