COMPMID-3203: Fix build failure with GCC 9.2

Rework convolve3x3 to avoid erroneous behavior by the register allocator.

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: Ifff2f4ae3a95b894462c7457ffba1f710cce0577
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/2839
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
diff --git a/src/core/NEON/kernels/NEDepthwiseConvolutionLayer3x3Kernel.cpp b/src/core/NEON/kernels/NEDepthwiseConvolutionLayer3x3Kernel.cpp
index 1dd05d2..03b9622 100644
--- a/src/core/NEON/kernels/NEDepthwiseConvolutionLayer3x3Kernel.cpp
+++ b/src/core/NEON/kernels/NEDepthwiseConvolutionLayer3x3Kernel.cpp
@@ -116,8 +116,7 @@
                 {
                     if(dilation == Size2D(1U, 1U))
                     {
-                        auto vres = detail::convolve_3x3(in_top, in_mid, in_low, vw_r0, vw_r1, vw_r2, stridex, input_offset);
-                        detail::store_results<stridex>(p_out, vres);
+                        detail::convolve_3x3<false>(in_top, in_mid, in_low, p_out, vw_r0, vw_r1, vw_r2, stridex, input_offset);
                     }
                     else
                     {