Prosecution Insights
Last updated: November 20, 2026
Application No. 18/298,315

INTEGRATED CIRCUITS, SYSTEMS, AND METHODS FOR MULTIPLE-PRECISION MULTIPLY-AND-ACCUMULATE OPERATION

Non-Final OA §103§112
Filed
Apr 10, 2023
Examiner
BUI, KENNY KIM
Art Unit
2182
Tech Center
2100 — Computer Architecture & Software
Assignee
Edgecortix Inc.
OA Round
1 (Non-Final)
62%
Grant Probability
Moderate
1-2
OA Rounds
6m
Est. Remaining
99%
With Interview

Examiner Intelligence

Grants 62% of resolved cases
62%
Career Allowance Rate
16 granted / 26 resolved
+6.5% vs TC avg
Strong +46% interview lift
Without
With
+46.2%
Interview Lift
resolved cases with interview
Typical timeline
4y 2m
Avg Prosecution
14 currently pending
Career history
49
Total Applications
across all art units

Statute-Specific Performance

§101
29.2%
-10.8% vs TC avg
§103
37.0%
-3.0% vs TC avg
§102
5.7%
-34.3% vs TC avg
§112
26.0%
-14.0% vs TC avg
Black line = Tech Center average estimate • Based on career data from 26 resolved cases

Office Action

§103 §112
DETAILED ACTION 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 . Specification The disclosure is objected to as failing to comply with 37 CFR 1.71(a) because of the following informalities: Par.50, ll.5, “floating poing” should read as “floating point” Appropriate correction is required. Claim Objections Claims 1-8 and 20 objected to as failing to comply with 37 CFR 1.75(a) because of the following informalities: Claim 1, ll.16, “among the a plurality of floating point values” should read as “among a plurality of floating point values”; Claims 2-8 inherits this issue. Claim 20, ll.6, “of shared multipliers” should read as “of the shared multipliers”. Claim 20, ll.7, “dedicated multipliers among the” should read as “the dedicated multipliers among the”. Appropriate correction is required. 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 8, 15, and 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. Regarding claims 8, 15, and 20, recites “a plurality of accumulation adders”, however each claim depends on an independent claim (hereinafter parent claim) that recites “an accumulation adder”. It is unclear if the plurality of accumulation adders are new additional adders used for the systolic array, or if the plurality of accumulation adders are the accumulation adder of their respective parent claim. For purposes of examination, the plurality of accumulation adders are the accumulation adder of their respective parent claim. Claim Rejections - 35 USC § 103 In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status. 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. The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows: 1. Determining the scope and contents of the prior art. 2. Ascertaining the differences between the prior art and the claims at issue. 3. Resolving the level of ordinary skill in the pertinent art. 4. Considering objective evidence present in the application indicating obviousness or nonobviousness. Claims 1-3, 6, 9-11, and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Kim et al. (US 12,223,289 B2), hereinafter Kim. Regarding claim 1, Kim discloses an integrated circuit comprising: a multiply-and-accumulate (MAC) unit configured to operate in an integer mode to perform computations on first data-width integer values to produce third data-width integer values and configured to operate in a floating point mode to perform computations on second data-width floating point values to produce third data-width floating point values, wherein the second data width is twice the first data width and the third data width is larger than the second data width ["it is assumed that the calculation circuit 20 performs an operation on 16-bit input data and includes a floating point multiplier that can perform a multiplication operation on 8-bit or 11-bit data, and the integer data is expressed in 8 bits." Col.18, ll.40-44; "calculation circuit 20 may operate in a floating point calculation mode... or an integer calculation mode..." Col.18,ll.62-67; "the input data…brain float 16 (BF16)-type format...the output data may have a floating point 32 (FP32)-type format… the input data… 8-bit integer (INT8) format...output data may have a 32-bit integer (INT32) format..." Col.6.,ll.41-60; See figures 8B/10 and 8C/11 for both modes, using the floating point multiplier 21 for both modes], the MAC unit including a first multiplier, an exponent adder, a comparator, a subtractor, and a shifter [Fig.8B, FP multiplier 21, Align shifter 23, and Sign and Exponent Calculator 22; Fig.10: MT (max tree circuit, the comparator) to determine max exponent, AC11 Exponent adder, and MC1 (multiplier); Fig.12, SC1-5, subtractor circuits, and SHT11-51 for shifting], wherein the MAC unit is configured to perform computations in the integer mode including multiplying two integer values among a plurality of integer values with the first multiplier to produce an intermediate integer value, each integer value among the plurality of integer values having the first data width, and wherein the MAC unit is further configured to perform computations in the floating point mode including ["The floating point multiplier 21 supports floating point operations on floating point data pairs when the calculation circuit 20 is in a floating point calculation mode and supports integer operations on integer data pairs when the calculation circuit 20 is in an integer calculation mode" Col.13-14,ll.55-23; “a floating point calculation mode… 20a receives the first through eighth input data… that are floating point data…” Col.22,ll.61-67; “an integer calculation mode of… 20a… receives the first through fourth input data … that are integer data” Col.23,ll.44-65]: extracting an exponent value and a mantissa value from each of a plurality of floating point values, each floating point value among the plurality of floating point values having the second data width [Fig. 8B+8C and 10/11, see the solid black line for the extractor, shows the extraction of exponents and mantissa and passing the integers values to the floating point multiplier in the integer mode], multiplying, with the first multiplier, mantissa values extracted from each of two floating point values among the plurality of floating point values to produce a first mantissa product value [See fig. 8B, and 10; “the first multiplier circuit MC1 outputs the first through fourth fraction multiplication results mul1 through mul4 by performing a multiplication operation on each pair of first and second fractions Fa and Fb” Col.23,ll.40-43], adding, with the exponent adder, exponent values extracted from each of the two floating point values among the plurality of floating point values to produce an exponent sum value [Fig.8: "In an embodiment, the sign and exponent calculator 22 generates exponent information EI and sign information SI by determining a sign and performing exponent addition operations" Col.19, ll.38-43; Fig.10: AC11 “the first add circuit AC11 outputs first through fourth exponent addition results E_add1 through E_add4 by performing an addition operation on each pair of first and second exponents…” Col.23, ll.26-30], determining, with the comparator, a largest exponent value among a plurality of exponent sum values produced from the MAC unit [“In an embodiment, the sign and exponent calculator 22 identifies the maximum value from the exponent addition results and a value of the exponent ACC1_E of the first cumulative data ACC1…" Col.19, ll.44-49; “the max tree circuit MT… compares the exponent addition results and the third exponent ACC1_E with each other, and outputs the largest exponent value as the maximum value