[PATCH} [Review Request] MCJIT PIC support for x86-64

[Kaylor, Andrew]

What we need for MCJIT is the ability to run a single test with a variety of command line options and have the XFAIL status be specific to the command line options used.

This goes beyond just running the tests once with JIT and once with MCJIT.  Within MCJIT I'd like to exercise a variety of code model and relocation model combinations as well as remote execution.  Remote execution is a particular problem because only a few of the tests make sense to run that way at the moment.

Within those constraints, I'd be happy to adopt whatever solution is available.

-Andy

From: daniel.dunbar at gmail.com [mailto:daniel.dunbar at gmail.com] On Behalf Of Daniel Dunbar
Sent: Friday, August 16, 2013 3:44 PM
To: Eli Bendersky
Cc: Kaylor, Andrew; Commit Messages and Patches for LLVM
Subject: Re: [PATCH} [Review Request] MCJIT PIC support for x86-64

On Fri, Aug 16, 2013 at 2:56 PM, Eli Bendersky <eliben at google.com<mailto:eliben at google.com>> wrote:
On Fri, Aug 16, 2013 at 11:40 AM, Kaylor, Andrew <andrew.kaylor at intel.com<mailto:andrew.kaylor at intel.com>> wrote:
Hi Eli,

Thanks for the suggestions regarding documentation.  I'll see what I can put together.  I believe I still have the diagrams you referred to, and I can put together something more to describe the control flow.

BTW, I wanted to let you know that your old blog posts describing the PIC relocation model were extremely helpful as I was putting this patch together.  Yours was the clearest description of PIC handling that I was able to find.

Glad it helped! I'll have to reread them myself before I review your patch ;-)


Also, I meant to ask what ever happened to the lit subtest patch that you submitted about a year and a half ago.  As far as I can tell it was never committed, but I couldn't tell why not.  The "new" tests I'm introducing are just duplications of other tests with variations in invocation arguments.  It would be nice to be able to simplify that ExecutionEngine testing tree.

I recall the patch was very close to being done, and Daniel Dunbar was OK with it, but somehow in the last stages of reviewing we both got distracted. The real problem is that the last versions of the patch were mailed directly to Daniel from... my old @intel.com<http://intel.com> address. So unless Daniel (CCd) still has them lying somewhere, I fear the patch is lost :-/ Not that it was a big deal, and it would have to be tweaked in light of the recent additions to FileCheck anyhow. I still believe it's a good thing to have and will be happy to help reviewing a new patch if anyone is willing to take it over.

I've attached the last version I got from you. I think if we were going to revisit adding this I might want to take a slightly different approach in light of some other directions I would like to take lit. Primarily I would like to consider lifting the subtest handling up to the generic lit layer instead of specific to the ShTest format. There are other test formats I am hoping to develop that could make use of some kind of explicit subtest support. This would also solve one of the outstanding issues with the patch which was how to compute an overall test result status in the presence of XFAILs, XPASS, etc.

I would also like to know how support for parameterized tests overlaps with this. For MCJIT specifically it sounds like what is really desired is some support for parameterized tests so that all the JIT tests can be run with and without the MCJIT. That feature has come up in the past as well, and could be really useful in cases where we want to, e.g., run a test for each configured target.

 - Daniel


Eli





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130816/59451e43/attachment.html>