Search

CN-121478706-B - Request broadcasting method, multi-chip device, storage medium, and computer program product

CN121478706BCN 121478706 BCN121478706 BCN 121478706BCN-121478706-B

Abstract

The application provides a broadcasting method for DVM requests of a multi-chip device, each chip in the multi-chip device comprises a plurality of crystal grains, each crystal grain in the plurality of crystal grains comprises a plurality of DN domains, each DN domain in the plurality of DN domains comprises a DN and a plurality of connection cores, each DN in each crystal grain comprises a main DN, the first connection core initiates a first DVM request to a first DN of a first DN domain to which the first connection core belongs, wherein the crystal grain to which the first DN belongs is a first crystal grain, the chip to which the first crystal grain belongs is a first chip, the first DN broadcasts the first DVM request to other DNs except the first DN of the first crystal grain, and the first main DN of the first crystal grain broadcasts the first DVM request to the main DNs of the other crystal grains except the first crystal grain of the first crystal grain. The application also provides a multi-chip device, a computer readable storage medium and a computer program product.

Inventors

  • YANG MENGXUE

Assignees

  • 深圳市中兴微电子技术有限公司

Dates

Publication Date
20260512
Application Date
20260112

Claims (12)

  1. 1. A broadcast method for distributed virtual memory, DVM, requests for a multi-chip device, each chip in the multi-chip device comprising a plurality of dies, each die in the plurality of dies comprising a plurality of DVM node, DN, fields, each DN field in the plurality of DN fields comprising a DN and a plurality of connection cores, a master DN included in a plurality of DNs for each die, the method comprising: A first connection core initiates a first DVM request to a first DN of a first DN domain to which the first connection core belongs, wherein a grain to which the first DN belongs is a first grain, and a chip to which the first grain belongs is a first chip; the first DN broadcasting the first DVM request to other DNs of the first die than the first DN; Broadcasting the first DVM request to the primary DNs of other dies of the first chip than the first die by the first primary DN of the first die; Broadcasting the first DVM request to other DNs of the grains to which the master DN belongs by the master DNs of the other grains except the first grain of the first chip; Each DN of the other die broadcasts the first DVM request to a plurality of connected cores of a DN domain to which the DN belongs.
  2. 2. The method according to claim 1, wherein the method further comprises: the first DN broadcasts the first DVM request to other connection cores of the first DN domain than the first connection core.
  3. 3. The method according to claim 1, wherein the method further comprises: The other DNs of the first die, except for the first DN, broadcast the first DVM request to a plurality of connection cores of a DN domain to which the DN belongs.
  4. 4. A method according to any one of claims 1 to 3, wherein the first master DN is a DN closest to the other die among a plurality of DNs of the first die.
  5. 5. The method of claim 1, wherein the plurality of DNs for each die includes a cross-chip transmit DN, the method further comprising: Broadcasting the first DVM request to at least one chip-to-chip node bound by the cross-chip send DN of the first die; each of the at least one chip-to-chip node broadcasts the first DVM request in N lanes to N chips of the multi-chip device other than the first chip, where N is a positive integer.
  6. 6. The method of claim 5, wherein each of the at least one chip-to-chip node binds M DNs within the first chip, wherein M is a positive integer, the method further comprising: Each of the at least one chip-to-chip node forwards a second DVM request received from at least one of the M chips other than the first chip to a corresponding DN of the M DNs.
  7. 7. The method of claim 6, wherein the method further comprises: A second DN receiving a second DVM request forwarded from a chip to a chip node broadcasts the second DVM request to other DNs except the second DN of a second die to which the second DN belongs; broadcasting the second DVM request to a plurality of connection cores of a DN domain to which the DN belongs by each DN of the second crystal grain; The master DN of the second die broadcasts the second DVM request to the master DNs of other dies of the first chip than the second die.
  8. 8. The method of claim 7, wherein the method further comprises: Broadcasting the second DVM request to other DNs of the grains of the first chip except the second crystal grain by using the main DNs of the other crystal grains of the first chip except the second crystal grain; Each DN of the other die broadcasts the second DVM request to a plurality of connected cores of a DN domain to which the DN belongs.
  9. 9. The method of any of claims 5-8, wherein the cross-chip transmit DN is a DN of the plurality of DNs that minimizes a communication path or network load.
  10. 10. A multi-chip device, each chip in the multi-chip device comprising a plurality of dies, each die of the plurality of dies comprising a plurality of distributed virtual memory DVM node DN domains, each DN domain of the plurality of DN domains comprising a DN and a plurality of connection cores, each die comprising a master DN, the multi-chip device being configured to perform the method of any of claims 1-9.
  11. 11. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, causes the processor to implement the method according to any of claims 1 to 9.
  12. 12. A computer program product, characterized in that the computer program product comprises a computer program which, when executed by a processor, causes the processor to implement the method according to any one of claims 1 to 9.

