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 .
DETAILED ACTION
This action is in response to the claimed listing amended in the preliminary amendment filed on 09/25/2024.
Claims 1-20 are pending.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 14-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
As per claims 14-17: Independent claim 14 in the method of claim 14-17 recite:
“constructing a data set, wherein the data set comprise at least one data entry, each of the data entry comprises: a third source code and a corresponding second compile optimization option”.
Referred to the limitation “a third source code”, it has no first and second source code established as an antecedent basis in the claim, and referred to the limitation “second compile optimization option”, there is no first corresponding compile optimization option recited in the claim.
It has the limitations unclear and logical error because of missing a gap between the limitations.
Therefore, the claims fail to particularly point out and distinctly claim the subject matter required under USC 112(b).
As per claims 18-20: Independent claim 14 in the method of claim 14-17 recite:
“constructing a data set, wherein the data set comprises at least one data entry, each of the data entry comprises: feature information corresponding to a third source code and a corresponding second compile optimization option;”.
Referred to the limitation “a third source code”, it has no first and second source code established as an antecedent basis in the claim, and referred to the limitation “a corresponding second compile optimization option”, there is no first corresponding compile optimization option recited in the claim.
It has the limitations unclear and logical error because of missing a gap between the limitations.
Therefore, the claims fail to particularly point out and distinctly claim the subject matter required under USC 112(b).
The limitations above in claims 14-20 would be interpreted as “source code” and “corresponding compile optimization option”, respectively.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-13 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
As per claim 1: It recites “A method”: it falls within one of the statutory categories of invention.
and regarding in bold below:
for determining a compile optimization option, comprising:
splitting a first source code into at least one second source code according to a preset splitting level ; obtaining feature information corresponding to each second source code;
(The limitations read both on a performance in human mind with “determining” and used with pen and paper - “splitting” could be implemented on paper ) ; and
inputting the feature information corresponding to each second source code into [a pre-trained prediction model to obtain a first compile optimization option] corresponding to each second source code. (The limitations encompass the performance with pen and paper)
[Step 2A, prong 1] With broadest reasonable interpretation, the limitations of the claim as a whole in bold above, encompass an observation along with judgments as it is set forth in the preamble “for determining a compile optimization option”. Within the steps, it covers a mental process that clearly can be performed in human mind and on paper with pen.
The courts consider a mental process (thinking) that “can be performed in the human mind, or by a human using a pen and paper” to be an abstract idea. The limitations of the claim recited in bold above meet all aspects considered by the court.
Thus, the method as a whole covers the performance in the mind and could be added with the pen and paper over a source code of a program. It has the method directed to a mental process, a judicial exception.
[Step 2A, prong 2]. The claim recites an additional element:
“a pre-trained prediction model to obtain a first compile optimization option” is merely applying generic computer components and generic of machine learning concepts for performing generic computer functions at a high level of generality. The additional element is considered not to be integrated into practical application (MPEP 2106.05(d)(f)).
[Step 2B] The additional element is merely applied using generic computing components to perform generic computing functions; it does not amount to significantly more (MPEP 2106.05(f)).
Thus, the claim is directed to a judicial exception without providing an inventive concept or significantly more; it is not eligible under 35 USC 101.
As per Claims 2-9, 12-13: The claims are depending on claim 1, they are within a mental step “determining a compile optimization option” as set forth in claim 1; the limitations of the claims encompass the performance in human mind and with pen and paper. They fail to cure the deficiencies of the above ineligible under 35 USC 101 of claim 1.
As per Claims 10-11: The claims recite a device and non-transitory computer-readable medium and perform the steps recited in claim . They fall within the statutory categories of invention; but they embed the abstract idea as addressed in claim 1. The claims are not eligible under 35 USC 101.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 14, 16, 18, 20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by
Yu et al., "Order Matters: Semantic-Aware Neural Networks for Binary Code Similarity Detection", 2020, Association for the Advancement of Artificial Intelligence, www.aaai.org, pages 1145-1152.
