Add FP32 Neon™ swish activation

Change-Id: Id37b59adbc8c4cbe218d1652aeb02a0b4ce42c66
Signed-off-by: Jonathan Deakin <jonathan.deakin@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8256
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
diff --git a/src/cpu/kernels/activation/generic/neon/impl.h b/src/cpu/kernels/activation/generic/neon/impl.h
index 4d4aa8d..4f39286 100644
--- a/src/cpu/kernels/activation/generic/neon/impl.h
+++ b/src/cpu/kernels/activation/generic/neon/impl.h
@@ -149,6 +149,9 @@
                 case ActivationLayerInfo::ActivationFunction::HARD_SWISH:
                     tmp = wrapper::vmul(vin, wrapper::vmul(const_inv_6, wrapper::vmin(const_6, wrapper::vmax(const_0, wrapper::vadd(vin, const_3)))));
                     break;
+                case ActivationLayerInfo::ActivationFunction::SWISH:
+                    tmp = wrapper::vmul(vin, wrapper::vinv(wrapper::vadd(const_1, wrapper::vexpq(wrapper::vneg(wrapper::vmul(va, vin))))));
+                    break;
 #ifdef __aarch64__
                 case ActivationLayerInfo::ActivationFunction::GELU:
                     tmp = wrapper::vmul(vin, wrapper::vmul(const_inv_2, wrapper::vadd(const_1, wrapper::verf(wrapper::vmul(vin, const_inv_sqrt_2)))));
@@ -208,6 +211,9 @@
                 case ActivationLayerInfo::ActivationFunction::HARD_SWISH:
                     tmp = in * ((std::min(std::max((in + 3), 0.0f), 6.0f)) * 0.166666667f);
                     break;
+                case ActivationLayerInfo::ActivationFunction::SWISH:
+                    tmp = in / (static_cast<T>(1) + std::exp(-a*in));
+                    break;
                 case ActivationLayerInfo::ActivationFunction::GELU:
                     tmp = in * static_cast<T>(0.5f * (1.0f + erff(static_cast<float>(in) / 1.41421356237f)));
                     break;