Milestones

Reini Urban rurban at x-ray.at
Thu Dec 20 17:31:17 UTC 2012


https://github.com/parrot/parrot/issues/milestones

threads:
I've updated our thread milestone to "threads stability".
Its is merged and usable for experimentation, but not yet stable for
production.
nqp will need some time to implement support for it.

"threads work on normal examples and passes all our tests,
but fails in production, when one thread tries to allocate memory from GC,
and another thread GC visits the same pointer.
GC does not need to be concurrent, just not pass the interp border to
other threads.
There might be also some more deadlocks lurking."

Is this text okay?
How long will we need? 5.0 or 5.3? I've set it to 5.0, due in 26 days.

m0:
github says it should be due in 12 days.
But I have not seen any public activity on m0. How is it going? How can I help?
I only know http://trac.parrot.org/parrot/wiki/LoritoRoadmap
Shouldn't this be brought over to our new wiki?

llvm:
I have improved bacek's ops2c_llvm in rurban/ops2c_llvm to be mergable.
Enables LLVM compilation for dynpmc/llvm_engine and llvm_extra.
I know http://blog.bacek.com/2011/03/ho-ho-ho-llvm-in-parrot-is-getting.html
But I fail to see the big picture regarding ops2c and it seems to be
similar to the m0 effort,
maybe easier to accomplish. Can someone please summarize where a ops2c
with llvm should go?
Should ops2c_llvm create nqp code for all ops and then jit it?

Maybe update https://github.com/parrot/parrot/wiki/Projects please.

What else?

I have a plan to radically change our directory layout, see my branch
new-layout.
We need to support relocatable executables, not necessarily our own.
perl6 needs that least on darwin and windows. commit
7b33fead08b76e6a9d3ad34dbebb7678152377c7

I want to get rid of the function call overhead (deoptimized signature
handling), and the deeply nested
sprintf orgies on most function calls.
Fixing this should bring us nearer to p5 vm performance. p5 has a weak
vm API design with
their heap-based stack, we can do better theoretically. But we are still behind.
-- 
Reini Urban
http://cpanel.net/   http://www.perl-compiler.org/


More information about the parrot-dev mailing list