[LLVMdev] ocaml+llvm

[Gordon Henriksen]

On Aug 14, 2007, at 06:24, Gordon Henriksen wrote:

> The two major problems I had really boil down to identifying GC  
> points in machine code and statically identifying live roots at  
> those GC points, both problems common to many collection  
> techniques. Looking at the problem from that perspective makes the  
> problem much more tractable, actually…

Chris,

This is much more generic than what I was originally thinking, but  
what do you think of providing a facility like this?

     enum GCPointKind {
       GCPSubroutineReturn,
       GCPFunctionExit,
       GCPSubroutineCall,
       GCPBackBranch
     };

     class ??? {
     public:
       // Allows the code generator to avoid inserting unnecessary GC  
points.
       bool RequiresGCPointsOfKind(GCPointKind Kind) const;

       // If false, instructs the code generator to keep all GC roots  
on the
       // stack at GC points.
       bool SupportsRegisterRoots() const; // false for now...

       // Visits a GC point, including information necessary to  
identify all
       // live GC roots in the stack frame.
       virtual bool VisitGCPoint(MachineInstruction *Point,
                                 GCPointKind Kind,
                                 intptr_t *StackRoots, size_t  
StackCount,
                                 int *RegisterRoots, size_t  
RegisterCount);
     };

That could enable several kinds of conservative collectors.

The code generator would need to cooperate by:

  1. Inserting nodes for GC points if required, recording the  
corresponding original Instruction*.
  2. Recording the physical location of each llvm.gcroot'd alloca.

Later, a MachineFunctionPass could walk the MachineFunction to lower  
these nodes. Live roots can be identified using domination information.

In addition to providing a stack and register map for building  
tables, such an interface offers a collector a chance to insert any  
necessary landing pads at exit/call/back-branch.

If the collector supports it, LLVM could be allowed to scalarize GC  
roots down the road.

— Gordon

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20070814/62d71db8/attachment.html>