Message description information transferring method between kernels

Abstract

The invention discloses a transferring method of an inner core message depict information. The inner core uses the ring type alignment to do the communication; it uses the reading pointer and the writing pointer control of the ring type alignment to transfer the receiving and sending of the message depict information of each message. The method comprises: setting the message finishing field for signing the finish of a message in the message depict information read and write field of each item entrance entry of the ring type alignment; writing the message depict information of each message in the ring type alignment by the message finishing field; reading the message depict information of each message in the ring type alignment by the message finishing field.

Description

technical field [0001] The invention relates to a communication method between cores in a multi-core processor, more specifically, the invention relates to a method for transmitting message description messages between cores in a multi-core processor. Background technique [0002] With the development of information technology, the data flow processed by the processor is increasing. Traditional processors with a single processing core can no longer meet the processing performance requirements. Therefore, processors with multiple processing cores have emerged. In such multi-core processors, it is customary to implement a specific functional module in one processing core. The cores of this type of processor use public dynamic random access Memory (DRAM, Dynamic Random-Access Memory), static random access memory (SRAM, Synchronous Random-Access Memory) and other resources store and process data packets, so that the transfer of data packets between different processing cores do...

AI Technical Summary

Problems solved by technology

[0020] 2) Release the buffer pointer that was received last time in the entry. If the buffer pointer is modified by the receiving end, the buffer pointer may be lost

[0025] First of all, according to the above entry structure, the message description message read and write control field includes R, D, S, E fields, and the processing of the R, D, S fields will cause the ring to hang abnormally

For example, if the receiving end kernel finds that read_ptr is not equal to write_ptr, it thinks that the message needs to be received, but the D field in the entry (due to misoperation) is not set to true, then the receiving end kernel will wait in the current entry, causing the ring to hang; otherwise, if If the R field is set incorrectly at the sending end, the processing will also stop

[0026] Secondly, in the above-mentioned prior art, if the processing of the R and D fields of the entry structure is abnormal or the buffer is not released, the buffer may be lost, causing problems in the processing of the message

[0027] Again, because the message buffer address buffer addr and R, D, S, E are set in a long word, during specific processing, the kernel may cause the long word to read and write the long word error (for example, the CPU reads and writes the long word through PCI ), so that the message cache address is lost

For example, in the operation of R, D, S, E and other fields, the error of buffer pointer may be caused (because in a long word, misoperation is unavoidable), and the cache buffer pointed to by the correct buffer pointer will be lost

Images