The URL is:-->http://peacekeeper.futuremark.com/run.action Actually i was testing this on android emulator(android default browser).
With V8-3.13.3.0:-->This peacekeeper Benchmark was working.
With V8-3.14.0.0:-->This peacekeeper Benchmark is not working.
i have even checked with latest V8-3.14.5.0 & peacekeeper Benchmark is not
working.Any Insights on this please..
I'm surprised that the default Android browser even compiles with a current
V8 version (without your changes it's using 3.9, right?). Usually there are
API changes every now and then that require modifications in the browser
and/or WebKit. Sorry, but I guess you're pretty much on your own with this
configuration.
If you find something that doesn't work in a configuration that's being
shipped (e.g. Chrome tip-of-tree, which includes an up-to-date V8 version),
we'd be very much interested!
Even, I have experienced this problem when upgrading to 3.14.5.1 from
3.11.10.23. The peacekeeper benchmark fails for the HTML5 - WebGl test case.
The link to individual test case is http://goo.gl/mDcYi The Fps reported as a result comes as NaN now instead of float.
I have also seen this issue with V8 3.13.7.5
This failure is seen in the below mentioned chrome Version also:-->
Google Chrome 22.0.1229.79 (Official Build 158531) m
OS Windows
WebKit 537.4 (@129177)
JavaScript V8 3.12.19.11
Flash 11.3.31.331
User Agent Mozilla/5.0 (Windows NT 5.1) AppleWebKit/537.4 (KHTML, like
Gecko) Chrome/22.0.1229.79 Safari/537.4
Command Line "D:\Documents and Settings\Nagarjuna Atluri\Local
Settings\Application Data\Google\Chrome\Application\chrome.exe"
--flag-switches-begin --flag-switches-end
Executable Path D:\Documents and Settings\Nagarjuna Atluri\Local
Settings\Application Data\Google\Chrome\Application\chrome.exe
Profile Path D:\Documents and Settings\Nagarjuna Atluri\Local
Settings\Application Data\Google\Chrome\User Data\Default
Sorry for duplicating the Issue, i thought this Issue was closed(as the
status indicated was close) hence logged one more bug.
I have Corned down the v8 versions which is causing this Issue
v8 Version 3.13.5.0:--> this Issue is not seen
V8 Version 3.13.6:--> this Issue is Reproducible
Between these 2 versions check ins done are as below:-->
Revision: 12448
Author: jkumme...@chromium.org
Date: Wednesday, September 05, 2012 10:14:50 PM
Message:
Merged r12434, r12435, r12440, r12441, r12443, r12444, r12445, r12446 into
trunk branch.
Support register as right operand in min/max support.
Fixed test expectation.
Add build system infrastructure for ENABLE_EXTRA_CHECKS flag (not used yet)
Fix missing colon in common.gypi
Push stacktrace and die if the receiver is of unknown type.
Disable accessor inlining (due to broken deopts)
Add empty-handle checks to API functions (#ifdef ENABLE_EXTRA_CHECKS)
Check the return value of API calls on ia32 and x64.
R=yang...@chromium.org
Review URL: https://chromiumcodereview.appspot.com/10910093 ----
Modified : /trunk/src/api.cc
Modified : /trunk/src/flag-definitions.h
Modified : /trunk/src/isolate.cc
Modified : /trunk/src/objects.cc
Modified : /trunk/src/version.cc
Modified : /trunk/build/common.gypi
Modified : /trunk/src/ia32/macro-assembler-ia32.cc
Modified : /trunk/src/x64/lithium-codegen-x64.cc
Modified : /trunk/src/x64/macro-assembler-x64.cc
Modified : /trunk/src/api.h
Modified : /trunk/src/isolate.h
Modified : /trunk/Makefile
Modified : /trunk/src/checks.h
Added : /trunk/test/mjsunit/regress/regress-crbug-145961.js
Revision: 12436
Author: verwa...@chromium.org
Date: Tuesday, September 04, 2012 3:28:32 PM
Message:
Merged r12413, r12414 into trunk branch.
Elements load depends on the type of the receiver.
Seems to be the above mentioned patches are causing this Issue.
Root Cause seems to be :--> The Fps reported as a result comes as NaN now
instead of float
Finally seems to be the root cause is found.
With 3.13.7.0 v8 version the peacekeeper site works(but later versions does
not work)
Changes made are:-->
1)es5_readonly flag to false(flag-definitions.h)
DEFINE_bool(es5_readonly, false,
"activate correct semantics for inheriting readonliness")
2)In EmitDeepCopy function removed static casting(lithium-codegen-arm.cc)
// We only support little endian mode...
int32_t value_low = value & 0xFFFFFFFF;
int32_t value_high = value >> 32;
Can any one let me know why these changes are causing this Issue
In each of your comments, you're saying something that contradicts your
previous statements. In comment #6, you said that 3.13.5 worked but 3.13.6
was broken. Now in comment #7 you're saying that 3.13.7 (which came right
after 3.13.6!) works, but all later versions are broken. Can you reproduce
even one of these findings reliably, or does it just randomly sometimes
fail and sometimes work? Could this be an issue on your system? Maybe the
Android simulator doesn't report/detect/simulate framerates correctly?
Maybe your graphics driver is broken?
Also, for the record, NaN is a valid floating-point value. Seeing NaNs when
you're expecting numbers *might* hint at a bug in V8, but might just as
well be a bug in the JavaScript application (e.g., the division 0/0 returns
NaN).
I guess u din't understand my Comments.
What i meant was after the changes done what i have mentioned in Comment#7
the peacekeeper site works fine.
i have basically backtracked the patches based on version & found that the
2 chnages mentioned are causing this Issue
I guess u din't understand my Comments.
What i meant was after the changes done what i have mentioned in Comment#7
the peacekeeper site works fine.
i have basically backtracked the patches based on version & found that the
2 changes mentioned are causing this Issue.Once these changes are done in
3.13.6.0 & 3.13.7.0 this Issue is not seen.
I guess this clarifies your doubt
I still don't know what you're doing -- Peacekeeper is a browser benchmark,
and you never even told us which browser you're using.
I still can't reproduce any issues.
The changes you listed that you claim made your issue go away don't seem
related at all.
Best guess: the Android emulator doesn't properly support measuring
framerates, or the Peacekeeper benchmark doesn't properly support whatever
the emulator does.
At this point, I don't see any indication that there's a V8 issue at play
here that the team should spend time on.
