GPU

Graphics Processing Unit (GPU)

The Graphics Processing Unit is controlled by the CPU through a direct interface of memory-mapped IO registers, and indirectly by parsing commands that the CPU has placed in memory. The Display interface and Blitter (block image transferrer) are controlled primarily by direct CPU register addresses, while the 3D and Media pipelines and the parallel Video Codec Engine (VCE) are controlled primarily through instruction lists in memory.

The subsystem contains an array of cores, or execution units, with a number of “shared functions”, which receive and process messages at the request of programs running on the cores. The shared functions perform critical tasks, such as sampling textures and updating the render target (usually the frame buffer). The cores themselves are described by an instruction set architecture, or ISA.

Memory Object Overview
Any memory data accessed by the device is considered part of a memory object of some memory object type.

Ring Buffer
Ring Buffer can exist anywhere in memory mapped via Global GTT. Ring buffer details are mentioned in the ring context area of LRCA (Ring Buffer - Start Address, Head Offset, Tail Pointer & Control Register) in Execution List mode of scheduling. Ring Buffer registers are directly programmed in Ring Buffer mode of scheduling.

Classical BLT Engine Functional Description
Blitter (block image transferrer)
The graphics controller provides a hardware-based BLT engine to off load the work of moving blocks of graphics data from the host CPU. Although the BLT engine is often used simply to copy a block of graphics data from the source to the destination, it also has the ability to perform more complex functions. The BLT engine is capable of receiving three different blocks of graphics data as input as shown in the figure below. The source data may exist in the frame buffer or the Graphics aperture. The pattern data always represents an 8x8 block of pixels that can be located in the frame buffer, Graphics aperture, or passed through a command packet. The pattern data must be located in linear memory. The data already residing at the destination may also be used as an input. The destination data can also be located in the frame buffer or graphics aperture.

Frame Buffer Compression (FBC) gives a lossless compression of the display frame buffer to save power by reducing system memory read bandwidth and increasing the time between display engine reads to system memory.

The display watermarks are used to control the display engine memory request behavior.

 

ref:

https://www.jianshu.com/p/87cf95b1faa0