PowerVRPerformanceRecommendations资源-CSDN文库

需积分: 10 93 浏览量 2020-11-17 11:08:56 上传评论收藏 1.24MB PDF 举报

recommendation and advice for developers who wish to get the best graphics performance from a PowerVR SGX or PowerVR Rogue enabled device 《PowerVR性能优化建议》是一份由Imagination Technologies公司发布的技术文档，主要针对希望从PowerVR SGX或Rogue系列图形处理器中获取最佳图形性能的开发者。这份文档包含了专有的信息，可能会随时间变化，且在没有保修的情况下提供。开发者可以自由分发此文档，但需注明来源。本文档的版本为PowerVR SDK REL_18.1@5080009a，发布日期为2018年5月31日。 1. 引言 - 文档概述：这份文档旨在为开发者提供一套黄金规则和最佳实践，以帮助他们充分利用PowerVR图形核心的潜力。 - 黄金规则：可能包括一些基础的优化原则，如避免不必要的计算，减少内存带宽需求等。 - 最优开发方法：强调了在设计阶段就应考虑性能优化，而非后期补救。 - 渲染瓶颈理解：开发者需要识别并解决可能导致性能下降的关键环节，如GPU负载、内存访问速度等。 - 针对PowerVR图形核心优化应用：建议开发者了解目标设备的特性，以便进行有针对性的优化。 1. 优化几何 - 几何复杂性：降低模型的多边形数量可以显著提高渲染效率，特别是在移动设备上。 - 基元类型：选择合适的绘制命令（如点、线、三角形）可以影响渲染效率。例如，三角形是最基本的图形元素，通常用于构建复杂的3D模型，而线和点的处理则相对简单。 2. 更深入的优化策略可能包括： - 顶点和片段着色器的优化：减少着色器中的计算量，使用高效的算法和数据结构。 - 合批处理：将多个相似的绘制调用组合在一起，减少GPU上下文切换的开销。 - 纹理压缩：减少纹理数据的存储需求，提高加载速度和内存利用率。 - 使用硬件特性：如硬件级别的光照、阴影和抗锯齿等，可以有效提升图形质量同时保持性能。 - 分批渲染：根据材质或动画状态对对象进行分组，减少渲染状态变更。 - 内存管理：合理分配和释放显存，避免内存碎片。 3. 其他关键点可能涉及： - 动态资源管理：避免频繁地创建和销毁资源，优化内存分配。 - GPU缓存利用：理解GPU缓存的工作原理，优化数据访问模式以提高命中率。 - 混合操作：优化混合操作，减少颜色缓冲区的读写次数。通过遵循这些性能推荐，开发者能够有效地提升PowerVR图形处理单元的性能，提供流畅的用户体验，并节省宝贵的系统资源。这是一份对任何使用PowerVR技术进行图形开发的人员来说至关重要的参考资料。

资源详情

资源评论

Public Imagination Technologies

PowerVR 1 Revision PowerVR SDK REL_18.1@5080009a

PowerVR

Performance Recommendations

Public. This publication contains proprietary information which is subject to change without notice

and is supplied 'as is' without warranty of any kind. Redistribution of this document is permitted

with acknowledgement of the source.

Filename : PowerVR.Performance Recommendations

Version : PowerVR SDK REL_18.1@5080009a External Issue

Issue Date : 31 May 2018

Author : Imagination Technologies Limited

Imagination Technologies Public

Revision PowerVR SDK REL_18.1@5080009a 2 Performance Recommendations

Contents

1. Introduction ................................................................................................................................. 5

1.1. Document Overview ........................................................................................................... 5

1.2. The Golden Rules ............................................................................................................... 5

1.3. Optimal Development Approach......................................................................................... 5

1.4. Understanding Rendering Bottlenecks ............................................................................... 5

1.5. Optimising Applications for PowerVR Graphics Cores....................................................... 6

1.5.1. Know the Target ....................................................................................................... 6

1.5.2. Analysing an Application‟s Performance .................................................................. 7

2. Optimising Geometry ................................................................................................................. 8