Max_E” Col.23, ll.16-25], subtracting, with the subtractor, the exponent sum value from the largest exponent value to produce a difference value, and shifting, with the shifter, the first mantissa product value based on the difference value to produce an intermediate mantissa value ["the first shift information SHI used by the align shifter 23 to align shift of each of the first through fourth fraction multiplication results...in a direction that corresponds to the sign of a difference between each of the exponent addition results and the exponent ACC1_E of the first cumulative data ACCl and the maximum value by the number of bits that corresponds to the difference therebetween" Col.19, ll.59-63; see Fig.12, SC1-5, subtractor circuits, and SHT11-51 shifting circuits]; an accumulation adder connected to two or more multipliers included in the MAC unit [Fig.8C shows the FP multiplier and int Multiplier connected to an adder 24], the accumulation adder configured to accumulate intermediate mantissa values of the MAC unit to produce an accumulation mantissa value in the floating point mode, and accumulate intermediate integer values of the MAC unit to produce an accumulation integer value in the integer mode ["in floating point calculation mode, the adder 24 adds the aligned fraction multiplication results of the floating point data items to the fraction of cumulative data and provides the addition result data to the first post adder 25… in integer calculation mode, the adder 24 adds integer multiplication results of the integer data items to cumulative data and provides the addition result to the second post adder 28." Col.14,ll.1-23; see figures 13A/B]; and normalizing accumulation FP adder and an INT adder connected to the accumulation adder, the normalizing accumulation adder configured to normalize the accumulation mantissa value with the largest exponent value to produce a third data-width floating point value in the floating point mode, and the INT adder produce a third data-width integer value without normalizing in the integer mode ["the second post adder 28 receives the first and second addition result data I_add1 and I_add2 and generates second cumulative data ACC2." Col.21, ll.61-63; "the normalization and rounding circuit 25_2 normalizes the shifted third addition result data according to the position of “1" and rounds the shifted third addition result data according to the number of bits in the format of data output from the calculation circuit 20." Col.20,ll.49-53; Fig.14, shows using the Max exponent] While Kim does not explicitly show an combined normalizing accumulation FP/INT adder. Kim discloses "In an embodiment, the first post adder 25 and the second post adder 28 are integrated with each other" on Col.14,ll.24-25. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim before him before the effective filing date of the claimed invention to implement the embodiment disclosed by Kim to have an integrated normalizing accumulation adder, allowing for sharing of hardware for different modes effectively reducing size and efficient accumulation/dot product operations [Kim: Col.12, ll.12-15, and Col.21, ll.54-60] Regarding claim 2, Kim disclose the invention substantially as claimed. See the discussion of claim 1 above. Kim further discloses a memory in communication with the MAC unit; wherein the computations in the integer mode further include reading the plurality of integer values from the memory; and wherein the computations in the floating point mode further include reading the plurality of floating point values from the memory [See fig.1: "the neural network device 120 receives a plurality of floating point data items FPD or integer data items ID from the memory 130 as input data items and performs dot product operations on the floating point data items FPD or the integer data items ID using the calculation circuit 20." Col.6, ll.26-34]. Regarding claim 3, Kim disclose the invention substantially as claimed. See the discussion of claim 1 above. a controller in communication with the MAC unit, the controller configured to operate in the integer mode to perform neural network inference on first data-width integer values and configured to operate in the floating point mode to perform neural network inference on second data-width floating point values [“the CPU 110 executes an application and controls the neural network device 120 to perform neural network-based tasks involved in the execution of the application.” Col.3,ll.58-61; "the neural network processor 30 includes a controller and a buffer that stores neural network parameters… the neural network processor 30 performs a neural network operation that includes a floating point calculation and an integer calculation according to embodiments, using the calculation circuit 20." Col.13,ll.27-36; See Fig.10 "IS_FLOAT" is an input to the calculation circuit 20]. Kim also discloses two types of high-level operations done by the system, i.e. training and inferencing [" According to an embodiment, the neural network system 100 trains a neural network or infers information from input data by analyzing the input data using the neural network” Col.3, ll.19-21]. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim before him before the effective filing date of the claimed invention to implement the embodiment of one of the two possibilities disclosed by Kim to perform neural network inferencing, in order to analyzing the input data using efficient hardware for the neural network operations [Kim: Col.3, ll.19-21, Col.12, ll.12-15, and Col.21, ll.54-60] Regarding claim 6, Kim disclose the invention substantially as claimed. See the discussion of claim 1 above. Kim discloses the MAC unit further includes a second multiplier, the MAC unit is further configured to, in the integer mode, multiply two integer values among the plurality of integer values with the second multiplier to produce an intermediate integer value, and the MAC unit is further configured to refrain from using the second multiplier in the floating point mode [Fig.8B+Fig.11, shows the integer multiplier gated during FP computation; "the integer multiplier 27 performs an integer calculation" Col.14,ll.10-16]. Regarding claim 9, Kim discloses: An integrated circuit comprising: a multiply-and-accumulate (MAC) unit configured to operate in an integer mode to perform computations on first data-width integer values to produce third data-width integer values and configured to operate in a floating point mode to perform computations on second data-width floating point values to produce third data-width floating point values, wherein the second data width is twice the first data width and the third data width is larger than the second data width ["it is assumed that the calculation circuit 20 performs an operation on 16-bit input data and includes a floating point multiplier that can perform a multiplication operation on 8-bit or 11-bit data, and the integer data is expressed in 8 bits." Col.18, ll.40-44; "calculation circuit 20 may operate in a floating point calculation mode... or an integer calculation mode..." Col.18,ll.62-67; "the input data…brain float 16 (BF16)-type format...the output data may have a floating point 32 (FP32)-type format… the input data… 8-bit integer (INT8) format...output data may have a 32-bit integer (INT32) format..." Col.6.,ll.41-60; See figures 8B/10 and 8C/11 for both modes, using the floating point multiplier 21 for both modes], the MAC unit including: an extractor configured to extract an exponent value and a mantissa value from each of a plurality of second data-width floating point values, the extractor further configured to pass through at