Oracle America, Inc. v. Google Inc.

Filing 509

DECLARATION of RUCHIKA AGRAWAL in Opposition to 496 MOTION in Limine No. 5, 494 MOTION in Limine No. 3, 492 MOTION in Limine No. 1, 493 MOTION in Limine NO. 2, 495 MOTION in Limine No. 4 filed byGoogle Inc.. (Attachments: # 1 Exhibit 1-1, # 2 Exhibit 1-2, # 3 Exhibit 1-3, # 4 Exhibit 1-4, # 5 Exhibit 1-5, # 6 Exhibit 1-6, # 7 Exhibit 1-7, # 8 Exhibit 1-8, # 9 Exhibit 1-9, # 10 Exhibit 1-10, # 11 Exhibit 1-11, # 12 Exhibit 1-12, # 13 Exhibit 2-1, # 14 Exhibit 2-2, # 15 Exhibit 2-3, # 16 Exhibit 2-4, # 17 Exhibit 2-5, # 18 Exhibit 2-6, # 19 Exhibit 2-7, # 20 Exhibit 2-8, # 21 Exhibit 2-9, # 22 Exhibit 2-10, # 23 Exhibit 2-11, # 24 Exhibit 2-12, # 25 Exhibit 2-13, # 26 Exhibit 2-14, # 27 Exhibit 2-15, # 28 Exhibit 2-16, # 29 Exhibit 2-17, # 30 Exhibit 3-1, # 31 Exhibit 3-2, # 32 Exhibit 3-3, # 33 Exhibit 3-4, # 34 Exhibit 3-5, # 35 Exhibit 3-6, # 36 Exhibit 3-7, # 37 Exhibit 3-8, # 38 Exhibit 3-9, # 39 Exhibit 3-10, # 40 Exhibit 3-11, # 41 Exhibit 4-1, # 42 Exhibit 4-2, # 43 Exhibit 4-3, # 44 Exhibit 5-1, # 45 Exhibit 5-2, # 46 Exhibit 5-3, # 47 Supplement 5-4)(Related document(s) 496 , 494 , 492 , 493 , 495 ) (Kamber, Matthias) (Filed on 10/7/2011)