2.1. Geometry Complexity ......................................................................................................... 8

2.2. Primitive Type ..................................................................................................................... 8

2.3. Data Types ......................................................................................................................... 8

2.4. Interleaving Attributes ......................................................................................................... 9

2.5. Vertex Buffer Objects – OpenGL ES .................................................................................. 9

2.6. Draw Call Size .................................................................................................................. 10

2.7. Triangle Size ..................................................................................................................... 10

2.8. Face Culling ...................................................................................................................... 10

2.9. Sorting Geometry.............................................................................................................. 10

2.9.1. Distance.................................................................................................................. 10

2.9.2. Render State .......................................................................................................... 10

2.10. Z Pre-pass ........................................................................................................................ 11

3. Optimising Textures ................................................................................................................. 12

3.1. Texture Size ...................................................................................................................... 12

3.1.1. Demystifying NPOT ................................................................................................ 12

3.2. Texture Compression ....................................................................................................... 13

3.2.1. PVRTexTool ........................................................................................................... 13

3.2.2. Why use PVRTC? .................................................................................................. 14

3.2.3. Image File Compression vs. Texture Compression ............................................... 15

3.3. Mipmapping ...................................................................................................................... 17

3.3.1. Advantages ............................................................................................................ 17

3.3.2. Generation .............................................................................................................. 17

3.3.3. Filtering ................................................................................................................... 17

3.4. Texture Sampling.............................................................................................................. 18

3.4.1. Texture Filtering ..................................................................................................... 18

3.4.2. Texel Fetch ............................................................................................................. 18

3.4.3. Dependent Texture Read ....................................................................................... 18

3.4.4. Wide Floating Point Textures ................................................................................. 19

3.5. Texture Uploading ............................................................................................................ 19

3.5.1. Texture Warm-up ................................................................................................... 19

3.5.2. Texture Formats and Precision .............................................................................. 19

3.6. Mathematical Look-ups .................................................................................................... 19

4. Optimising Shaders .................................................................................................................. 21

4.1. PVRShaderEditor ............................................................................................................. 21

4.1.1. GLSL Optimiser ...................................................................................................... 21

4.2. Choose the Right Algorithm .............................................................................................. 22

4.3. Know Your Spaces ........................................................................................................... 22

4.4. Flow Control ...................................................................................................................... 22

4.4.1. Discard ................................................................................................................... 22

4.4.2. Shader Group Vote – OpenGL ES ......................................................................... 22

4.5. Demystifying Precision ..................................................................................................... 23

4.5.1. Highp ...................................................................................................................... 23

4.5.2. Mediump ................................................................................................................. 23

4.5.3. Lowp ....................................................................................................................... 24

4.5.4. Swizzling................................................................................................................. 24

Public Imagination Technologies

PowerVR 3 Revision PowerVR SDK REL_18.1@5080009a

4.5.5. Attributes ................................................................................................................ 24

4.5.6. Varyings.................................................................................................................. 24

4.5.7. Samplers ................................................................................................................ 25

4.5.8. Uniforms ................................................................................................................. 25

4.5.9. Conversion Costs ................................................................................................... 25

4.6. Scalar Operations ............................................................................................................. 25

4.7. Constant Data in Shaders ................................................................................................ 26

4.8. Geometry / Tessellation Shaders ..................................................................................... 26

5. Optimising Specific Techniques ............................................................................................. 27

5.1. Using Multiple Render Targets Efficiently ........................................................................ 27

5.1.1. Recommended HDR Texture Formats ................................................................... 27

5.2. Preferred Lighting Solution ............................................................................................... 29

5.3. Preferred Shadowing Solution .......................................................................................... 29

5.4. MSAA Performance .......................................................................................................... 29

5.5. Preferred Analytical AA Solution ...................................................................................... 30

5.6. Screen Space Ambient Occlusion .................................................................................... 30

5.7. Ray-Marching ................................................................................................................... 30

5.8. Separable Kernels ............................................................................................................ 30

5.9. Efficient Sprite Rendering ................................................................................................. 31

