Implement 1D Adaptive Workload Splitting in CPPScheduler

Resolves COMPMID-4649
Change-Id: I941d2f8a40737ff05c49f6695a42884731ef2dc9
Signed-off-by: Dana Zlotnik <dana.zlotnik@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6656
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 13 insertions, 0 deletions

diff --git a/arm_compute/runtime/IScheduler.h b/arm_compute/runtime/IScheduler.h
index 3759fee8a8..f67357dcfd 100644
--- a/arm_compute/runtime/IScheduler.h
+++ b/arm_compute/runtime/IScheduler.h
@@ -215,6 +215,19 @@ protected:
      */
     void schedule_common(ICPPKernel *kernel, const Hints &hints, const Window &window, ITensorPack &tensors);
 
+    /** Adjust the number of windows to the optimize performance
+     * (used for small workloads where smaller number of threads might improve the performance)
+     *
+     * @param[in] window           Window to use for kernel execution
+     * @param[in] split_dimension  Axis of dimension to split
+     * @param[in] init_num_windows Initial number of sub-windows to split
+     * @param[in] kernel           Kernel to execute
+     * @param[in] cpu_info         The CPU platform used to create the context.
+     *
+     * @return Adjusted number of windows
+     */
+    std::size_t adjust_num_of_windows(const Window &window, std::size_t split_dimension, std::size_t init_num_windows, const ICPPKernel &kernel, const CPUInfo &cpu_info);
+
 private:
     unsigned int _num_threads_hint = {};
 };