Minor fixes for dtypes and client examples

client_example/22_im2col_col2im/column_to_image.cpp

@@ -72,7 +72,7 @@ int main()
                                                                              ImageLayout,
                                                                              InDataType,
                                                                              OutDataType,
-                                                                             ColumnToImageOp>;
+                                                                             ColumnToImage>;
 
     // get device op instances
     const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<

client_example/22_im2col_col2im/image_to_column.cpp

@@ -66,13 +66,13 @@ int main()
     SimpleDeviceMem in(sizeof(InDataType) * N * Hi * Wi * G * C);
     SimpleDeviceMem out(sizeof(OutDataType) * N * Ho * Wo * Y * X * C);
 
-    using ImageToColumnOp = ck::conv_tensor_rearrange_op::ImageToColumn;
+    using namespace ck::conv_tensor_rearrange_op;
 
     using DeviceOp = ck::tensor_operation::device::DeviceConvTensorRearrange<NumDimSpatial,
                                                                              ImageLayout,
                                                                              InDataType,
                                                                              OutDataType,
-                                                                             ImageToColumnOp>;
+                                                                             ImageToColumn>;
 
     // get device op instances
     const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<

library/src/tensor_operation_instance/gpu/image_to_column/device_image_to_column_nhwc_1d_instance.cpp

@@ -15,28 +15,36 @@ void add_device_image_to_column_nhwc_1d_bf16_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<1, GNWC, BF16, BF16, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_BF16
     add_device_operation_instances(instances, device_image_to_column_bf16_instances<1, GNWC>{});
+#endif
 }
 
 void add_device_image_to_column_nhwc_1d_f16_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<1, GNWC, F16, F16, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_FP16
     add_device_operation_instances(instances, device_image_to_column_f16_instances<1, GNWC>{});
+#endif
 }
 
 void add_device_image_to_column_nhwc_1d_f32_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<1, GNWC, F32, F32, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_FP32
     add_device_operation_instances(instances, device_image_to_column_f32_instances<1, GNWC>{});
+#endif
 }
 
 void add_device_image_to_column_nhwc_1d_i8_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<1, GNWC, int8_t, int8_t, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_INT8
     add_device_operation_instances(instances, device_image_to_column_i8_instances<1, GNWC>{});
+#endif
 }
 
 } // namespace instance

library/src/tensor_operation_instance/gpu/image_to_column/device_image_to_column_nhwc_2d_instance.cpp

@@ -15,21 +15,27 @@ void add_device_image_to_column_nhwc_2d_bf16_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<2, GNHWC, BF16, BF16, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_BF16
     add_device_operation_instances(instances, device_image_to_column_bf16_instances<2, GNHWC>{});
+#endif
 }
 
 void add_device_image_to_column_nhwc_2d_f16_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<2, GNHWC, F16, F16, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_FP16
     add_device_operation_instances(instances, device_image_to_column_f16_instances<2, GNHWC>{});
+#endif
 }
 
 void add_device_image_to_column_nhwc_2d_f32_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<2, GNHWC, F32, F32, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_FP32
     add_device_operation_instances(instances, device_image_to_column_f32_instances<2, GNHWC>{});
+#endif
 }
 
 void add_device_image_to_column_nhwc_2d_i8_instances(
@@ -37,7 +43,9 @@ void add_device_image_to_column_nhwc_2d_i8_instances(
         std::unique_ptr<DeviceConvTensorRearrange<2, GNHWC, int8_t, int8_t, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_INT8
     add_device_operation_instances(instances, device_image_to_column_i8_instances<2, GNHWC>{});
+#endif
 }
 
 } // namespace instance

library/src/tensor_operation_instance/gpu/image_to_column/device_image_to_column_nhwc_3d_instance.cpp

@@ -15,21 +15,27 @@ void add_device_image_to_column_nhwc_3d_bf16_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<3, GNDHWC, BF16, BF16, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_BF16
     add_device_operation_instances(instances, device_image_to_column_bf16_instances<3, GNDHWC>{});
+#endif
 }
 
 void add_device_image_to_column_nhwc_3d_f16_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<3, GNDHWC, F16, F16, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_FP16
     add_device_operation_instances(instances, device_image_to_column_f16_instances<3, GNDHWC>{});
+#endif
 }
 
 void add_device_image_to_column_nhwc_3d_f32_instances(
     std::vector<std::unique_ptr<DeviceConvTensorRearrange<3, GNDHWC, F32, F32, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_FP32
     add_device_operation_instances(instances, device_image_to_column_f32_instances<3, GNDHWC>{});
+#endif
 }
 
 void add_device_image_to_column_nhwc_3d_i8_instances(
@@ -37,7 +43,9 @@ void add_device_image_to_column_nhwc_3d_i8_instances(
         std::unique_ptr<DeviceConvTensorRearrange<3, GNDHWC, int8_t, int8_t, ImageToColumn>>>&
         instances)
 {
+#ifdef CK_ENABLE_INT8
     add_device_operation_instances(instances, device_image_to_column_i8_instances<3, GNDHWC>{});
+#endif
 }
 
 } // namespace instance