Compute optimizations for low precision machine learning operations
One embodiment provides a general-purpose graphics processing unit comprising a dynamic precision floating-point unit including a control unit having precision tracking hardware logic to track an available number of bits of precision for computed data relative to a target precision, wherein the dynamic precision floating-point unit includes computational logic to output data at multiple precisions.
1. A general-purpose graphics processing unit comprising:
a dynamic precision floating-point unit including a controller having precision tracking hardware to track an available number of bits of precision for computed data relative to a target precision, wherein the dynamic precision floating-point unit includes functional units to perform variable precision floating-point operations, the functional units including native floating-point hardware to perform the variable precision floating-point operations, wherein hardware of the dynamic precision floating-point unit includes a significand block to perform a significand portion of a floating-point computation and an exponent block to perform an exponent portion of the floating-point computation.
2. The general-purpose graphics processing unit as in claim 1 , wherein the dynamic precision floating-point unit includes a set of registers to store input data and intermediate data at multiple precisions, wherein the set of registers is to store the significand portion of an input data value separately from the exponent portion of the input data value.
3. The general-purpose graphics processing unit as in claim 2 , wherein the set of registers includes an error accumulator to track an accumulated error over a set of floating-point operations.
4. The general-purpose graphics processing unit as in claim 1 , wherein the significand block includes a first dynamic precision adder configurable via the controller to add or subtract input data at multiple precisions.
5. The general-purpose graphics processing unit as in claim 4 , the significand block including a dynamic precision multiplier configurable via the controller to add or multiply or divide input data at multiple precisions.
6. The general-purpose graphics processing unit as in claim 5 , wherein the exponent block includes a second dynamic precision adder configurable via the controller to add or subtract exponents of input data at multiple precisions.
7. The general-purpose graphics processing unit as in claim 6 , the exponent block and the significand block to perform a first floating-point operation to output a first output value having 16-bits of precision.
8. The general-purpose graphics processing unit as in claim 7 , the exponent block and the significand block to perform a second operation to output a second output value having 32-bits of precision.
9. The general-purpose graphics processing unit as in claim 8 , the exponent block and the significand block to perform a third floating-point operation on input data having 32-bit values to output a third output value having a 32-bit data type, the third output value generated at 16-bits of precision.
10. The general-purpose graphics processing unit as in claim 9 , the exponent block including an 8-bit multiplier, and wherein the exponent block and the significand block are configurable via the controller to perform a dual 8-bit integer operation.
11. A data processing system comprising:
a general-purpose graphics processing unit comprising a dynamic precision floating-point unit including a controller having precision tracking hardware to track an available number of bits of precision for computed data relative to a target precision, wherein the dynamic precision floating-point unit includes functional units to perform variable precision floating-point operations, the functional units including native floating-point hardware to perform the variable precision floating-point operations, wherein hardware of the dynamic precision floating-point unit includes a significand block to perform a significand portion of a floating-point computation and an exponent block to perform an exponent portion of the floating-point computation; and
a memory coupled to the general-purpose graphics processing unit.
12. The data processing system as in claim 11 , wherein the dynamic precision floating-point unit includes a set of registers to store input data and intermediate data at multiple precisions, wherein the set of registers is to store the significand portion of an input data value separately from the exponent portion of the input data value.
13. The data processing system as in claim 12 , wherein the set of registers includes an error accumulator to track an accumulated error over a set of floating-point operations.
14. The data processing system as in claim 11 , wherein the significand block includes a first dynamic precision adder configurable via the controller to add or subtract input data at multiple precisions and a dynamic precision multiplier configurable via the controller to add or multiply or divide input data at multiple precisions.
15. The data processing system as in claim 14 , wherein the exponent block includes a second dynamic precision adder configurable via the controller to add or subtract exponents of input data at multiple precisions.
16. An electronic device comprising:
a general-purpose graphics processing unit comprising a dynamic precision floating-point unit, the dynamic precision floating-point unit including a controller having precision tracking hardware to track an available number of bits of precision for computed data relative to a target precision; and
wherein the dynamic precision floating-point unit includes functional units to perform variable precision floating-point operations, the functional units including native floating-point hardware to perform the variable precision floating-point operations, wherein hardware of the dynamic precision floating-point unit includes a significand block to perform a significand portion of a floating-point computation and an exponent block to perform an exponent portion of the floating-point computation.
17. The electronic device as in claim 16 , wherein the dynamic precision floating-point unit includes a set of registers to store input data and intermediate data at multiple precisions, wherein the set of registers is to store the significand portion of an input data value separately from the exponent portion of the input data value and includes an error accumulator to track an accumulated error over a set of floating-point operations.
18. The electronic device as in claim 16 , wherein:
the significand block includes a first dynamic precision adder configurable via the controller to add or subtract input data at multiple precisions and a dynamic precision multiplier configurable via the controller to add, multiply, or divide input data at multiple precisions; and
the exponent block includes a second dynamic precision adder configurable via the controller to add or subtract exponents of input data at multiple precisions.
19. The electronic device as in claim 18 , wherein the exponent block and the significand block are to perform a first floating-point operation to output a first output value having 16-bits of precision, a second operation to output a second output value having 32-bits of precision, and a third floating-point operation on input data having 32-bit values to output a third output value having a 32-bit data type, the third output value generated at 16-bits of precision.
20. The electronic device as in claim 18 , the exponent block including an 8-bit multiplier, and wherein the exponent block and the significand block are configurable via the controller to perform a dual 8-bit integer operation.