First of all, I'd like to thank the entire LLVM team for this opportunity.
Torvald, google only displays the abstract on the results page, but I do know
about Tanger and it was mentioned in both the proposal and the comments
that followed, I'll write an email to you later with that info in a
bit, and I do intend
to use as much available infrastructure as possible, and tanger/tinystm are
certainly going to be of great interest to the project.
I'm kind of on a deadline right now, but I'd like to arrange a meeting
on irc or
something like that some time if you can do it. Do you hang on #llvm ?
Duncan, I was writing an email to you, but then I noticed that you're
baldrick
from IRC, so I figured I might as well just contact you there :)
Cheers,
--lf
On Tue, Apr 22, 2008 at 4:58 PM, Torvald Riegel
<torvald at se.inf.tu-dresden.de> wrote:> On Tuesday 22 April 2008, Chris Lattner wrote:
> > [llvmir] "Software Transactional Memory (STM) support in
LLVM" by
> > Luis Felipe Strano
>
> Hi,
>
> I'd like to know more about the directions for the STM project.
>
> Somewhat contrary to what the project's abstract states, there are
open C/C++
> based implementations for compiler support for memory transactions.
Besides
> some efforts to modify GCC (AFAIK the respective groups haven't
published
> source code yet), there is also an open-source tool that is implemented as
an
> LLVM pass (Tanger, see http://tinystm.org/tanger).
>
> I'm part of a group that does research on transactional memory. We are
working
> on many aspects of (S)TM (STM algorithms and implementations, compiler
> support, infrastructure, ...) and collaborate with several other research
> groups. For example, I have worked on STM optimizations based on pointer
> analysis (see
> http://wwwse.inf.tu-dresden.de/ABSTRACTS/riegel2008partitioning.html and
> http://wwwse.inf.tu-dresden.de/ABSTRACTS/riegel2008objbased.html). These
> optimizations are implemented on top of our LLVM pass (see above) and use
> DSA.
>
> Coming back to the compiler support, I would suggest that the project
should
> try to use existing infrastructure (tools, code, ...) as much as possible.
> Many of the subproblems do have simple solutions, but integrating them can
be
> difficult because there is a lot of interaction, and design choices can
often
> have consequences that are hard to spot unless you have broader knowledge
> about TM.
>
> We can also offer support and advice regarding TM and the project's
topic if
> there should be interest in that.
>
>
> Torvald
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
--
"I am not a number, I am a free man" Number Six