least two first data-width integer values among a plurality of first data-width integer values [Fig. 8B+8C and 10, see the solid black line for the extractor, shows the extraction of exponents and mantissa and passing the integers values to the floating point multiplier in the integer mode; Fig.10 shows that the fractional part or the integer is sent to the multiplier, see col.23, ll.31-65], a first multiplier connected to the extractor, the first multiplier configured to multiply mantissa values extracted from each of two floating point values among the plurality of floating point values to produce a first mantissa product value, the first multiplier further configured to multiply two of the at least two first data-width integer values among the plurality of first data-width integer values to produce a first integer product value [Fig,8B and 8C; "The floating point multiplier 21 supports floating point operations on floating point data pairs when the calculation circuit 20 is in a floating point calculation mode and supports integer operations on integer data pairs when the calculation circuit 20 is in an integer calculation mode" Col.13-14,ll.55-23; “a floating point calculation mode… 20a receives the first through eighth input data… that are floating point data…” Col.22,ll.61-67; “an integer calculation mode of… 20a… receives the first through fourth input data … that are integer data” Col.23,ll.44-65], an exponent adder connected to the extractor, the exponent adder configured to add exponent values extracted from each of the two floating point values among the plurality of floating point values to produce an exponent sum value[Fig.8: "In an embodiment, the sign and exponent calculator 22 generates exponent information EI and sign information SI by determining a sign and performing exponent addition operations" Col.19, ll.38-43; Fig.10: AC11 “the first add circuit AC11 outputs first through fourth exponent addition results E_add1 through E_add4 by performing an addition operation on each pair of first and second exponents…” Col.23, ll.26-30], a comparator connected to the exponent adder, the comparator configured to determine a largest exponent value among a plurality of exponent sum values produced from the MAC unit [“In an embodiment, the sign and exponent calculator 22 identifies the maximum value from the exponent addition results and a value of the exponent ACC1_E of the first cumulative data ACC1…" Col.19, ll.44-49; “the max tree circuit MT… compares the exponent addition results and the third exponent ACC1_E with each other, and outputs the largest exponent value as the maximum value Max_E” Col.23, ll.16-25], a subtractor connected to the exponent adder and the comparator, the subtractor configured to subtract the exponent sum value from the largest exponent value to produce a difference value, and a shifter connected to the subtractor and the first multiplier, the shifter configured to shift the first mantissa product value based on the difference value to produce an intermediate mantissa value, the shifter further configured to pass through the first integer product value as an intermediate integer value ["the first shift information SHI used by the align shifter 23 to align shift of each of the first through fourth fraction multiplication results...in a direction that corresponds to the sign of a difference between each of the exponent addition results and the exponent ACC1_E of the first cumulative data ACC1 and the maximum value by the number of bits that corresponds to the difference there between" Col.19, ll.59-63; see Fig.12, SC1-5, subtractor circuits, and SHT11-51 shifting circuits; “When the calculation circuit 20 is in integer calculation mode,… the floating point multiplier 21 performs an integer calculation and directly provides second integer multiplication results to the first add circuit 24_1 by bypassing the align shifter 23” Col.14, ll.10-16]; an accumulation adder configured to accumulate intermediate mantissa values of the MAC unit to produce an accumulation mantissa value, the accumulation adder further configured to accumulate intermediate integer values of the MAC unit to produce an accumulation integer value ["in floating point calculation mode, the adder 24 adds the aligned fraction multiplication results of the floating point data items to the fraction of cumulative data and provides the addition result data to the first post adder 25… in integer calculation mode, the adder 24 adds integer multiplication results of the integer data items to cumulative data and provides the addition result to the second post adder 28." Col.14,ll.1-23; see figures 13A/B]; and a normalizing accumulation adder connected to the accumulation adder, the normalizing accumulation adder configured to normalize the accumulation mantissa value with the largest exponent value to produce a third data-width floating point value, the normalizing accumulation adder further configured to produce a third data-width integer value without normalizing ["the second post adder 28 receives the first and second addition result data I_add1 and I_add2 and generates second cumulative data ACC2." Col.21, ll.61-63; "the normalization and rounding circuit 25_2 normalizes the shifted third addition result data according to the position of “1" and rounds the shifted third addition result data according to the number of bits in the format of data output from the calculation circuit 20." Col.20,ll.49-53; Fig.14, shows using the Max exponent] While Kim does not explicitly show an combined normalizing accumulation FP/INT adder. Kim discloses "In an embodiment, the first post adder 25 and the second post adder 28 are integrated with each other" on Col.14,ll.24-25. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim before him before the effective filing date of the claimed invention to implement the embodiment disclosed by Kim to have an integrated normalizing accumulation adder, allowing for sharing of hardware for different modes effectively reducing size and efficient accumulation/dot product operations [Kim: Col.12, ll.12-15, and Col.21, ll.54-60]. Claim 10 is directed to claim 2. Wherein Kim disclose the invention substantially as claimed. See the discussion of claims 9 and 2 above. Claim 11 is directed to claim 3. Wherein Kim disclose the invention substantially as claimed. See the discussion of claims 9 and 3 above. Claim 14 is directed to claim 6 with less limitations. Wherein Kim disclose the invention substantially as claimed. See the discussion of claims 9 and 6 above. Claims 4-5, and 12-13 are rejected under 35 U.S.C. 103 as being unpatentable over Kim, and in view of Pillai et al. (US 2022/0382514 A1), hereinafter Pillai. Regarding claim 4, Kim disclose the invention substantially as claimed. See the discussion of claim 1 above. Kim discloses the use of activation layers in neural networks [“The neural network NN further includes an activation layer and may further include other layers that perform other types of operations” Col.8, ll.5-23] Kim also discloses outputting data in the third data-width and storing the final data in memory after the operation based on the input format ["the input data… 8-bit integer (INT8) format...output data may have a 32-bit integer (INT32) format..." Col.6.,ll.41-60;; and see figure 4] However, Kim does not explicitly disclose an integer activation pipeline in communication with the memory, the integer activation pipeline configured to activate the third data-width integer value to produce an activated first data-width integer value. In the analogous art of integer and floating point operations in neural network architectures, Pillai teaches an