I think you're making this sound more uncertain than it actually is. I'm not sure what the deal with PyPy is, but there is no inherent problem with JIT and garbage collection – Julia has never had any major problems with GC. The main issue our current garbage collector has is that since it's non-incremental and stop-the-world, it's not good for real-time applications.
On Thu, Aug 28, 2014 at 4:14 PM, Ivar Nesje <[email protected]> wrote: > There are also a PR for some changes to the GC, #5227 > <https://github.com/JuliaLang/julia/pull/5227>. > > It is hard to know the results in such cases before you have tried. Using > a unfinished language like Julia, is different from an established language > like Python, or a commercial product. The positive side is that we can fix > issues "the right way" even if it means breaking backwards compatibility. > The negative side is that there are much bigger chance of hitting an issue, > and that your code might break (in subtle and obvious ways) when you update > Julia. > > Programming in Julia is really fun, and everybody here will be exited if > you make an attempt to port your project. If you manage to create a small > reproducible example where we perform poorly (after you have read and > understood the performance tricks in the manual), please post it so we can > look at it to see if there is something simple we can fix. > > Ivar > > kl. 20:33:46 UTC+2 torsdag 28. august 2014 skrev Stefan Karpinski følgende: >> >> Julia uses a simple stop-the-world mark and sweep collector. There is no >> reference counting, so deallocation and finalization may happen at a >> somewhat arbitrary time after an object becomes garbage, but it will happen >> eventually. It's not an award winning GC, but it's quite reliable, partly >> due to its simplicity. >> >> On Thu, Aug 28, 2014 at 2:08 PM, V.D. Veksler <[email protected]> wrote: >> >>> >>> i'm thinking of moving a python project into julia, but before i make >>> such a commitment, i need to know about julia garbage collection. >>> how does it work? >>> >>> why i'm asking: >>> my code has a non-vanilla linked list class that provides the >>> functionality to drop elements during iteration. >>> this works fine with CPython, but there's something not quite right >>> about gc in PyPy's JIT compiler which causes a memory leak. >>> >>> can i trust that Julia will not have the same problem as PyPy? >>> or should i just move to C for speed and manual deallocation? >>> >>> >>
