COMPMID-3565: Exposes interface to enable thread binding

Expose `set_num_threads_with_affinity` as an interface to the `IScheduler` to
allow binding of threads to given logical cores.

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I062db7caafb0101972ba45d31ee9e61b26800127
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3481
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

2 files changed, 19 insertions, 30 deletions

diff --git a/arm_compute/runtime/CPP/CPPScheduler.h b/arm_compute/runtime/CPP/CPPScheduler.h
index 855535ebce..9d55ed448e 100644
--- a/arm_compute/runtime/CPP/CPPScheduler.h
+++ b/arm_compute/runtime/CPP/CPPScheduler.h
@@ -39,16 +39,6 @@ public:
     CPPScheduler();
     /** Default destructor */
     ~CPPScheduler();
-    /** Sets the number of threads the scheduler will use to run the kernels.
-     *
-     * @param[in] num_threads If set to 0, then the maximum number of threads supported by C++11 will be used, otherwise the number of threads specified.
-     */
-    void set_num_threads(unsigned int num_threads) override;
-    /** Returns the number of threads that the CPPScheduler has in his pool.
-     *
-     * @return Number of threads available in CPPScheduler.
-     */
-    unsigned int num_threads() const override;
 
     /** Access the scheduler singleton
      *
@@ -56,27 +46,12 @@ public:
      * @return The scheduler
      */
     static CPPScheduler &get();
-    /** Multithread the execution of the passed kernel if possible.
-     *
-     * The kernel will run on a single thread if any of these conditions is true:
-     * - ICPPKernel::is_parallelisable() returns false
-     * - The scheduler has been initialized with only one thread.
-     *
-     * @param[in] kernel Kernel to execute.
-     * @param[in] hints  Hints for the scheduler.
-     */
+
+    // Inherited functions overridden
+    void set_num_threads(unsigned int num_threads) override;
+    void set_num_threads_with_affinity(unsigned int num_threads, BindFunc func) override;
+    unsigned int num_threads() const override;
     void schedule(ICPPKernel *kernel, const Hints &hints) override;
-    /** Multithread the execution of the passed kernel if possible.
-     *
-     * The kernel will run on a single thread if any of these conditions is true:
-     * - ICPPKernel::is_parallelisable() returns false
-     * - The scheduler has been initialized with only one thread.
-     *
-     * @param[in] kernel  Kernel to execute.
-     * @param[in] hints   Hints for the scheduler.
-     * @param[in] inputs  Vector that contains the input tensors.
-     * @param[in] outputs Vector that contains the output tensors.
-     */
     void schedule_op(ICPPKernel *kernel, const Hints &hints, const InputTensorMap &inputs, const OutputTensorMap &outputs) override;
 
 protected:
diff --git a/arm_compute/runtime/IScheduler.h b/arm_compute/runtime/IScheduler.h
index 9382c20b29..fff77274bd 100644
--- a/arm_compute/runtime/IScheduler.h
+++ b/arm_compute/runtime/IScheduler.h
@@ -47,6 +47,13 @@ public:
         DYNAMIC, /**< Split the workload dynamically using a bucket system */
     };
 
+    /** Function to be used and map a given thread id to a logical core id
+     *
+     * Mapping function expects the thread index and total number of cores as input,
+     * and returns the logical core index to bind against
+     */
+    using BindFunc = std::function<int(int, int)>;
+
     /** When arm_compute::ISchedular::Hints::_split_dimension is initialized with this value
      * then the schedular is free to break down the problem space over as many dimensions
      * as it wishes
@@ -137,6 +144,13 @@ public:
      */
     virtual void set_num_threads(unsigned int num_threads) = 0;
 
+    /** Sets the number of threads the scheduler will use to run the kernels but also using a binding function to pin the threads to given logical cores
+     *
+     * @param[in] num_threads If set to 0, then one thread per CPU core available on the system will be used, otherwise the number of threads specified.
+     * @param[in] func        Binding function to use.
+     */
+    virtual void set_num_threads_with_affinity(unsigned int num_threads, BindFunc func);
+
     /** Returns the number of threads that the SingleThreadScheduler has in his pool.
      *
      * @return Number of threads available in SingleThreadScheduler.