Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Detail Action
In the amendment filed on 03/17/2026, Claims 1-20 are pending and claims 1-4, 8-10, 13-16, and 20 are amended.
This is a Final Action.
Response to Arguments
Applicant’s arguments with respect to claims 1-8 and 13-20 have been considered but are moot because the new ground of rejection.
Claim 9-12:
Regarding claim 9, applicant argued that Sur does not teaches a break point between layers.
Examiner disagrees. Sur teaches this limitation because Sur teaches check whether to pause the operation when a layer of Li-1 is done. (see Sur p0042) The check for pause serves a breaking point between Layers of Li. Therefore, Sur teaches a break point between layers.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 6, 7, 13, and 18-19 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946.
18/364,742
Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946.
Claim 1
A method performed by deep learning accelerator (DLA) hardware for task preemption, comprising:
Rosemarine p0018-p0028;
executing a first task by using a neural network of multiple layers on a given input, wherein the neural network is described by a directed acyclic graph (DAG);
Rosemarine p0019;
Sur teaches schedule processing of tensors layer by based on a DAG. See p0017;
It would have been obvious at the time of the invention for a person ordinary skill in the art (POSITA) to include Sur method with method of Rosemarine such that time is not wasted when image data is available for processing.
Receiving a stop command from a DLA driver indicating a second task waiting for execution by the DLA, the second task having a higher priority than the first task;
Rosemarine teaches a job stop request from by the TCI driver; see p0068;
in response to the stop command, completing execution of at least a current sublayer of the neural network before sending an interrupt request (IRQ) to the DLA driver, wherein the current sublayer is represented by a node in the DAG;
Rosemarine p0031-p0035;
Rosemarine p0068-p0120;
Sur teaches pause between network of Layer; Fig. 10-Fig. 12, p0042-p0051; and Sur teaches a send operation, which includes pause command, may not be issued until all the received in other operations have executed; p0042;
It would have been obvious at the time of the invention for a person ordinary skill in the art (POSITA) to include Sur’s teaching with method of Rosemarine in order to optimize the schedule to prioritize upper layer.
receiving a second task from the DLA driver; and executing the second task to completion before resuming the execution of the first task.
Rosemarine p0031-p0035;
Rosemarine p0068-p0120;
Claim 6
The method of claim 1, further comprising: receiving a restored context of the first task from the DLA driver; and
resuming the execution of the first task using the restored context.
Rosemarine p0031-p0035;
Claim 7
The method of claim 1, further comprising: saving, by the DLA hardware, states of the first task during the execution of the first task; and retrieving the saved states of the first task to resume the execution of the first task.
Rosemarine p0031-p0035;
As per claim 13, it is rejected under the same rationale as claim 1. See rejection above.
As per claims 18-19, they are rejected under the same rationale as claims 6-7. See rejections above.
Claims 2, 4, 14 and 16 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946 and Kwong’783 US2022/0207783.
18/364,742
Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946 and Kwong’783 US2022/0207783
Claim 2
The method of claim 1, wherein in response to the stop command, the method further comprises completing a current layer of the neural network before sending the IRQ to the DLA driver, and wherein the current layer is represented by a subgraph in the DAG.
Rosemarine p0032-p0057;
Rosemarine teaches layer of network; p0104-p0105
Kwong’783 teaches prioritize different layer of neural network layers p0055-p0057;
It would have been obvious at the time of the invention for a person ordinary skill in the art (POSITA) to include Kwong’783 method with method of Rosemarine such that time is not wasted when image data is available for processing.
Claim 4
The method of claim 1, wherein before resuming the execution of the first task, the method further comprises:
Completing the second task by executing all of the multiple layers of the neural network with respect to the second task.
Rosemarine p0032-p0057;
Rosemarine teaches layer of network; p0104-p0105
Kwong’783 teaches prioritize different layer of neural network layers p0055-p0056;
It would have been obvious at the time of the invention for a person ordinary skill in the art (POSITA) to include Kwong’783 method with method of Rosemarine such that time is not wasted when image data is available for processing.
As per claims 14 and 16, they are rejected under the same rationale as claims 2 and 4. See rejection above.
Claims 5 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946 and Di Gregorio US Publication 2006/0161924.
18/364,742
Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946 and Di Gregorio US Publication 2006/0161924;
Claim 5
The method of claim 1, further comprising: detecting, by the DLA hardware, a predetermined register value that indicates the stop command issued by the DLA driver.
