Re: [ANN] LuaJIT Roadmap 2012/2013

  • From: Paulo Matias <syscoder@xxxxxxxxx>
  • To: luajit@xxxxxxxxxxxxx
  • Date: Tue, 12 Jun 2012 13:23:16 -0300

On Tue, Jun 12, 2012 at 10:53 AM, Mike Pall <mike-1206@xxxxxxxxxx> wrote:
> Robert G. Jakabosky wrote:
>> On Monday 11, Mike Pall wrote:
>> > Adam Strzelecki wrote:
>> > > for i = 0, 1000 -- some tight loop
>> > >   local value = model[i] * view
>> > >   gl.UniformMatrix4fv(location, 1, gl.TRUE,
>> > > end
>> >
>> > Well, no, because 'value' *does* escape. There's no way for the
>> > compiler to know that gl.UniformMatrix4fv() doesn't store the
>> > 'value' pointer somewhere else.
>> If gl.UniformMatrix4fv() did store the value and the value wasn't also stored
>> somewhere on the Lua side, then there is already a bug since the GC would 
>> free
>> the value which might still be in use by the C code.
> No, there's a guarantee that arguments to a C function call are
> not collected before the C call returns.

But - correct if I'm wrong - if the C side stored a 'value' pointer
anywhere, there wouldn't be any guarantee the GC wouldn't free the
value *after* the FFI call returns. So, the only way one could prevent
the value from being freed by the GC after one iteration of the loop
would be to store a reference to it on the Lua side. But then 'value'
goes out of scope and no reference was stored to it anywhere on the
Lua side.

It seems to me that, in this case, it could be possible that some sort
of liveness analysis would detect 'value' is not used after the
iteration on which it was created, and thus would mark it as movable
to preallocated memory. However, I don't know if such an analysis
would be easy to implement in the actual codebase.

Other related posts: