Add support for non-constant weights and biases in CpuFullyConnected

Changing the approach for specifying that weights and biases tensors are
non-constant by making it a member of TensorInfo rather than an option
of the functions.

Resolves: COMPMID-4222, COMPMID-4811

Signed-off-by: Giorgio Arena <giorgio.arena@arm.com>
Change-Id: I9b0081ccbcf8271ce029ba6755563d64c59e1d32
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6313
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 6 insertions, 3 deletions

diff --git a/src/core/NEON/kernels/arm_gemm/quantize_wrapper.hpp b/src/core/NEON/kernels/arm_gemm/quantize_wrapper.hpp
index 1e2a9acc1d..ce727032e6 100644
--- a/src/core/NEON/kernels/arm_gemm/quantize_wrapper.hpp
+++ b/src/core/NEON/kernels/arm_gemm/quantize_wrapper.hpp
@@ -179,13 +179,16 @@ public:
         return _subgemm->get_B_pretransposed_array_size() + col_sum_size();
     }
 
+    void requantize_bias(void *in_buffer, const To *B, const int ldb, const int B_multi_stride) override {
+        _col_sums = reinterpret_cast<int32_t *>(in_buffer);
+        col_sums_pretransposed(B, ldb, B_multi_stride);
+    }
+
     void pretranspose_B_array(void *buffer, const To *B, const int ldb, const int B_multi_stride) override {
         uintptr_t buffer_int = reinterpret_cast<uintptr_t>(buffer);
         _subgemm->pretranspose_B_array(reinterpret_cast<void *>(buffer_int + col_sum_size()), B, ldb, B_multi_stride);
 
-        _col_sums = reinterpret_cast<int32_t *>(buffer);
-
-        col_sums_pretransposed(B, ldb, B_multi_stride);
+        requantize_bias(buffer, B, ldb, B_multi_stride);
     }
 
     void set_pretransposed_B_data(void *buffer) override {