Tracking long ghv in high performance out-of-order superscalar processors

A method of tracking global history vector in high performance processors, comprising:

• receiving a fetch group comprising a group of instructions for execution on a processor;
• tagging the fetch group with a youngest bit location of a circular buffer, the circular buffer implementing a global history vector comprising a plurality of bits, each of the bits for associating with a corresponding fetch group;
• receiving branch prediction associated with the fetch group, predicted using at least the global history vector;
• storing the branch prediction in one or more next entries to the youngest bit location of the circular buffer;
• and
• updating the youngest bit location of the circular buffer based on a number of bits inserted in the circular buffer,
• wherein the global history vector represents an execution path.

An apparatus for tracking global history vector in high performance processors, comprising:

• a shift register comprising a plurality of bits implementing the global history vector;
• a shift counter associated with the global history vector, the shift counter storing current shift count of the global history vector;
• and
• a global history vector mechanism operable to receive a fetch group comprising a group of instructions for execution on a processor,
• the global history vector mechanism further operable to tag the fetch group with the current shift count of the global history vector,
• the global history vector mechanism further operable to receive branch prediction associated with the fetch group, predicted using at least the global history vector;
• the global history vector mechanism further operable to shift the global history vector by a number of bits to a direction of older bit position,
• the global history vector mechanism further operable to store the branch prediction associated with the fetch group in most recent bit position of the global history vector,
• the global history vector mechanism further operable to update the current shift count of the global history vector based on the number of bits,
• wherein the global history vector represents an execution path and each of the plurality of bits are allocated for a corresponding fetch group.

An apparatus for tracking global history vector in high performance processors, comprising:

• a circular buffer implementing the global history vector comprising a plurality of bits, each of the bits for associating with a corresponding fetch group;
• a younger pointer hardware component tracking a youngest bit location of the circular buffer;
• an oldest pointer hardware component tracking an oldest bit location of the circular buffer;
• and
• a global history vector mechanism operable to receive a fetch group comprising a group of instructions for execution on a processor,
• the global history vector mechanism further operable to tag the fetch group with the youngest bit location of the circular buffer,
• the global history vector mechanism further operable to receive branch prediction associated with the fetch group, predicted using at least the global history vector,
• the global history vector mechanism further operable to store the branch prediction in one or more next entries to the youngest bit location of the circular buffer, and
• the global history vector mechanism further operable to update the youngest bit location of the circular buffer based on a number of bits inserted in the circular buffer,
• wherein the global history vector represents an execution path.