COMPMID-661 Add optimal lws heuristics for the Bifrost direct_convolution kernels #45
Change-Id: I9e7ec5ed937fb4e8cab44a11c49a93f3aa01bedb
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/110877
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
diff --git a/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp b/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp
index e1901af..3c5799f 100644
--- a/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp
+++ b/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp
@@ -131,6 +131,23 @@
unsigned int num_elems_written_per_iteration_x = 0;
unsigned int num_elems_written_per_iteration_y = 0;
+ // Through extensive experimentation with over 30 representative tensor
+ // shapes, we found a small number of local work size configurations
+ // that result in nearly optimal execution times. Selecting the right
+ // lws for a given shape, however, required a complex decision tree,
+ // until we constructed a simple feature as described below.
+ //
+ // We started from the number of multiply-accumulate operations for a
+ // convolution layer, which is equal to the product of the input
+ // dimensions 0..2 and the weights dimensions 0..2. Unfortunately,
+ // this resulted in ties between distinct shapes that required distinct
+ // lws configurations. Replacing the width of the input with the kernel
+ // size, however, resulted in nearly optimal predictions. We use underscores
+ // in variable names to indicate when they are intentionally misleading.
+ const size_t product_of_weights_dimensions = weights->info()->dimension(0) * weights->info()->dimension(1) * weights->info()->dimension(2);
+ const size_t product_of_input_dimensions_ = input->info()->dimension(0) * weights->info()->dimension(1) * input->info()->dimension(2);
+ const float mega_ops_ = 1e-6 * product_of_weights_dimensions * product_of_input_dimensions_;
+
switch(kernel_size)
{
case 1:
@@ -139,6 +156,18 @@
num_elems_read_per_iteration_y = 4;
num_elems_written_per_iteration_x = 4;
num_elems_written_per_iteration_y = 4;
+ if(mega_ops_ < 1.f)
+ {
+ _lws_hint = cl::NDRange(1, 1, 8);
+ }
+ else if(mega_ops_ < 7.f)
+ {
+ _lws_hint = cl::NDRange(1, 1, 4);
+ }
+ else
+ {
+ _lws_hint = cl::NDRange(1, 1, 2);
+ }
break;
}
case 3:
@@ -147,6 +176,22 @@
num_elems_read_per_iteration_y = 5;
num_elems_written_per_iteration_x = 4;
num_elems_written_per_iteration_y = 3;
+ if(mega_ops_ < 1.f)
+ {
+ _lws_hint = cl::NDRange(1, 1, 8);
+ }
+ else if(mega_ops_ < 13.f)
+ {
+ _lws_hint = cl::NDRange(2, 1, 4);
+ }
+ else if(mega_ops_ < 50.f)
+ {
+ _lws_hint = cl::NDRange(3, 1, 4);
+ }
+ else
+ {
+ _lws_hint = cl::NDRange(2, 1, 6);
+ }
break;
}
case 5:
@@ -155,6 +200,14 @@
num_elems_read_per_iteration_y = 6;
num_elems_written_per_iteration_x = 4;
num_elems_written_per_iteration_y = 2;
+ if(mega_ops_ < 2.f || mega_ops_ > 80.f)
+ {
+ _lws_hint = cl::NDRange(2, 1, 4);
+ }
+ else
+ {
+ _lws_hint = cl::NDRange(2, 1, 8);
+ }
break;
}
default: