Specialized fixed function hardware for efficient convolution

A method comprising:

• storing an instruction in an instruction cache of a graphics processor, wherein the instruction is a single instruction to perform a multi-dimensional parallel matrix multiply operation;
• scheduling, via a scheduler of the graphics processor, a plurality of threads for execution of the instruction;
• and
• performing, via a compute block of the graphics processor, parallel multiply-accumulate operations based on the instruction, wherein performing the parallel multiply-accumulate operations includes processing, in parallel via the plurality of threads, a corresponding plurality of matrix elements to multiply a first matrix and a second matrix, and wherein multiplying the first matrix and the second matrix includes multiplying data elements in a plurality of rows of the first matrix by corresponding data elements in a plurality of columns of the second matrix to generate a plurality of products.

A general-purpose parallel processor comprising:

• an instruction cache to store an instruction, wherein the instruction is a single instruction to perform a multi-dimensional parallel matrix multiply operation;
• and
• a compute block configured to perform multiply-accumulate operations in response to execution of the instruction, the compute block including:
• a scheduler to schedule a plurality of threads for execution of the instruction;
• and
• multiply-accumulate circuitry configured to execute the instruction via the plurality of threads, wherein the multiply-accumulate circuitry includes a plurality of functional units specialized to perform matrix multiply and accumulate operations, the plurality of functional units configured to process, in parallel via the plurality of threads, a corresponding plurality of matrix elements to multiply a first matrix and a second matrix, and to multiply the first matrix and the second matrix includes to multiply data elements in a plurality of rows of the first matrix by corresponding data elements in a plurality of columns of the second matrix to generate a plurality of products.