Dynamic distribution for distributed arrays and related rules

1. A non-transitory computer-readable medium to store instructions, the instructions comprising:

one or more instructions that, when executed by at least one processor, cause the at least one processor to:

receive a call to a function in a function library:

the function library including a plurality of functions,

the function including a characteristic, and

the function operating on a distributable array;

determine a distribution scheme for the distributable array,

the one or more instructions to determine the distribution scheme including:

one or more instructions to evaluate the distributable array, and

one or more instructions to evaluate the characteristic of the function; and

partition the distributable array into parts based on the determined distribution scheme.

2. The computer-readable medium of claim 1 , where the distributable array is dynamically partitioned into a plurality of parts.

3. The computer-readable medium of claim 2 , where the plurality of parts are provided to a plurality of processing elements for processing.

4. The computer-readable medium of claim 3 , where the plurality of parts are processed simultaneously.

5. The computer-readable medium of claim 3 , where the plurality of processing elements are arranged in a hierarchical relationship.

6. The computer-readable medium of claim 3 , where the plurality of processing elements reside on a processor.

7. The computer-readable medium of claim 3 , where the plurality of processing elements are distributed amongst two or more processors.

8. The computer-readable medium of claim 3 , where the plurality of processing elements are distributed amongst two or more computers.

9. The computer-readable medium of claim 2 , where at least two of the plurality of parts are unequal in size.

10. The computer-readable medium of claim 1 , further comprising:

one or more instructions that, when executed by the at least one processor, cause the at least one processor to:

receive at least a portion of the distributable array from a local memory location.

11. The computer-readable medium of claim 1 , further comprising:

one or more instructions that, when executed by the at least one processor, cause the at least one processor to:

receive the distributable array from a global memory location; and

store the distributable array in a local memory location.

12. The computer-readable medium of claim 1 , where the one or more instructions to determine the distribution scheme include:

one or more instructions that, when executed by the at least one processor, cause the at least one processor to:

account for a location of the distributable array; and

use information associated with the location of the distributable array to determine the distribution scheme.

13. The computer-readable medium of claim 1 , where the one or more instructions to evaluate the distributable array include:

one or more instructions that, when executed by the at least one processor, cause the at least one processor to:

account for a size of the distributable array; and

use information associated with the size of the distributable array to evaluate the distributable array.

14. The computer-readable medium of claim 1 , where the characteristic of the function is a category of the function.

15. The computer-readable medium of claim 1 , where the one or more instructions to evaluate the distributable array include:

one or more instructions that, when executed by the at least one processor, cause the at least one processor to:

evaluate a shape of the distributable array; and

use information associated with the shape of the distributable array to evaluate the distributable array.

16. The computer-readable medium of claim 1 , where the distribution scheme is a default scheme.

17. The computer-readable medium of claim 1 , where the distribution scheme is an optimized distribution scheme.

18. The computer-readable medium of claim 1 , where the distribution scheme increases memory throughput.

19. The computer-readable medium of claim 1 , where the one or more instructions to determine the distribution scheme include:

one or more instructions that, when executed by the at least one processor, cause the at least one processor to:

receive a criterion from a user; and

determine the distribution scheme based on the received criterion.

20. A device comprising:

a memory to store instructions; and

a processor to execute the instructions to:

receive a call to a function in a function library:

the function library including a plurality of functions,

the function including a characteristic, and

the function operating on a distributable array;

determine a distribution scheme for the distributable array,

the processor, when determining the distribution scheme, being to:

evaluate the distributable array, and

evaluate the characteristic of the function; and

partition, based on the determined distribution scheme, the distributable array into parts.

21. A method comprising:

receiving an input portion,

the input portion being part of a distributable input array,

the input portion being received from a storage location, and

the receiving the input portion being performed by a computing device;

receiving a function,

the function being retrieved from a function library,

the function including a characteristic, and

the receiving the function being performed by the computing device;

evaluating the characteristic of the function,

the evaluating being performed by the computing device;

processing, based on the evaluated characteristic of the function, the input portion to produce an output portion,

the processing being performed by the computing device; and

sending the output portion to the storage location,

the output portion being part of a distributable output array, and

the sending being performed by the computing device.

22. The method of claim 21 , further comprising:

providing the input portion from a memory location; and

receiving the output portion at the memory location.

23. The method of claim 22 , where the input portion is in first memory location of a plurality of memory locations.

24. The method of claim 21 , where the output portion is further produced based on:

reallocating at least a portion of the input portion prior to the processing.

25. The method of claim 21 , where the input portion is provided from a local memory location.

26. The method of claim 25 , comprising:

receiving the input portion from a global memory location; and

storing the input portion in the local memory location.

27. The method of claim 21 , where

the input portion includes a plurality of input portions, and

the processing the input portion comprises:

simultaneously processing the plurality of input portions.

28. A non-transitory computer-readable medium to store instructions, the instructions comprising:

one or more instructions that, when executed by at least one processor, cause the at least one processor to:

receive an input portion,

the input portion being part of a distributable input array, and

the input portion being received from a storage location;

receive a function,

the function being retrieved from a function library, and

the function including a characteristic;

evaluate the characteristic of the function;

process, based on the evaluated characteristic of the function, the input portion to produce an output portion; and

send the output portion to the storage location,

the output portion being part of a distributable output array.