5.10. Physically Based Rendering & Per-Pixel LOD – Rogue Performance ............................. 32

6. OpenGL ES Specific Optimisations ........................................................................................ 34

6.1. glClears & glColorMask .................................................................................................... 34

6.1.1. Invalidating Frame Buffer Attachments .................................................................. 34

6.2. Draw*Indirect and MultiDraw*IndirectEXT ........................................................................ 34

6.2.1. Draw*Indirect .......................................................................................................... 34

6.2.2. MultiDraw*IndirectEXT ........................................................................................... 35

6.2.3. Instancing ............................................................................................................... 35

6.3. PBO Texture Uploads ....................................................................................................... 35

6.3.1. Optimal Texture Updates with PBOs ..................................................................... 35

6.4. Rogue Specific.................................................................................................................. 36

6.4.1. Using glTexStorage2D & glTexStorage3D ............................................................. 36

6.5. VAOs, UBOs, Transform Feedback Buffers and SSBOs in OpenGL ES ......................... 36

6.5.1. Vertex Array Objects (VAOs) ................................................................................. 36

6.5.2. Uploading Uniforms (Uniform Buffer Objects) ........................................................ 36

6.5.3. Transform Buffer Objects ....................................................................................... 37

6.5.4. SSBOs – Shader Storage buffer Objects ............................................................... 37

6.6. Synchronisation ................................................................................................................ 37

6.6.1. Multithreading in OpenGL ES ................................................................................ 38

6.7. Frame-buffer Down Sampling ........................................................................................... 38

6.8. Pixel Local Storage Extension .......................................................................................... 38

7. Vulkan Specific Optimisations ................................................................................................ 40

7.1. Memory Types .................................................................................................................. 40

7.2. Pipelines ........................................................................................................................... 41

7.2.1. Pipeline Barriers ..................................................................................................... 41

7.2.2. Pipeline Caching .................................................................................................... 42

7.2.3. Derivative Pipelines ................................................................................................ 42

7.3. Descriptor Sets ................................................................................................................. 42

7.3.1. Multiple Descriptor Sets ......................................................................................... 42

7.3.2. Pooled Descriptor Sets ........................................................................................... 42

7.4. Push Constants ................................................................................................................ 42

7.5. Queues ............................................................................................................................. 43

7.6. Command Buffers ............................................................................................................. 43

7.6.1. Command Buffer Usage Flags ............................................................................... 43

7.6.2. Transient Command Buffers .................................................................................. 43

7.6.3. Secondary Command Buffers ................................................................................ 43

7.7. Render Pass ..................................................................................................................... 44

7.7.1. Sub-Passes ............................................................................................................ 44

7.7.2. Load & Store Ops ................................................................................................... 44

Imagination Technologies Public

Revision PowerVR SDK REL_18.1@5080009a 4 Performance Recommendations

7.7.3. Transient Attachments ........................................................................................... 44

7.7.4. Optimal Number of Attachments ............................................................................ 44

7.7.5. Attachment Order ................................................................................................... 45

7.8. MSAA ................................................................................................................................ 45

7.9. Image Layout .................................................................................................................... 45

8. Contact Details .......................................................................................................................... 46

List of Figures

Figure 1. Cyclical profiling ....................................................................................................................... 5

Figure 2. PowerVR TBDR Graphics Architecture ................................................................................... 6

Figure 3. Optimisation Process for Graphics Applications ...................................................................... 7

Figure 4. PVRTexTool GUI ................................................................................................................... 14

Figure 5. Image file compression vs. texture compression ................................................................... 16

Figure 6. PVRShaderEditor GUI ........................................................................................................... 21

Figure 7. Increasing complexity and reducing processing .................................................................... 31

Figure 8. PBR Theory ........................................................................................................................... 32

Public Imagination Technologies

PowerVR 5 Revision PowerVR SDK REL_18.1@5080009a

1. Introduction

PowerVR SGX and PowerVR Rogue are Graphics Core architectures from Imagination Technologies

designed specifically for shader-based APIs such as OpenGL ES 2.0, 3.x, and Vulkan (compatibility

depends on the graphics core and available driver). Due to their scalable architectures, the PowerVR

family spans a huge performance range.

