US 12,405,827 B2
Cognitive allocation of specialized hardware resources
Clement Decrop, Arlington, VA (US); Abhishek Malvankar, White Plains, NY (US); John M. Ganci, Jr., Raleigh, NC (US); and Thomas Jefferson Sandridge, Tampa, FL (US)
Assigned to INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Jan. 7, 2022, as Appl. No. 17/570,659.
Prior Publication US 2023/0221992 A1, Jul. 13, 2023
Int. Cl. G06F 9/50 (2006.01); G06F 9/48 (2006.01); G06F 16/242 (2019.01); G06F 16/2455 (2019.01); G06N 3/042 (2023.01); G06N 3/045 (2023.01)
CPC G06F 9/5027 (2013.01) [G06F 9/4881 (2013.01); G06F 16/242 (2019.01); G06F 16/2455 (2019.01); G06N 3/042 (2023.01); G06N 3/045 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A computer implemented method comprising:
receiving source code input by a user to a user interface of an integrated development environment (IDE);
sending a tokenized segment of the source code to at least one of a code analysis module and a runtime machine learning module;
sending a search query comprising a hardware requirement associated with the tokenized code segment to a knowledge base, wherein the search query is generated based on an analysis of the tokenized code segment by a first trained neural network;
obtaining a result from the search query indicating a prediction of a specialized hardware resource associated with the hardware requirement;
issuing an instruction associated with the tokenized code segment to a hardware scheduler, wherein the instruction causes the hardware scheduler to allocate the predicted specialized hardware resource to the tokenized code segment during a predicted time allotment; and
executing the tokenized code segment using the allocated specialized hardware resource during the predicted time allotment, wherein the predicted time allotment is generated using a second trained neural network and is based on the tokenized code segment.