native precision pipeline in communication with the memory wherein the pipeline combines both an quantization pipeline for converting data back to native precision and an activation pipeline for applying activation functions on the input data [Fig.13, shows the accumulated value coming from memory; “Fig.13 shows a quantization process… In a quantization operation, the multiformat controller reads the accumulated values… which are accumulated with extended precision and quantize the accumulated values to native precisions of the accumulated values based on the direction provided by the software per channel” par.87; Fig.14 shows the input from memory or quantization; “The activation pipeline 590 may generally be implemented with the embodiments described herein… The activation pipeline 590 may be programmed as a standalone pipe, or operate in conjunction with a quantization output... Input data may be read from either memory or from a quantization pipe” par.89; par.2 discloses INT8/BF16 as potential input precisions] It would have been obvious to one of ordinary skill in the art, having the teaching of Kim and Pillai before him before the effective filing date of the claimed invention to include an activation/quantization pipeline taught by Pillai into the system of Kim, in order to implement activation and quantization functions, as required for neural networks, in a pipelined fashion allowing for increased efficiency and support for various precisions [Pillai: par.24-26, and 54-55]. The combination of Kim and Pillai discloses disclose an integer activation pipeline in communication with the memory, the integer activation pipeline configured to activate the third data-width integer value to produce an activated first data-width integer value. Regarding claim 5, Kim disclose the invention substantially as claimed. See the discussion of claim 1 above. Kim discloses the use of activation layers in neural networks [“The neural network NN further includes an activation layer and may further include other layers that perform other types of operations” Col.8, ll.5-23] Kim also discloses outputting data in the third data-width and storing the final data in memory after the operation based on the input format ["the input data…brain float 16 (BF16)-type format...the output data may have a floating point 32 (FP32)-type format” Col.6.,ll.41-60;; and see figure 4] However, Kim does not explicitly disclose a floating point activation pipeline in communication with the memory, the floating point activation pipeline configured to activate the third data-width floating point value to produce an activated second data-width floating point value. In the analogous art of integer and floating point operations in neural network architectures, Pillai teaches an native precision pipeline in communication with the memory wherein the pipeline combines both an quantization pipeline for converting data back to native precision and an activation pipeline for applying activation functions on the input data [Fig.13, shows the accumulated value coming from memory; “Fig.13 shows a quantization process… In a quantization operation, the multiformat controller reads the accumulated values… which are accumulated with extended precision and quantize the accumulated values to native precisions of the accumulated values based on the direction provided by the software per channel” par.87; Fig.14 shows the input from memory or quantization; “The activation pipeline 590 may generally be implemented with the embodiments described herein… The activation pipeline 590 may be programmed as a standalone pipe, or operate in conjunction with a quantization output... Input data may be read from either memory or from a quantization pipe” par.89; par.2 discloses INT8/BF16 as potential input precisions] It would have been obvious to one of ordinary skill in the art, having the teaching of Kim and Pillai before him before the effective filing date of the claimed invention to include an activation/quantization pipeline taught by Pillai into the system of Kim, in order to implement activation and quantization functions, as required for neural networks, in a pipelined fashion allowing for increased efficiency and support for various precisions [Pillai: par.24-26, and 54-55]. The combination of Kim and Pillai discloses disclose a floating point activation pipeline in communication with the memory, the floating point activation pipeline configured to activate the third data-width floating point value to produce an activated second data-width floating point value. Claim 12 is directed to claim 4. Wherein Kim and Pillai disclose the invention substantially as claimed. See the discussion of claim 9 and 4 above. Claim 13 is directed to claim 5. Wherein Kim and Pillai disclose the invention substantially as claimed. See the discussion of claim 9 and 5 above. Claims 7, and 16-18 are rejected under 35 U.S.C. 103 as being unpatentable over Kim, and in view of Wang et al. (US 2024/0329933 A1), hereinafter Wang. Regarding claim 7, Kim disclose the invention substantially as claimed. See the discussion of claim 1 above. Kim discloses the MAC unit is further configured to perform the computations in the floating point mode for the first multiplier, wherein the first multiplier is grouped with an instance of the exponent adder, an instance of the comparator, an instance of the subtractor, and an instance of the shifter for computing floating point operations [See Fig.8B for FP operation mode, and Fig.10: MT (max tree circuit, the comparator), AC11 Exponent adder, and MC1 (multiplier); Fig.12, SC1-5, subtractor circuits, and SHT11-51 for shifting] And Pair of the first multiplier and the second multiplier [Fig.8B/8C, shows the two multipliers] However, Kim does not explicitly disclose the MAC unit includes a plurality of instances of pairs of the first multiplier and the second multiplier, and the MAC unit is further configured to perform the computations in the floating point mode for each first multiplier. In the analogous art of shared FP multipliers and dedicated integer multipliers, Wang discloses a plurality of instances of the first multipliers and the second multipliers and the MAC unit is further configured to perform the computations in the floating point mode for each first multiplier [Fig.1 and 6; "multiplier circuit MULA is capable of performing multiplication in a float-point mode or in an integer mode while multiplier MULB is capable of performing multiplication only in the integer mode." par.58; " all of multipliers 604 may be active in the integer mode whereas only a subset of multipliers 604 may be active during the floating-point mode" par.75]. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim and Wang before him before the effective filing date of the claimed invention to modify the multiplier pair of Kim include additional multipliers as taught by Wang, in order to allow for additional parallel operations to occur and accelerate MAC operations [Wang: par.66, 71, 73-75, 79, and 82]. The combination of Kim and Wang discloses a plurality of instances of pairs of the first multiplier and the second multiplier, and the MAC unit is further configured to perform the computations in the floating point mode for each first multiplier. Regarding claim 16, Kim discloses: An integrated circuit comprising: a multiplier group configured to operate in an integer mode to perform computations on first data-width integer values to produce third data-width integer values and configured to operate in a floating point mode to perform computations on second data-width floating point values to produce third data-width floating point values, wherein the second data width is twice the first data width and the third data width is larger than the second data width ["it is assumed that the calculation circuit 20 performs an operation on 16-bit input data and includes a floating point multiplier that can perform a multiplication operation on 8-bit or 11-bit data, and the integer data is expressed in 8 bits." Col.18, ll.40-44; "calculation circuit 20 may operate in a floating point calculation mode... or an integer calculation mode..." Col.18,ll.62-67; "the input data…brain float 16 (BF16)-type format...the output data may have a floating point 32 (FP32)-type format… the input data… 8-bit integer (INT8) format...output data may have a 32-bit integer (INT32) format..." Col.6.,ll.41-60; See figures 8B/10 and 8C/11 for both modes, using the floating point multiplier 21 for both modes], multiplier group including: an extractor configured to extract an exponent value and a mantissa value from each of a plurality of second data-width floating point values, the extractor further configured to pass through at least two first data-width integer values among a plurality of first data-width integer values [Fig. 8B+8C and 10, see the solid black line for the extractor, shows the extraction of exponents and mantissa and passing the integers values to the floating point multiplier in the integer mode; Fig.10 shows that the fractional part or the integer is sent to the multiplier, see col.23, ll.31-65], a group multiplier connected to the extractor, the group multiplier configured to multiply mantissa values extracted from each of two floating point values among the plurality of floating point values to produce a first mantissa product value, the group multiplier further configured to multiply two of the at least two first data-width integer values among the plurality of first data-width integer values to produce a first integer product value [Fig,8B and 8C; "The floating point multiplier 21 supports floating point operations on floating point data pairs when the calculation circuit 20 is in a floating point calculation mode and supports integer operations on integer data pairs when the calculation circuit 20 is in an integer calculation mode" Col.13-14,ll.55-23; “a floating point calculation mode… 20a receives the first through eighth input data… that are floating point data…” Col.22,ll.61-67; “an integer calculation mode of… 20a… receives the first through fourth input data … that are integer data” Col.23,ll.44-65], an exponent adder connected to the extractor, the exponent adder configured to add exponent values extracted from each of the two floating point values among the plurality of floating point values to produce an exponent sum value [Fig.8: "In an embodiment, the sign and exponent calculator 22 generates exponent information EI and sign information SI by determining a sign and performing exponent addition operations" Col.19, ll.38-43; Fig.10: AC11 “the first add circuit AC11 outputs first through fourth exponent addition results E_add1 through E_add4 by performing an addition operation on each pair of first and second exponents…” Col.23, ll.26-30], a comparator connected to the exponent adder, the comparator configured to determine a largest exponent value among a plurality of exponent sum values produced from the plurality of multiplier groups [“In an embodiment, the sign and exponent calculator 22 identifies the maximum value from the exponent addition results and a value of the exponent ACC1_E of the first cumulative data ACC1…" Col.19, ll.44-49; “the max tree circuit MT… compares the exponent addition results and the third exponent ACC1_E with each other, and outputs the largest exponent value as the maximum value Max_E” Col.23, ll.16-25], a subtractor connected to the exponent adder and the comparator, the subtractor configured to subtract the exponent sum value from the largest exponent value to produce a difference value, and a shifter connected to the subtractor and the group multiplier, the shifter configured to shift the first mantissa product value based on the difference value to produce an intermediate mantissa value, the shifter further configured to pass through the first integer product value as an intermediate integer value ["the first shift information SHI used by the align shifter 23 to align shift of each of the first through fourth fraction multiplication results...in a direction that corresponds to the sign of a difference between each of the exponent addition results and the exponent ACC1_E of the first cumulative data ACC1 and the maximum value by the number of bits that corresponds to the difference there between" Col.19, ll.59-63; see Fig.12, SC1-5, subtractor circuits, and SHT11-51 shifting circuits; “When the calculation circuit 20 is in integer calculation mode,… the floating point multiplier 21 performs an integer calculation and directly provides second integer multiplication results to the first add circuit 24_1 by bypassing the align shifter 23” Col.14, ll.10-16]; a dedicated multiplier, the dedicated multiplier configured to multiply two first data-width integer values among the plurality of first data-width integer values to produce a second integer product value [Fig 8C integer multiplier 27; "the integer multiplier 27 performs an integer calculation" Col.14,ll.10-16; "the integer multiplier 27 generates fifth through eighth integer multiplication results" Col.21, ll.38-53]; an accumulation adder connected to shared multipliers among the plurality of multiplier groups and dedicated multipliers among the plurality of dedicated multipliers [Fig 8B+8C, Adder 24; "one of the first and second add circuits 24_1 and 24_2 of the adder 24 is shared with an integer multiplier that performs an integer multiplication..." Col.21, ll.54-60], the accumulation adder configured to accumulate the intermediate mantissa values of the multiplier group to produce an accumulation mantissa value, the accumulation adder further configured to accumulate the intermediate integer values of the multiplier group and the dedicated multiplier to produce an accumulation integer value ["in floating point calculation mode, the adder 24 adds the aligned fraction multiplication results of the floating point data items to the fraction of cumulative data and provides the addition result data to the first post adder 25… in integer calculation mode, the adder 24 adds integer multiplication results of the integer data items to cumulative data and provides the addition result to the second post adder 28." Col.14,ll.1-23; see figures 13A/B]; and a normalizing accumulation adder configured to normalize the accumulation mantissa value with the largest exponent value to produce a third data-width floating point value, the normalizing accumulation adder further configured to produce a third data-width integer value without normalizing ["the second post adder 28 receives the first and second addition result data I_add1 and I_add2 and generates second cumulative data ACC2." Col.21, ll.61-63; "the normalization and rounding circuit 25_2 normalizes the shifted third addition result data according to the position of “1" and rounds the shifted third addition result data according to the number of bits in the format of data output from the calculation circuit 20." Col.20,ll.49-53; Fig.14, shows using the Max exponent] While Kim does not explicitly show an combined normalizing accumulation FP/INT adder. Kim discloses "In an embodiment, the first post adder 25 and the second post adder 28 are integrated with each other" on Col.14,ll.24-25. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim before him before the effective filing date of the claimed invention to implement the embodiment disclosed by Kim to have an integrated normalizing accumulation adder, allowing for sharing of hardware for different modes effectively reducing size and efficient accumulation/dot product operations [Kim: Col.12, ll.12-15, and Col.21, ll.54-60]. However, Kim does not explicitly disclose a plurality of multiplier groups, a plurality of dedicated multipliers, and an accumulation adder connected to shared multipliers among the plurality of multiplier groups and dedicated multipliers among the plurality of dedicated multipliers. In the analogous art of shared FP multipliers and dedicated integer multipliers, Wang discloses a plurality of instances of the FP/INT multipliers and the dedicated INT multipliers and the MAC unit is further configured to perform the computations in the floating point mode for each FP/INT multipliers [Fig.1 and 6; "multiplier circuit MULA is capable of performing multiplication in a float-point mode or in an integer mode while multiplier MULB is capable of performing multiplication only in the integer mode." par.58; " all of multipliers 604 may be active in the integer mode whereas only a subset of multipliers 604 may be active during the floating-point mode" par.75]. And discloses an adder-shifter, with shifter bypass for integer mode, connected to shared multipliers among the plurality of FP/INT multipliers and dedicated integer multipliers among the plurality of dedicated integer multipliers [Fig.6, shows 446; “the multiplied output may bypass shifter circuits in adder-shifter 446, for example, when in the integer mode the multiplied outputs… may bypass shifter circuits” par.59] It would have been obvious to one of ordinary skill in the art, having the teaching of Kim and Wang before him before the effective filing date of the claimed invention to modify the multiplier pair and the accumulator adder of Kim include additional multipliers and related adder as taught by Wang, in order to allow for additional parallel operations to occur and accelerate MAC operations [Wang: par.66, 71, 73-75, 79, and 82]. The combination of Kim and Wang discloses a plurality of multiplier groups and a plurality of dedicated multipliers, and an accumulation adder connected to shared multipliers among the plurality of multiplier groups and dedicated multipliers among the plurality of dedicated multipliers. Regarding claim 17, Kim and Wang disclose the invention substantially as claimed. See the discussion of claim 16 above. Kim further discloses a memory in communication with the multiplier group and the dedicated multiplier; wherein the computations in the integer mode further include reading the plurality of integer values from the memory; and wherein the computations in the floating point mode further include reading the plurality of floating point values from the memory for the multiplier group [See fig.1: "the neural network device 120 receives a plurality of floating point data items FPD or integer data items ID from the memory 130 as input data items and performs dot product operations on the floating point data items FPD or the integer data items ID using the calculation circuit 20." Col.6, ll.26-34; see Fig.8B/C for the input types]. However, Kim does not explicitly disclose the memory in communication with each multiplier group among the plurality of multiplier groups and each dedicated multiplier among the plurality of multipliers; In the analogous art of shared FP multipliers and dedicated integer multipliers, Wang discloses a plurality of instances of the FP/INT multipliers and the dedicated INT multipliers and the MAC unit is further configured to perform the computations in the floating point mode for each FP/INT multipliers [Fig.1 and 6; "multiplier circuit MULA is capable of performing multiplication in a float-point mode or in an integer mode while multiplier MULB is capable of performing multiplication only in the integer mode." par.58; " all of multipliers 604 may be active in the integer mode whereas only a subset of multipliers 604 may be active during the floating-point mode" par.75]. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim and Wang before him before the effective filing date of the claimed invention to modify the multiplier pair of Kim include additional multipliers as taught by Wang, in order to allow for additional parallel operations to occur and accelerate MAC operations [Wang: par.66, 71, 73-75, 79, and 82]. The combination of Kim and Wang discloses the memory in communication with each multiplier group among the plurality of multiplier groups and each dedicated multiplier among the plurality of multipliers; Regarding claim 18, Kim and Wang disclose the invention substantially as claimed. See the discussion of claim 16 above. Kim discloses a controller in communication with the MAC unit, the controller configured to operate in the integer mode to perform neural network inference on first data-width integer values and configured to operate in the floating point mode to perform neural network inference on second data-width floating point values [“the CPU 110 executes an application and controls the neural network device 120 to perform neural network-based tasks involved in the execution of the application.” Col.3,ll.58-61; "the neural network processor 30 includes a controller and a buffer that stores neural network parameters… the neural network processor 30 performs a neural network operation that includes a floating point calculation and an integer calculation according to embodiments, using the calculation circuit 20." Col.13,ll.27-36; See Fig.10 "IS_FLOAT" is an input to the calculation circuit 20]. Kim also discloses two types of high-level operations done by the system, i.e. training and inferencing [" According to an embodiment, the neural network system 100 trains a neural network or infers information from input data by analyzing the input data using the neural network” Col.3, ll.19-21]. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim before him before the effective filing date of the claimed invention to implement the embodiment of one of the two possibilities disclosed by Kim to perform neural network inferencing, in order to analyzing the input data using efficient hardware for the neural network operations [Kim: Col.3, ll.19-21, Col.12, ll.12-15, and Col.21, ll.54-60]. However, Kim does not explicitly disclose the controller in communication with the plurality of multiplier groups. In the analogous art of shared FP multipliers and dedicated integer multipliers, Wang discloses a plurality of instances of the FP/INT multipliers and the dedicated INT multipliers and the MAC unit is further configured to perform the computations in the floating point mode for each FP/INT multipliers [Fig.1 and 6; "multiplier circuit MULA is capable of performing multiplication in a float-point mode or in an integer mode while multiplier MULB is capable of performing multiplication only in the integer mode." par.58; " all of multipliers 604 may be active in the integer mode whereas only a subset of multipliers 604 may be active during the floating-point mode" par.75]. