Currently, in NetworkStats, there are many methods to manipulate
the records. However, some methods are similar and ambiguous,
such as addEntry, addValues, setValues, addIfaceValues,
combineValues and combineAllValues.
Thus, properly grouping and renaming methods are necessary.
In this change, for methods that add one record unconditionally,
name them insertEntry.
setIfaceValues -> insertEntry
addEntry -> insertEntry
Test: atest FrameworksNetTests ImsPhoneCallTrackerTest TetheringTests
Test: m doc-comment-check-docs
Bug: 148895143
Change-Id: I801ddc49e283a254b9586700c9b597c0adb0d459
In order to let external module report their network stats,
expose necessary APIs to construct NetworkStats object.
Test: atest FrameworksNetTests CtsUsageStatsTestCases
Test: m doc-comment-check-docs
Bug: 130855321
Change-Id: Id3ec8aaff3df67948c25eac2319a74cf33a27979
This reverts commit 95aa6d446f.
Reason for revert: This change has been implicated in 4-way deadlocks as seen in b/134244752.
Bug: 134244752
Change-Id: I2f1839d7776a613ca571af8a542755ddc5fc8760
Merged-In: Ibdaad3a4cbf0d8ef1ed53cfab1e454b9b878bae9
This reverts commit 917327b193.
Reason for revert: This change has been implicated in 4-way deadlocks as seen in b/134244752.
Bug: 134244752
Change-Id: Ibdaad3a4cbf0d8ef1ed53cfab1e454b9b878bae9
VPN uid.
(cherry picked from commit 95aa6d446f)
Bug: 113122541
Bug: 120145746
Test: atest FrameworksNetTests
Test: Manually verified on device that stats from VPN UID are moved
appropriately based on its declared underlying network set.
Test: vogar --mode app_process --benchmark NetworkStatsBenchmark.java
Change-Id: I7f368c5970b2dcb969fe0daf5ef44edb1f51d09d
In order to notify netd to swap eBPF maps before pulling the
networkStats from eBPF maps, NetworkStatsFactory need to use the
NetdServices to issue binder calls. So it need to be moved from
framework/base/core to framework/base/service since object in
framework/base/core cannot get any system services. This change is also
necessary for setting up a lock inside NetworkStatsFactory to prevent
racing between two netstats caller since the lock need to be hold before
netd trigger the map swap.
Also fix the compile problem caused by moving the NetworkStatsFactory
and the related tests. Rename the packages and the jni functions to a
more proper name.
Bug: 124764595
Bug: 128900919
Test: NetworkStatsFactoryTest
android.app.usage.cts.NetworkUsageStatsTest
android.net.cts.TrafficStatsTest
Change-Id: Ifcfe4df81caf8ede2e4e66a76552cb3200378fa8
In order to notify netd to swap eBPF maps before pulling the
networkStats from eBPF maps, NetworkStatsFactory need to use the
NetdServices to issue binder calls. So it need to be moved from
framework/base/core to framework/base/service since object in
framework/base/core cannot get any system services. This change is also
necessary for setting up a lock inside NetworkStatsFactory to prevent
racing between two netstats caller since the lock need to be hold before
netd trigger the map swap.
Also fix the compile problem caused by moving the NetworkStatsFactory
and the related tests. Rename the packages and the jni functions to a
more proper name.
Bug: 124764595
Bug: 128900919
Test: NetworkStatsFactoryTest
android.app.usage.cts.NetworkUsageStatsTest
android.net.cts.TrafficStatsTest
Merged-In: Ifcfe4df81caf8ede2e4e66a76552cb3200378fa8
Change-Id: Ifcfe4df81caf8ede2e4e66a76552cb3200378fa8
VPN uid.
Bug: 113122541
Bug: 120145746
Test: atest FrameworksNetTests
Test: Manually verified on device that stats from VPN UID are moved
appropriately based on its declared underlying network set.
Test: vogar --mode app_process --benchmark NetworkStatsBenchmark.java
Change-Id: I9d8d0cc58d18002c1c96f8ddff780ef8dc452d21
Moving ResourcesBenchmark.java to the proper android.perftests infrastructure.
Test: ran the benchmark locally.
Change-Id: Ia981274e1e3c167a2a8900498fc40b7a03508a74
We've been advocating as little usage of streams as possible inside
the system_server, and this is a benchmark showing the costs of using
streams compared to rolling manual logic.
Results on a taimen show that direct route is 2.2x faster (!).
Trial Report (1 of 2):
Experiment {instrument=runtime, benchmarkMethod=timeDirect, vm=app_process, parameters={}}
Results:
runtime(ns): min=47234.28, 1st qu.=47234.28, median=47234.28, mean=47234.28, 3rd qu.=47234.28, max=47234.28
Trial Report (2 of 2):
Experiment {instrument=runtime, benchmarkMethod=timeStreams, vm=app_process, parameters={}}
Results:
runtime(ns): min=105660.39, 1st qu.=105660.39, median=105660.39, mean=105660.39, 3rd qu.=105660.39, max=105660.39
Bug: 77808546
Test: vogar --mode app_process --benchmark frameworks/base/core/tests/benchmarks/src/android/util/StreamsBenchmark.java
Change-Id: Iac43a75970eee2c91593316a1b5a45d2c96c476b
Also extend API to accept a "count" argument of exactly how many
bytes to copy, and return the actual number of copied bytes.
Improve docs.
