Make it possible to use online profile merging ("%m" mode) with
continuous sync ("%c" mode).
To implement this, the merged profile is locked in the runtime
initialization step and either a) filled out for the first time or b)
checked for compatibility. Then, the profile can simply be mmap()'d with
MAP_SHARED set. With the mmap() in place, counter updates from every
process which uses an image are mapped onto the same set of physical
pages assigned by the filesystem cache. After the mmap() is set up, the
profile is unlocked.
Differential Revision: https://reviews.llvm.org/D69586
Add support for continuously syncing profile counter updates to a file.
The motivation for this is that programs do not always exit cleanly. On
iOS, for example, programs are usually killed via a signal from the OS.
Running atexit() handlers after catching a signal is unreliable, so some
method for progressively writing out profile data is necessary.
The approach taken here is to mmap() the `__llvm_prf_cnts` section onto
a raw profile. To do this, the linker must page-align the counter and
data sections, and the runtime must ensure that counters are mapped to a
page-aligned offset within a raw profile.
Continuous mode is (for the moment) incompatible with the online merging
mode. This limitation is lifted in https://reviews.llvm.org/D69586.
Continuous mode is also (for the moment) incompatible with value
profiling, as I'm not sure whether there is interest in this and the
implementation may be tricky.
As I have not been able to test extensively on non-Darwin platforms,
only Darwin support is included for the moment. However, continuous mode
may "just work" without modification on Linux and some UNIX-likes. AIUI
the default value for the GNU linker's `--section-alignment` flag is set
to the page size on many systems. This appears to be true for LLD as
well, as its `no_nmagic` option is on by default. Continuous mode will
not "just work" on Fuchsia or Windows, as it's not possible to mmap() a
section on these platforms. There is a proposal to add a layer of
indirection to the profile instrumentation to support these platforms.
rdar://54210980
Differential Revision: https://reviews.llvm.org/D68351
to reflect the new license.
We understand that people may be surprised that we're moving the header
entirely to discuss the new license. We checked this carefully with the
Foundation's lawyer and we believe this is the correct approach.
Essentially, all code in the project is now made available by the LLVM
project under our new license, so you will see that the license headers
include that license only. Some of our contributors have contributed
code under our old license, and accordingly, we have retained a copy of
our old license notice in the top-level files in each project and
repository.
llvm-svn: 351636
Introduces a 'owner' struct to include the overridable write
method and the write context in C.
This allows easy introdution of new member API to help reduce
profile merge time in the follow up patch.
llvm-svn: 306432
This reverts commit r268840, as it breaks Thumb2 self-hosting. There is something
unstable in the profiling for Thumb2 that needs to be sorted out before we continue
implementing these changes to the profiler. See PR27667.
llvm-svn: 268864
Compiler-rt miscalculates the number of entries in the __llvm_prf_data section
on i386 Darwin. This results in a number of test failures (which we started
catching after r261344).
The fix we attempted earlier is insufficient (r261683). It caused some tests to
start passing again, but that hid the fact that we drop some data entries.
This patch should fix the real problem. It fixes the way we compute DataSize by
taking into account the way the Darwin linker lays out __llvm_prf_data.
Differential Revision: http://reviews.llvm.org/D17623
llvm-svn: 261957
(patch suggested by silvas)
With this patch, the IO information is wrapped in struct
ProfDataIOVec, and interface of writerCallback takes a vector
of IOVec and a pointer to writer context pointer.
Differential Revision: http://reviews.llvm.org/D14859
llvm-svn: 253764
This change adds extends the data structures and adds in the routines
for handling runtime calls for value profiling. The profile data format
is modified and the version number is incremented.
llvm-svn: 253483
- This is to handle a corner case where profile lib is linked
in but non of the modules are instrumented (On linux, since
we avoided the overhead to emit runtime hook use functions so
this is the side effect of that size optimization).
- Added a profile runtime test case to cover all scenarios of
shared library builds.
Differential Revision: http://reviews.llvm.org/D14468
llvm-svn: 253098
This patch introduces a well defined header struct
to represent raw profile header instead of using raw array.
Previously the raw array is used in two different files and
is very error prone when header structure is re-organized.
This is a small cleanup with NFC.
llvm-svn: 250561
On Darwin, compiler_rt uses magic linker symbols to find the profile
counters in the __DATA segment. This is a reasonable method for
normal, hosted, userspace programs. However programs with custom
memory layouts, such as the kernel, will need to tell compiler_rt
explicitly where to find these sections.
Patch by Lawrence D'Anna. Thanks!
llvm-svn: 223840
The naming scheme we're using for counters in profile data shares a
prefix with some fixed names we use for the runtime, notably
__llvm_profile_data_begin and _end. Embarrassingly, this means a
function called begin() can't be instrumented.
This modifies the runtime names so as not to collide with the
instrumentation.
llvm-svn: 217166
Change the API of the instrumented profiling library to work with shared
objects.
- Most things are now declared hidden, so that each executable gets
its own copy.
- Initialization hooks up a linked list of writers.
- The raw format with shared objects that are profiled consists of a
concatenated series of profiles. llvm-profdata knows how to deal
with that since r208938.
<rdar://problem/16918688>
llvm-svn: 208940
GCC -pedantic warns that the initialization of Header is not constant:
InstrProfilingFile.c:31:5: error: initializer element is not computable at load time [-Werror]
LLVM defaults to enabling -pedantic. If this warning is unhelpful, we
can consider revisiting that decision.
llvm-svn: 207784
This is a bit of a stab in the dark as I'm not sure I've got these
source files compiling correctly locally. (and the warning only
reproduces on a 32bit build anyway)
llvm-svn: 204521
Write __llvm_profile_write_buffer(), which uses the same logic as
__llvm_profile_write_file(), but writes directly to a provided `char*`
buffer instead.
<rdar://problem/15943240>
llvm-svn: 204499
It was misguided to plan to rely on __llvm_profile_write_buffer() in
__llvm_profile_write_file(). It's less complex to duplicate the writing
logic than to mmap the file.
Since it's here to stay, move `FILE*`-based writing logic into
InstrProfilingFile.c.
<rdar://problem/15943240>
llvm-svn: 204498
Move functions around to prepare for some other changes.
- Merge InstrProfilingExtras.h with InstrProfiling.h. There's no
benefit to having these split.
- Rename InstrProfilingExtras.c to InstrProfilingFile.c.
- Split actual buffer writing code out of InstrProfiling.c into
InstrProfilingBuffer.c.
- Drive-by corrections of a couple of header comments.
<rdar://problem/15943240>
llvm-svn: 204497
Vedant Kumar