Nim's Memory Management

Author:Andreas Rumpf
Version:2.3.1

"The road to hell is paved with good intentions."

Multi-paradigm Memory Management Strategies

Nim offers multiple different memory management strategies. To choose the memory management strategy use the --mm: switch.

Hint: The recommended switch for newly written Nim code is `--mm:orc`.

ARC/ORC

ORC is the default memory management strategy. It is a memory management mode primarily based on reference counting. Reference cycles are handled by a cycle collection mechanism based on "trial deletion". Since algorithms based on "tracing" are not used, the runtime behavior is oblivious to the involved heap and stack sizes.

The reference counting operations (= "RC ops") do not use atomic instructions and do not have to -- instead entire subgraphs are moved between threads. The Nim compiler also aggressively optimizes away RC ops and exploits move semantics.

Nim performs a fair share of optimizations for ARC/ORC; you can inspect what it did to your time critical function via --expandArc:functionName. Likewise, you can inspect the whole module via --expandArc:fileName.

--mm:arc uses the same mechanism as --mm:orc, but it leaves out the cycle collector. Both ARC and ORC offer deterministic performance for hard realtime systems, but ARC can be easier to reason about for people coming from Ada/C++/C -- roughly speaking the memory for a variable is freed when it goes "out of scope".

We generally advise you to use the acyclic annotation in order to optimize away the cycle collector's overhead but --mm:orc also produces more machine code than --mm:arc, so if you're on a target where code size matters and you know that your code does not produce cycles, you can use --mm:arc. Notice that the default async implementation produces cycles and leaks memory with --mm:arc, in other words, for async you need to use --mm:orc.

Other MM modes

Note: The refc GC is incremental, thread-local and not "stop-the-world".
--mm:refc
It's a deferred reference counting based garbage collector with a simple Mark&Sweep backup GC in order to collect cycles. Heaps are thread-local. This document contains further information.
--mm:markAndSweep
Simple Mark-And-Sweep based garbage collector. Heaps are thread-local.
--mm:boehm
Boehm based garbage collector, it offers a shared heap.
--mm:go
Go's garbage collector, useful for interoperability with Go. Offers a shared heap.
--mm:none
No memory management strategy nor a garbage collector. Allocated memory is simply never freed. You should use --mm:arc instead.

Here is a comparison of the different memory management modes:

Memory ManagementHeapReference CyclesStop-The-WorldAtomicValgrind compatibleCommand line switch
ORCSharedCycle CollectorNoNoYes--mm:orc
ARCSharedLeakNoNoYes--mm:arc
Atomic ARCSharedLeakNoYesYes--mm:atomicArc
RefCLocalCycle CollectorNoNoNo--mm:refc
Mark & SweepLocalCycle CollectorNoNoNo--mm:markAndSweep
BoehmSharedCycle CollectorYesNoNo--mm:boehm
GoSharedCycle CollectorYesNoNo--mm:go
NoneManualManualManualManualManual--mm:none

JavaScript's garbage collector is used for the JavaScript and NodeJS compilation targets. The NimScript target uses the memory management strategy built into the Nim compiler.