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim and Wang before him before the effective filing date of the claimed invention to modify the multiplier pair of Kim include additional multipliers as taught by Wang, in order to allow for additional parallel operations to occur and accelerate MAC operations [Wang: par.66, 71, 73-75, 79, and 82]. The combination of Kim and Wang disclose the controller in communication with the plurality of multiplier groups. Claims 8 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Kim, and in view of Mellempudi et al. (US 2022/0318013 A1), hereinafter Mellempudi. Regarding claim 8, Kim disclose the invention substantially as claimed. See the discussion of claim 1 above. Kim discloses the MAC unit wherein the MAC unit is configured to perform, in the floating point mode, determining, with the comparator, the largest exponent value among the plurality of exponent sum values of the input data and accumulated data. [See fig.10, MT; "In an embodiment, the sign and exponent calculator 22 identifies the maximum value from the exponent addition results and a value of the exponent ACC1_E of the first cumulative data ACC1…" Col.19, ll.44-49] A accumulation adder connected to two or more of any combination of multipliers included in the plurality of instances of MAC units [See fig. 8C, for the adder connected to two multipliers] wherein the plurality of accumulation adders are collectively configured to accumulate the intermediate mantissa values of the plurality of MAC units to produce the accumulation mantissa value in the floating point mode, and accumulate the intermediate integer values of the plurality of MAC units to produce the accumulation integer value in the integer mode ["The floating point multiplier 21 supports floating point operations on floating point data pairs when the calculation circuit 20 is in a floating point calculation mode and supports integer operations on integer data pairs when the calculation circuit 20 is in an integer calculation mode" Col.13-14,ll.55-23; "in floating point calculation mode, the adder 24 adds the aligned fraction multiplication results of the floating point data items to the fraction of cumulative data and provides the addition result data to the first post adder 25… in integer calculation mode, the adder 24 adds integer multiplication results of the integer data items to cumulative data and provides the addition result to the second post adder 28." Col.14,ll.1-23; see figures 13A/B]. wherein the normalizing accumulation adder is included with the accumulation adder [See fig.8B/8C] However, Kim does not explicitly disclose a systolic array including a plurality of instances of the MAC unit, and a plurality of accumulation adders, each accumulation adder connected to two or more of any combination of multipliers included in the plurality of instances of MAC units and preceding accumulation adders. In the analogous art of systolic architecture of multipliers, Mellempudi discloses a plurality of instances of MAC unit and a plurality of accumulation adders each accumulation adder connected to two or more of any combination of multipliers included in the plurality of instances of MAC units and preceding accumulation adders [Fig.17A, 17B, wherein it shows multiple multipliers and accumulating the results with one adder and accumulating further in the next adder; Fig. 32 shows using the last adder to output the final result in the destination format, see par.351]. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim and Mellempudi before him before the effective filing date of the claimed invention to modify the system of Kim include the systolic architecture for dot product and accumulate operations as taught by Mellempudi, in order to allow for a chaining of multiplier units to efficiently perform dot-product iterations [Mellempudi: par.236-238]. Regarding claim 15 is directed to claim 8. Kim disclose the invention substantially as claimed. See the discussion of claim 9 and 8 above. Claim 19 is rejected under 35 U.S.C. 103 as being unpatentable over Kim, and Wang, and further in view of Pillai. Regarding claim 19, Kim and Wang disclose the invention substantially as claimed. See the discussion of claim 16 above. Kim discloses the use of activation layers in neural networks [“The neural network NN further includes an activation layer and may further include other layers that perform other types of operations” Col.8, ll.5-23] Kim also discloses outputting data in the third data-width and storing the final data in memory after the operation based on the input format ["the input data…brain float 16 (BF16)-type format...the output data may have a floating point 32 (FP32)-type format” Col.6.,ll.41-60;; and see figure 4] However, Kim and Wang does not explicitly disclose a floating point activation pipeline in communication with the memory, the floating point activation pipeline configured to activate the third data-width floating point value to produce an activated second data-width floating point value. In the analogous art of integer and floating point operations in neural network architectures, Pillai teaches an native precision pipeline in communication with the memory wherein the pipeline combines both an quantization pipeline for converting data back to native precision and an activation pipeline for applying activation functions on the input data [Fig.13, shows the accumulated value coming from memory; “Fig.13 shows a quantization process… In a quantization operation, the multiformat controller reads the accumulated values… which are accumulated with extended precision and quantize the accumulated values to native precisions of the accumulated values based on the direction provided by the software per channel” par.87; Fig.14 shows the input from memory or quantization; “The activation pipeline 590 may generally be implemented with the embodiments described herein… The activation pipeline 590 may be programmed as a standalone pipe, or operate in conjunction with a quantization output... Input data may be read from either memory or from a quantization pipe” par.89; par.2 discloses INT8/BF16 as potential input precisions] It would have been obvious to one of ordinary skill in the art, having the teaching of Kim and Pillai before him before the effective filing date of the claimed invention to include an activation/quantization pipeline taught by Pillai into the system of Kim, in order to implement activation and quantization functions, as required for neural networks, in a pipelined fashion allowing for increased efficiency and support for various precisions [Pillai: par.24-26, and 54-55]. The combination of Kim and Pillai discloses disclose a floating point activation pipeline in communication with the memory, the floating point activation pipeline configured to activate the third data-width floating point value to produce an activated second data-width floating point value. Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Kim, and Wang, and further in view of Mellempudi. Regarding claim 20, Kim and Wang disclose the invention substantially as claimed. See the discussion of claim 16 above. Kim discloses the multiplier group wherein the multiplier group is configured to perform, in the floating point mode, determining, with the comparator, the largest exponent value among the plurality of exponent sum values of the input data and accumulated data. [See fig.10, MT; "In an embodiment, the sign and exponent calculator 22 identifies the maximum value from the exponent addition results and a value of the exponent ACC1_E of the first cumulative data ACC1…" Col.19, ll.44-49] a dedicated multiplier, [Fig 8C integer multiplier 27; "the integer multiplier 27 performs an integer calculation" Col.14,ll.10-16; "the integer multiplier 27 generates fifth through eighth integer multiplication results" Col.21, ll.38-53]; an accumulation adder connected to shared multipliers among the plurality of multiplier groups and dedicated multipliers among the plurality of dedicated multipliers [Fig 8B+8C, Adder 24; "one of the first and second add circuits 24_1 and 24_2 of the adder 24 is shared with an integer multiplier that performs an integer