https://www.youtube.com/watch?v=XOGIDMJThto

https://www.khronos.org/assets/uploads/developers/library/2016-vulkan-devday-uk/9-Asynchonous-compute.pdf

https://docs.microsoft.com/en-us/windows/win32/direct3d12/user-mode-heap-synchronization

https://gpuopen.com/concurrent-execution-asynchronous-queues/

通过queue的并行 增加GPU的并行

并发性 concurrency

Radeon™ Fury X GPU consists of 64 Compute Units (CUs), each of those containing 4 Single-Instruction-Multiple-Data units (SIMD) and each SIMD executes blocks of 64 threads, which we call a “wavefront”.

Since latency for memory access can cause significant stalls in shader execution, up to 10 wavefronts can be scheduled on each SIMD simultaneously to hide this latency.

GPU有64个CU

每个CU 4个SIMD

每个SIMD 64blocks ----- 一个wavefront

ps的计算在里面

GPU提升并发性 减小GPU idel

async compute

  • Copy Queue(DirectX 12) / Transfer Queue (Vulkan): DMA transfers of data over the PCIe bus
  • Compute queue (DirectX 12 and Vulkan): execute compute shaders or copy data, preferably within local memory
  • Direct Queue (DirectX 12) / Graphics Queue (Vulkan):  this queue can do anything, so it is similar to the main device in legacy APIs

这三种queue对应metal里面三种encoder 是为了增加上文所述并发性

对GPU底层的 操作这种可行性是通过这里的queue体现的

vulkan对queue的个数有限制 可以query

dx12没有这种个数限制

更多部分拿出来用cs做异步计算

看图--技能点还没点

problem shooting

  • If resources are located in system memory accessing those from Graphics or Compute queues will have an impact on DMA queue performance and vice versa.
  • Graphics and Compute queues accessing local memory (e.g. fetching texture data, writing to UAVs or performing rasterization-heavy tasks) can affect each other due to bandwidth limitations  带宽限制 数据onchip
  • Threads sharing the same CU will share GPRs and LDS, so tasks that use all available resources may prevent asynchronous workloads to execute on the same CU
  • Different queues share their caches. If multiple queues utilize the same caches this can result in more cache thrashing and reduce performance

Due to the reasons above it is recommended to determine bottlenecks for each pass and place passes with complementary bottlenecks next to each other:

  • Compute shaders which make heavy use of LDS and ALU are usually good candidates for the asynchronous compute queue
  • Depth only rendering passes are usually good candidates to have some compute tasks run next to it
  • A common solution for efficient asynchronous compute usage can be to overlap the post processing of frame N with shadow map rendering of frame N+1
  • Porting as much of the frame to compute will result in more flexibility when experimenting which tasks can be scheduled next to each other
  • Splitting tasks into sub-tasks and interleaving them can reduce barriers and create opportunities for efficient async compute usage (e.g. instead of “for each light clear shadow map, render shadow, compute VSM” do “clear all shadow maps, render all shadow maps, compute VSM for all shadow maps”)

然后给异步计算的功能加上开关

看vulkan这个意思 它似乎没有metal2 那种persistent thread group 维持数据cs ps之间传递时还可以 on tile

vulkan asynchronous compute的更多相关文章

  1. Vulkan在Android使用Compute shader

    oeip 相关功能只能运行在window平台,想移植到android平台,暂时选择vulkan做为图像处理,主要一是里面有单独的计算管线且支持好,二是熟悉下最新的渲染技术思路. 这个 demo(git ...

  2. android下vulkan与opengles纹理互通

    先放demo源码地址:https://github.com/xxxzhou/aoce 06_mediaplayer 效果图: 主要几个点: 用ffmpeg打开rtmp流. 使用vulkan Compu ...

  3. 剖析虚幻渲染体系(13)- RHI补充篇:现代图形API之奥义与指南

    目录 13.1 本篇概述 13.1.1 本篇内容 13.1.2 概念总览 13.1.3 现代图形API特点 13.2 设备上下文 13.2.1 启动流程 13.2.2 Device 13.2.3 Sw ...

