Method of transferring data between external devices and an array processor
A method for coordinating the transfer of data between external memory and an array of data processors using address generators and local memory includes loading a plurality of groups of operands into local memory, processing the plurality of groups of operands on a single processor, and then returning the processed results to the external memory.
1. A method, comprising:
providing a processing device that includes:
an array of data processors;
a global internal memory for the array of data processors;
a plurality of address generators, wherein a data processor of the array of data processors is adapted to receive operands from the internal memory and to provide outputs to said internal memory according to addresses provided by corresponding ones of said plurality of address generators; and
a function control unit operatively connected to the array of data processors and to the address generators;
the function control unit receiving instructions from an external source;
responsive to a first set of said instructions, causing the internal memory to receive from an external memory at least two different groups of operands;
responsive to a second set of said instructions, causing the at least two different groups of operands to be forwarded to respective ones of said data processors according to the addresses of the corresponding ones of said plurality of address generators, thereby causing results of processing of the at least two groups of operands to be stored in a result data block in said internal memory; and
responsive to a third set of said instructions, causing the result data block to be transmitted from the internal memory to said external memory.
2. The method according to claim 1 further comprising generating said instructions with a compiler.
3. The method according to claim 2 wherein said compiler comprises a high level language compiler.
4. The method according to claim 3 wherein said high level language compiler is C.
5. The method of claim 1 wherein the at least two different groups of operands are received from the external source according to a program loop in said instructions.
6. The method of claim 1 wherein the loading of the at least two groups of operands from external memory, the causing processing of said at least two groups of operands, and the causing of transmission of the result data block to said external memory comprise at least three different groups of one or more instructions.
7. The method of claim 1 wherein the addresses of operands in internal memory are linear.