The main issue was due to multiple recompilations of functions. Now
code objects are grouped by function using SFI object address.
JSFunction objects are no longer tracked, instead we track SFI object
moves. To pick a correct code version, we now sample return addresses
instead of JSFunction addresses.
tools/{linux|mac|windows}-tickprocessor scripts differentiate
between code optimization states for the same function
(using * and ~ prefixes introduced earlier).
DevTools CPU profiler treats all variants of function code as
a single function.
ll_prof treats each optimized variant as a separate entry, because
it can disassemble each one of them.
tickprocessor.py not updated -- it is deprecated and will be removed.
BUG=v8/1087,b/3178160
TEST=all existing tests pass, including Chromium layout tests
Review URL: http://codereview.chromium.org/6551011
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@6902 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The problem appeared due to a fact that stubs doesn't create a stack
frame, reusing the stack frame of the caller function. When building
stack traces, the current function is retrieved from PC, and its
callees are retrieved by traversing the stack backwards. Thus, for
stubs, the stub itself was discovered via PC, and then stub's caller's
caller was retrieved from stack.
To fix this problem, a pointer to JSFunction object is now captured
from the topmost stack frame, and is saved into stack trace log
record. Then a simple heuristics is applied whether a referred
function should be added to decoded stack, or not, to avoid reporting
the same function twice (from PC and from the pointer.)
BUG=553
TEST=added to mjsunit/tools/tickprocessor
Review URL: http://codereview.chromium.org/546089
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3673 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
As this is only needed for internal profiling (not for DevTools),
the following approach had been chosen:
- during snapshot creation, positions of serialized objects inside
a snapshot are logged;
- then during V8 initialization, positions of deserealized objects
are logged;
- those positions are used for retrieving code objects names from
snapshot creation log, which needs to be supplied to tick processor
script.
Positions logging is controlled with the new flag: --log_snapshot_positions.
This flag is turned off by default, and this adds no startup penalty.
To plug this fix to Golem, the following actions are needed:
- logs created using 'mksnapshot' need to be stored along with VM images;
- tick processor script needs to be run with '--snapshot-log=...' cmdline
argument.
BUG=571
Review URL: http://codereview.chromium.org/551062
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3635 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Turned on with '--log-producers' flag, also needs '--noinline-new' (this is temporarily), '--log-code', '--log-gc'. Not all allocations are traced (I'm investigating.)
Stacks are stored using weak handles. Thus, when an object is collected, its allocation stack is deleted.
Review URL: http://codereview.chromium.org/267077
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3069 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
MSVS names '.map' file using only module's name, so both 'a.exe' and 'a.dll' will have 'a.map' file. To distinguish an originating module, we're now checking for image base which is always 00400000 for .exe files, and not 00400000 for .dlls.
Verified that windows-tick-processor can now process logs from Chromium using .map file generated for 'chrome.dll', an that it still works for V8's 'shell.exe'.
Review URL: http://codereview.chromium.org/172044
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2699 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
'nm' is now called with an option to report function code sizes. Static code entries are restricted to the sizes reported, and the remaining unnamed code is attributed to a library as a whole. This makes reports more accurate, as some functions are tiny, but has chunks of unnamed code behind them.
This change doesn't affect reporting on Windows, as in .map files function code sizes aren't specified.
Review URL: http://codereview.chromium.org/149513
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2455 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Without an explicit check if a function belongs to shared library address space, "finishing" a library symbols processing with 'addPrevEntry(libEnd);' can cause emission of code entries which cover almost the entire address space, shadowing other code.
Review URL: http://codereview.chromium.org/131033
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2221 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The main problem was due to the following: after Erik had fixed the logger to report library addresses, tickprocessor started to add to the code map entries that covered almost entire memory. This happened because tickprocessor contains a heuristic to bias addresses of functions from dynamic libraries:
if (funcInfo.start < libStart && funcInfo.start < libEnd - libStart) {
funcInfo.start += libStart;
}
And, as tickprocessor tried to process all symbols from the library, including data entries, which can be outside reported library addresses range, the second condition failed, and funcInfo.start remained unbiased.
Review URL: http://codereview.chromium.org/125192
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2194 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This is a trivial per-row compression:
- short aliases are introduced for events and code creation tags;
- in tick events, offsets are used instead of absolute addresses;
- removed 'code-allocation' event, as it seems not used.
The first two options are depend on the new flag: 'compress-log', which is off by default.
On benchmarks run w/o snapshot, this gives 45% log size reduction.
Review URL: http://codereview.chromium.org/119304
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@2122 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This is an effort to reuse profiler data processing code both in
TickProcessor and Dev Tools Profiler. The old Python implementation
will be removed.
The new TickProcessor works almost identical to the previous one.
However, it has some differences:
1. Not very useful "Call profile" section is replaced with a new
WebKit-like "Bottom up (heavy) profile" which shows the most
expensive functions together with their callers. I used it
personally in order to find and remove bottlenecks in the
tickprocessor script itself, and found it quite helpful.
2. Code entries with duplicate names (they occur for RegExes, stubs
and sometimes for anonymous Function objects) are now distinguished
by adding an occurence number inside curly brackets.
3. (Address -> code entry) mapping is more precise in boundary cases.
4. Windows version no more requires specifying .map file location.
5. Works faster.
Review URL: http://codereview.chromium.org/99054
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@1802 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
