CVS-174585: Memory map shared weights when possible

[ericcraw]

https://jira.devtools.intel.com/browse/CVS-174585

[Copilot]

Pull Request Overview

This PR introduces memory mapping functionality for shared weights in the OpenVINO execution provider to improve performance by avoiding unnecessary memory copies. The changes enable direct memory mapping of external weight files to device tensors when possible.

• Added version check macro for OpenVINO to conditionally enable features based on version
• Implemented memory mapping support for weight files with device-specific optimizations
• Refactored tensor creation to use memory-mapped weights when available, falling back to traditional copying

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
onnxruntime/core/providers/openvino/ov_interface.h	Added version check macro for OpenVINO compatibility
onnxruntime/core/providers/openvino/contexts.h	Extended WeightsFile class with device mapping functionality
onnxruntime/core/providers/openvino/backend_utils.cc	Implemented memory mapping logic and refactored tensor creation

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

Pull Request Overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

Pull Request Overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

Corrected spelling of 'mmaped' to 'mmapped' in variable name and comment.

Suggested change

	// CPU/virtual device case, create a CPU tensor memory mapped from file
	auto&& mmaped_tensor = ov::read_tensor_data(file_path_);
	it->second = MappingContainer{.ptr_ = mmaped_tensor.data(), .tensor_ = mmaped_tensor};
	// CPU/virtual device case, create a CPU tensor memory mmapped from file
	auto&& mmapped_tensor = ov::read_tensor_data(file_path_);
	it->second = MappingContainer{.ptr_ = mmapped_tensor.data(), .tensor_ = mmapped_tensor};

[Copilot]

Corrected spelling of 'mmaped' to 'mmapped' in variable name.

Suggested change

	uint8_t* mmaped_weights = static_cast<uint8_t*>(weights.TryGetOrCreateDeviceMapping(opt_remote_ctx));
	if (mmaped_weights) {
	// We have memory mapped weights. Create a Tensor view into it for this value.
	ORT_ENFORCE(value.data_offset < weights.Size() &&
	value.size <= weights.Size() &&
	(value.data_offset <= weights.Size() - value.size),
	"File offset + size outside of external initializer file");
	void* mmapped_offset = static_cast<void*>(mmaped_weights + value.data_offset);
	uint8_t* mmapped_weights = static_cast<uint8_t*>(weights.TryGetOrCreateDeviceMapping(opt_remote_ctx));
	if (mmapped_weights) {
	// We have memory mapped weights. Create a Tensor view into it for this value.
	ORT_ENFORCE(value.data_offset < weights.Size() &&
	value.size <= weights.Size() &&
	(value.data_offset <= weights.Size() - value.size),
	"File offset + size outside of external initializer file");
	void* mmapped_offset = static_cast<void*>(mmapped_weights + value.data_offset);

[Copilot]

Corrected spelling of 'mmaped' to 'mmapped' in variable name.

Suggested change

	uint8_t* mmaped_weights = static_cast<uint8_t*>(weights.TryGetOrCreateDeviceMapping(opt_remote_ctx));
	if (mmaped_weights) {
	// We have memory mapped weights. Create a Tensor view into it for this value.
	ORT_ENFORCE(value.data_offset < weights.Size() &&
	value.size <= weights.Size() &&
	(value.data_offset <= weights.Size() - value.size),
	"File offset + size outside of external initializer file");
	void* mmapped_offset = static_cast<void*>(mmaped_weights + value.data_offset);
	uint8_t* mmapped_weights = static_cast<uint8_t*>(weights.TryGetOrCreateDeviceMapping(opt_remote_ctx));
	if (mmapped_weights) {
	// We have memory mapped weights. Create a Tensor view into it for this value.
	ORT_ENFORCE(value.data_offset < weights.Size() &&
	value.size <= weights.Size() &&
	(value.data_offset <= weights.Size() - value.size),
	"File offset + size outside of external initializer file");
	void* mmapped_offset = static_cast<void*>(mmapped_weights + value.data_offset);

[Copilot]

Corrected spelling of 'mmaped' to 'mmapped' in variable name (note: mmapped_offset is already correct).

Suggested change

	uint8_t* mmaped_weights = static_cast<uint8_t*>(weights.TryGetOrCreateDeviceMapping(opt_remote_ctx));
	if (mmaped_weights) {
	// We have memory mapped weights. Create a Tensor view into it for this value.
	ORT_ENFORCE(value.data_offset < weights.Size() &&
	value.size <= weights.Size() &&
	(value.data_offset <= weights.Size() - value.size),
	"File offset + size outside of external initializer file");
	void* mmapped_offset = static_cast<void*>(mmaped_weights + value.data_offset);
	uint8_t* mmapped_weights = static_cast<uint8_t*>(weights.TryGetOrCreateDeviceMapping(opt_remote_ctx));
	if (mmapped_weights) {
	// We have memory mapped weights. Create a Tensor view into it for this value.
	ORT_ENFORCE(value.data_offset < weights.Size() &&
	value.size <= weights.Size() &&
	(value.data_offset <= weights.Size() - value.size),
	"File offset + size outside of external initializer file");
	void* mmapped_offset = static_cast<void*>(mmapped_weights + value.data_offset);