slightly increases the performance. It also moderately improves
the nr of cases where helgrind can provide the stack trace of the old
access (when using the same amount of memory for the OldRef entries).
The patch also provides a new helgrind monitor command to show
the recorded accesses for an address+len, and adds an optional argument
lock_address to the monitor command 'info locks', to show the info
about just this lock.
Currently, oldref are maintained in a sparse WA, that points to N
entries, as specified by --conflict-cache-size=N.
For each entry (associated to an address), we have the last 5 accesses.
Old entries are recycled in an exact LRU order.
But inside an entry, we could have a recent access, and 4 very
old accesses that are kept 'alive' by a single thread accessing
repetitively the address shared with the 4 other old entries.
The attached patch replaces the sparse WA that maintains the OldREf
by an hash table.
Each OldRef now also only maintains one single access for an address.
As an OldRef now maintains only one access, all the entries are now
strictly in LRU mode.
Memory used for OldRef
-----------------------
For the trunk, an OldRef has a size of 72 bytes (on 32 bits archs)
maintaining up to 5 accesses to the same address.
On 64 bits arch, an OldRef is 104 bytes.
With the patch, an OldRef has a size of 32 bytes (on 32 bits archs)
or 56 bytes (on 64 bits archs).
So, for one single access, the new code needs (on 32 bits)
32 bytes, while the trunk needs only 14.4 bytes.
However, that is the worst case, assuming that the 5 entries in the
accs array are all used.
Looking on 2 big apps (one of them being firefox), we see that
we have very few OldRef entries that have the 5 entries occupied.
On a firefox startup, of the 5x1,000,000 accesses, we only have
1,406,939 accesses that are used.
So, in average, the trunk uses in reality around 52 bytes per access.
The default value for --conflict-cache-size has been doubled to 2000000.
This ensures that the memory used for the OldRef is more or less the
same as the trunk (104Mb for OldRef entries).
Memory used for sparseWA versus hashtable
-----------------------------------------
Looking on 2 big apps (one of them being firefox), we see that
there are big variations on the size of the WA : it can go in a few
seconds from 10MB to 250MB, or can decrease back to 10 MB.
This all depends where the last N accesses were done: if well localised,
the WA will be small.
If the last N accesses were distributed over a big address space,
then the WA will be big: the last level of WA (the biggest memory consumer)
uses slightly more than 1KB (2KB on 64 bits) for each '256 bytes' memory
zone where there is an oldref. So, in the worst case, on 32 bits, we
need > 1_000_000_000 sparseWA memory to keep 1_000_000 OldRef.
The hash table has between 1 to 2 Word overhead per OldRef
(as the chain array is +- doubled each time the hash table is full).
So, unless the OldRef are extremely localised, the overhead of the
hash table will be significantly less.
With the patch, the core arena total alloc is:
5299535/1201448632 totalloc-blocks/bytes
The trunk is
6693111/3959050280 totalloc-blocks/bytes
(so, around 1.20Gb versus 3.95Gb).
This big difference is due to the fact that the sparseWA repetitively
allocates then frees Level0 or LevelN when OldRef in the region covered
by the Level0/N have all been recycled.
In terms of CPU
---------------
With the patch, on amd64, a firefox startup seems slightly faster (around 1%).
The peak memory mmaped/used decreases by 200Mb.
For a libreoffice test, the memory decreases by 230Mb. CPU also decreases
slightly (1%).
In terms of correctness:
-----------------------
The trunk could potentially show not the most recent access
to the memory of a race : the first OldRef entry matching the raced upon
address was used, while we could have a more recent access in a following
OldRef entry. In other words, the trunk only guaranteed to find the
most recent access in an OldRef, but not between the several OldRef that
could cover the raced upon address.
So, assuming it is important to show the most recent access, this patch
ensures we really show the most recent access, even in presence of overlapping
accesses.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@15289
Activating this hint using --sim-hints=no-nptl-pthread-stackcache
means the glibc nptl stack cache will be disabled.
Disabling this stack/tls cache avoids helgrind false positive race conditions
errors when using __thread variables.
Note: disabling the stack cache is done by a kludge, dependent on
internal knowledge of glibc code, and using libpthread debug info.
So, this kludge might be broken with newer glibc version.
This has been tested on various platforms and various
glibc versions 2.11, 2.16 and 2.18
To check if the disabling works, you can do:
valgrind --tool=helgrind --sim-hints=no-nptl-pthread-stackcache -d -v ./helgrind/tests/tls_threads |& grep kludge
If you see the below 2 lines, then hopefully the stack cache has been disabled.
