its members into the base class.
First, to help motivate this kind of change, understand that in
a self-link, LLD creates 5.5 million defined regular symbol bodies (and
6 million symbol bodies total). A significant portion of its time is
spent allocating the memory for these symbols, and befor ethis patch
the defined regular symbol body objects alone consumed some 420mb of
memory during the self link.
As a consequence, I think it is worth expending considerable effort to
make these objects as memory efficient as possible. This is the first of
several components of that. This change starts with the goal of removing
the virtual functins from SymbolBody so that it can avoid having a vptr
embedded in it when it already contains a "kind" member, and that member
can be much more compact than a vptr.
The primary way of doing this is to sink as much of the logic that we
would have to dispatch for into data in the base class. As part of this,
I made the various flags bits that will pack into a bitfield with the
kind tag. I also sank the Name down to eliminate the dispatch for that,
and used LLVM's RTTI-style dispatch for everything else (most of which
is cold and so doesn't matter terribly if we get minutely worse lowering
than a vtable dispatch).
As I was doing this, I wanted to make the RTTI-dispatch (which would
become much hotter than before) as efficient as possible, so I've
re-organized the tags somewhat. Notably, the common case (regular
defined symbols) is now zero which we can test for faster.
I also needed to rewrite the comparison routine used during resolving
symbols. This proved to be quite complex as the semantics of the
existing one were very subtle due to the back-and-forth virtual dispatch
caused by re-dispatching with reversed operands. I've consolidated it to
a single function and tried to comment it quite a bit more to help
explain what is going on. However, this may need more comments or other
explanations. It at least passes all the regression tests. I'm not
working on Windows, so I can't fully test it.
With all of these changes, the size of a DefinedRegular symbol on
a 64-bit build goes from 80 bytes to 64 bytes, and we save approximately
84mb or 20% of the memory consumed by these symbol bodies during the
link.
The link time appears marginally faster as well, and the profile hotness
of the memory allocation subsystem got a bit better, but there is still
a lot of allocation traffic.
Differential Revision: http://reviews.llvm.org/D10792
llvm-svn: 241001
This uses a single cast and test to get the section for the symbol, and
uses the cast_or_null<> pattern throughout to handle the known type but
unknown non-null-ness.
No functionality changed.
Differential Revision: http://reviews.llvm.org/D10791
llvm-svn: 241000
Set debug location for terminator instruction in loop backedge block
(which is an unconditional jump to loop header). We can't copy debug
location from original backedges, as there can be several of them,
with different debug info locations. So, we follow the approach of
SplitBlockPredecessors, and copy the debug info from first non-PHI
instruction in the header (i.e. destination block).
This is yet another change for PR23837.
llvm-svn: 240999
This flattens the entire liveness walk from a recursive mark approach to
a worklist approach. It also sinks the worklist management completely
out of the SectionChunk and into the Writer by exposing the ability to
iterato over children of a chunk and over the symbol bodies of relocated
symbols. I'm not 100% happy with the API names, so suggestions welcome
there.
This allows us to use a single worklist for the entire recursive walk
and would also be a natural place to take advantage of parallelism at
some future point.
With this, we completely inline away the GC walk into the
Writer::markLive function and it makes it very easy to profile what is
slow. Currently, time is being wasted checking whether a Chunk isa
SectionChunk (it essentially always is), finding (or skipping)
a replacement for a symbol, and chasing pointers between symbols and
their chunks. There are a bunch of things we can do to fix this, and its
easier to do them after this change IMO.
This change alone saves 1-2% of the time for my self-link of lld.exe
(which I'm running and benchmarking on Linux ironically).
Perhaps more notably, we'll no longer blow out the stack for large
links. =]
Just as an FYI, at this point, I/O is starting to really dominate the
profile. Well over 10% of the time appears to be inside the kernel doing
page table silliness. I think a decent chunk of this can be nuked as
well, but it's a little odd as cross-linking in this way isn't really
the primary goal here.
Differential Revision: http://reviews.llvm.org/D10790
llvm-svn: 240995
This matches the implementation of the gcc support for the same
feature, including checking the values set up by libgcc at runtime.
