H2D write operation acceleration method and apparatus, compute device and storage medium

Abstract

The invention relates to an H2D write operation acceleration method and apparatus, a computer device and a storage medium. The method includes the steps of designing an FIFO queue for storing a Flashaddress; obtaining an H2D write command of a host; analyzing and processing the H2D write command; moving the data into the buffer space of a memory; obtaining a preset Flash address to be written from a first in first out queue; writing data into a Flash address to be written; recycle buffer space for memory. As the FIFO array is arranged and the Flash address to be written obtain by an FTL algorithm is stored in advance, so when the hardware needs to write data with the Flash address to be written, the Flash address to be written is directly called without waiting for the FTL algorithm to provide the Flash address to be written in real time, thus simplifying the software operation steps, greatly reducing the CPU overhead, fully releasing the hardware performance, and improving the SSD writing performance.

Description

technical field [0001] The present invention relates to an H2D writing operation method, and more specifically refers to an H2D writing operation acceleration method, device, computer equipment, and storage medium. Background technique [0002] The existing H2D write command method is as follows figure 1 As shown, it includes: hardware obtains H2D write commands; firmware waits for commands (including query or interrupt mode); firmware parses commands; firmware applies for buffer space; firmware generates NVME PRD commands (ie non-volatile memory host controller interface specification PRD command) and send it to the NVME circuit for execution; the firmware FTL algorithm generates the Flash address to be written; the NVME circuit reads data from the host according to the PRD command and stores it in the buffer; the firmware waits for the data transmission to complete and the FTL algorithm generates the flash address to be written After entering the Flash address, generate a...

AI Technical Summary

Problems solved by technology

Among them, 6 steps are pure software actions, 5 steps are software and hardware interactions, and only 2 steps are pure hardware actions. Due to the existence of many software actions, the CPU overhead is large; there are many software and hardware interactions, and there is time overhead. Often have to wait for the software trigger to start working, resulting in low efficiency of H2D write operations and low SSD read and write performance

Images