diff --git a/onnxruntime/contrib_ops/cpu/cpu_contrib_kernels.cc b/onnxruntime/contrib_ops/cpu/cpu_contrib_kernels.cc
index 31b1b12d38..acb79b8076 100644
--- a/onnxruntime/contrib_ops/cpu/cpu_contrib_kernels.cc
+++ b/onnxruntime/contrib_ops/cpu/cpu_contrib_kernels.cc
@@ -79,6 +79,7 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1,
 // ******** End: Quantization ******************* //
 
 #ifdef MLAS_F16VEC_INTRINSICS_SUPPORTED
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MLFloat16, FusedConv);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MLFloat16, NhwcFusedConv);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSInternalNHWCDomain, 12, MLFloat16, MaxPool);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSInternalNHWCDomain, 11, MLFloat16, AveragePool);
@@ -158,7 +159,7 @@ Status RegisterNchwcKernels(KernelRegistry& kernel_registry) {
 #ifdef MLAS_F16VEC_INTRINSICS_SUPPORTED
 Status RegisterFp16Kernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
-      BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MLFloat16, FusedConv)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MLFloat16, NhwcFusedConv)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSInternalNHWCDomain, 12, MLFloat16, MaxPool)>,
       BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSInternalNHWCDomain, 11, MLFloat16, AveragePool)>,