Add optimization for global pooling in pooling_layer.cl
- Simplify the implementation when the pooling size has the same spatial
dimensions of the input tensor
- Rework the heuristic for F32/F16
- Add test for validating the global pooling path
- Fix compare_dimensions in validation. The validation fails because we have different
number of dimensions for NCHW and NHWC (e.g. 1,1,2,1(NCHW) -> 2,1,1,1(NHWC)
Change-Id: Iba680cb30bf2a5d0952265a4cc9794f368549ca5
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5510
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
diff --git a/tests/validation/Validation.h b/tests/validation/Validation.h
index a75562b..f1ce0fe 100644
--- a/tests/validation/Validation.h
+++ b/tests/validation/Validation.h
@@ -159,11 +159,13 @@
{
// In case a 1D/2D shape becomes 3D after permutation, the permuted tensor will have two/one dimension(s) more and the first (two) value(s) will be 1
// clang-format off
- if((dimensions1.num_dimensions() != dimensions2.num_dimensions()) &&
- ((dimensions1.num_dimensions() != (dimensions2.num_dimensions() + 1)) || (dimensions1.x() != 1)) &&
- ((dimensions1.num_dimensions() != (dimensions2.num_dimensions() + 2)) || (dimensions1.x() != 1) || (dimensions1.y() != 1)))
+ const auto max_dims = std::max(dimensions1.num_dimensions(), dimensions2.num_dimensions());
+ for(unsigned int i = 3; i < max_dims; ++i)
{
- return false;
+ if(dimensions1[i] != dimensions2[i])
+ {
+ return false;
+ }
}
// clang-format on
@@ -171,14 +173,6 @@
{
return false;
}
-
- for(unsigned int i = 3; i < dimensions1.num_dimensions(); ++i)
- {
- if(dimensions1[i] != dimensions2[i])
- {
- return false;
- }
- }
}
return true;