As per Claim 14: Regarding limitations in bold below:
(Note: the limitation “third source code” is first introduced: It will be interpreted as “source code”, and “second compile optimization option” will be as “compile optimization option”,
Yu discloses:
14. (New) A training method, comprising:
constructing a data set, wherein the data set comprises at least one data entry, each of the data entry comprises (Yu, p. 1149, Table 1, and in Datasets): a third source code and a corresponding second compile optimization option (Yu, p. 1147-1148, in Semantic-aware Modeling “First, we can extract block embedding from different CFGs which generated from different platforms, different architectures, different compile optimization options based on the same model.”,
p. 1149, in Datasets: “compile on gcc with O2 and O3 optimization options”);
obtaining feature information corresponding to the third source code in each data entry (Yu, p. 1147, Figure 3, blocks and matrix read as data entry); and
performing model training according to the feature information corresponding to the third source code (Figure 3 and BERT in p. 1146-1147, and Dataset in Table 1 read on model training; blocks and matrix, tasks read as source code) and the corresponding second compile optimization option to obtain the prediction model, wherein in the model training process, the feature information corresponding to the third source code is used as an input of the prediction model, and the second compile optimization option is used as an output of the prediction model (Figure 3, the output from Concat & MLP reads on ‘output of the prediction model’, and in BERT, p. 1146-1147).
As per Claim 16: Regarding limitations in bold below:
16. (New) The training method according to claim 14, wherein the data entry further comprises: target runtime environment information corresponding to the third source code (e.g. table 1 in p. 1149), and performing model training according to the feature information corresponding to the third source code and the corresponding second compile optimization option to obtain the prediction model comprises (BERTs or CNN) :
performing model training according to the feature information corresponding to the third source code and the corresponding second compile optimization option, as well as the target runtime environment information corresponding to the third source code, to obtain the prediction model (See Figure 3) , wherein
in the model training process, the feature information corresponding to the third
source code and the target runtime environment information corresponding to the
third source code are used as inputs of the prediction model, and the second
compile optimization option is used as an output of the prediction model (See Figure 3) .
As per Claims 18, 20: Claims 18 and 20 recite the limitations corresponding to the limitations and functionality in claims 14 and 16 above. The Claims would be addressed with the same rationales as in Claims 14 and 16.
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 1-13 are rejected under 35 U.S.C. 103 as being unpatentable over
Yu et al., "Order Matters: Semantic-Aware Neural Networks for Binary Code Similarity Detection", 2020, Association for the Advancement of Artificial Intelligence, www.aaai.org, pages 1145-1152 in view of Lin et al., "Improving Code Summarization with Block-wise Abstract Syntax Tree Splitting", 2021, arXiv, 12 pages.
As per Claim 1: Yu discloses the limitations in bold below:
1. A method for determining a compile optimization option (Yu, p.1149, in Datasets “…and compile on gcc with O2 and O3 optimization options. The second task is graph classification, in this task we classify the optimization option of the graph embeddings”), comprising:
splitting a first source code into at least one second source code according to a preset splitting level;
Per above limitations, Yu, p.1151, in Model variants for order-aware modeling, in the right column “…On different platforms, the code is compiled into different CFGs: node 3 in the left graph is split into node 3 and 4 in the right graph.”, as a whole, the source code as the whole program that is chunked as blocks/nodes, in graph-level tasks
but does not explicitly address italics limitations above, especially splitting.
Lin discloses “splitting a first source code into at least one second source code according to a preset splitting level;” (Lin, p. 3, Fig. 2, and Abstract: “BASTS splits the code of a method based on the blocks in the dominator tree of the Control Flow Graph, and generates a split AST for
each code split. Each split AST is then modeled by a Tree-LSTM using a pre-training strategy to capture local non-linear syntax encoding.”. Thus it shows a source code is split in small chunk of nodes according to dominator tree
Therefore, it would be obvious to an ordinary in the art before the effective filing of the application to incorporate the teaching of Yu for determining a compile optimization option using chunks/nodes generated from a program according to graph-level tasks with the teaching of Lin for spiling source code to provide tree-form embedded as inputs into pre-training model.
