Vulkan 1.4.332 is out today as the latest weekly update to this high performance graphics and compute API specification. Besides a number of documentation clarifications/corrections, there is one new extension this week in the name of AI / machine learning.
This week’s new Vulkan API extension is a Qualcomm vendor extension: VK_QCOM_data_graph_model. The VK_QCOM_data_graph_model is for enhancing their machine learning capabilities with the Vulkan API. Vulkan for AI/ML continues to be a hot topic with even NVIDIA exploring and working on it.
The VK_QCOM_data_graph_model extension is described as:
“This extension supports new VkPhysicalDevice…
Vulkan 1.4.332 is out today as the latest weekly update to this high performance graphics and compute API specification. Besides a number of documentation clarifications/corrections, there is one new extension this week in the name of AI / machine learning.
This week’s new Vulkan API extension is a Qualcomm vendor extension: VK_QCOM_data_graph_model. The VK_QCOM_data_graph_model is for enhancing their machine learning capabilities with the Vulkan API. Vulkan for AI/ML continues to be a hot topic with even NVIDIA exploring and working on it.
The VK_QCOM_data_graph_model extension is described as:
“This extension supports new VkPhysicalDeviceDataGraphProcessingEngineTypeARM, and VkPhysicalDeviceDataGraphOperationTypeARM types for data graph pipelines added in VK_ARM_data_graph.
A new pipeline cache type is also added to seamlessly import ML models such as ONNX through QNN workflow, and run them on the device or an external compute engine.“
More details and information on today’s Vulkan 1.4.332 specification update can be found via GitHub.