4 ms·
Will calling “free” or “delete” in C/C++ release the memory to the system?
- mkhn 7y agoAlways appreciate your posts. Perhaps good to explain the sbrk/mmap threshold for this and maybe also why the memset() is required to fault in pages :-)
- Iwan-Zotow 7y agoYou could trim memory as much as you want, but next malloc call will result in (a lot of) syscalls, and they are NOT free and NOT cheap
- olliej 7y agoTLDR: no more so than standard mallocs, all of which for obvious reasons cache allocated pages.
- _bxg1 7y agoInteresting. It calls to mind garbage collection, actually, where the "collection" of memory happens at some indeterminate time in the future based on heuristics about how much there is. Of course it doesn't have the same performance impact, probably because the collection process consists of just freeing known chunks instead of iteratively searching for chunks that can be freed.
- DmitryOlshansky 7y agoGC doesn’t search for chunks that can be freed. Also “iteratively” - what do you mean by this? There is no known algorithm for GC that searches for garbage to delete, the way GCs work is exactly opposite - find all live objects (reachable from roots - registers,stack, statics/globals) and the rest of the heap by definition is garbage.
- _bxg1 7y agohttps://en.m.wikipedia.org/wiki/Tracing_garbage_collection https://en.m.wikipedia.org/wiki/Tracing_garbage_collection > tracing garbage collection is a form of automatic memory management that consists of determining which objects should be deallocated ("garbage collected") by tracing which objects are reachable by a chain of references from certain "root" objects, and considering the rest as "garbage" and collecting them So it starts at the root and, in a sense, "iterates" down through the reference tree to see which allocations are reachable, and then the inverse of that is freed. In the OP case, by contrast, "which allocations to free" is known at "collection" time, but there's still an extra step of actually doing the collection.
- DmitryOlshansky 7y agoThe point is that tracing GC searches for things to keep and assumes the rest is free. Which I believe is the important distinction, and quite often it doesn’t know or care “which allocations to free” because it does it in bulk. That allows it to be efficient.
- _bxg1 7y agoNo, the important distinction is that typically GC iterates over something (incurring a cost of O(N)), whereas the C++ case iterates over nothing (incurring a cost of O(1)). But my original point was just that it's interesting that there is an extra step at all in the C++ case, where we tend to think of collection as happening "for free". It has a constant-time overhead, which is not the same as zero overhead.
- DmitryOlshansky 7y ago> important distinction is that typically GC iterates over something (incurring a cost of O(N)) During _collection_ and N is size of live set, so divided by allocation the cost is amortized and can safely considered as O(1) much like appending a dynamic array where costly O(N) resize is amortized over N appends. Typically GCs allocate via bump a pointer allocation which is actually faster then what pretty much all of libc malloc implementations would do. > C++ case iterates over nothing (incurring a cost of O(1)). Not quite, take a look at jemalloc paper for instance: https://www.facebook.com/notes/facebook-engineering/scalable-memory-allocation-using-jemalloc/480222803919/ https://www.facebook.com/notes/facebook-engineering/scalable... A lookup to get the metadata for this specific allocation could be anything from O(1) to O(lgN) depending on size and malloc strategy.
- UncleMeat 7y agoTracing gc works this way. But reference counting gc exists, even if it isn’t widely used in gced languages. Ref counting finds garbage directly.
- ncmncm 7y agoThere are usually very sound reasons not to release memory back to the OS, particularly in a multithreaded program. Each such release causes a "TLB shootdown", in which threads on other cores are blocked while the cores' "translation lookaside buffers", caches of page mappings, are cleared, and further stalls as their entries are re-filled. This is another reason to prefer single-threaded processes, which are less subject to such shootdowns, and less-coupled forms of parallelism. Besides the TLB potholes, releasing memory means that the next time memory is requested, the OS is obliged to zero it before the process gets to see it again. Furthermore, each page will be marked read-only, causing a trap the first time it is touched, and then zeroed lazily. As a result, freeing memory to the OS should only be essayed with the support of a great deal of measurement of the consequences.
- Ace17 7y agoAll else being equal, could the efficiency of malloc_trim depend on heap fragmentation? Could the heap end up with "gaps" of free'd memory that couldn't be returned to the OS independently?
- flohofwoe 7y agoNot an expert in the matter, but: In my simple mental model of how memory management works under the hood (fixed-size memory pages, and a virtual- to physical-address indirection through a page table), only "clean" pages could be returned to the OS which don't have any allocations left in them. So fragmentation might prevent a 'mostly clean' page from being returned to the OS because there's a single tiny allocation left which pins it into the process' address space. However: Most (all?) general memory allocators have small-allocation buckets for minimizing this effect, which basically groups allocations of specific sizes into different memory pages. Despite that, virtual address space fragmentation in 32-bit applications is definitely a thing. Memory might still become fragmented over time so that "big" allocations (bigger than a memory page) are failing because there's no big-enough gap left in the processes virtual address space, despite having enough free physical memory available for mapping. This usually isn't a problem in 64-bit processes, because there's always enough room at the "front" (new memory allocations would basically move like a tidal wave through the 64-bit address space, leaving a mess of fragmented address space behind). The best way to prevent memory fragmentation of course is to minimize allocation frequency in long-running applications (ideally only allocate memory when the program starts up).
- flqn 7y agoI'm guessing this is because the runtime keeps a few pages around for free-lists and the like? Interesting that there's a hump in the middle there.