Cache to cache data transfer acceleration techniques

A method comprising:

• receiving a memory request from a first processing node;
• generating a prediction of whether data targeted by the memory request resides in a cache subsystem of another processing node of a plurality of processing nodes;
• and
• causing a speculative probe to be sent to a second processing node responsive to predicting that the data targeted by the memory request resides in a cache subsystem of the second processing node;
• the second processing node:
• receiving the speculative probe;
• retrieving the data responsive to determining that the data targeted by the speculative probe is cached in a cache subsystem of the second processing node;
• storing the retrieved data in a buffer until a demand probe arrives;
• starting a timer;
• and
• discarding the data if the timer expires before a demand probe targeting the data is received.

An apparatus comprising:

• a first processing node comprising one or more processors and a cache subsystem;
• and
• a first coherent master unit comprising a first prediction unit, wherein the first prediction unit is configured to:
• receive a memory request from the first processing node;
• cause a speculative probe to be sent to a second processing node responsive to predicting that data targeted by the memory request resides in a cache subsystem of the second processing node;
• wherein the second processing node is configured to:
• receive the speculative probe;
• retrieve the data responsive to determining that the data targeted by the speculative probe is cached in a cache subsystem of the second processing node;
• store the retrieved data in a buffer until a demand probe arrives;
• start a timer;
• and
• discard the data if the timer expires before a demand probe targeting the data is received.