网络计算技术加速 GPU 应用.pdf

编号：29476

PDF 68页 2.75MB 下载积分：VIP专享

下载报告请您先登录！

网络计算技术加速 GPU 应用.pdf

1、GTC CHINA 2020 Qingchun Song December 2020 IN-NETWORK COMPUTING PROGRAMMING FOR GPU APPLICATIONS 2 REMOTE DIRECT MEMORY ACCESS (RDMA) GPU DIRECT RDMA (GDR) GPU DIRECT STORAGE (GDS) SCALABLE HIERARCHICAL AGGREGATION AND REDUCTION PROTOCOL (SHARP) NCCL SHARP AGENDA 3 REMOTE DIRECT MEMORY ACCESS (RDMA)

2、 4 INFINIBANDS LAYERED ARCHITECTURE 4 5 RDMA DATA TRANSFER MODEL Queue Pair (QP) QPs are in pairs (Send/Receive) Work Request (WR) Work items that the HW should perform Work Completion (completion) When a WR is completed, it may create a Work Completion which provides information about the ended WR

3、Work Queue (WQ) A queue which contains WRs Scheduled by the HW Can be either Send or Receive Queue Completion Queue (CQ) 6 RDMA OPERATION: SEND The responder Post Receive Requests (before data is received) The requester Post Send Request Only data is sent over the wire ACK is sent only in reliable transport types 6 7 RDMA OPERATION: RDMA WRITE The requester Post Send Request Data and remote memory

友情提示

1、下载报告失败解决办法
2、PDF文件下载后，可能会被浏览器默认打开，此种情况可以点击浏览器菜单，保存网页到桌面，就可以正常下载了。
3、本站不支持迅雷下载，请使用电脑自带的IE浏览器，或者360浏览器、谷歌浏览器下载即可。
4、本站报告下载后的文档和图纸-无水印,预览文档经过压缩，下载后原文更清晰。

本文（网络计算技术加速 GPU 应用.pdf）为本站（X-iao）主动上传，三个皮匠报告文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若此文所含内容侵犯了您的版权或隐私，请立即通知三个皮匠报告文库（点击联系客服），我们立即给予删除！

温馨提示：如果因为网速或其他原因下载失败请重新下载，重复下载不扣分。