  4. GPUImage移植总结

    项目github地址: aoce 我是去年年底才知道有GPUImage这个项目,以前也一直没有在移动平台开发过,但是我在win平台有编写一个类似的项目oeip(不要关注了,所有功能都移植或快移植到ao ...

  5. Compute Resource Consolidation Pattern 计算资源整合模式

    Consolidate multiple tasks or operations into a single computational unit. This pattern can increase ...

  6. 论文笔记之:Asynchronous Methods for Deep Reinforcement Learning

    Asynchronous Methods for Deep Reinforcement Learning ICML 2016 深度强化学习最近被人发现貌似不太稳定,有人提出很多改善的方法,这些方法有很 ...

  7. Vulkan Tutorial 13 Render passes

    操作系统:Windows8.1 显卡:Nivida GTX965M 开发工具:Visual Studio 2017 Setup 在我们完成管线的创建工作,我们接下来需要告诉Vulkan渲染时候使用的f ...

  8. Vulkan Tutorial 16 Command buffers

    操作系统:Windows8.1 显卡:Nivida GTX965M 开发工具:Visual Studio 2017 诸如绘制和内存操作相关命令,在Vulkan中不是通过函数直接调用的.我们需要在命令缓 ...

  9. Vulkan Tutorial 29 Loading models

    操作系统:Windows8.1 显卡:Nivida GTX965M 开发工具:Visual Studio 2017 Introduction 应用程序现在已经可以渲染纹理3D模型,但是 vertice ...

随机推荐

  1. js 中json遍历 添加 修改 类型转换

    <!DOCTYPE html> <html> <head lang="en"> <meta charset="UTF-8&quo ...

  2. Flask Bug记录之The innermost block that needs to be closed is 'block'.

    源码 <!DOCTYPE html> <title>{% block title %}{% endblock title %} - Flask</title> &l ...

  3. ros msg和srv使用

    在包文件中新建文件夹srv和msg,在这两个文件夹中新建test.msg,test.srv 修改apckage.xml 添加以下内容 <build_depend>:message_gene ...

  4. 第7章:LeetCode--算法:递归问题

    70. Climbing Stairs This problem is a Fibonacci problem.F(n)=F(n-1)+F(n-2);Solving this problem by r ...

  5. python-day4(正式学习)

    数据类型 不同种类的变量值用不同的数据类型来描述 数字类型 整形:int 定义方式: name(变量名)=(赋值符号)180(变量值) name=int(180) 当变量值为小数时,自动向下取整 运算 ...

  6. java 内部类、匿名内部类

    一:内部类 1:什么是内部类? 大部分时候,类被定义成一个独立的程序单元.在某些情况下,也会把一个类放在另一个类的内部定义,这个定义在其他类内部的类就被称为内部类(有些地方也叫做嵌套类),包含内部类的 ...

  7. 【Trie】Nikitosh 和异或

    [参考博客]: LOJ#10051」「一本通 2.3 例 3」Nikitosh 和异或(Trie [题目链接]: https://loj.ac/problem/10051 [题意]: 找出两个不相交区 ...

  8. Jmeter博文索引~基础知识和实践操作汇总

    所有Jmeter笔记的目录/索引 一,基础操作和常用操作 Jmeter入门(一)理论基础 Jmeter安装及配置(含JDK安装) Jmeter之设置线程组运行次数/时间 Jmeter之参数化(4种设置 ...

  9. Excel2016 保存\复制 卡死问题解决

    遇到的问题: 工作中经常碰到一些Excel表, 复制一行, 再粘贴要等5s以上才能显示成功. 保存一下文档, 也会出现页面白屏卡死的情况, 经过网上多个帖子进行操作依旧无解, 最后找到了自己的方法得以 ...

  10. spring-boot-plusV1.2.3发布,CentOS快速安装环境/构建/部署/启动项目

    spring-boot-plusV1.2.3发布,CentOS快速安装环境/构建/部署/启动项目 [V1.2.3-RELEASE] 2019.09.09