The combination would yield predictable results because splitting code is a standard and requirement process in determination of an optimization, and always used by users for simplify the analysis, and processed by any compilers, large language model.
obtaining feature information corresponding to each second source code; and
(Yu, p. 1146, in Our contribution are as follow: “1) We propose a general framework to learn graph embeddings of CFGs, which could learn semantic information, structural information, and order information.” )
inputting the feature information corresponding to each second source code into a pre-trained prediction model to obtain a first compile optimization option corresponding to each second source code.
(Yu, p. 1146, in Our contribution are as follow: “ 2) In semantic-aware modeling, we adopt BERT to pre-train token embeddings and block embeddings with masked language model task (MLM) and adjacent node prediction task (ANP). Additionally we add two graph-level tasks to learn block representation better, which are block inside graph task (BIG) and graph classification task (GC). .”, p. 1146, Figure 3)
As per Claim 2: Yu and combining Lin, where Yu further discloses:
2. The method for determining a compile optimization option according to claim 1, further comprising:
performing code optimization on each second source code according to the first compile optimization option corresponding to the second source code (p. 1146, in top left column, “The other is to distinguish which platform/optimization the block belongs to, which is named graph classification task (GC). We find that the additional tasks could help extract more semantic information and learn block representation better. After pre-training the block embeddings, we fine-tune them on graph-level tasks.”; and Figure 3).
As per Claim 3: Yu and combining Lin, where Yu further discloses:
3. The method for determining a compile optimization option according to claim 1, wherein before splitting the first source code into at least one second source code according to the preset splitting level, the method further comprises:
constructing a data set, wherein the data set comprises at least one data entry, each of the data entry comprises: a third source code and a corresponding second compile optimization option (Yu, p. 1147-1148, in Semantic-aware Modeling “First, we can extract block embedding from different CFGs which generated from different platforms, different architectures, different compile optimization options based on the same model.”, p. 1149, in Datasets: “compile on gcc with O2 and O3 optimization options”);
obtaining feature information corresponding to the third source code in each data entry (Yu, p. 1147, Figure 3, blocks and matrix read as data entry); and
performing model training according to the feature information corresponding to the third source code (Yu, Figure 3 and BERT in p. 1146-1147, and Dataset in Table 1 read on model training; blocks and matrix, tasks read as source code) and the corresponding second compile optimization option to obtain the prediction model, wherein in the model training process, the feature information corresponding to the third source code is used as an input of the prediction model, and the second compile optimization option is used as an output of the prediction model (Yu, Figure 3, the output from Concat & MLP reads on ‘output of the prediction model’, and in BERT p. 1146-1147).
As per Claim 4: Yu and combining Lin, where
4. The method for determining a compile optimization option according to claim 1, wherein before splitting the first source code into at least one second source code according to the preset splitting level, the method further comprises:
Yu discloses determining a compile optimization option, not explicitly address:
“before splitting the first source code into at least one second source code” (Lin, p. 3, Fig. 2, and Abstract: “BASTS splits the code of a method based on the blocks in the dominator tree of the Control Flow Graph, and generates a split AST for each code split. Each split AST is then modeled by a Tree-LSTM using a pre-training strategy to capture local non-linear syntax encoding”.
Therefore, it would be obvious to an ordinary in the art before the effective filing of the application to incorporate the teaching of Yu for determining a compile optimization option using chunks/nodes generated from a program according to graph-level tasks with the teaching of Lin for spiling source code to provide tree-form embedded as inputs into pre-training model.
The combination would yield predictable results because splitting code is a standard and requirement process in determination of an optimization, and always used by users for simplify the analysis, and processed by any compilers, large language model.
Regarding, Yu discloses,
constructing a data set, wherein the data set comprises at least one data entry, each of the data entry comprises: feature information corresponding to a third source code and a corresponding second compile optimization option; and
performing model training according to the feature information corresponding to the third source code and the corresponding second compile optimization option to obtain the prediction model, wherein in the model training process, the feature information corresponding to the third source code is used as an input of the prediction model, and the second compile optimization option is used as an output of the prediction model.