Download PDF
EXHIBIT 2-11 Highly Confidential - Attorneys' Eyes Only 1 UNITED STATES DISTRICT COURT 2 NORTHERN DISTRICT OF CALIFORNIA 3 SAN FRANCISCO DIVISION 4 5 ------------------------- 6 ORACLE AMERICA, INC., ) 7 Plaintiff, ) 8 vs. ) No. CV 10-03561 WHA 9 GOOGLE, INC., ) VOLUME I 10 11 Defendant. ) ------------------------- 12 13 HIGHLY CONFIDENTIAL - ATTORNEYS' EYES ONLY 14 15 Videotaped Deposition of ROBERT VANDETTE, 16 taken at 42 Chauncy Street, Boston, 17 Massachusetts, commencing at 10:02 a.m., 18 Wednesday, September 7, 2011, before 19 Jill Shepherd, RPR, MA-CSR No. 148608, 20 NH-CSR No. 128, CA-CSR No. 13275, CLR, 21 and Notary Public. 22 23 24 25 PAGES 1 - 124 Page 1 Veritext National Deposition & Litigation Services 866 299-5127 EXHIBIT 2-11 Highly Confidential - Attorneys' Eyes Only 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 preserve the objection. I don't believe this was produced to Oracle, but perhaps you can confirm that after this deposition and we can take it off line. Q. Are you familiar -- have you seen this presentation before? A. (Witness reviewing document). MS. AGRAWAL: Objection. Form. MR. FRANCIS: I would note that this presentation is cited in Oracle's claim charts. MS. AGRAWAL: The video? MR. FRANCIS: So you should be familiar with it. MS. AGRAWAL: The video or the actual presentations? MR. FRANCIS: Both, I believe. MS. AGRAWAL: All right. Let's take it off line. A. (Witness reviewing document). I may have seen a presentation similar to this. I can't confirm that this is the exact content that I have seen or whether I read through the entire presentation. Q. Okay. Just for a second, jump to slide 11, 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Q. All I asked if you see this third line, and if you understand what it seems to be saying. I'm not asking if you agree with it. A. I do not agree with -MS. AGRAWAL: You've got to let me object. The court reporter has to be able to take it down. Sorry. Objection. Form. A. I do see it and I do not agree with it. Q. Have you run any tests to determine what percentage of the time the CPU spends executing byte codes as opposed to natively compiled code? MS. AGRAWAL: Objection. Form. A. I have performance analysis in the past on our own virtual machines, and it very much depends on the byte codes and the program that you are running whether it spends little or a lot of time in the JIT -- or in executing byte codes, I'm sorry. Q. Maybe we can clarify just a little bit. Your performance report is measuring the performance of the Dalvik Virtual Machine, but not Android operating system Page 42 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 there's a diagram that's labeled "Dalvik Trace JIT Flow"? A. Okay. Q. Have you seen this before? A. No, I have not. Q. Now looking at slide five, the third point that's listed here, it says, "Typically, less than a third of time spent in the interpreter." Do you see that? A. This is very subjective. Doing what? Q. Do you understand generally what it means? I'm not asking if you agree, but do you understand what it's saying here? A. I would just like to come to your point -MS. AGRAWAL: Objection. Form. A. -- with the data on this slide that shows that running the checkers, that you're running 93 percent of the time in JIT code cache. So you are using almost 100 percent of the CPU when you are running checkers, so how do you conclude, then, that one-third of the time you are, on average, in interpreter? Page 44 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 overall -MS. AGRAWAL: Objection. Form. Q. -- is that correct? MS. AGRAWAL: Objection. Form. A. It's difficult to answer that question because the Dalvick Virtual Machine is part of the Android operating system, so which part are you -Q. Is it part of the prior Android operating system? MS. AGRAWAL: Objection. Form. A. My report states that I disabled much of the Android platform so the CPU was available for executing these benchmarks, so... Q. In a normal environment, is much of the Android platform disabled? MS. AGRAWAL: Objection. Form. A. In the normal Android platform, there is many Dalvick Virtual Machines running, which could have interfered with my results. Q. In a normal environment, is there anything other than a Dalvik Virtual Machine running on the Android operating system? MS. AGRAWAL: Objection. Form. A. It's running on top of the Linux kernel, but Page 43 Page 45 12 (Pages 42 - 45) Veritext National Deposition & Litigation Services 866 299-5127 EXHIBIT 2-11 Highly Confidential - Attorneys' Eyes Only 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Q. I'm not asking you for a specific number. Give me a general ballpark figure. MS. AGRAWAL: Objection. Form. A. You are asking me to describe changes or incremental performance improvements in groups that I really wasn't involved in. Q. Turning to page eight, paragraph 28 of your report, you discuss the modifications that you made to conduct your experiments; is that correct? A. Yes. Q. And you created these modifications based on what you were told by Professor Mitchell and Peter Kessler; is that correct? MS. AGRAWAL: Objection. Form. A. Yes. We discussed the functionality, and Peter I both looked through the sources to try to find out how to properly disable this functionality, and we came to a consensus. Q. It appears that you attempted two out of three possible scenarios here? A. That's correct. MS. AGRAWAL: Sorry, objection. Form. Q. The first scenario in paragraph 28 is 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 A. It would impact the results potentially, since I'd be adding additional functionality to Dalvick that it doesn't currently have. Q. It would, however, be technically possible for someone to do so? MS. AGRAWAL: Objection. Form. A. Let's see. It may be technically possible to build a system that does quickening without side tables, but it would involve adding additional overhead that Dalvick doesn't currently have. Q. In paragraph 36, you state that you did not try running the trace compiler; is that correct? A. Oh, paragraph -- sorry. That is correct, for the same reason that we didn't do the quickening alone. Q. What is the trace compiler? MS. AGRAWAL: Objection. Form. A. That is Dalvick's implementation of a