Dominik Seichter wrote: > I definitly have to try massif. I sounds very interesting! I am currently > working on a Java profiling project and we are basing our software on HProf > (a profiler example ship with every VM by Sun) and I really miss valgrind. > Even if valgrind is slow, the output is far superior to hprof and there so > many more tools for valgrind.
Didn't you hear that Java programmers don't have to worry about memory management? :-P It's an attitude I find ... frustrating ... and probably what got Java it's "bloated" repuatation. Well, that, the early JREs, and of course Swing. I'm going to be doing some work in Java at my job, once I'm largely finished with the database design & implementation side. It'll mostly be GUI and DB interface code, the real work happens in the DB. Given that, if you have any tips about particularly good Java tools or favourite libraries etc I'm all ears. > I will try to find a paper describing libstdc++ memory management. Maybe this > gives us more insight. Sounds good. It's not like we have any sort of urgent problem, but the memory handling side of things is something we can benefit from putting a bit of attention into over time. > I think PdfReference is most of the time allocated as part of a PdfObject (ok > sometimes also as a value to a key in PdfDictionary) and it is not allocated > on the heap there but as part of the larger PdfObject. That > PdfRefCountedBuffer does not give us benefits is also a surprise for me. I would expect that at most 1/2 of all PdfReference instances would be allocated as members of PdfObject. After all, every PDF object should be referenced at least once, and that other reference is going to be stored in a PdfVariant, probably as a dictionary key or part of a PdfArray - so it'll be on the heap. In any case, the instances of PdfReference that are value members in PdfObject wouldn't show up in any testing since, as you say, they're not directly heap allocated. >> I need to do some testing on Windows, too, but that's harder due to the >> fact that I don't have the same sort of tools available. > > Are there any (besides Purify which is commercial AFAIK)? Unfortunately none that I am aware of. *hugs valgrind* There's always a trial of Purify I guess. Even without that it's possible to learn a bit by just watching overall memory consumption & execution time. At this point, that was more what I had in mind. -- Craig Ringer ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Podofo-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/podofo-users
