Why does scrolling allocate more and more memory?

[peterh92014343]

I have two pdfs (7MB and 14MB), both with watermarks and page size changes. Obvisouly, neither file is huge. When I load them (and do nothing else) Acrobat Reader takes 10-25% of the cpu for fifteen minutes and allocates up to 117MB then drops to 53MB. After 20 minutes, the cpu finally stops with an allocation of 63MB.

Now, if I start scrolling up and down the pdfs -- that's it, just scrolling -- I can get the memory allocation to jump to 600MB.

Now I think that's ridiculous for 21MB worth of data, but a coworker with an Acronat Pro Subscription had 63GB allocated before his machine crashed.

Are these bad files? Is Acrobat leaking memory (again)? Is this normal?

I am running the newest versions of Reader and Windows 10, and I have performed a Save As on the files. But I haven't done anything else to them. These are new client files and I have no certainty as to how they were made.

Just looking for insight.

Thanks

[Test Screen Name]

I'm puzzled by the reference to 63 GB (as opposed to MB). In Windows the most Acrobat can use is 2 GB. What you describe doesn't sound very surprising. PDF files are very well compressed, so reading parts of it into memory will often use MUCH more than the file size.

[peterh92014343]

It was GB. But without access to that machine, I can't guarantee what else was messing up memory. It just that those two users have never seen memory issues until loading this newest bunch of pdfs.

Either way, assuming the 20 minutes of significant CPU time was uncompressing the file (which is a LOT), why would compression have anything to do with scrolling?

[ZBalling]

But that is just impossible for windows as Acrobat is 32 bit app there (only 4 GB for one process)! Maybe for MacOS Catalina 😉

[peterh92014343]

Just to reiterate. The whole file loads. The memory allocated stops at 100MB. I scroll to the bottom of the screen and back to the top. The amount of memory allocated climbs by about 100MB. And stabalizes there. Then I scroll up and down and it goes up another 100MB. Each time I go back to the file and scroll it climbs another 100MB. 

If I have already viewed the entire file, why does the memory allocation go up another 100MB each time?

[ZBalling]

And... It does not happen for me. please give pdfs and what OS you are using.

[peterh92014343]

Sorry, I can't share the files. They're proprietary. As stated above, all machines in question are running the newest patch levels of Windows 10 and the newest patches of Reader/Acrobat -- well, as of when I initially posted.

So when you open watermarked pdfs using the newest copy of Reader or Acroabt, the memory allocation doesn't go up each time you scroll? Okay then, that helps. I'll see if I can convert the files and convert them back to pdfs and see if the same thing happens.

[Dov Isaacs]

Some observations:

Yes, both Adobe Acrobat Reader and Adobe Acrobat Standard/Pro are 32-bit applications in Windows. This limits them to an address space of 3+GB of virtual memory.

Acrobat does have the ability to handle very large PDF files. I've personally opened very complex, multiple GB PDF files from graphic arts sources.

Problems such as you encountered are most often due to very complex or maybe better stated as overly complex PDF content. Example include content from CAD software where the generated PDF is unnecessarily complex. In some cases, we've seen a long dotted line segment generated as via single PDF line segments for each dot in the line. Or perhaps transparency is being unnecessarily applied to objects.

However, without being able to examine the file in question (or perhaps a comparable file exhibiting the same issues), it really is impossible to definitively dignose the problem. I do understand that this may be proprietary information you don't want to publicly post, but if you are willing to let us examine such a file at Adobe, we might be able to assist by determining the real source of the problem and help (1) you and you clients avoid these issues and/or (2) apply any appropriate fixes in Acrobat's code. Contact me via private message if this might be possible.