JIT. Q. Are you saying that for your performance benchmark regarding the '104 patent you had to disable the JIT? MS. AGRAWAL: Form. A. That's correct. Page 82 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 building side tables, but not quickening instructions, and the second scenario is not building side tables or quickening instructions; is that correct? A. That's correct. Q. Is there a third possibility of not building side tables, but building quickening instructions? MS. AGRAWAL: Objection. Form. A. The quickening was dependent upon the side table for its implementation in order to avoid, you know, any possible error in the results. We did not want to substantially modify Dalvick in order to try to attempt that. We wanted to restrict our changes to just simple commenting out of code that would provide the before and after. Q. So if it got too complicated, you did not attempt it? MS. AGRAWAL: Objection. Form. A. It's not an issue of complication. It's an issue of possibly altering the Dalvick to the point where I wouldn't be measuring what I wanted to measure. Q. It would be -- Page 84 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Q. Do you understand that the JIT is not part of the accused functionality of the '104 patent? MS. AGRAWAL: Objection. Form. A. I'm not certain that it isn't somehow involved in some of the claims, but we focused on turning off the functionality in a mode that was possible. Q. If, in fact, JIT is not part of the accused functionality, then wouldn't disabling it affect the performance of this benchmark? MS. AGRAWAL: Objection. Form. A. I do believe that the numbers would be slightly different; however, the overhead of having to re-resolve all of the classes, fields, and methods is a fixed overhead that the JIT could not compensate for. So I believe the performance reduction or degradation would still be substantial. Q. Despite fixed overhead, you are referring to other aspects of the benchmarking programs might execute faster if the JIT was enabled; is that correct? MS. AGRAWAL: Objection. Form. A. They would be severely diminished by the Page 83 Page 85 22 (Pages 82 - 85) Veritext National Deposition & Litigation Services 866 299-5127 EXHIBIT 2-11 Highly Confidential - Attorneys' Eyes Only 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 JIT compiler? MS. AGRAWAL: Objection. Form, beyond the scope. A. I understand that HotSpot method-based JIT compiler and Android is a trace-based JIT compiler. Q. If Android was using a method-based JIT compiler, is it your belief that it would infringe the patent? MS. AGRAWAL: Objection. Form, beyond the scope. A. You'd have to show me and my team the implementation in order to determine that. Q. Does HotSpot practice -- strike that. Does the HotSpot Just-In-Time compiler practice the '205 patent? MS. AGRAWAL: Objection. Form, beyond the scope. A. It's my understanding that this patent was issued around the time of early Java, but we had alternative -- we had a -- you know, a pre-computer HotSpot compiler, so it's hard to say. My guess, I would believe it would. Q. The current HotSpot Just-In-Time compiler practices the '205 patent? 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 -Xint:fast mode and the -Xint:jit mode for the interpreter versus the JIT-enabled results. Q. Paragraph 49, you say, "These tests show the performance difference that JIT provides above and beyond interpreter only"; is that correct? A. Yes. Q. Is the accused functionality the entire JIT or only a specific portion within the JIT? MS. AGRAWAL: Objection. Form. A. As I understand it, it's the technique used to store the results of the JIT and such. If you are unable to store the results of the trace JIT, you wouldn't have a JIT; therefore, disabling the JIT is comparable to disabling the patent. Q. Are there other ways to store the results of the JIT? MS. AGRAWAL: Objection. Form, beyond the scope. A. I don't know. You are asking: Are there other ways to store the results that are not infringing? Is that what you are asking me? Q. I'm asking you: Is the only way to Page 98 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 MS. AGRAWAL: Objection. Form, beyond the scope, and calls for a legal conclusion. A. From my understanding, I believe it does. Q. Did you try comparing the performance of a current HotSpot Just-In-Time compiler with one that existed before the '205 patent? MS. AGRAWAL: Objection. Form, beyond the scope. A. From my report, I measured the current HotSpot implementation. Q. Looking at page 18 of your report, the chart here is entitled "Android CaffeineMark JIT Improvement Results." Does this reflect the difference between running Android with and without a JIT? MS. AGRAWAL: Objection. Form. I also just note for the record that we produced this to Google in color, and so this isn't the original that was -- the report wasn't what was given to Google; but you can answer the question. A. The command that I used to execute is in the report. It's on paragraph 49. I used Page 100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 implement a JIT is by using the '205 patent? MS. AGRAWAL: Objection. Form, beyond the scope. A. I don't know. Q. If there was a way to implement a JIT without practicing the '205 patent, would it make sense to benchmark the performance between that JIT and the current Android JIT that Oracle alleges infringes the '205 patent? MS. AGRAWAL: Objection. Form, beyond the scope. A. You are asking me to speculate on something which I have already stated that I don't know how you would do. So, again, the answer is, I don't know. Q. In paragraph 53, you say, "Before starting each benchmark run, the script cleans out the dalvik-cache." Do you see that? A. Yes. Q. What is in the dalvik-cache? MS. AGRAWAL: Objection. Form. A. The dalvik-cache contains an optimized version of the dex file, and if you run -- Page 99 Page 101 26 (Pages 98 - 101) Veritext National Deposition & Litigation Services 866 299-5127 EXHIBIT 2-11

Disclaimer: Justia Dockets & Filings provides public litigation records from the federal appellate and district courts. These filings and docket sheets should not be considered findings of fact or liability, nor do they necessarily reflect the view of Justia.


Why Is My Information Online?