Multi-level classification method for transaction address conflicts for ensuring efficient ordering in a two-level snoopy cache architecture

A method of classification of address conflicts between an operation occurring first in time and one or more operations occuring second in time, in a multiprocessor system having a plurality of nodes coupled by an interconnecting communications pathway comprised of a central hardware device which is capable of storing information regarding the location and state of data within the system, each node having at least one cache, a memory device local to the node and at least one processor device, the memory and processor device being coupled to form a complete subsystem, the processor device within each node being capable of accessing data from the local memory device, the local cache, or over the interconnecting communications pathway from a non local memory device, or a non local cache, the method including the steps of

• classification of a first in time operation;
• classification of a second in time operation;
• comparing the classification of said first operation with said second operation;
• selecting a dependency criteria from a dependency release table based on said classification of said first and said second operation;
• and
• releasing said second operation based on said above release criteria,
• wherein said dependency release criteria is comprised of:
• a first class wherein the said first operation is placed in an ordered processor bus queue before the said second operation can proceed;
• a second class wherein said first operation is placed in an ordered memory queue before said second operation can proceed;
• and
• a third class wherein said first operation must have all required dependencies on that transaction released before said second operation can proceed in the system.

In a multiprocessor system having a plurality of nodes coupled by an interconnecting communications pathway comprised of a central hardware device which is capable of storing information regarding the location and state of data within the system, each node having at least one cache, a memory device local to the node and at least one processor device, the memory and processor device being coupled to form a complete subsystem, the processor device within each node being capable of accessing data from the local memory device, the local cache, or over the interconnecting communications pathway from a non local memory device, or a non local cache, wherein such system classifies transactions within the system in part in accordance with the address of the transaction, and one or more transactions occurring later in time may conflict with an address of with a transaction previous in time, a method of handling conflicts between such transactions including the steps of

• placing conflicting transactions later in time a ordering queue;
• holding said conflicting transactions in said ordering queue until any first in tune transactions with which said conflicting transactions conflict have progressed to a point defined by a predetermined classification of the relationship between said first in time and said second in time transactions;
• and
• releasing said second in time transactions from said ordering queue,
• wherein said predetermined, classification includes:
• a first class wherein said first in time transaction is placed in an ordered processor bus of said at least one processor device;
• a second class wherein said first in time transaction is placed in an ordered memory queue;
• and
• a third class wherein said first in time transaction has all required dependencies on that transaction throughout the system released before any conflicting second in time transaction can proceed in the system.

An article executable in a multiprocessor system, the system having a plurality of interconnected nodes, each node having at least one local memory device and at least one processor device capable of accessing data from both the local memory device of said node and the local memory device of another node, the article comprising:

• a classification of a first in time transaction to be executed by one of the processors, said classification being based at least in part on location of data to be accessed during execution of the transaction;
• a classification of a second in time transaction to be executed by the same or another one of the processors, said classification being based at least in part on location of data to be accessed during execution of the transaction;
• an execution dependency criterion based on the classifications;
• a deferral of the second in time transaction based on the criterion;
• and
• a release of the second in time transaction for execution based at least in part on the criterion and on execution of the first in time transaction,
• wherein the classification of the first in time transaction is further based on one or more factors selected from the group consisting of a source from which the transaction was initiated, and a type of transaction,
• wherein the nodes are interconnected by a central hardware device storing information regarding location of data within the system, and
• wherein the classification of the first in time transaction is further based on one or more factors selected from the group consisting of:
• a result of a cache snoop;
• a response of the central hardware device;
• and whether the central hardware device requires an acknowledgment.

A computer system comprising:

• a plurality of interconnected nodes, each node having at least one local memory device and at least one processor device capable of accessing data from both the local memory device of said node and the local memory device of another node;
• a classification of a first in time transaction to be executed by one of the processors, said classification being based at least in part on location of data to be accessed during execution of the transaction;
• a classification of a second in time transaction to be executed by the same or another one of the processors, said classification being based at least in part on location of data to be accessed during execution of the transaction;
• an execution dependency criterion based on the classifications;
• a deferred execution queue for the second in time transaction based on the criterion, and
• a release of the second in time transaction for execution based at least in part on the criterion and on execution of the first in time transaction,
• wherein the execution dependency criterion comprises a criterion selected from the group consisting of placement of the first in time transaction in an ordered processor bus queue;
• placement of the first in time transaction in an ordered memory queue;
• and release of all required dependencies of the first in time transaction.