Memory organization and access for efficient matrix operations

A system, comprising:

• memory having executable instructions stored thereon;
• and
• one or more processors configured, individually or collectively, to execute the executable instructions in order to cause the system to:
• organize a plurality of physical memory banks having a base size into a plurality of logical memory banks that comprise a first set of logical memory banks having a size equal to the base size and a second set of logical memory banks having a size double that of the base size;
• receive a request to execute operations on the plurality of physical memory banks, wherein the request to execute the operations comprises a request to interact with data having a sample width based on the base size and a request to read the data from the plurality of physical memory banks;
• and
• responsive to receiving the request to execute the operations, execute the operations on one or more logical memory banks of the plurality of logical memory banks via a memory crossbar shared across the plurality of logical memory banks, wherein an amount of the data on which the operations are executed is a multiple of the sample width, wherein each logical memory bank of the one or more logical memory banks has a size based on the base size and a multiplier value, wherein in order to execute the operations, the one or more processors are configured, individually or collectively, to cause the system to retrieve the data from at least one first logical memory bank of the second set of logical memory banks via the shared memory crossbar, and wherein in order to retrieve the data from the at least one first logical memory bank of the second set of logical memory banks, the one or more processors are configured, individually or collectively, to cause the system to retrieve the data from logical memory banks of the first set of logical memory banks that form the at least one first logical memory bank of the second set of logical memory banks.

A system, comprising:

• means for organizing a plurality of physical memory banks having a base size into a plurality of logical memory banks that comprise a first set of logical memory banks having a size equal to the base size and a second set of logical memory banks having a size double that of the base size;
• means for receiving a request to execute operations on the plurality of physical memory banks, wherein the request to execute the operations comprises a request to interact with data having a sample width based on the base size and a request to read the data from the plurality of physical memory banks;
• and
• means for executing, responsive to receiving the request to execute the operations, the operations on one or more logical memory banks of the plurality of logical memory banks via a memory crossbar shared across the plurality of logical memory banks, wherein an amount of the data on which the operations are executed is a multiple of the sample width, wherein each logical memory bank of the one or more logical memory banks has a size based on the base size and a multiplier value, wherein the means for executing the operations comprises means for retrieving the data from at least one first logical memory bank of the second set of logical memory banks via the shared memory crossbar, and wherein the means for retrieving the data from the at least one first logical memory bank of the second set of logical memory banks comprises means for retrieving the data from logical memory banks of the first set of logical memory banks that form the at least one first logical memory bank of the second set of logical memory banks.

A non-transitory computer-readable medium having instructions stored thereon which, when executed by one or more processors individually or collectively, perform an operation comprising:

• organizing a plurality of physical memory banks having a base size into a plurality of logical memory banks that comprise a first set of logical memory banks having a size equal to the base size and a second set of logical memory banks having a size double that of the base size;
• receiving a request to execute operations on the plurality of physical memory banks, wherein the request to execute the operations comprises a request to interact with data having a sample width based on the base size and a request to read the data from the plurality of physical memory banks;
• and
• responsive to receiving the request to execute the operations, executing the operations on one or more logical memory banks of the plurality of logical memory banks via a memory crossbar shared across the plurality of logical memory banks, wherein an amount of the data on which the operations are executed is a multiple of the sample width, wherein each logical memory bank of the one or more logical memory banks has a size based on the base size and a multiplier value, wherein executing the operations comprises retrieving the data from at least one first logical memory bank of the second set of logical memory banks via the shared memory crossbar, and wherein retrieving the data from the at least one first logical memory bank of the second set of logical memory banks comprises retrieving the data from logical memory banks of the first set of logical memory banks that form the at least one first logical memory bank of the second set of logical memory banks.