[PATCH v2 1/1] libcamera: add method to set thread affinity

[Cheng-Hao Yang]

Hi Laurent,

On Sun, Oct 13, 2024 at 11:58 PM Laurent Pinchart
<laurent.pinchart at ideasonboard.com> wrote:
>
> Hi Harvey, Han-Lin,
>
> Thank you for the patch.
>
> On Sat, Oct 12, 2024 at 12:47:57PM +0000, Harvey Yang wrote:
> > From: Han-Lin Chen <hanlinchen at chromium.org>
> >
> > Add method to set thread affinity to Thread class.
> >
> > Signed-off-by: Han-Lin Chen <hanlinchen at chromium.org>
> > Co-developed-by: Harvey Yang <chenghaoyang at chromium.org>
> > Signed-off-by: Harvey Yang <chenghaoyang at chromium.org>
> > ---
> >  include/libcamera/base/thread.h |  6 +++++
> >  src/libcamera/base/thread.cpp   | 47 +++++++++++++++++++++++++++++++++
> >  2 files changed, 53 insertions(+)
> >
> > diff --git a/include/libcamera/base/thread.h b/include/libcamera/base/thread.h
> > index 4f33de63d..c74fe102b 100644
> > --- a/include/libcamera/base/thread.h
> > +++ b/include/libcamera/base/thread.h
> > @@ -35,6 +35,8 @@ public:
> >       void exit(int code = 0);
> >       bool wait(utils::duration duration = utils::duration::max());
> >
> > +     void setThreadAffinity(const std::vector<int> &cpus);
> > +
> >       bool isRunning();
> >
> >       Signal<> finished;
> > @@ -54,6 +56,8 @@ private:
> >       void startThread();
> >       void finishThread();
> >
> > +     void setThreadAffinityInternal();
> > +
> >       void postMessage(std::unique_ptr<Message> msg, Object *receiver);
> >       void removeMessages(Object *receiver);
> >
> > @@ -67,6 +71,8 @@ private:
> >
> >       std::thread thread_;
> >       ThreadData *data_;
> > +
> > +     std::vector<int> cpus_;
>
> You can move this to ThreadData so it won't affect the API.

Got it. Done.

>
> >  };
> >
> >  } /* namespace libcamera */
> > diff --git a/src/libcamera/base/thread.cpp b/src/libcamera/base/thread.cpp
> > index 8735670b8..de5423eda 100644
> > --- a/src/libcamera/base/thread.cpp
> > +++ b/src/libcamera/base/thread.cpp
> > @@ -254,6 +254,8 @@ void Thread::start()
> >       data_->exit_.store(false, std::memory_order_relaxed);
> >
> >       thread_ = std::thread(&Thread::startThread, this);
> > +
> > +     setThreadAffinityInternal();
> >  }
> >
> >  void Thread::startThread()
> > @@ -410,6 +412,51 @@ bool Thread::wait(utils::duration duration)
> >       return hasFinished;
> >  }
> >
> > +/**
> > + * \brief Set the CPU affinity mask of the thread
> > + * \param[in] cpus The list of CPU indices that the thread is set affinity to
>
> As there's error handling in the function, you should describe what
> happens in case of errors.

Added description.

>
> > + */
> > +void Thread::setThreadAffinity(const std::vector<int> &cpus)
>
> Make it a span so the caller can also use an array instead of a vector.

Done.

>
> > +{
> > +     cpus_.clear();
> > +
> > +     const unsigned int num_cpus = std::thread::hardware_concurrency();
> > +     for (const int &cpu : cpus) {
> > +             if (cpus_.size() == CPU_SETSIZE) {
> > +                     LOG(Thread, Error) << "cpus_ already contains " << CPU_SETSIZE
> > +                                        << " cpus. Ignoring the rest.";
> > +                     break;
> > +             }
> > +
> > +             if (cpu >= static_cast<int>(num_cpus) || cpu < 0) {
>
> Why is the argument to the function a vector of int if entries can't be
> negative ? Can't it be a vector of unsigned int instead ? You'll be able
> to drop the cast here and drop the second check.

Right, I thought CPU_SET only takes int, while I was wrong.

>
> > +                     LOG(Thread, Error) << "Ignore an invalid cpu index: " << cpu;
> > +                     continue;
> > +             }
> > +
> > +             cpus_.push_back(cpu);
> > +     }
> > +
> > +     MutexLocker locker(data_->mutex_);
> > +     if (data_->running_)
> > +             setThreadAffinityInternal();
> > +}
> > +
> > +void Thread::setThreadAffinityInternal()
> > +{
> > +     if (cpus_.empty())
> > +             return;
> > +
> > +     auto handle = thread_.native_handle();
> > +
> > +     cpu_set_t cpuset;
> > +     CPU_ZERO(&cpuset);
> > +
> > +     for (auto cpu : cpus_)
> > +             CPU_SET(cpu, &cpuset);
>
> Can't you compute the cpu_set_t and store it in the ThreadData class,
> instead of storing a vector ? This can be done without holding the lock.

Done, while I still think that it should only be updated when holding the
lock, to avoid race conditions.

>
> > +
> > +     pthread_setaffinity_np(handle, sizeof(cpu_set_t), &cpuset);
>
> sizeof(cpu_set_t)

Sorry, I don't get it: I think I already used `sizeof(cpu_set_t)`.


>
> > +}
> > +
> >  /**
> >   * \brief Check if the thread is running
> >   *
>
> --
> Regards,
>
> Laurent Pinchart