Description

Request broadcasting method, multi-chip device, storage medium, and computer program product Technical Field Embodiments of the present application relate to the field of virtual memory management, and in particular, to a broadcasting method for distributed virtual memory requests of a multi-chip device, a computer-readable storage medium, and a computer program product. Background In the field of virtual memory, in order to solve the problems of complex memory access, difficult programming, low resource utilization rate and the like in a distributed system, a distributed virtual memory (distributed virtual memory, DVM) technology has been widely adopted to uniformly manage the memory resources of a processor. With the development of packaging technology and the increasing demand for heterogeneous computing, multi-Die (Die) and multi-Chip (Chip) architectures are becoming an important trend for high-performance Chip designs. In such architecture, efficient inter-die and inter-chip communication is one of the key technical difficulties in system design. How to realize the effective communication of the DVM mechanism between the multiple dies and the multiple chips has become a problem to be solved with emphasis. Disclosure of Invention The embodiment of the application provides a broadcasting method for a distributed virtual memory request of a multi-chip device, the multi-chip device, a computer readable storage medium and a computer program product, which are used for effectively relieving load pressure of cross-grain communication, improving processing efficiency of cross-chip broadcasting, reducing occupation of system bandwidth and improving consistency maintenance efficiency and performance of a system. The embodiment of the application provides a broadcasting method for DVM requests of a multi-chip device, wherein each chip in the multi-chip device comprises a plurality of crystal grains, each crystal grain in the plurality of crystal grains comprises a plurality of DVM Node (DN) domains, each DN domain in the plurality of DN domains comprises one DN and a plurality of connecting cores, one main DN is included in the plurality of DNs of each crystal grain, the method comprises the steps that a first connecting core initiates a first DVM request to a first DN of a first DN domain which the first connecting core belongs to, wherein the crystal grain which the first DN belongs to is a first crystal grain, the chip which the first crystal grain belongs to is a first chip, the first DN broadcasts the first DVM request to other DNs except the first crystal grain, and the first main DN of the first crystal grain broadcasts the first DVM request to the main DNs of the other crystal grains except the first crystal grain. Embodiments of the present application also provide a multi-chip device, each chip in the multi-chip device including a plurality of dies, each die in the plurality of dies including a plurality of DN domains, each DN domain in the plurality of DN domains including a DN and a plurality of connection cores, each die including a master DN in the plurality of DNs, the multi-chip device configured to perform a method of broadcasting a DVM request according to an embodiment of the present application. The embodiment of the application also provides a computer readable storage medium, on which a computer program is stored, the computer program, when executed by a processor, causes the processor to implement a method for broadcasting DVM requests according to the embodiment of the application. Embodiments of the present application also provide a computer program product comprising a computer program which, when executed by a processor, causes the processor to implement a method of broadcasting a DVM request according to an embodiment of the present application. According to the technical scheme provided by the embodiment of the application, the cross-grain broadcasting is performed by introducing the main DN, so that the communication path can be effectively shortened, the delay can be reduced, the bandwidth occupation can be reduced, and the problem of network congestion can be relieved. Drawings The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application. The above and other features and advantages will become more readily apparent to those skilled in the art by describing in detail exemplary embodiments with reference to the attached drawings, in which: FIG. 1 shows a schematic block diagram of one die in a multi-chip device; FIG. 2 illustrates a schematic diagram of cross-die DVM request broadcasting in the related art; FIG. 3 shows a schematic block diagram of one die in a multi-chip device according to an embodiment of the application; FIG. 4 illustrates a schematic diagr