Re: [MLton] MLton Object Root Sets

On Tue, Sep 8, 2015 at 12:41 AM, muyuanli <[email protected]> wrote:
> Hi,
> I'm writing a simple real time mark & sweep GC for MLton and I
> encountered some problems:
>
> 1. In the Mark & Sweep phase, it looks like the GC wind up collecting
> some objects that are still in use. I wonder if I missed some objects
> in the root sets for DFS mark. The code logic looks like:
>
> performGC:
>     GC_stack currentStack = getStackCurrent(s);
>     foreachGlobalObjptr (s, markObjects);
>     foreachObjptrInObject(s, (pointer) currentStack, markObjects, FALSE);
>
>     // sweep & clean mark

In general, that looks ok.  But, some comments:

 * foreachGlobalObjptr will visit the current stack (by way of
visiting the current thread); see
https://github.com/MLton/mlton/blob/master/runtime/gc/foreach.c#L18

 * What do you mean by "markObjects"?  If this simply sets the mark
bit of the object, then that is insufficient --- it won't perform the
tracing that marks all reachable objects.  You really need to use a
specialization of dfsMarkByMode
(https://github.com/MLton/mlton/blob/master/runtime/gc/dfs-mark.c#L29).

> 2. In the object allocation phase, I saw some arrays allocated with
> length 0. What is the purpose with these arrays?

"Array.array (0, 42)" will allocate a length 0 array.  Such arrays
still need to be allocated in order to have identity (for polymorphic
equality of arrays) and to have a length field in the header metadata
(for Array.length).  Also vectors are initially allocated as arrays
and then cast to vectors.  So "Vector.tabulate (0, fn _ => raise
Subscript)" will allocate a length 0 array.

------------------------------------------------------------------------------
Monitor Your Dynamic Infrastructure at Any Scale With Datadog!
Get real-time metrics from all of your servers, apps and tools
in one place.
SourceForge users - Click here to start your Free Trial of Datadog now!
http://pubads.g.doubleclick.net/gampad/clk?id=241902991&iu=/4140