Hello Simon, Tuesday, April 19, 2005, 4:15:53 PM, you wrote:
1) can you add disableGC and enableGC procedures? this can significantly improve performance in some cases
SM> Sure. I imagine you want to do this to avoid a major collection right SM> at the peak of a residency spike. SM> You probably only want to disable major collections though: it's safe SM> for minor collections to happen. no, in that particular case i have very simple and fast algorithm, which allocates plenty of memory. minor GC's in such situation is just waste of time. so i want to do: disableGC result <- eatMemory enableGC with a effect that all memory allocated in 'eatMemory' procedure will be garbage collected only after return from this procedure. currently i have this stats: INIT time 0.01s ( 0.00s elapsed) MUT time 0.57s ( 0.60s elapsed) GC time 1.41s ( 1.41s elapsed) EXIT time 0.00s ( 0.00s elapsed) Total time 1.99s ( 2.01s elapsed) %GC time 70.8% (70.1% elapsed) Alloc rate 171,249,142 bytes per MUT second Productivity 28.7% of total user, 28.4% of total elapsed as you see, it is very inefficient
2) if, for example, program's data before GC is 80 mb and after GC is 60 mb then the program will occupy after GC the whole 140 mb and ALL this space will be marked by OS as used! if there's a memory shortage, old program data even can be swapped to disk despite the fact that we absolutely don't need them! that behaviour significantly enlarge memory needs of GHC-compiled programs
if this unused memory will be returned to OS or marked as unneeded after GC then will problem will go on. preferably this must be done during the time of GC, on each page which contents has been already compacted. in this case such program will never use more than 80mb of real memory (+ 1 page + memory for GC)
SM> I guess you're proposing using madvise(M_FREE) (or whatever the SM> equivalent is on your favourite OS). This would certainly be a good SM> idea if the program is swapping, but might impose an overhead when SM> running in memory. I don't know, I haven't tried. i don't see resons why this can be slower. we will be a "good citizens" - return memory what is not used at current moment and reallocate memory when needed. current implementation only allows memory usage to grow and that is not perfect too. imho it will be better to release unneeded memory after major GC and perform next major GC after allocating fixed amount of memory or, say, after doubling used memory area SM> Also, you might be better off using +RTS -c or +RTS -M<size> to avoid SM> swapping in the first place. it seems to me that we can join benefits of compacting and copying algorithms. at least this combined algorithm will be better than compacting in all cases and better than copying in almost all cases. am i right, right or right? :))) currently, compacting algorithm works "in place" and copying alg. use additional memory. this alg. will use additional logical memory but works almost in-place in terms of physical memory -- Best regards, Bulat mailto:bulatz@HotPOP.com