commit f77b969d3375f412ff236586d2b6eb2297c6b96d
author ramy.elgammal@arm.com <ramy.elgammal@arm.com> Mon Aug 07 17:07:02 2023 +0100
committer Ramy Elgammal <ramy.elgammal@arm.com> Tue Aug 08 13:51:06 2023 +0000
tree 8d82c6ebdee3ffeaf17046d529bd0e87746c1545
parent 78da34c101d22e13f657e07e56e4baf52c88a016

Avoid using CLMatMul in CLFullyConnected when GPUTarget is Midgard

Resolves: COMPMID-6428
Change-Id: I255a46e894bafe1d3a26f0aebb828911bbfd750b
Signed-off-by: ramy.elgammal@arm.com <ramy.elgammal@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10070
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

src/gpu/cl/operators/ClFullyConnected.cpp

diff --git a/src/gpu/cl/operators/ClFullyConnected.cpp b/src/gpu/cl/operators/ClFullyConnected.cpp
index f71bf41..5845bbc 100644
--- a/src/gpu/cl/operators/ClFullyConnected.cpp
+++ b/src/gpu/cl/operators/ClFullyConnected.cpp
@@ -298,6 +298,7 @@
                                  FullyConnectedLayerInfo fc_info)
 {
     ARM_COMPUTE_ERROR_ON_NULLPTR(src, weights, dst);
+    const GPUTarget gpu_target = get_arch_from_target(CLScheduler::get().target());
 
     // Perform validate step
     ARM_COMPUTE_ERROR_THROW_ON(ClFullyConnected::validate(src, weights, biases, dst, fc_info));
@@ -316,7 +317,7 @@
     // 2. MatMul does not support broadcasting batch dimension, and therefore is disabled if fc is batched.
     // 3. When FC is after convolution and src tensor data layout does not match weights trained data layout (weights conversion kernel is required)
     const bool is_batched_fc_layer = dst->dimension(1) > 1;
-    _use_matmul                    = !weights->are_values_constant() && !is_batched_fc_layer && !(src->num_dimensions() > 1 && (src->data_layout() != fc_info.weights_trained_layout));
+    _use_matmul                    = gpu_target != GPUTarget::MIDGARD && !weights->are_values_constant() && !is_batched_fc_layer && !(src->num_dimensions() > 1 && (src->data_layout() != fc_info.weights_trained_layout));
     _dynamic_gemm                  = !weights->are_values_constant() && _transpose_weights && !_use_matmul;
 
     // With the Fully Connected layer we can have 4 different cases:
@@ -432,6 +433,7 @@
     ARM_COMPUTE_RETURN_ERROR_ON(weights->num_dimensions() > 2);
     ARM_COMPUTE_RETURN_ERROR_ON(fc_info.activation_info.enabled() && is_data_type_quantized(src->data_type()) && fc_info.activation_info.activation() != ActivationLayerInfo::ActivationFunction::RELU
                                 && fc_info.activation_info.activation() != ActivationLayerInfo::ActivationFunction::BOUNDED_RELU && fc_info.activation_info.activation() != ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU);
+    const GPUTarget gpu_target = get_arch_from_target(CLScheduler::get().target());
 
     const bool transpose_weights = fc_info.transpose_weights ? !fc_info.are_weights_reshaped : false;
     bool       is_fc_after_conv  = true;
@@ -439,7 +441,7 @@
     // When using dynamic weights - use matmul kernels.
     // Note: MatMul does not support broadcasting so fallback with batched cases.
     const bool is_batched_fc_layer = dst->dimension(1) > 1;
-    const bool use_matmul          = !weights->are_values_constant() && !is_batched_fc_layer && !(src->num_dimensions() > 1 && (src->data_layout() != fc_info.weights_trained_layout));
+    const bool use_matmul          = gpu_target != GPUTarget::MIDGARD && !weights->are_values_constant() && !is_batched_fc_layer && !(src->num_dimensions() > 1 && (src->data_layout() != fc_info.weights_trained_layout));
 
     const ITensorInfo &flatten_src       = TensorInfo(src->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(compute_flatten_shape(src)).set_data_layout(DataLayout::NCHW));
     const ITensorInfo &reshaped_weights  = TensorInfo(weights->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(compute_transposed_shape(*weights)));