Application load times by caching shader binaries in a persistent storage

Abstract

A method for compiling a shader for execution by a graphics processor. The method comprises selecting a shader for execution. A key is computed for the selected shader. A memory is searched for a copy of the computed key. A shader binary stored in the memory is passed to the graphics processor for execution if the copy of the computed key is located in the memory. Otherwise, the shader is compiled to produce the shader binary for execution by the graphics processor and storing the shader binary in the memory. The shader binary is associated with the computed key and the copy of the computed key.

Description

CROSS-REFERENCES TO RELATED APPLICATIONS[0001]This application claims the benefit of Provisional Application No. 61 / 585,620, filed on Jan. 11, 2012, titled “GRAPHICS PROCESSOR CLOCK SCALING, APPLICATION LOAD TIME IMPROVEMENTS, AND DYNAMICALLY ADJUSTING RESOLUTION OF RENDER BUFFER TO IMPROVE AND STABILIZE FRAME TIMES OF A GRAPHICS PROCESSOR,” by Swaminathan Narayanan, et al., which is herein incorporated by reference.TECHNICAL FIELD[0002]Embodiments of the present disclosure relate generally to the field of graphics processing and more specifically to the field of improved shader binary caching and execution for efficient graphics processing.BACKGROUND[0003]High level graphics languages (e.g., OpenGL and DirectX) allow applications to specify the execution of particular shaders. Shaders are instruction sets that define how certain pieces of geometry or fragments are processed by a graphics processor. These shader instruction sets can be quite long and detailed in what they do, and of...

AI Technical Summary

Benefits of technology

This patent presents solutions to the challenge of compiling shaders efficiently during runtime. The invention provides a method for compiling a shader for execution by a graphics processor. This method involves selecting a shader, computing a key for the shader, searching the memory for a copy of the key, and passing a shader binary stored in the memory to the graphics processor for execution if the key is found in the memory. The shader binary is associated with the computed key and the copy of the computed key. This invention also includes a computer system comprising a processor, graphics processor, and memory. A compiler is operable to compile and link shader source code to create a shader binary, which is stored in the memory. The graphics driver module is operable to select one or more shders for execution by the graphics processor and to compute a key for a selected shader. The memory is operable to search for a copy of the key. A shader binary is passed from the memory for execution by the graphics processor if the copy of the key is located in the memory. However, if the key cannot be found, the compiler is operable to compiled and link the shader to create a shader binary for execution by the graphics processor and storing it in the memory. The shader binary is associated with the computed key and the copy of the computed key. The technical effects of this patent are improved efficiency in compiling shaders during runtime and improved performance of graphics processing.

Problems solved by technology

Shader compilation times on mobile hardware can take a significant portion of frame render time.

When the application compiles a handful of shaders, which take on average 3-5 ms to compile on current mobile hardware, the shader compilation time can easily exceed the frame time and unfortunately cause visible stuttering on the screen.

An application attempting to compile shaders during runtime risks frame hitches or a gap of time when rendering stops or slows as shaders are compiled and programs linked together.

Such visible stuttering or frame rate hitches are undesirable.

It would also be difficult for an application vendor to supply a shader binary library that contains all of the possible binaries that would need to be stored so as to avoid compiling.

Even so, should unexpected changes occur, the stored binaries would then be out of date.

In other words, a large binary library will not provide the required flexibility to update the executable shader binaries to reflect any changes to hardware and software.

Images