Tile-based rendering

All Mali GPU architectures are based on the principle of tile-based rendering, a technique designed to improve system-wide energy efficiency by minimizing the external DDR memory bandwidth needed for framebuffer access. The key principles of the Mali GPU tile-based rendering approach are:

All primitives are vertex shaded and assigned to screen-space tiles based on pixel coverage.
Fragment shading renders into an intermediate tile-buffer stored in a local memory that is tightly coupled to each shader core, and only written back to main memory at the end of rendering for that screen region. All frame buffer accesses during shading, such as depth testing or alpha blending, are energy efficient accesses made to this local memory instead of external DDR.
The use of local memory for intermediate storage enables energy efficient blending, zero bandwidth transient attachments, and almost free 4x multi-sample anti-aliasing.

Over time the Mali GPU architecture has evolved with new features to meet the needs of the latest standard APIs — such as OpenGL ES, OpenCL and Vulkan — and to provide further improvements in area efficiency and energy efficiency.

Transaction Elimination

Transaction Elimination (TE) is a key bandwidth saving feature of the Mali GPU architecture which allows for significant energy savings on a System on Chip (SoC). When performing TE, the GPU compares the current frame buffer with the previously rendered frame and performs a partial update only to the particular parts of it that have been modified, thus significantly reducing the amount of data that need to be transmitted per frame to external memory. The comparison is done on a per tile basis, using a Cyclic Redundancy Check (CRC) signature to determine if the tile has been modified. Tiles with the same CRC signature are identical; therefore eliminating them has no impact to the resulting image quality. TE can be used by every application for all frame buffer formats supported by the GPU, irrespective of the frame buffer precision requirements.

Configurability

The Mali GPU architectures provide a configurable set of graphics and GPU compute IP, allowing a silicon implementor to select the shader core count and level two cache size. This configurability allows each implementation to find the appropriate balance of performance, power, and silicon area that meets its requirements.

Popular Community posts