Direct IO access from a CPU's instruction stream

摘要

A method for network access of remote memory directly from a local instruction stream using conventional loads and stores. In cases where network IO access (a network phase) cannot overlap a compute phase, a direct network access from the instruction stream greatly decreases latency in CPU processing. The network is treated as yet another memory that can be directly read from, or written to, by the CPU. Network access can be done directly from the instruction stream using regular loads and stores. Example scenarios where synchronous network access can be beneficial are SHMEM (symmetric hierarchical memory access) usages (where the program directly reads/writes remote memory), and scenarios where part of system memory (for example DDR) can reside over a network and made accessible by demand to different CPUs.

主权项

1. A method for accessing data comprising the steps of:
(a) issuing, by a CPU, a LOAD command including a virtual address; (b) deriving network access information based on said virtual address, said network access information including one or more references to a remote memory; (c) checking, using said one or more references to said remote memory, if access is allowed to said remote memory; (d) if access is denied to said remote memory:
(i) then
(A) notifying the CPU said LOAD command failed; and(B) invoking management software to handle said failure of said LOAD command,(ii) else attempt to load data from said remote memory; (e) if said attempt is successful:
(i) then returning said data to the CPU;(ii) else
(A) notifying the CPU said LOAD command failed; and(B) invoking management software to handle said failure of said LOAD command.