Rosemarine p0032-p0057;
Di Gregorio teaches register value assigned to command; (see Di Gregorio p0091-p0122)
It would have been obvious at the time of the invention for a person ordinary skill in the art (POSITA) to include Di Gregorio’s teaching with method of Rosemarine in order to use register to implement scheduler command.
As per claims 17, it is rejected under the same rationale as claim 5. See rejection above.
Claims 8 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946 and Kwong’596 US Publication 2024/0111596.
18/364,742
Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946 and Kwong’596 US Publication 2024/0111596;
Claim 8
The method of claim 1, wherein the second task has a higher frame-per-second (FPS) requirement than the first task.
Rosemarine p0032-p0057;
Kwong’596 teaches examining FPS in priority determination; (see Kwong’596 p0093-p0133)
It would have been obvious at the time of the invention for a person ordinary skill in the art (POSITA) to include Kwong’596 ’s teaching with method of Rosemarine in order to include FPS metric in the module.
As per claim 20, it is rejected under the same rationale as claim 20. See rejection above.
Claims 3 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946 and Lee US Publication 2023/0143270.
18/364,742
Rosemarine US Publication 2024/0256332 in view Kwong Sur US Publication 2019/0042946 and Lee US Publication 2023/0143270.
Claim 3
The method of claim 1, wherein before resuming the execution of the first task, the method further comprises:
Completing the second task be executing all layer of a second neural network that is different the neural network.
Rosemarine p0054-p0073;
Rosemarine teaches layer of network; p0104-p0105
Lee teaches prioritizes a second neural network Fig .5 p0115;
It would have been obvious at the time of the invention for a person ordinary skill in the art (POSITA) to include Lee method with method of Rosemarine to operate multiple neural networks.
As per claim 15, it is rejected under the same rationale as claim 3. See rejection above.
Claims 9, 11-12 are rejected under 35 U.S.C. 103 as being unpatentable over Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946.
18/364,742
Rosemarine US Publication 2024/0256332 in view of Sur US Publication 2019/0042946
Claim 9
A method performed by deep learning accelerator (DLA) hardware for task preemption, comprising:
Rosemarine p0018-p0028;
executing a first task by using a neural network of multiple layers on a given input, wherein the first task has been modified by a DLA driver to include a breakpoint at an end of each layer of the neural network;
Rosemarine p0019;
Sur teaches pause between network of Layer; Fig. 10-Fig. 12, p0042-p0051;
It would have been obvious at the time of the invention for a person ordinary skill in the art (POSITA) to include Sur’s teaching with method of Rosemarine in order to optimize the schedule to prioritize upper layer.
sending an interrupt request (IRQ) to the DLA driver when execution of the first task reaches the breakpoint of a given layer of the neural network;
waiting for an instruction from the DLA driver after sending the IRQ; and
when receiving from the DLA driver the instruction indicating a second task waiting in a queue,
executing the second task to completion before resuming execution of the first task, wherein the second task has a higher priority than the first task.
Rosemarine p0031-p0035;
Claim 11
The method of claim 9, wherein the DLA driver backs up the first task before modifying the first task, and restores the first task after the DLA hardware completes the execution of the first task.
Rosemarine p0031-p0035;
Rosemarine p0068-p0120;
Claim 12
The method of claim 9, wherein an interrupt bit is inserted at the end of each layer of the neural network to indicate the breakpoint.
Rosemarine p0031-p0035;
Sur teaches pause between network of Layer; Fig. 10-Fig. 12, p0042-p0051;
Claims 9, 11-12 are rejected under 35 U.S.C. 103 as being unpatentable over Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946 and Lee US Publication 2023/0143270.
18/364,742
Rosemarine US Publication 2024/0256332 in view Sur US Publication 2019/0042946 and Lee US Publication 2023/0143270.
Claim 10
The method of claim 9, further comprising: completing the second task by executing all layers of a second neural network that is different from the neural network, or by executing all of the multiple layers of the neural network with respect to the second.
Rosemarine p0031-p0035;
Sur teaches pause between network of Layer; Fig. 10-Fig. 12, p0042-p0051;
Lee teaches prioritizes a second neural network Fig .5 p0115;
It would have been obvious at the time of the invention for a person ordinary skill in the art (POSITA) to include Lee method with method of Rosemarine to operate multiple neural networks.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PENG KE whose telephone number is (571)272-4062. The examiner can normally be reached M-F 6:30-5:00.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kevin Young can be reached at (571) 270-3180. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
PENG KE
Primary Examiner
Art Unit 2194
/PENG KE/Primary Examiner, Art Unit 2194
Prosecution Insights