Cascaded accelerator functions
View Patent ↗Accelerator functions are cascaded, such that a result of one accelerator function is directly forwarded to another accelerator function, bypassing the processor requesting the functions to be performed. The cascading may be provided during compilation of a program specifying the functions to be performed, but can be dynamically reversed during runtime of the program.
1. A computer program product for facilitating processing within a computing environment, the computer program product comprising:
a non-transitory storage medium readable by a processing circuit and storing instructions for execution by the processing circuit for performing a method comprising:
determining whether a result of an accelerator function is to be forwarded to another accelerator function, the accelerator function encountered in a program and to be processed by an accelerator; and
based on the determining indicating the result is to be forwarded to the another accelerator function, generating machine code to provide the result directly to the another accelerator function, bypassing a processor to execute the program.
2. The computer program product of claim 1 , wherein the method further comprises:
deciding, during runtime of the program, that the processor is not to be bypassed; and
receiving by the processor, based on the deciding, the result, wherein the result is provided to the processor rather than directly to the another accelerator function.
3. The computer program product of claim 2 , wherein the method further comprises forwarding the result from the processor to the another accelerator function.
4. The computer program product of claim 2 , wherein the deciding includes determining that the result is to be subject to new processing.
5. The computer program product of claim 2 , wherein the deciding comprises having an indication that a link coupled the accelerator function and the another accelerator function is unavailable.
6. The computer program product of claim 5 , wherein the link is unavailable due to one of link failure, congestion on the link, lack of timeliness on the link, or unacceptable performance on the link.
7. The computer program product of claim 1 , wherein the another accelerator function is on another accelerator, the another accelerator being physically distinct from and coupled to the accelerator.
8. The computer program product of claim 1 , wherein the determining includes deciding that a compiler to compile the program can meet a specified performance metric based on input to the compiler.
9. The computer program product of claim 1 , wherein the method further comprises:
determining whether a computation module is used in processing the result; and
including the computation module on an accelerator to execute the another accelerator function, said computation module to be used in processing the result.
10. A computer system for facilitating processing within a computing environment, said computer system comprising:
a memory; and
a processor in communications with the memory, wherein the computer system is capable of performing a method, said method comprising:
determining whether a result of an accelerator function is to be forwarded to another accelerator function, the another accelerator function encountered in a program and to be processed by an accelerator; and
based on the determining indicating the result is to be forwarded to the another accelerator function, generating machine code to provide the result directly to the another accelerator function, bypassing a processor to execute the program.
11. The computer system of claim 10 , wherein the method further comprises:
deciding, during runtime of the program, that the processor is not to be bypassed; and
receiving by the processor, based on the deciding, the result, wherein the result is provided to the processor rather than directly to the another accelerator function.
12. The computer system of claim 11 , wherein the method further comprises forwarding the result from the processor to the another accelerator function.
13. The computer system of claim 10 , wherein the another accelerator function is on another accelerator, the another accelerator being physically distinct from and coupled to the accelerator.
14. The computer system of claim 10 , wherein the determining includes deciding that a compiler to compile the program can meet a specified performance metric based on input to the compiler.
15. The computer system of claim 10 , wherein the method further comprises:
determining whether a computation module is used in processing the result; and
including the computation module on an accelerator to execute the another accelerator function, said computation module to be used in processing the result.
16. A method of facilitating processing within a computing environment, said method comprising:
determining, by a first processor, whether a result of an accelerator function is to be forwarded to another accelerator function, the accelerator function encountered in a program and to be processed by an accelerator; and
based on the determining indicating the result is to be forwarded to the another accelerator function, generating machine code to provide the result directly to the another accelerator function, bypassing a second processor to execute the program.
17. The method of claim 16 , wherein the first processor and the second processor are the same processor.
18. The method of claim 17 , wherein the first processor and the second processor are different processors.
19. The method of claim 16 , further comprising:
deciding, during runtime of the program, that the second processor is not to be bypassed; and
receiving by the second processor, based on the deciding, the result, wherein the result is provided to the second processor rather than directly to the another accelerator function.
20. The method of claim 16 , further comprising:
determining whether a computation module is used in processing the result; and
including the computation module on an accelerator to execute the another accelerator function, said computation module to be used in processing the result.