create quantized_linear_per_tensor_out in cpu

Author: Zonglin Peng

backends/cadence/reference/operators/operators.h

@@ -14,8 +14,9 @@ namespace native {
 namespace {
 using ::executorch::runtime::getLeadingDims;
 
+
 #define ET_FORALL_CADENCE_QUANTIZED_TYPES(_) \
-  _(uint8_t, Byte)                           \
+  _(uint8_t, Byte)                          \
   _(int8_t, Char)
 
 inline __attribute__((always_inline)) void linear_(
@@ -37,7 +38,12 @@ inline __attribute__((always_inline)) void linear_(
 
   // Given an N-dimensional input [d0, d1, d2, ..., d_{N-2}, d_{N-1}], the
   // leading dimensions is d0 * d1 * ... * d_{N-2}
+<<<<<<< HEAD
   int64_t leading_dims = getLeadingDims(input, input.dim() - 1);
+=======
+  int64_t leading_dims =
+      getLeadingDims(input, input.dim() - 1);
+>>>>>>> 5ebcf7f03 (create quantized_linear_per_tensor_out in cpu)
 
   for (int i = 0; i < leading_dims; ++i) {
     for (int j = 0; j < M; ++j) {

backends/cadence/reference/operators/quantized_ops.h

@@ -87,8 +87,13 @@ inline __attribute__((always_inline)) void quantized_linear_per_tensor_(
             (int32_t)weight_data[j * in_dim + k] - (int32_t)weight_zero_point;
         sum += x * w;
       }
+<<<<<<< HEAD
       out_data[i * out_dim + j] = ::impl::reference::kernels::quantize<T>(
           sum, requant_scale, out_zero_point);
+=======
+      out_data[i * out_dim + j] =
+          ::impl::reference::kernels::quantize<T>(sum, requant_scale, out_zero_point);
+>>>>>>> 5ebcf7f03 (create quantized_linear_per_tensor_out in cpu)
     }
   }
 }