llvm dev - Jan 2012 - [LLVMdev] Implementing the MC-JIT execution path for ELF, with debugging

Hello,

I'm part of a team at Intel that works on implementing the MC-JIT execution
path for ELF objects with built-in support for debugging the JITted code with
GDB. We currently have a working implementation passing all ExecutionEngine
tests on MC-JIT for ELF, as well as several proprietary workloads. The current
implementation focuses on Linux, although with some trivial modifications it can
also be made to work on Windows (pending on some changes to Triple, as discussed
in this thread:
http://lists.cs.uiuc.edu/pipermail/llvmdev/2012-January/046583.html).

I was granted "commit after approval" rights to the repo by Chris
Lattner, and will start delivering patches to llvm-commits today. To make the
review process manageable, we intend to split the patches into a few incremental
parts:


1.       Modifications (more precisely, extension) to ELFObjectFile (from
libObject) to support dynamic loading

2.       Implementation of MC-JIT for ELF, including a new RuntimeDyldELF class
(that parallels the existing RuntimeDyldMachO) which implements the dynamic
loading of the ELF image into memory, its relocation and execution.

3.       An implementation of a MC-JIT - GDB bridge which allows to fully debug
the JITted code with GDB.

These patches will be a superset of an earlier patch sent by Daniel Malea
(http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20111219/133702.html).
We'll be very happy to have the patches reviewed, and see a functional
MC-JIT implementation for ELF in LLVM trunk soon.

Eli
---------------------------------------------------------------------
Intel Israel (74) Limited

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20120109/56b9b1ca/attachment.html>

Hi Eli,

This is GREAT news!
So finally there will be JIT support for AVX machines on Linux, thank 
you so much! :)

Cheers,
Ralf


On 1/9/12 4:07 PM, Bendersky, Eli wrote:
> Hello,
>
> I’m part of a team at Intel that works on implementing the MC-JIT
> execution path for ELF objects with built-in support for debugging the
> JITted code with GDB. We currently have a working implementation passing
> all ExecutionEngine tests on MC-JIT for ELF, as well as several
> proprietary workloads. The current implementation focuses on Linux,
> although with some trivial modifications it can also be made to work on
> Windows (pending on some changes to Triple, as discussed in this thread:
> http://lists.cs.uiuc.edu/pipermail/llvmdev/2012-January/046583.html).
>
> I was granted “commit after approval” rights to the repo by Chris
> Lattner, and will start delivering patches to llvm-commits today. To
> make the review process manageable, we intend to split the patches into
> a few incremental parts:
>
> 1.Modifications (more precisely, extension) to ELFObjectFile (from
> libObject) to support dynamic loading
>
> 2.Implementation of MC-JIT for ELF, including a new RuntimeDyldELF class
> (that parallels the existing RuntimeDyldMachO) which implements the
> dynamic loading of the ELF image into memory, its relocation and execution.
>
> 3.An implementation of a MC-JIT – GDB bridge which allows to fully debug
> the JITted code with GDB.
>
> These patches will be a superset of an earlier patch sent by Daniel
> Malea
>
(http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20111219/133702.html).
>
> We’ll be very happy to have the patches reviewed, and see a functional
> MC-JIT implementation for ELF in LLVM trunk soon.
>
> Eli
>
> ---------------------------------------------------------------------
> Intel Israel (74) Limited
>
> This e-mail and any attachments may contain confidential material for
> the sole use of the intended recipient(s). Any review or distribution
> by others is strictly prohibited. If you are not the intended
> recipient, please contact the sender and delete all copies.
>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev