User-transparent operation core parallel printing order-preserving agent method

Abstract

The invention discloses a user-transparent operation core parallel printing order-preserving agent method, and the method comprises the following steps: expanding and replacing printing output statements on an operation core, and converting the printing output statements into a thread printing processing function in a calling runtime system; performing parameter analysis on the thread printing processing function, applying for a data transfer buffer area, and performing data transfer according to the parameter type; enabling the operation core to request the control core to carry out printingorder-preserving agent, and transmitting the head address of the applied data transfer buffer area to the control core; enabling the control core to receive the interrupt signal, transferring the signal to interrupt processing, executing an embedded assembly instruction to fill printing function parameters according to the parameter type of the data transfer buffer, and calling a printf agent to realize printing output of the operation core. According to the invention, the compiler and the runtime are combined to automatically complete parallel and ordered output of all operation cores; the method is transparent to a user and does not need to additionally manage the printout of the operation core; and only a basic compiler of the operation core is required to perform simple program transformation on an operation core program.

Description

technical field [0001] The invention belongs to the technical field of heterogeneous many-cores, and in particular relates to a user-transparent computing core parallel printing order-preserving agent method. Background technique [0002] The computing cores of heterogeneous many-cores do not support operating systems in the usual sense. There is only a simple management library, and the support for user program printing functions is limited. When multiple computing cores call the printing function at the same time, the multiple computing cores The screen output will be interlaced, and useful information cannot be identified, which will affect the debugging and operation control of the program. figure 1 As shown in , it is a schematic representation of the printed information of the eight computing cores. It can be seen that the printed information is intertwined and chaotic, and useful information cannot be effectively identified. The inability of multiple computing cores ...

AI Technical Summary

Problems solved by technology

This I/O proxy technology has the following deficiencies: (1) This I/O proxy technology must require a parallel compiler to process parallel programs from source to source. Complicated analysis and program transformation are required to solve the problems of thi

Images