--12624-- deactivate nptl pthread stackcache via kludge: found symbol stack_cache_actsize at addr 0x3AF178
--12624:1:sched pthread stack cache size disabling done via kludge
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@14313
of memcheck and helgrind in a common module:
pub_tool_addrinfo.h pub_core_addrinfo.h m_addrinfo.c
At the same time, the factorised code is made usable by other
tools also (and is used by the gdbserver command 'v.info location'
which replaces the helgrind 'describe addr' introduced 1 week ago
and which is now callable by all tools).
The new address description code can describe more addresses
(e.g. for memcheck, if the block is not on the free list anymore,
but is in an arena free list, this will also be described).
Similarly, helgrind address description can now describe more addresses
when --read-var-info=no is given (e.g. global symbols are
described, or addresses on the stack are described as
being on the stack, freed blocks in the arena free list are
described, ...).
See e.g. the change in helgrind/tests/annotate_rwlock.stderr.exp
or locked_vs_unlocked2.stderr.exp
The patch touches many files, but is basically a lot of improvements
in helgrind output files.
The code changes are mostly refactorisation of existing code.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@13965
- Helgrind GDB server monitor command 'describe <address>'
allowing to describe an address (e.g. where it was allocated).
- Helgrind GDB server monitor command 'info locks' giving
the list of locks, their location, and their status.
In a further patch, it is intended to
1. factorise the describe address code between memcheck and helgrind
2. generalise the describe address gdbsrv command so as to make
it usable for all tools.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@13930
In case a lock order violation is detected in a multi lock cycle,
then the current code cannot produce the set of locks and the
stack traces involved in the cycle.
However, it is still possible to produce the stack trace of
the new lock and the other lock between which a cycle was discovered.
Also, add a comment in the code clarifying why the set of locks
establishing the required order cannot (currently) be produced.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@13281
#define N_FRAMES 8
(defined in libhb_core.c:3888)
implies that 'other thread' stack traces are limited to 8,
even with a bigger --num-callers.
=> document this in the manual to avoid that a user believes this is a
bug in the stack trace logic of Valgrind.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@12767
- Actually remove the dead docs/images/massif*.png files (this was meant to
happen in r10720).
- Inline $TOOL/docs/Makefile.am into $TOOL/Makefile.am for all 10 tools. 10
fewer Makefile.am files FTW!
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@10721
- Use "heap blocks" rather than "malloc'd blocks" as heap blocks covers
calloc, realloc, new, new[], memalign, etc.
- Used "GDB" and "GCC" throughout rather than "gcc" and "gdb".
- Made various tag uses more consistent.
- Greatly clarified the instructions on --xml=yes and its friends.
- Lots of other little improvements and fixes to out-of-date things and
Linux-centric things, mostly in Section 2.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@10701
line options. This commit changes them to all <option>.
Also make consistent how options with multiple names (eg. -h --help) are
shown.
Also, remove section describing --help and --version in Callgrind's chapter;
these aren't necessary and are presumably a hangover from when Callgrind was
a separate tool.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@10659
- There were detailed descriptions of all the tools in the Quick Start
Guide, the Manual introduction, and the start of each tool chapter. To
avoid duplication/overlap, I removed these altogether from the Quick Start
Guide, and shortened them in the intro.
- Improved the description of what errors Memcheck can find.
- Made all tool chapters start with "Overview" section, for consistency.
- Made the "run with --tool=XXX" bit consistent in each tool chapter.
- Made all tool chapter titles match the description given when running them.
- Added BBV to the User Manual intro.
- Generally clarified, updated, and future-proofed various bits of text in
the Quick Start Guide and User Manual introduction.
Also:
- Changed Nulgrind's start-up description to "the minimal Valgrind tool".
- Fixed some punctuation in the usage message.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@10652
converted by Donna. Hooked it into the build system so they are only
built when specifically asked for, and when doing "make dist".
They're not perfect; in particular, there are the following problems:
- The plain-text FAQ should be built from FAQ.xml, but this is not
currently done. (The text FAQ has been left in for now.)
- The PS/PDF building doesn't work -- it fails with an incomprehensible
error message which I haven't yet deciphered.
Nonetheless, I'm putting it in so others can see it.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@3153
dist". This avoids the need to put it in "EXTRA_DIST", and saves a few lines
in all the docs Makefile.am files.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@2614
the places that normal users will see:
- command line: --tool=foo (although --skin=foo still works)
- docs: removed all traces (included renaming coregrind_skins.html to
coregrind_tools.html)
- in the usage messages
- in error messages
Also did in in some places that I judged were unlikely to cause clashes with
existing workspaces:
- in the header comments of many files (eg. "This file is part of Memcheck, a
Valgrind tool for...")
- in the regtests script
- in the .supp files
- in AUTHORS
- in README_MISSING_SYSCALL_OR_IOCTL
Also update the AUTHORS file to mention Jeremy.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@2027
where I _should_ have put them in the first place, and fix up the
Makefile.am's accordingly. 'make' and 'make install' now work.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@1292
