Add device type

crates/openvino/src/core.rs

@@ -4,12 +4,12 @@
 use crate::error::LoadingError;
 use crate::{cstr, drop_using_function, try_unsafe, util::Result};
 use crate::{model::CompiledModel, Model};
-use crate::{SetupError, Tensor};
+use crate::{DeviceType, SetupError, Tensor};
 use openvino_sys::{
     self, ov_core_compile_model, ov_core_create, ov_core_create_with_config, ov_core_free,
     ov_core_read_model, ov_core_read_model_from_memory_buffer, ov_core_t,
 };
-
+use std::ffi::CString;
 /// See [`Core`](https://docs.openvino.ai/2023.3/api/c_cpp_api/group__ov__core__c__api.html).
 pub struct Core {
     ptr: *mut ov_core_t,
@@ -68,13 +68,14 @@ impl Core {
     }
 
     /// Compile a model to `CompiledModel`.
-    pub fn compile_model(&mut self, model: &Model, device: &str) -> Result<CompiledModel> {
+    pub fn compile_model(&mut self, model: &Model, device: DeviceType) -> Result<CompiledModel> {
+        let device: CString = device.into();
         let mut compiled_model = std::ptr::null_mut();
         let num_property_args = 0;
         try_unsafe!(ov_core_compile_model(
             self.ptr,
             model.as_ptr(),
-            cstr!(device),
+            device.as_ptr(),
             num_property_args,
             std::ptr::addr_of_mut!(compiled_model)
         ))?;

crates/openvino/src/device_type.rs

@@ -0,0 +1,87 @@
+use std::borrow::Cow;
+use std::convert::Infallible;
+use std::ffi::CString;
+use std::fmt::{Display, Formatter};
+use std::str::FromStr;
+
+/// `DeviceType` represents accelerator devices.
+#[derive(Ord, PartialOrd, Eq, PartialEq, Hash, Debug)]
+pub enum DeviceType<'a> {
+    /// [CPU Device](https://docs.openvino.ai/2024/openvino-workflow/running-inference/inference-devices-and-modes/cpu-device.html)
+    CPU,
+    /// [GPU Device](https://docs.openvino.ai/2024/openvino-workflow/running-inference/inference-devices-and-modes/gpu-device.html)
+    GPU,
+    /// [NPU Device](https://docs.openvino.ai/2024/openvino-workflow/running-inference/inference-devices-and-modes/npu-device.html)
+    NPU,
+    /// [GNA Device](https://docs.openvino.ai/2023.3/openvino_docs_OV_UG_supported_plugins_GNA.html)
+    #[deprecated = "Deprecated since OpenVINO 2024.0; use NPU device instead"]
+    GNA,
+    /// Arbitrary device.
+    Other(Cow<'a, str>),
+}
+
+impl DeviceType<'_> {
+    /// Creates a device type with owned string data.
+    pub fn to_owned(&self) -> DeviceType<'static> {
+        match self {
+            DeviceType::CPU => DeviceType::CPU,
+            DeviceType::GPU => DeviceType::GPU,
+            DeviceType::NPU => DeviceType::NPU,
+            #[allow(deprecated)]
+            DeviceType::GNA => DeviceType::GNA,
+            DeviceType::Other(s) => DeviceType::Other(Cow::Owned(s.clone().into_owned())),
+        }
+    }
+}
+
+impl AsRef<str> for DeviceType<'_> {
+    fn as_ref(&self) -> &str {
+        match self {
+            DeviceType::CPU => "CPU",
+            DeviceType::GPU => "GPU",
+            DeviceType::NPU => "NPU",
+            #[allow(deprecated)]
+            DeviceType::GNA => "GNA",
+            DeviceType::Other(s) => s,
+        }
+    }
+}
+
+impl<'a> From<&'a DeviceType<'a>> for &'a str {
+    fn from(value: &'a DeviceType) -> Self {
+        value.as_ref()
+    }
+}
+
+impl<'a> From<DeviceType<'a>> for CString {
+    fn from(value: DeviceType) -> Self {
+        CString::new(value.as_ref()).expect("a valid C string")
+    }
+}
+
+impl<'a> From<&'a str> for DeviceType<'a> {
+    fn from(s: &'a str) -> Self {
+        match s {
+            "CPU" => DeviceType::CPU,
+            "GPU" => DeviceType::GPU,
+            "NPU" => DeviceType::NPU,
+            #[allow(deprecated)]
+            "GNA" => DeviceType::GNA,
+            s => DeviceType::Other(Cow::Borrowed(s)),
+        }
+    }
+}
+
+impl FromStr for DeviceType<'static> {
+    type Err = Infallible;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        Ok(DeviceType::from(s).to_owned())
+    }
+}
+
+impl Display for DeviceType<'_> {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        f.write_str(self.into())
+    }
+}

