Software systems and methods for advanced output-affecting linear pathways
View Patent ↗Unlike conventional or otherwise known decomposition methods like standard functional decomposition, time-affecting linear pathway (TALP) decomposition, or output-affecting linear pathways (OALPs) from the decomposition of TALPs, advanced output-affecting linear pathway (A-OALP) decomposition from algorithms separate the output variables of each execution pathway such that there is only one output variable per A-OALP, converting an algorithm into a set of process groups. A-OALPs extend parallelization by combining task-like parallelism with dynamic loop parallelism. The lightweight nature of A-OALPs allows for persistent thread and code management.
1. A method for optimization of software or one or more algorithms, comprising:
receiving the software or the one or more algorithms, and test data, from an operator;
decomposing the software or the one or more algorithms into one or more executable and analyzable advanced output-affecting linear pathways (A-OALPs), wherein each of the one or more A-OALPs includes runtime information generation for one or more instances;
executing an A-OALP hardware controller to:
receive one or more input variable attribute values;
compare the one or more input variable attribute values that correspond to a pathway selection entry in an A-OALP selection table, wherein the A-OALP selection table uses the one or more input variable attribute values and certain value ranges;
select at least one A-OALP that corresponds to the pathway selection;
identify the selected at least one A-OALP for further execution;
determine which of the one or more input variable attribute values vary a processing time of the selected at least one A-OALP;
determine which of the one or more input variable attribute values vary memory allocation of the selected at least one A-OALP; and
determine a maximum number of parallel cores usable by the selected at least one A-OALP for a current one or more input variable attribute values;
executing in parallel multiple A-OALPs on multiple separate servers while concurrently executing in parallel multiple instances of each of the multiple A-OALPs on multiple separate cores of a server; and
limiting A-OALP cross-communication to multiple cores on the server and eliminating a need for A-OALP cross-communication between the multiple separate servers.
2. The method of claim 1 , further comprising determining a maximum number of parallel cores usable by the selected at least one A-OALP for a currently available number of cores or a user-indicated maximum number of cores.
3. The method of claim 1 , further comprising optimizing energy consumption or processing performance for executing the software or one or more algorithms using the selected at least one A-OALP.
4. The method of claim 3 , wherein at least one of cost, water use, and carbon footprint data is converted into either the energy consumption or the processing performance.
5. The method of claim 1 , further comprising generating a temporal profile for each of the one or more A-OALPs using the one or more input variable attribute values that vary processing time.
6. The method of claim 5 , further comprising generating an advanced time prediction curve fit for each A-OALP of the one or more A-OALPS using the temporal profile.
7. The method of claim 5 , further comprising generating an advanced speedup prediction curve fit for each A-OALP of the one or more A-OALPs using a scaled version of the temporal profile.
8. The method of claim 1 , further comprising generating a spatial profile for each of the one or more A-OALPs, using one or more variable attribute values that affect the memory allocation.
9. The method of claim 8 , further comprising generating an advanced space complexity curve fit for each A-OALP of the one or more A-OALPs using the spatial profile.
10. The method of claim 8 , further comprising generating an advanced freeup prediction curve fit for each A-OALP of the one or more A-OALPs using a scaled version of the spatial profile.
11. The method of claim 1 , further comprising automatically constructing and using one or more persistent thread pools with persistent A-OALPs to remove thread construction, thread destruction, thread pool construction, or thread pool destruction overhead.
12. The method of claim 11 , further comprising linking one or more particular physical processing cores to one or more particular processing threads.
13. The method of claim 1 , further comprising performing super-linear processing from A-OALP-based dynamic all-loop unrolling parallelization.
14. The method of claim 1 , further comprising performing dynamic loop unrolling on at least one of the one or more A-OALPs to determine multiple instances of the at least one of the one or more A-OALPs.
15. The method of claim 1 , further comprising increasing code performance by removing extraneous code for task-like parallelism.
16. The method of claim 1 , further comprising facilitating cross communication between multiple addressable hardware socket components.
17. The method of claim 16 , wherein each of the multiple addressable hardware socket components comprise multiple processing cores.
18. A system for optimization of software or one or more algorithms, comprising:
a memory; and
a processor operatively coupled with the memory, wherein the processor is configured to execute a program code to:
receive the software or the one or more algorithms, and test data, from an operator;
decompose the software or the one or more algorithms into one or more executable and analyzable advanced output-affecting linear pathways (A-OALPs), wherein each of the one or more A-OALPs includes runtime information generation for one or more instances;
execute an A-OALP hardware controller to:
receive one or more input variable attribute values;
compare the one or more input variable attribute values that correspond to a pathway selection entry in an A-OALP selection table, wherein the A-OALP selection table uses the one or more input variable attribute values and certain value ranges;
select at least one A-OALP that corresponds to the pathway selection;
identify the selected at least one A-OALP for further execution;
determine which of the one or more input variable attribute values vary a processing time of the selected at least one A-OALP;
determine which of the one or more input variable attribute values vary memory allocation of the selected at least one A-OALP; and
determine a maximum number of parallel cores usable by the selected at least one A-OALP for a current one or more input variable attribute values;
executing in parallel multiple A-OALPs on multiple separate servers while concurrently executing in parallel multiple instances of each of the multiple A-OALPs on multiple separate cores of a server; and
limiting A-OALP cross-communication to multiple cores on the server and eliminating a need for A-OALP cross-communication between the multiple separate servers.