Oracle America, Inc. v. Google Inc.

Filing 497

Declaration of DANIEL PURCELL in Support of #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, #2 Exhibit 2, #3 Exhibit 3, #4 Exhibit 4, #5 Exhibit 5, #6 Exhibit 6, #7 Exhibit 7, #8 Exhibit 8, #9 Exhibit 9, #10 Exhibit 10, #11 Exhibit 11, #12 Exhibit 12, #13 Exhibit 13, #14 Exhibit 14, #15 Exhibit 15, #16 Exhibit 16, #17 Exhibit 17, #18 Exhibit 18, #19 Exhibit 19, #20 Exhibit 20, #21 Exhibit 21, #22 Exhibit 22, #23 Exhibit 23, #24 Exhibit 24, #25 Exhibit 25, #26 Exhibit 26, #27 Exhibit 27, #28 Exhibit 28, #29 Exhibit 29, #30 Exhibit 30, #31 Exhibit 31, #32 Exhibit 32, #33 Exhibit 33, #34 Exhibit 34, #35 Exhibit 35, #36 Exhibit 36, #37 Exhibit 37, #38 Exhibit 38, #39 Exhibit 39, #40 Exhibit 40)(Related document(s) #496 , #494 , #492 , #493 , #495 ) (Kamber, Matthias) (Filed on 10/7/2011)

EXHIBIT 7

UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF CALIFORNIA
SAN FRANCISCO DIVISION

ORACLE AMERICA, INC.

Case No. CV 10-03561 WHA

Plaintiff,
v.
GOOGLE INC.
Defendant.

SUMMARY AND REPORT OF ROBERT (“BOB”) G. VANDETTE
SUBMITTED ON BEHALF OF PLAINTIFF
ORACLE AMERICA, INC.

IV.

PERFORMANCE ANALYSIS OF THE ’104 PATENT ON ANDROID
A.

26.

Google Android benefits from RE38,104

I ran experiments to disable the Android functionality that Oracle accuses of

infringing the ’104 patent.
27.

The baseline for these experiments was the Froyo release of Android, pulled from

the Google git repository with the Google repo commands
$ repo init -u git://android.git.kernel.org/platform/manifest.git -b froyo
$ repo sync
to initialize and sync from the Google repository.
28.

I then made two additional copies of the repository
(1) building side tables of resolved constant pool entries, but not quickening
instructions
(2) not building side tables of resolved constant pool entries, nor quickening
instructions

29.

Modifications were made to the source in the copies to implement the

experiments.
30.

The modification for the first experiment (side tables but no quickening) are

restricted to
dalvik/vm/analysis/DexOptimize.c
and consist of bracketing with #ifdef and #endif code that rewrites instructions to their QUICK
forms. The modifications for the first experiment are shown in ’104 Appendix 1.
31.

The modifications for the second experiment (neither side tables nor quickening)

included all the changes for the first experiment, plus additional changes to
dalvik/vm/analysis/DexOptimize.c
and some changes to
dalvik/vm/oo/Resolve.c

8

32.

The changes for the second experiment also consist of bracketing with #ifdef and

#endif code that builds the side tables of resolved constant pool entries. The modifications for
the second experiment are shown in ’104 Appendix 2.
33.

Each workspace was compiled with
$ make -j2

34.

The benchmarks used to test the performance of the workspaces were
(1) CaffeineMark™ 3.0 * (http://www.benchmarkhq.ru/cm30)
(2) SciMark 2.0 (http://math.nist.gov/scimark2)
(3) kBench (Sun/Oracle Internal Benchmark)

35.

All the benchmarks were converted from standard jar files to dex jar files with,

for example
$ dx --dex --verbose --output=scimark-dex.jar scimark.jar
36.

I tested the performance of the repositories by running on the supplied Android

emulator with
# dalvikvm -Xint:fast -cp /data/app/cm3-dex.jar CaffeineMarkEmbeddedApp
# dalvikvm -Xint:fast -cp /data/app/scimark-dex.jar scimark 2.0
# dalvikvm -Xint:fast -cp /data/app/kBench-dex.jar RunAll
I did not try running the trace compiler, because an examination of the source code indicated that
it would not run without the side tables of resolved constant pool entries.
37.

Each of the three workspaces ran all three of the benchmarks. I ran 10 iterations

of each benchmark.
38.

Separately, I built new workspaces with the modified source files, and ran them

on a Beagleboard.
39.

The accompanying charts attached as Exhibit E record the results of runs with the

-Xint:fast command line option on the Beagleboard.

9