crates/openvino/src/lib.rs

@@ -26,6 +26,7 @@
 )]
 
 mod core;
+mod device_type;
 mod dimension;
 mod element_type;
 mod error;
@@ -42,6 +43,7 @@ mod tensor;
 mod util;
 
 pub use crate::core::Core;
+pub use device_type::DeviceType;
 pub use dimension::Dimension;
 pub use element_type::ElementType;
 pub use error::{InferenceError, LoadingError, SetupError};

crates/openvino/tests/classify-alexnet.rs

@@ -5,7 +5,9 @@ mod util;
 
 use anyhow::Ok;
 use fixtures::alexnet::Fixture;
-use openvino::{prepostprocess, Core, ElementType, Layout, ResizeAlgorithm, Shape, Tensor};
+use openvino::{
+    prepostprocess, Core, DeviceType, ElementType, Layout, ResizeAlgorithm, Shape, Tensor,
+};
 use std::fs;
 use util::{Prediction, Predictions};
 
@@ -45,7 +47,7 @@ fn classify_alexnet() -> anyhow::Result<()> {
     let new_model = pre_post_process.build_new_model()?;
 
     // Compile the model and infer the results.
-    let mut executable_model = core.compile_model(&new_model, "CPU")?;
+    let mut executable_model = core.compile_model(&new_model, DeviceType::CPU)?;
     let mut infer_request = executable_model.create_infer_request()?;
     infer_request.set_tensor("data", &tensor)?;
     infer_request.infer()?;

crates/openvino/tests/classify-inception.rs

@@ -5,7 +5,9 @@ mod util;
 
 use anyhow::Ok;
 use fixtures::inception::Fixture;
-use openvino::{prepostprocess, Core, ElementType, Layout, ResizeAlgorithm, Shape, Tensor};
+use openvino::{
+    prepostprocess, Core, DeviceType, ElementType, Layout, ResizeAlgorithm, Shape, Tensor,
+};
 use std::fs;
 use util::{Prediction, Predictions};
 
@@ -42,7 +44,7 @@ fn classify_inception() -> anyhow::Result<()> {
     let new_model = pre_post_process.build_new_model()?;
 
     // Compile the model and infer the results.
-    let mut executable_model = core.compile_model(&new_model, "CPU")?;
+    let mut executable_model = core.compile_model(&new_model, DeviceType::CPU)?;
     let mut infer_request = executable_model.create_infer_request()?;
     infer_request.set_tensor("input", &tensor)?;
     infer_request.infer()?;

crates/openvino/tests/classify-mobilenet.rs

@@ -5,7 +5,9 @@ mod fixtures;
 mod util;
 
 use fixtures::mobilenet::Fixture;
-use openvino::{prepostprocess, Core, ElementType, Layout, ResizeAlgorithm, Shape, Tensor};
+use openvino::{
+    prepostprocess, Core, DeviceType, ElementType, Layout, ResizeAlgorithm, Shape, Tensor,
+};
 use std::fs;
 use util::{Prediction, Predictions};
 
@@ -45,7 +47,7 @@ fn classify_mobilenet() -> anyhow::Result<()> {
     let new_model = pre_post_process.build_new_model()?;
 
     // Compile the model and infer the results.
-    let mut executable_model = core.compile_model(&new_model, "CPU")?;
+    let mut executable_model = core.compile_model(&new_model, DeviceType::CPU)?;
     let mut infer_request = executable_model.create_infer_request()?;
     infer_request.set_tensor("input", &tensor)?;
     infer_request.infer()?;