qthreads

qthreads copied to clipboard

[WIP] Re-enable CI testing on Mac

1

Re-enables CI testing on Mac OS.

janciesko

As reported by @ronawho > I'm seeing failures for a couple types of tests. > > One failure for tests that sleep, [here](https://github.com/chapel-lang/chapel/blob/4c48d6191fa2ea46e2dcb8964a6b2d2327f865a4/runtime/src/tasks/qthreads/tasks-qthreads.c#L950-L983) it seems like when doing sleeps from...

janciesko

Thread Sanitizer Interop

4

This adds the API calls needed to help the thread sanitizer correctly carry its happens-before relationships across context swaps.

insertinterestingnamehere

Race Conditions in `qutil`

2

I'm seeing a handful of thread sanitizer errors about race conditions in the `qutil` code. One example: apparently the macro invoked at https://github.com/sandialabs/qthreads/blob/a2d1dbadb80a161a4a287c1056348808b022ba04/src/qutil.c#L152, https://github.com/sandialabs/qthreads/blob/a2d1dbadb80a161a4a287c1056348808b022ba04/src/qutil.c#L156, and https://github.com/sandialabs/qthreads/blob/a2d1dbadb80a161a4a287c1056348808b022ba04/src/qutil.c#L160 includes both a read...

insertinterestingnamehere

Explicit Atomics Needed In FEB Implementation

3

The current implementation of FEB synchronization depends on the guarantees of x86 loads and stores. For example, when running the `hello_world_multi` test with thread sanitizer enabled numerous data races show...

insertinterestingnamehere

The thread sanitizer is flagging several issues with the Sherwood threadqueue too. Here are a few of them: Non-atomic write: https://github.com/sandialabs/qthreads/blob/a2d1dbadb80a161a4a287c1056348808b022ba04/src/threadqueues/sherwood_threadqueues.c#L322 Non-atomic read: https://github.com/sandialabs/qthreads/blob/a2d1dbadb80a161a4a287c1056348808b022ba04/src/threadqueues/sherwood_threadqueues.c#L845 Something non-atomic is still happening in...

insertinterestingnamehere

The thread sanitizer is flagging various issues in the distrib threadqueue. A few examples: Non-atomic write: https://github.com/sandialabs/qthreads/blob/a2d1dbadb80a161a4a287c1056348808b022ba04/src/threadqueues/distrib_threadqueues.c#L207 Non-atomic read: https://github.com/sandialabs/qthreads/blob/a2d1dbadb80a161a4a287c1056348808b022ba04/src/threadqueues/distrib_threadqueues.c#L265 Non-atomic read: https://github.com/sandialabs/qthreads/blob/a2d1dbadb80a161a4a287c1056348808b022ba04/src/threadqueues/distrib_threadqueues.c#L216 Non-atomic write: https://github.com/sandialabs/qthreads/blob/a2d1dbadb80a161a4a287c1056348808b022ba04/src/threadqueues/distrib_threadqueues.c#L374 There are several more.

insertinterestingnamehere

Deliberate Symbol Visibility

2

It's generally a good a idea to set the default symbol visibility to "hidden" while compiling a library so that internal functions don't get automatically included in the interface exposed...

insertinterestingnamehere

Syncvar Nees Atomic Loads and Stores

6

Our syncvar implementation also needs to be updated to use explicit atomic reads and writes instead of just relying on the x86 memory consistency guarantees. One example race: https://github.com/sandialabs/qthreads/blob/d6ce514a70c65b74c5e04906615ec51c7f288e0f/src/syncvar.c#L175 and...

insertinterestingnamehere

Drop Support for Unused/Abandoned Architectures

1

We currently have a ton of old untested ifdefs for a bunch of crazy architectures that don't really exist in practice anymore (e.g. Tilera). We should get rid of that...

insertinterestingnamehere