Test: bit FrameworksCoreTests:android.os.FileUtilsTest
Test: vogar --mode app_process --benchmark frameworks/base/core/tests/benchmarks/src/android/os/FileUtilsBenchmark.java
Bug: 71932978
Change-Id: I8d255e4f97462838c02a8ecb6d2d221188c4eff0
There are several places across the OS where Java code is simply
copying data between two points, which requires bringing that data
out into userspace before going back into the kernel. (That's pretty
lame.) The patches for the recent Meltdown/Spectre security issues
have made this overhead even worse, so it's finally time to move this
copying directly into the kernel.
This change adds a couple new FileUtils.copy() methods which inspect
the given streams/FDs, and attempt to do as much optimization as
possible before falling back to a slower userspace-based copy.
Benchmarks are showing typical improvements of 44% for 32KB files,
50% for 32MB files, and 35% for 32MB pipes.
Plenty of tests are included, and there's a simple kill-switch that
can be used to enable/disable the feature if it starts causing any
trouble. (A future CL will enable the optimizations.)
Test: bit FrameworksCoreTests:android.os.FileUtilsTest
Test: vogar --mode app_process --benchmark frameworks/base/core/tests/benchmarks/src/android/os/FileUtilsBenchmark.java
Bug: 71932978
Change-Id: I52518d529da5d961610998b9f61399064d8025cd
Add the native method used to read the detail information of network
stats from bpf maps. The native method of NetworkStatsFactory should
choose the correct implementation to get the stats detail depending on
the kernel version. Currently the bpf result is printed as a reference
and the actual behavior of NetworkStatsFactory should not change.
Test: NetworkStatsFactory related cts test should not fail.
Bug: 30950746
Change-Id: I4715a23559b5b2306bd556cea0431f0ed172a993
* No new functionality was added, this was already possible via
-Djava.class.path=classpath
* However, this makes it a bit more compatible with other tools
that use -cp to invoke the main class.
* Also update the benchmarks README for correct up-to-date vogar
instructions.
Test: vogar --mode app_process --benchmark frameworks/base/core/tests/benchmarks/src/android/os/ParcelBenchmark.java
Bug: 31807538
Change-Id: Idb04600fed3dd955437ccac832617dcfd1b52b63
SpannableStringBuilder used to return the result of getSpans in the
order of start indices because of the interval tree it uses. However,
style spans has to be applied in the order of insertion to get
predictable results. Sorted the results of getSpans ordered first by
priority flag and then by insertion time. Moreover improved the
performance of SpannableStringInternal copy constructor.
Bug: 26240132
Change-Id: I0b0fa7bb30a3bd9ca37dedca66d8993718586027
After previous change removed LOCAL_NO_STANDARD_LIBRARIES:=true
it is no longer necessary to explicitly add framework to
LOCAL_JAVA_LIBRARIES as it is added automatically by the build.
Change-Id: I68bf700375e34a10302f76d7463110755b815f85
Added an Android.mk file to core/tests/benchmarks to build the
benchmarks against caliper-api-target to ensure that they build
correctly. Changed the implementations to work with the new
version of Caliper.
Bug: 24848946
Change-Id: Ib28026816dbc5c1d54f65974f4a1abfee299c43f
On modern versions of Android running in AOT mode
FloatMath is slower than Math. Calls to Math.sqrt(),
etc. are replaced by intrinsics which can be as small
as a single CPU opcode.
When running in interpreted mode the new
implementation is unfortunately slower, but I'm
judging this acceptable and likely to be improved
over time. This change saves a small amount of native
code.
Example timings:
Mako AOSP AOT:
Method: Original / New / Direct call to Math
ceil: 596ns / 146.ns / 111ns
sqrt: 694ns / 56ns / 25ns
Mako AOSP interpreted:
Method: Original / New / Direct call to Math
ceil: 1900ns / 2307ns / 1485ns
sqrt: 1998ns / 2603ns / 1788ns
Other calls Mako AOT:
Method: Original / New
cos: 635ns / 270ns
exp: 566ns / 324ns
floor: 604ns / 150ns
hypot: 631ns / 232ns
pow: 936ns / 643ns
sin: 641ns / 299ns
The advice to use Math directly, in preference to
FloatMath, is still good. FloatMath will be deprecated
separately.
Bug: https://code.google.com/p/android/issues/detail?id=36199
Change-Id: If07fcbd78543d13bc6d75f9743f999860e8d58d7
Every incoming Binder call ends up triggering a BlockGuardPolicy
update, which would thrash between new AndroidBlockGuardPolicy
instances and BlockGuard.LAX_POLICY, causing GC churn.
This change avoids the extra allocations by recycling a single
AndroidBlockGuardPolicy in a ThreadLocal. Worst-case thrashing case
is now 10% faster; from 7.46us to 6.65us.
Bug: 9424568
Change-Id: I9c3b1c097a2aecc9b1f109a824cf3ea319fb3393
Current timings on a mako listed below, which show minimal overhead
compared to writing directly. The increased code readability and
automatic indenting is worth the overhead.
Also worth noting is that writing concatenated strings is
substantially faster than printing components separately.
benchmark us linear runtime
ComplexDirect 10712.7 =========================
ComplexIndenting 12623.7 ==============================
PairIndenting 45.8 =
PairRaw 82.1 =
SimpleDirect 282.2 =
SimpleIndenting 294.4 =
Change-Id: I7c38690c14b017fa46248ebb0be69f6beff03371
Switch to parsing detailed network stats with native code, which
is 71% faster than ProcFileReader.
Change-Id: I2525aaee74d227ce187ba3a74dd08a2b06514deb
Deprecate transport layer statistics, leaving only the summarized
network layer statistics.
Improve documentation to be clear about layers where measurements
occur, and their behavior since boot. Under the hood, move to using
xt_qtaguid UID statistics.
Bug: 6818637, 7013662
Change-Id: I9f26992e5fcdebd88c671e5765bd91229e7b0016
