I am running a server which automatically texes (pdflatex) a lot of documents (which are provided by a handfull of users, and in some cases source codes are generated automatically).

Speed is an issue, in particular for texing many very small documents.

I have so far been perfectly fine with a quite small installation, and only recently had a problem with a missing fdsymbol. I now could just install texlive-fonts-extra and hope that there will be no more missing packages; or I could solve the package problems once and for all and get texlive-full.

But I am slightly worried that this could impact performance. Is this realistic? Of course most of the (4GB?) installation will just sit idly on the harddisk and do no harm. But the search path will be much bigger, so might it be slower to find a package if the system first has to sift through many thousands of directories?

This problem might be affect especially very small documents; large ones will presumably have a large compile times compared to the package loading time anyway.

I assume that in most cases Kpathsea is used (is it?), so the system will not actually traverse these directories, but still the database lookup will be slower for larger databases. (And filesystem lookups are cached as well, so I guess in the end this will not make so much difference?)

Of course the obvious answer is: ``Try it out and time it'', but I am hesitant to play around too much with the running system...

To summarize: Is there a (measurable, quantifiable) decrease in performance when installing texlive-full?

  • 2
    as long as you are using mktexlsr (texhash) the filesystem isn't used to search for files as all the locations are hashed in advance, so search time isn't really affected by a large search tree. – David Carlisle Jan 22 at 14:33

As David said, no, it is not slower. The large trees are all by default hashed into ls-R files and thus the only increase in time is the loading of the larger ls-R file, which can be disregarded.

|improve this answer|||||

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.