(See above in claim 3)
As per Claim 5: Yu and combining Lin, where Yu further discloses:
5. The method for determining a compile optimization option according to claim 3, wherein the data entry further comprises: target runtime environment information corresponding to the third source code (Yu, e.g. table 1 in p. 1149),, and performing model training according to the feature information corresponding to the third source code and the corresponding second compile optimization option to obtain the prediction model comprises(BERTs or CNN) :
performing model training according to the feature information corresponding to the third source code and the corresponding second compile optimization option, as well as the target runtime environment information corresponding to the third source code, to obtain the prediction model, wherein in the model training process, the feature information corresponding to the third source code and the target runtime environment information corresponding to the third source code are used as inputs of the prediction model, and the second compile optimization option is used as an output of the prediction model.
(See rationales address in claim 16)
As per Claim 6: Yu and combining Lin, where Yu further discloses:
6. The method for determining a compile optimization option according to claim 3, wherein the third source code (Yu, node /block number in Figures, 1,2, or 3) comprises:
source codes in the same file; or source codes in the same function (Figure 1 or 2, or 3).
As per Claim 7: Yu and combining Lin, where
Lin further discloses:
7. The method for determining a compile optimization option according to claim 1, wherein
the preset splitting level comprises a file level, and the second source code comprises source codes in the same file.
(Lin: in p. 3, see Fig. 2)
Therefore, it would be obvious to an ordinary in the art before the effective filing of the application to incorporate the teaching of Yu for determining a compile optimization option using chunks/nodes generated from a program according to graph-level tasks with the teaching of Lin for spiling source code to provide tree-form embedded as inputs into pre-training model.
The combination would yield predictable results conforming to a standard source code file.
As per Claim 8: Yu and combining Lin, where Yu further discloses:
8. The method for determining a compile optimization option according to claim 1, wherein inputting the feature information corresponding to each second source code into a pre-trained prediction model to obtain the first compile optimization option corresponding to each second source code comprises (Yu, Figure 3):
inputting the feature information corresponding to each second source code and target runtime environment information corresponding to the second source code into a pre-trained prediction model, to obtain the first compile optimization option corresponding to each second source code (Yu, Figure 3):
As per Claim 9: Yu and combining Lin, where Yu further discloses:
9. The method for determining a compile optimization option according to claim 8, wherein the target runtime environment information comprises: hardware information and operating system environment information.
(Yu: Table 1, information in Data set column)
As per Claim 12: Yu and combining Lin, where Yu further discloses:
12. (New) The method for determining a compile optimization option
according to claim 4, wherein the data entry further comprises:
target runtime environment information corresponding to the third source code, and
performing model training according to the feature information corresponding to the third source code and the corresponding second compile optimization option to obtain the prediction model comprises:
performing model training according to the feature information corresponding to the third source code and the corresponding second compile optimization option, as well as the target runtime environment information corresponding to the third source code, to obtain the prediction model, wherein in the model training process, the feature information corresponding to the third source code and the target runtime environment information corresponding to the third source code are used as inputs of the prediction model, and the second compile optimization option is used as an output of the prediction model.
(See the rejections addressed in claims 14, 18)
As per Claim 13: Yu and combining Lin, where Yu further discloses:
13. (New) The method for determining a compile optimization option
according to claim 1, wherein the preset splitting level comprises a function level, and the second source code comprises source codes in the same function.
(Claim has the functionality addressed in claim 7: Same rejection addressed in claim 7)
As per Claims 10-11: Claims are directed to an electronic device, and a non-transitory computer-readable medium, respectively. The claims have the same rejection addressed in claim 1.
Allowable Subject Matter
Claims 15, 17, and 19 are rejected under 35 USC 112(b). They are objected to under the prior art addressed above, but would be allowable if rewritten in independent form, including all the limitations of the base claim and any intervening claims, and provided with being amended to overcome the 112(b) issue.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Ted T Vo whose telephone number is (571)272-3706. The examiner can normally be reached 8am-4:30pm ET.
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, Wei Y Mui can be reached at (571) 272-3708. 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.
TTV
September 30, 2026
/Ted T. Vo/
Primary Examiner, Art Unit 2191
Prosecution Insights