DynamoRIO backend and C++11 thread

[gokcehan kara]

Hello,

I have got two questions..

I have been using DynamoRIO for binary instrumentation lately and I saw Google trademark in some of the source code. Since it is being supported by Google, are there any efforts/plans to port ThreadSanitizer to DynamoRIO or implement a brand new race detection tool in DynamoRIO? (there are none atm afaik, am I right?)

Secondly, I was wondering the current state of C++11 thread support in ThreadSanitizer as well as other race tools that you're aware (helgrind, drd, intel tc..)

hope this is not the wrong place to ask.

thanks in advance,
gokcehan

[Konstantin Serebryany]

gokcehan, 

We don't have plans to port ThreadSanitizer to DynamoRIO in the near future because we are concentrated on 

the new version based on LLVM. http://clang.llvm.org/docs/ThreadSanitizer.html

Neither version supports C++11 threading today, but it should be more or less trivial to add such support.

We will do it eventually, but you are welcome to help. :) 

--kcc 

--
You received this message because you are subscribed to the Google Groups "data-race-test" group.
To view this discussion on the web visit https://groups.google.com/d/msg/data-race-test/-/FMq1iMZY_wQJ.
To post to this group, send email to data-ra...@googlegroups.com.
To unsubscribe from this group, send email to data-race-tes...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/data-race-test?hl=en.

[Dmitry Vyukov]

On Wed, Oct 10, 2012 at 8:08 AM, Konstantin Serebryany <konstantin....@gmail.com> wrote:

gokcehan, 

We don't have plans to port ThreadSanitizer to DynamoRIO in the near future because we are concentrated on 

the new version based on LLVM. http://clang.llvm.org/docs/ThreadSanitizer.html

Neither version supports C++11 threading today, but it should be more or less trivial to add such support.

What exactly C++11 primitives are you interested in?

It supports C++11 atomic ops (at least with clang std lib, for others it depends on impl), it should also support thread/mutex if they are wrapped around pthread and/or use supported atomic ops.

[gokcehan kara]

well, I don't really know as I barely even used C++11 threads. I was just looking for a term project for my verification class. let me know if you have something I could be interested. I would be glad to contribute..

[Dmitry Vyukov]

On Wed, Oct 10, 2012 at 3:23 PM, gokcehan kara <gokceh...@gmail.com> wrote:

well, I don't really know as I barely even used C++11 threads. I was just looking for a term project for my verification class. let me know if you have something I could be interested. I would be glad to contribute..

Hi,

There are 2 separable pieces of ThreadSanitizer that come to mind:

1. So called scheduler shaker, that is, insert some delays into program execution to trigger/reveal more bugs. The interesting parts here are: (a) how to choose delay values and (b) how many new bugs we find with this approach (potentially as a function of something).

2. Modelling of C1x/C++11 relaxed memory model. This allows to reveal more bugs and find bugs that can happen mostly on POWER/ARM on x86 platform.

[Konstantin Serebryany]

On Wed, Oct 10, 2012 at 3:53 PM, Dmitry Vyukov <dvy...@google.com> wrote:

On Wed, Oct 10, 2012 at 3:23 PM, gokcehan kara <gokceh...@gmail.com> wrote:

well, I don't really know as I barely even used C++11 threads. I was just looking for a term project for my verification class. let me know if you have something I could be interested. I would be glad to contribute..

Hi,

There are 2 separable pieces of ThreadSanitizer that come to mind:

1. So called scheduler shaker, that is, insert some delays into program execution to trigger/reveal more bugs. The interesting parts here are: (a) how to choose delay values and (b) how many new bugs we find with this approach (potentially as a function of something).

2. Modelling of C1x/C++11 relaxed memory model. This allows to reveal more bugs and find bugs that can happen mostly on POWER/ARM on x86 platform.

3. writing positive (racy) and negative (correct) tsan tests using c++11