Hybrid ML is a software library and device driver architecture that addresses key industry challenges by bridging gaps between media API and graphics hardware on all classes of embedded devices, ranging from software-only and FPGA implementations to low-power programmable GPU and high-performance SoC processors with graphics cores.
Embedded OpenGL and Programmable Media Device Support
The Hybrid Media Library provides an embedded software library that enables developers and device manufacturers to take advantage of full profiles of the OpenGL SC, OpenGL ES 1.x, OpenGL ES 2.x, and OpenVG 1.x graphics APIs and additional programmable features on all classes of media accelerated chipsets and devices.
Available for all popular desktop and real time operating systems, Hybrid ML combines a flexible set of software components that support programmable media controllers of all types, including:
lightweight pure software implementations
media driven FPGAs
low power and high performance fully programmable graphics cores found in SoC processors from ARM, Intel, Freescale, TI and Fujitsu
Hybrid ML offers GPU virtualization, supports partitioned RTOS operation, and provides multi-context support and multiple GPU support within a single API interface.
DO-178B Certification Support
Hybrid ML for OpenGL SC includes a standardized API with DO-178B certifiable soft implementations of GPU stages, enabling intelligent insertion of hardware acceleration based on programmable pipeline stages (reliable shaders) or direct fixed function rendering on a wide array of devices. This embedded graphics solution has been architected with abstracted and reusable certifiable modules, which reduces the time and effort required to develop a DO-178B certifiable embedded system.
The Hybrid ML Architecture
Hybrid ML is comprised of a set of software modules that can be configured to intelligently layer implementations of different APIs on virtually any graphics enabling device by using various combinations of software rendering, FPGA, programmable shader interfaces, or fixed function commands. The Hybrid ML architecture consists of several components that together provide reliable media and graphics acceleration, including:
optimized footprint of the OpenGL SC, OpenGL ES and OpenVG media API.
a set of components that provide trusted implementations of API features such as geometry caching, state tracking, memory management, and context control.
reliable shader implementations of key OpenGL API for programmable systems.
a set of streamlined ‘micro-interfaces’ to software, FPGA, SoC, or discrete hardware media accelerators as required.
targeted extensions that provide shader capabilities within the Hybrid ML architecture for specific application requirements such as digital mapping and medical imaging.
control of a set of virtual contexts that allow multiple processes to access several logical GPUs on a single media accelerator, a single process to address multiple GPUs, and even dissimilar GPUs as needed.
CoreGL for 2D graphics
For embedded systems that require only 2D graphics support, ALT Software offers CoreGL, a scaled down version of Hybrid ML that fully satisfies the requirements of many enhanced 2D GUI. CoreGL includes a 2D graphics subset of the OpenGL SC and OpenGL ES API. It offers basic rasterization capabilities, a small footprint and optimized 2D rendering paths.
Hybrid ML Advantages
ALT’s Hybrid ML offers an innovative architecture for deploying low to mid-range accelerated 2D and 3D graphics applications.
Modular architecture reduces development costs and time to market
Hybrid ML’s modular architecture enables device manufacturers to create advanced user interfaces that can take advantage of the underlying API features. It can be easily configured and integrated into a variety of different software architectures, allowing an existing code base to be extended to support GPU capabilities with minimal impact.
The CoreGL 2D API subset can significantly reduce the cost of verification and the level of integration required for safety-critical applications by excluding support for unneeded media rendering functions included in the full OpenGL ES and OpenGL SC profiles.
Supports practical alternatives to discrete GPU
Hybrid ML provides an upgrade path from a discrete legacy GPUs to low-cost alternatives such as software-only rendering, FPGA, or SoC processors with graphics IP cores.
Flexible design supports the full range of media requirements
Hybrid ML addresses challenges posed by high-end GPUs such as managing programmable pipelines, reducing driver code size, and supporting multiple security/criticality levels. It can be used in application areas where a standard GPU can pose integration problems, such as systems that require video overlay or alpha-channel blended outputs, as well as in partitioned RTOS systems.