发明名称 Fencing network direct memory access data transfers in a parallel active messaging interface of a parallel computer
摘要 Fencing direct memory access (‘DMA’) data transfers in a parallel active messaging interface (‘PAMI’) of a parallel computer, the PAMI including data communications endpoints, each endpoint including specifications of a client, a context, and a task, the endpoints coupled for data communications through the PAMI and through DMA controllers operatively coupled to a deterministic data communications network through which the DMA controllers deliver data communications deterministically, including initiating execution through the PAMI of an ordered sequence of active DMA instructions for DMA data transfers between two endpoints, effecting deterministic DMA data transfers through a DMA controller and the deterministic data communications network; and executing through the PAMI, with no FENCE accounting for DMA data transfers, an active FENCE instruction, the FENCE instruction completing execution only after completion of all DMA instructions initiated prior to execution of the FENCE instruction for DMA data transfers between the two endpoints.
申请公布号 US9075759(B2) 申请公布日期 2015.07.07
申请号 US201012940282 申请日期 2010.11.05
申请人 International Business Machines Corporation 发明人 Blocksome Michael A.;Mamidala Amith R.
分类号 G06F15/16;G06F15/173;G06F7/38;H04L29/06;G06F9/54 主分类号 G06F15/16
代理机构 Kennedy Lenart Spraggins LLP 代理人 Kennedy Lenart Spraggins LLP
主权项 1. A parallel computer that fences direct memory access (DMA) data transfers in a parallel active messaging interface (‘PAMI’), the parallel computer comprising a plurality of compute nodes that execute a parallel application, the compute node comprising: computer processors; and computer memory coupled to the computer processors having disposed within it computer program instructions that, when executed by the computer processors, cause the parallel computer to function by: receiving, at an origin endpoint in the PAMI, a DMA PUT instruction from an origin application, wherein the DMA PUT instruction specifies a data transfer between the origin endpoint and a target endpoint in the PAMI, wherein the DMA PUT instruction, upon execution, transmits data from the origin endpoint directly into a target random access memory (RAM) for the target endpoint, wherein each endpoint specifies a client, a context, and a task; receiving, at an origin endpoint, a FENCE instruction corresponding the data transfer between the origin endpoint and the target endpoint, wherein the FENCE instruction advises the application when the transfers between the origin endpoint and the target endpoint are complete, and wherein the transfers between the origin endpoint and the target endpoint occur across a deterministic network connecting the origin endpoint and the target endpoint; posting the DMA PUT instructions and the FENCE instruction to a work queue of the context for the origin endpoint, wherein the FENCE instruction is posted at a location in the ordered sequence such that the FENCE instruction is processed after the plurality of DMA PUT instructions is processed; and executing, through the PAMI, by the origin endpoint, an advance function that executes instructions in the work queue strictly in the ordered sequence, wherein execution of the FENCE instruction indicates to the origin application that the execution of the DMA PUT instruction is complete.
地址 Armonk NY US