1.1. Document Overview

The purpose of this document is to serve as recommendation and advice for developers who wish to

get the best graphics performance from a PowerVR SGX or PowerVR Rogue enabled device.

Throughout the document, the specific recommendations for PowerVR SGX and PowerVR Rogue are

marked as appropriate.

1.2. The Golden Rules

The golden rules are a set of more generic performance recommendations that developers should

seek to implement, as well as observe as many of the techniques and principles mentioned to help

produce well-behaved, high performance graphics applications. These rules are detailed in the

document entitled “PowerVR Performance Recommendations: The Golden Rules”, which is supplied

with the PowerVR SDK.

1.3. Optimal Development Approach

It is crucial to adopt the practices identified in this document from the very start of development in

order to save much time and effort later. Once an application is implemented to a near-final state, the

process of iteration depicted in Figure 1 should be adopted. The main benefit of this approach is that

time is not wasted and graphics quality is not comprised by making changes that do not benefit

performance.

Profile the

application

Identify a

bottleneck

Optimise the

bottleneck

Test the

optimisation

Figure 1. Cyclical profiling

1.4. Understanding Rendering Bottlenecks

It is a common misconception that the same actions can speed up any application. For example:

 Polygon count reduction: If the bottleneck of the application is fragment processing or texture

bandwidth then the only result of this action will be to reduce the graphical quality of the

application without improving rendering speed. In fact, if simpler models cause more of the

render target to be covered by a material with complex fragments, then this can actually slow

down an application.

 Reduce rendering resolution: In this case, if the fragment processing workload of your

application is not the bottleneck then this will also only serve to reduce the quality of the

graphics in your application without improving performance.

剩余45页未读，继续阅读

评论收藏

内容反馈

PowerVR Performance Recommendations

评论0

最新资源

PowerVR Performance Recommendations

评论0

最新资源

相关推荐

PowerVR Performance Recommendations.The Golden Rules

POWERVR SGX OpenGL ES 2.0 Application Development Recommendations

PowerVR GDC

PowerVR MBX.3D Application Development Recommendations.1.0.67a.External.pdf

PowerVR Technology Overview.1.0.2e.External.pdf

安卓下使用OpenCL进行PowerVR GPU编程

Imagination发布PowerVR NNA神经网络加速器.pdf

PowerVR GR6500：GPU.pdf

PowerVR.Supported+Extensions.OpenGL+ES.EGL.pdf

Telechips选择PowerVR GPU开发车用芯片.pdf

PowerVR Series7 GPU提升可穿戴设备性能.pdf

Powervr_SGX530_OpenGL_ES2.0_demo程序.7z

Powervr.pdf

开启移动AI大门 解读PowerVR第九代图形、神经网络产品.pdf

Powervr.docx

DENSO获得Imagination MIPS CPU和PowerVR GPU授权.pdf

Telechips选择PowerVR GPU开发车用芯片 (1).pdf

全新Wizard系列PowerVR光线追踪GPU.pdf

PowerVR 2NX NNA引领神经网络加速器的未来.pdf

Imagination PowerVR GPU率先通过Khronos的OpenVX 1.1一致性测试.pdf

PowerVR G6230等：GPU.pdf

Imagination新款PowerVR GPU结合先进图形与优异视觉技术.pdf

Imagination发布PowerVR Series3NX神经网络加速器,为嵌入式人工智能市场带来多核可扩展性.pdf

联发科技采用Imagination PowerVR Series6 GPU开发出新款异构多重处理SoC.pdf

Imagination 发布PowerVR Series3NX 神经网络加速器.pdf

Imagination推出全新PowerVR光线追踪GPU.pdf

PowerVR Series3NX神经网络加速器为嵌入式人工智能带来多核可扩展性.pdf

Imagination发布PowerVR NNA神经网络加速器 (2).pdf

大麦抢票+源代码+辅助工具+详细文档教程

vdhcoapp-windows-x86-64-installer 2.0.19

开启移动AI大门解读PowerVR第九代图形、神经网络产品.pdf