Apparatuses, methods, and systems for instructions for matrix multiplication instructions
Techniques for matrix multiplication are described. In some examples, decode circuitry is to decode a single instruction having fields for an opcode, an indication of a location of a first source operand, an indication of a location of a second source operand, and an indication of a location of a destination operand, wherein the opcode is to indicate that execution circuitry is to at least convert data elements of the first and second source operands from a first floating point representation to a second floating point representation, perform matrix multiplication with the converted data elements, and accumulate results of the matrix multiplication in the destination operand in the first floating point representation; and the execution circuitry is to execute to the decoded instruction as specified by the opcode.
1. An apparatus comprising:
decode circuitry to decode a single instruction having fields for an opcode, an indication of a location of a first source operand, an indication of a location of a second source operand, and an indication of a location of a destination operand, wherein the opcode is to indicate that execution circuitry is to at least convert data elements of the first and second source operands from a first floating point representation to a second floating point representation, perform matrix multiplication with the converted data elements, and accumulate results of the matrix multiplication in the destination operand in the first floating point representation, wherein the first floating point representation is single precision floating point and the second floating point representation is a 19-bit floating-point representation; and
the execution circuitry to execute to the decoded instruction as specified by the opcode.
2. The apparatus of claim 1 , wherein the execution circuitry is further to execute the decoded instruction to convert the data elements from the first and second source operands from a signaling not-a-number value to a quiet not-a-number value.
3. The apparatus of claim 1 , wherein the 19-bit floating-point representation has a sign bit, 8 exponent bits, and 10 mantissa bits that are corresponding most significant bits of a single precision floating number.
4. The apparatus of claim 1 , wherein the execution circuitry is further to execute the decoded instruction to zero any remaining data elements of the destination operand.
5. A non-transitory machine readable medium storing at least an instance of a single instruction, wherein a processor is to respond to the single instruction to perform a method comprising:
decoding the single instruction having fields for an opcode, an indication of a location of a first source operand, an indication of a location of a second source operand, and an indication of a location of a destination operand, wherein the opcode is to indicate that execution circuitry is to at least convert data elements of the first and second source operands from a first floating point representation to a second floating point representation, perform matrix multiplication with the converted data elements, and accumulate results of the matrix multiplication in the destination operand in the first floating point representation, wherein the first floating point representation is single precision floating point and the second floating point representation is a 19-bit floating point representation;
executing the decoded instruction to perform operations as specified by the opcode.
6. The non-transitory machine readable medium of claim 5 , wherein the executing further comprises converting the data elements from the first and second source operands from a signaling not-a-number value to a quiet not-a-number value.
7. The non-transitory machine readable medium of claim 5 , wherein the 19-bit floating-point representation has a lower 13 bits set to zero.
8. The non-transitory machine readable medium of claim 5 , wherein the executing further comprises zeroing any remaining data elements of the destination operand.
9. The non-transitory machine readable medium of claim 5 , further comprising:
translating the single instruction into one or more instructions of a different instruction set, wherein the executing the decoded instruction as specified by the opcode comprises executing the one or more instructions of the different instruction set to perform the operations as specified by the opcode of the single instruction.
10. An apparatus comprising:
decode circuitry to decode a single instruction having fields for an opcode, an indication of a location of a first source operand, an indication of a location of a second source operand, and an indication of a location of a destination operand, wherein the opcode is to indicate that execution circuitry is to at least convert data elements of the first and second source operands from a first floating point representation to a second floating point representation, perform matrix multiplication with a transpose of the converted data elements of the first source operand and non-transposed, converted data elements of the second source operand, and accumulate results of the matrix multiplication in the destination operand in the first floating point representation, wherein the first floating point representation is single precision floating point and the second floating point representation is a 19-bit floating point representation; and
the execution circuitry to execute to the decoded instruction as specified by the opcode.
11. The apparatus of claim 10 , wherein the execution circuitry is further to execute the decoded instruction to convert the data elements from the first and second source operands from a signaling not-a-number value to a quiet not-a-number value.
12. The apparatus of claim 10 , wherein the 19-bit floating-point representation has a lower 13 bits set to zero.
13. The apparatus of claim 10 , wherein the execution circuitry is further to execute the decoded instruction to zero any remaining data elements of the destination operand.
14. A non-transitory machine-readable medium storing at least an instance of a single instruction, wherein a processor is to respond to the single instruction to perform a method comprising:
decoding the single instruction having fields for an opcode, an indication of a location of a first source operand, an indication of a location of a second source operand, and an indication of a location of a destination operand, wherein the opcode is to indicate that execution circuitry is to at least convert data elements of the first and second source operands from a first floating point representation to a second floating point representation, perform matrix multiplication with a transpose of the converted data elements of the first source operand and non-transposed, converted data elements of the second source operand, and accumulate results of the matrix multiplication in the destination operand in the first floating point representation, wherein the first floating point representation is single precision floating point and the second floating point representation is a 19-bit floating point representation;
executing the decoded instruction as specified by the opcode.