Method and apparatus for accelerated shared data migration

Abstract

A method and apparatus for accelerated shared data migration between cores, Using an Always Migrate protocol, when a migratory probe hits a directory entry in either modified or owned state, the entry is transitioned to an owned state, and a source done command is sent without sending cache block ownership or state information to the directory.

What is claimed is: 
     
         1 . A method for shared data management in a processor comprising:
 migrating ownership of a data block in accordance with a directory protocol to a requestor responsive to a data request for data within the data block from the requestor,   wherein the directory protocol is an always migrate protocol,   wherein the always migrate protocol is a directory and broadcast protocol optimized for migratory sharing which always forces ownership of the data block to migrate to a current requester regardless of a probe target processor's state; and   sending a source done command to indicate a completed transaction without sending information for a directory update.   
     
     
         2 . The method of  claim 1  further comprising:
 performing atomic directory lookup corresponding to the data request; and 
 performing atomic update of the directory based on the atomic directory lookup. 
 
     
     
         3 . The method of  claim 1 , wherein the always migrate protocol is associated with a plurality of caches. 
     
     
         4 . A memory subsystem comprising:
 a controller configured to receive, from a requestor, a data request for data within a data block;   the memory subsystem configured to migrate ownership of the data block in accordance with a directory protocol to the requestor responsive to receiving the data request,   wherein the directory protocol is an always migrate protocol,   wherein the always migrate protocol is a directory and broadcast protocol optimized for migratory sharing which always forces ownership of the data block to migrate to a current requester regardless of a probe target processor's state; and   sending a source done command to indicate a completed transaction without sending information for a directory update.   
     
     
         5 . The memory subsystem of  claim 4  further comprising:
 a coherence directory performing atomic directory lookup corresponding to the data request; and 
 a coherence directory performing atomic update of the directory corresponding to the atomic directory lookup. 
 
     
     
         6 . The method of  claim 5 , wherein the always migrate protocol is associated with a plurality of caches. 
     
     
         7 . A non-transitory computer-readable storage medium storing design code representing a hardware design of a processor that is to perform a method to provide shared data management, the method comprising:
 migrating ownership of a data block to a requestor responsive to a data request for data within the data block from the requestor,   wherein the directory protocol is an always migrate protocol,   wherein the always migrate protocol is a directory and broadcast protocol optimized for migratory sharing which always forces ownership of the data block to migrate to a current requester regardless of a probe target processor's state; and   sending a source done command to indicate a completed transaction without sending information for a directory update.   
     
     
         8 . The non-transitory computer-readable storage medium of  claim 7 , wherein design code comprises hardware description language (HDL) instructions used for the manufacture of a device including the processor.