To run JS in separate cooperative threads, we need to split up per-thread state
from per-runtime state. This patch does that for XPConnect.
MozReview-Commit-ID: 407SlJ7nR6v
It's fairly straightforward, and measures the important parts of:
- Sampler, PseudoStack, ProfileBuffer, ThreadInfo.
- LUL, PriMap, SecMap
Coverage isn't perfect, but it gets the major things I found via DMD on Linux.
Example output in about:memory:
├──151.21 MB (49.73%) -- profiler
│ ├──141.49 MB (46.53%) ── lul
│ └────9.72 MB (03.20%) ── sampler
This patch removes checking of all the callback calls in memory reporter
CollectReport() functions, because it's not useful.
The patch also does some associated clean-up.
- Replaces some uses of nsIMemoryReporterCallback with the preferred
nsIHandleReportCallback typedef.
- Replaces aCallback/aCb/aClosure with aHandleRepor/aData for CollectReports()
parameter names, for consistency.
- Adds MOZ_MUST_USE/[must_use] in a few places in nsIMemoryReporter.idl.
- Uses the MOZ_COLLECT_REPORT macro in all suitable places.
Overall the patch reduces code size by ~300 lines and reduces the size of
libxul by about 37 KiB on my Linux64 builds.
This patch makes the following changes on many in-class methods.
- NS_METHOD F() override; --> NS_IMETHOD F() override;
- NS_METHOD F() override {...} --> NS_IMETHOD F() override {...}
- NS_METHOD F() final; --> NS_IMETHOD F() final;
- NS_METHOD F() final {...} --> NS_IMETHOD F() final {...}
Using NS_IMETHOD is the preferred way of marking in-class virtual methods.
Although these transformations add an explicit |virtual|, they are safe --
there's an implicit |virtual| anyway because |override| and |final| only work
with virtual methods.
This patch makes most Run() declarations in subclasses of nsIRunnable have the
same form: |NS_IMETHOD Run() override|.
As a result of these changes, I had to add |override| to a couple of other
functions to satisfy clang's -Winconsistent-missing-override warning.
We have some oddities in our jemalloc stats reporting.
- "heap-overhead-ratio" is a strange measurement: overhead / non-overhead,
expressed as a percentage. And it omits "bin_unused", which appears to be an
oversight.
- "heap-committed" also omits "bin_unused".
- There are some minor errors in memory report descriptions.
This patch fixes these and improves the heap reporting. It makes the following
reporting changes:
- "heap-allocated": Duplicated as "heap-committed/allocated". (We keep
"heap-allocated" because that's a special value used in the computation of
"heap-unclassified".)
- "heap-committed/overhead": Added; it's the same as the sum of the
"explicit/heap-overhead/*" values. Together with "heap-committed/allocated"
it shows clearly what fraction of the heap is overhead and what fraction is
useful.
- "heap-committed": Removed; now implicit as the "heap-committed/" node.
- "heap-overhead-ratio":
- Removed from memory reports; now shown as the percentage of the new
"heap-committed/overhead" node.
- Still available as a distinguished amount (because it's useful in
isolation) but renamed to heapOverheadFraction, and the telemetry ID is
renamed as MEMORY_HEAP_OVERHEAD_FRACTION.
- "heap-chunks": Removed; it's not that interesting, and can be manually
computed as "heap-mapped" / "heap-chunksize" if necessary.
The WindowsAddressSpaceReporter generates one memory report per segment, and
there can be 10,000+ segments.
This patch changes things so that one memory report is generated per segment
*kind* -- at most a couple of dozen -- rather than one per *segment*.
The bulk of this commit was generated with a script, executed at the top
level of a typical source code checkout. The only non-machine-generated
part was modifying MFBT's moz.build to reflect the new naming.
CLOSED TREE makes big refactorings like this a piece of cake.
# The main substitution.
find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \
xargs perl -p -i -e '
s/nsRefPtr\.h/RefPtr\.h/g; # handle includes
s/nsRefPtr ?</RefPtr</g; # handle declarations and variables
'
# Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h.
perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h
# Handle nsRefPtr.h itself, a couple places that define constructors
# from nsRefPtr, and code generators specially. We do this here, rather
# than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename
# things like nsRefPtrHashtable.
perl -p -i -e 's/nsRefPtr/RefPtr/g' \
mfbt/nsRefPtr.h \
xpcom/glue/nsCOMPtr.h \
xpcom/base/OwningNonNull.h \
ipc/ipdl/ipdl/lower.py \
ipc/ipdl/ipdl/builtin.py \
dom/bindings/Codegen.py \
python/lldbutils/lldbutils/utils.py
# In our indiscriminate substitution above, we renamed
# nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up.
find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \
xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g'
if [ -d .git ]; then
git mv mfbt/nsRefPtr.h mfbt/RefPtr.h
else
hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h
fi
|getReportsForThisProcess| differs from |getReports| in that it is limited to current process and is synchronous. When asynchronous memory reporters are added the function will no longer be able tobe synchronous. There isn't much utility in only measuring the current process, so we can remove the function and switch existing users to |getReports|.
The calculation of |explicit| relies on the synchronous |getReportsForThisProcess|, once we have asynchronous reporters this will no longer work. As it is currently referenced in the about::memory tests we can just remove it.
|getReportsForThisProcess| differs from |getReports| in that it is limited to current process and is synchronous. When asynchronous memory reporters are added the function will no longer be able tobe synchronous. There isn't much utility in only measuring the current process, so we can remove the function and switch existing users to |getReports|.
The calculation of |explicit| relies on the synchronous |getReportsForThisProcess|, once we have asynchronous reporters this will no longer work. As it is currently referenced in the about::memory tests we can just remove it.
Jemalloc 4 purges dirty pages regularly during free() when the ratio of dirty
pages compared to active pages is higher than 1 << lg_dirty_mult. We set
lg_dirty_mult in jemalloc_config to limit RSS usage, but it also has an impact
on performance.
So instead of enforcing a high ratio to force more pages being purged, we keep
jemalloc's default ratio of 8, and force a regular purge of all dirty pages,
after cycle collection.
Keeping jemalloc's default ratio avoids cycle-collection-triggered purge to
have to go through really all dirty pages when there are a lot, in which case
the normal jemalloc purge during free() will already have kicked in. It also
takes care of everything that doesn't run the cycle collector still having
a level of purge, like plugins in the plugin-container.
At the same time, since jemalloc_purge_freed_pages does nothing with jemalloc 4,
repurpose the MEMORY_FREE_PURGED_PAGES_MS telemetry probe to track the time
spent in this cycle-collector-triggered purge.
