how to enlarger Java heap size?
in [Java Help]
|
From: Dan Polansky on 19 Sep 2009 12:26 Following up on the subject title "how to enlarger Java heap size?" rather than on the recent discussion: Is there a way to tell Java on the command line to take as much heap size as the physical memory or the virtual memory of the computer allows? I have a launcher written in C for Windows that runs a Java application. I can tell the launcher to pass "-Xmx1024M" to "javaw ", but then the launcher won't work on computers that have less memory, which is why I use "-Xmx256M", to be on the safe side. What I would really like to tell JRE is: "Look, take as much memory for the heap as the physical memory or the virtual memory of the computer allows". I do not know in advance what the physical memory of user's computer will be. --Dan
From: Peter Duniho on 19 Sep 2009 13:04 On Sat, 19 Sep 2009 09:26:37 -0700, Dan Polansky <dan.polansky(a)gmail.com> wrote: > Following up on the subject title "how to enlarger Java heap size?" > rather than on the recent discussion: > > Is there a way to tell Java on the command line to take as much heap > size as the physical memory or the virtual memory of the computer > allows? No, not solely with the command line arguments. You could write a helper application that detects the installed physical memory and then passes that along as it starts the Java process. However, that's not actually a very useful thing to do; physical memory has nothing at all to do with how much memory is actually available to the process. At best, it's a potential performance optimization, and in reality because your Java process has no way to monopolize the physical memory, it's an optimization that may or may not (and most likely not) have any effect. As far as "virtual memory" goes, the first issue is what you mean by "virtual memory". A process has two limits: virtual address space, and virtual memory. The former is limited in part architecturally, but also by overhead within the process. The latter is limited both by virtual address space and by available disk space (whichever is smaller). On 32-bit architectures, normally (these days) disk space is much larger than the virtual address space, so it's really just virtual address space that you are concerned with. But while the actual virtual address space is known (e.g. 2GB on 32-bit Windows), as with the physical memory, there's no reliable way to predict how much of that is usable by your process. Due to various other overhead in the process, on Windows you can realistically get up to about 1.6GB or so before a memory allocation fails. Depending on the actual allocation pattern, this can be significantly lower. So, as with physical memory, while it is possible to know the theoretical upper bound (i.e. the architecturally defined limit), that's not a useful number. There are other variables that are unknown. Note that all of the above is even worse if you're dealing with a 64-bit OS. Physical memory is even less relevant to your process, and the theoretical upper bound of how much you can allocate becomes huge, and impractical to try to reach. Which is all a long way of saying, no...not only can you not have Java determine this automatically, it is not even practical for you to use external information to configure a Java process in that way. The maximums you can measure aren't actually useful in terms of controlling performance of your process. > I have a launcher written in C for Windows that runs a Java > application. I can tell the launcher to pass "-Xmx1024M" to "javaw ", > but then the launcher won't work on computers that have less memory, "Less" what kind of "memory"? Physical memory? Irrelevant; you can start with 1GB of Java heap, even if the computer only has 128MB of RAM. Virtual memory? It's true that if you have a user with less than 1GB of disk space left, they may not be able to start a process that starts out by trying to allocate a 1GB block of memory. But how often do you really run into that these days? Why isn't it more useful to just tell the user "hey, you need to have at least 1GB of disk space for my process to run". > which is why I use "-Xmx256M", to be on the safe side. If your program can run in 256MB of heap, why would allocating more help? > What I would > really like to tell JRE is: "Look, take as much memory for the heap as > the physical memory or the virtual memory of the computer allows". I > do not know in advance what the physical memory of user's computer > will be. Personally, I think it's a bit archaic that you have to specify the upper bound of allocation up-front for Java programs. Java _ought_ to be able to adjust the heap size as necessary, within the limits available on the OS, according to usage. Your process ought to be able to start out with (for example) a 64MB heap, but have that heap grow as much as is supported by the OS should the Java program require additional memory. But, AFAIK that's just not possible right now. So, I think you're pretty much stuck figuring out what the largest heap your program will need, and just always asking for that. In theory, you could provide the user with additional start-up options, but it seems to me that in reality, you're not going to run into many computers where specifying the max you'll use won't actually work. Just specify 1GB all the time, and leave it at that. Pete
|
Pages: 1 Prev: confused with LayoutManagement: JSplitPane Next: License |