Add function declarations for quantized_conv2d_nchw/nhwc for vision

backends/cadence/aot/functions_vision.yaml

@@ -195,6 +195,16 @@
     - arg_meta: null
       kernel_name: impl::vision::native::quantized_conv_out
 
+- func: cadence::quantized_conv2d_nchw.out(Tensor input, Tensor weight, Tensor bias, int[] stride, SymInt[] padding, int[] dilation, int groups, int input_zero_point, Tensor weight_zero_point, Tensor bias_scale, float out_scale, int out_zero_point, Tensor out_multiplier, Tensor out_shift, *, Tensor(a!) out) -> Tensor(a!)
+  kernels:
+    - arg_meta: null
+      kernel_name: impl::vision::native::quantized_conv2d_nchw_out
+
+- func: cadence::quantized_conv2d_nhwc.out(Tensor input, Tensor weight, Tensor bias, int[] stride, SymInt[] padding, int[] dilation, int groups, int input_zero_point, Tensor weight_zero_point, Tensor bias_scale, float out_scale, int out_zero_point, Tensor out_multiplier, Tensor out_shift, *, Tensor(a!) out) -> Tensor(a!)
+  kernels:
+    - arg_meta: null
+      kernel_name: impl::vision::native::quantized_conv2d_nhwc_out
+
 - func: cadence::quantized_layer_norm.out(Tensor input, Tensor in_scale, Tensor in_zero_point, int[] normalized_shape, Tensor weight, Tensor bias, float eps, float output_scale, int output_zero_point, *, Tensor(a!) out) -> Tensor(a!)
   kernels:
     - arg_meta: null

backends/cadence/vision/operators/op_quantized_conv_out.cpp

@@ -603,6 +603,80 @@ void quantized_conv2d_nhwc_per_tensor_out(
       out);
 }
 
+void quantized_conv2d_nchw_out(
+    KernelRuntimeContext& ctx,
+    const Tensor& input,
+    const Tensor& weight,
+    const Tensor& bias,
+    IntArrayRef stride,
+    IntArrayRef padding,
+    IntArrayRef dilation,
+    int64_t groups,
+    int64_t in_zero_point,
+    const Tensor& weight_zero_point,
+    const Tensor& bias_scale,
+    double output_scale,
+    int64_t output_zero_point,
+    const Tensor& out_multiplier,
+    const Tensor& out_shift,
+    Tensor& out) {
+  quantized_conv_out(
+      ctx,
+      input,
+      weight,
+      bias,
+      stride,
+      padding,
+      dilation,
+      groups,
+      in_zero_point,
+      weight_zero_point,
+      bias_scale,
+      output_scale,
+      output_zero_point,
+      out_multiplier,
+      out_shift,
+      false, // channel_last = false for NCHW
+      out);
+}
+
+void quantized_conv2d_nhwc_out(
+    KernelRuntimeContext& ctx,
+    const Tensor& input,
+    const Tensor& weight,
+    const Tensor& bias,
+    IntArrayRef stride,
+    IntArrayRef padding,
+    IntArrayRef dilation,
+    int64_t groups,
+    int64_t in_zero_point,
+    const Tensor& weight_zero_point,
+    const Tensor& bias_scale,
+    double output_scale,
+    int64_t output_zero_point,
+    const Tensor& out_multiplier,
+    const Tensor& out_shift,
+    Tensor& out) {
+  quantized_conv_out(
+      ctx,
+      input,
+      weight,
+      bias,
+      stride,
+      padding,
+      dilation,
+      groups,
+      in_zero_point,
+      weight_zero_point,
+      bias_scale,
+      output_scale,
+      output_zero_point,
+      out_multiplier,
+      out_shift,
+      true, // channel_last = true for NHWC
+      out);
+}
+
 } // namespace native
 } // namespace vision
 } // namespace impl

backends/cadence/vision/operators/operators.h

@@ -55,6 +55,42 @@ inline __attribute__((always_inline)) void linear_(
   }
 }
 
+void quantized_conv2d_nchw_out(
+    ::executorch::runtime::KernelRuntimeContext& ctx,
+    const ::executorch::aten::Tensor& input,
+    const ::executorch::aten::Tensor& weight,
+    const ::executorch::aten::Tensor& bias,
+    ::executorch::aten::IntArrayRef stride,
+    ::executorch::aten::IntArrayRef padding,
+    ::executorch::aten::IntArrayRef dilation,
+    int64_t groups,
+    int64_t in_zero_point,
+    const ::executorch::aten::Tensor& weight_zero_point,
+    const ::executorch::aten::Tensor& bias_scale,
+    double output_scale,
+    int64_t output_zero_point,
+    const ::executorch::aten::Tensor& out_multiplier,
+    const ::executorch::aten::Tensor& out_shift,
+    ::executorch::aten::Tensor& out);
+
+void quantized_conv2d_nhwc_out(
+    ::executorch::runtime::KernelRuntimeContext& ctx,
+    const ::executorch::aten::Tensor& input,
+    const ::executorch::aten::Tensor& weight,
+    const ::executorch::aten::Tensor& bias,
+    ::executorch::aten::IntArrayRef stride,
+    ::executorch::aten::IntArrayRef padding,
+    ::executorch::aten::IntArrayRef dilation,
+    int64_t groups,
+    int64_t in_zero_point,
+    const ::executorch::aten::Tensor& weight_zero_point,
+    const ::executorch::aten::Tensor& bias_scale,
+    double output_scale,
+    int64_t output_zero_point,
+    const ::executorch::aten::Tensor& out_multiplier,
+    const ::executorch::aten::Tensor& out_shift,
+    ::executorch::aten::Tensor& out);
+
 } // namespace native
 } // namespace vision
 } // namespace impl