multiplication..." Col.21, ll.54-60], the accumulation adder configured to accumulate the intermediate mantissa values of the multiplier group to produce an accumulation mantissa value, the accumulation adder further configured to accumulate the intermediate integer values of the multiplier group and the dedicated multiplier to produce an accumulation integer value [ Fig. 8B/C; "in floating point calculation mode, the adder 24 adds the aligned fraction multiplication results of the floating point data items to the fraction of cumulative data and provides the addition result data to the first post adder 25… in integer calculation mode, the adder 24 adds integer multiplication results of the integer data items to cumulative data and provides the addition result to the second post adder 28." Col.14,ll.1-23; see figures 13A/B]; and wherein the normalizing accumulation adder is included with the accumulation adder [See fig.8B/8C] However, Kim does not explicitly disclose a systolic array including the plurality of multiplier groups, the plurality of dedicated multipliers, and a plurality of accumulation adders, each accumulation adder connected to two or more of any combination of shared multipliers among the plurality of multiplier groups, dedicated multipliers among the plurality of dedicated multipliers, and preceding accumulation adders. In the analogous art of shared FP multipliers and dedicated integer multipliers, Wang discloses a plurality of instances of the FP/INT multipliers and the dedicated INT multipliers and the MAC unit is further configured to perform the computations in the floating point mode for each FP/INT multipliers [Fig.1 and 6; "multiplier circuit MULA is capable of performing multiplication in a float-point mode or in an integer mode while multiplier MULB is capable of performing multiplication only in the integer mode." par.58; " all of multipliers 604 may be active in the integer mode whereas only a subset of multipliers 604 may be active during the floating-point mode" par.75]. And discloses an adder-shifter, with shifter bypass for integer mode, connected to shared multipliers among the plurality of FP/INT multipliers and dedicated integer multipliers among the plurality of dedicated integer multipliers [Fig.6, shows 446; “the multiplied output may bypass shifter circuits in adder-shifter 446, for example, when in the integer mode the multiplied outputs… may bypass shifter circuits” par.59] It would have been obvious to one of ordinary skill in the art, having the teaching of Kim and Wang before him before the effective filing date of the claimed invention to modify the multiplier pair and the accumulator adder of Kim include additional multipliers and related adder as taught by Wang, in order to allow for additional parallel operations to occur and accelerate MAC operations [Wang: par.66, 71, 73-75, 79, and 82]. The combination of Kim and Wang discloses a plurality of multiplier groups and a plurality of dedicated multipliers, and an accumulation adder connected to two or more of any combination of shared multipliers among the plurality of multiplier groups, dedicated multipliers among the plurality of dedicated multipliers. However, Kim and Wang does not explicitly disclose, a systolic array including the plurality of multiplier groups, the plurality of dedicated multipliers, and a plurality of accumulation adders, each accumulation adder connected to two or more of any combination of shared multipliers among the plurality of multiplier groups, dedicated multipliers among the plurality of dedicated multipliers, and preceding accumulation adders. In the analogous art of systolic architecture of multipliers, Mellempudi discloses a plurality of instances of multipliers and a plurality of accumulation adders each accumulation adder connected to two or more of any combination of multipliers included in the plurality of instances of multipliers and preceding accumulation adders [Fig.17A, 17B, wherein it shows multiple multipliers and accumulating the results with one adder and accumulating further in the next adder; Fig. 32 shows using the last adder to output the final result in the destination format, see par.351]. It would have been obvious to one of ordinary skill in the art, having the teaching of Kim, Wang, and Mellempudi before him before the effective filing date of the claimed invention to modify the system of the combination of Kim and Wang include the systolic architecture for dot product and accumulate operations as taught by Mellempudi, in order to allow for a chaining of multiplier units to efficiently perform dot-product iterations [Mellempudi: par.236-238]. The combination of Kim, Wang, and Mellempudi discloses disclose, a plurality of accumulation adders each accumulation adder, each accumulation adder connected to two or more of any combination of shared multipliers among the plurality of multiplier groups, dedicated multipliers among the plurality of dedicated multipliers, and preceding accumulation adders. Conclusion The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Bittner et al. (US 10,528,321 B2) discloses pairs of multipliers in a systolic array using one accumulator adder, see fig.4 and 5. Parra et al. (US 12,189,571 B2) discloses pair of multipliers in a systolic array using one accumulator adder, see fig.29 and 30 Lu et al. (US 12,613,675 B2) discloses an multiplier capable of computing using BF16 or INT8, see fig.2. Bratt et al. (US 4,525,780 A) discloses an detailed extractor capable of extracting FP and integer values onto two lines, one for exponent, and one for integer/mantissa, see fig.256 and Col.263-265. Nagel et al. (NPL: “A White Paper on Neural Network Quantization”) discloses requantization of the output of an activation function into the input data type, see fig.4. Diamant et al. (US 12,499,353 B1) discloses a pipeline wherein activation processing is done prior to quantization, see fig.3A Mills (US 11,580,353 B2) discloses shared FP/INT circuits, see fig.9A/9B. Urbanski et al. (US 12,254,061 B2) discloses mode controllers and pipelining multiple multipliers, see at least fig.10 Any inquiry concerning this communication or earlier communications from the examiner should be directed to Kenny K. Bui whose telephone number is (571)270-0604. The examiner can normally be reached 8:00 am to 3:00 pm on Monday, 8:00 am to 4:00 pm on Tuesday to Friday 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, Andrew T Caldwell can be reached at (571)272-3702. 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. /KENNY K. BUI/Patent Examiner, Art Unit 2182 (571)270-0604 /ANDREW CALDWELL/Supervisory Patent Examiner, Art Unit 2182
Read full office action

Prosecution Timeline

Apr 10, 2023
Application Filed
Oct 21, 2026
Non-Final Rejection mailed — §103, §112
Nov 17, 2026
Interview Requested

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12801689
Matrix Multiply Accelerator for Variable Bitwidth Operands
4y 7m to grant Granted Nov 17, 2026
Patent 12801348
FIXED ASYMMETRY COMPENSATION FOR MULTIPLY AND ACCUMULATE OPERATIONS
3y 11m to grant Granted Nov 17, 2026
Patent 12769701
IN-NETWORK PARALLEL PREFIX SCAN
5y 1m to grant Granted Oct 20, 2026
Patent 12769805
Matrix Multiply Accelerator For Variable Bitwidth Operands
5y 0m to grant Granted Oct 20, 2026
Patent 12743251
FLOATING POINT FUSED MULTIPLY ADD WITH MERGED 2'S COMPLEMENT AND ROUNDING
4y 8m to grant Granted Sep 22, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

1-2
Expected OA Rounds
62%
Grant Probability
99%
With Interview (+46.2%)
4y 2m (~6m remaining)
Median Time to Grant
Low
PTA Risk
Based on 26 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month