The structure looks like this:
unsigned int __cpu_vendor;
unsigned int __cpu_type;
unsigned int __cpu_subtype;
unsigned int __cpu_features[1];
with a set of enums to match various fields that are field out after
parsing the output of the cpuid instruction.
This also adds a set of errors checking for valid input (and cpu).
compiler-rt support for this and the other builtins in this family
(__builtin_cpu_init and __builtin_cpu_is) are forthcoming.
llvm-svn: 240994
This fixes test issues with building lldb/test/api/multithreaded and a few other tests that build against the LLDB.framework in our build directory.
llvm-svn: 240993
Summary: This patch fixes the cases of sext/zext constant folding in DAG combiner where constans do not fit 64 bits. The fix simply removes un$
Test Plan: New regression test included.
Reviewers: RKSimon
Reviewed By: RKSimon
Subscribers: RKSimon, llvm-commits
Differential Revision: http://reviews.llvm.org/D10607
llvm-svn: 240991
We failed to see that we should have deferred the creation of a type
which references a type currently under construction because of atomic
sugar.
This fixes PR23985.
llvm-svn: 240989
- Avoid sending the qfThreadInfo, qsThreadInfo packets if we have a stop reply packet with the threads already (save 2 round trip packets)
- Include the qname, qserial and qkind in the JSON info
- Report the qname, qserial and qkind to the thread so it can cache it to avoid many packets on MacOSX and iOS
- Don't clear all discoverable settings when we exec, just the ones we need to saves 1-5 packets for each exec.
llvm-svn: 240988
Summary: Adding the flag to C++ source files emits a warning, hence we set the compile flag depending on the file's language.
Reviewers: grosser
Subscribers: Meinersbur, pollydev, llvm-commits
Projects: #polly
Differential Revision: http://reviews.llvm.org/D10809
llvm-svn: 240986
At the suggestion of Chandler Carruth, I've removed the timestamp macro,
_KMP_BUILD_TIME, that cmake currently sets to "No_Timestamp" and replaced it with standard
__DATE__ and __TIME__ macros inside kmp_version.c.
llvm-svn: 240985
method to get a SymbolBody and into the callers, and kill now dead
includes.
This removes the need to have the SymbolBody definition when we're
defining the inline method and makes it a better inline method. That was
the only reason for a lot of header includes here. Removing these and
using forward declarations actually uncovers a bunch of cross-header
dependencies that I've fixed while I'm here, and will allow me to
introduce some *important* inline code into Chunks.h that requires the
definition of ObjectFile.
No functionality changed at this point.
Differential Revision: http://reviews.llvm.org/D10789
llvm-svn: 240982
CMake files and should not be by both build systems and also the targets
were also installed by the CMake build system which they should not be.
The problem was that
- the CMake build of LLVM installs and exports the gtest library
targets. We should not being doing this, these are not part of LLVM.
- the Autoconf/Makefile build of LLVM still had gtest libraries in the
installed LLVMConfig.cmake.
These problems would cause problems for an external project because when
calling llvm_map_components_to_libnames(XXX all) ${XXX} would to contain
LLVM's internal gtest libraries.
llvm-svn: 240981
- Hexagon options were physically next to to ones that had a
preceding comment saying "Double dash options", which they aren't.
- The 'ld' tool classes are named Linker, not Link.
llvm-svn: 240980
Fixes include:
- use FileSystem::Unlink() instead of a direct call to ::unlink(...) when deleting files when iterating through the current directory
- save directories from current directory in a list and iterate through those _after_ the current directory has been iterated
- Use new FileSpec::ForEachItemInDirectory() instead of manually iterating across directories with opendir()/readdir()/closedir()
We should switch all code over to using FileSpec::ForEachItemInDirectory(...) in the near future and get rid of FileSpec::EnumerateDirectory().
This is a follow up patch to:
http://reviews.llvm.org/D10787
llvm-svn: 240978
If a function containing a local class is instantiated, instantiate
all of local class member, including default arguments and exception
specifications.
This change fixes PR21332 and thus implements DR1484.
Differential Revision: http://reviews.llvm.org/D9990
llvm-svn: 240974
Avoid listing inclusions (like `!projects/LLVMBuild.txt`) for files
directly underneath `projects/` in `.gitignore`. Instead, change the
`projects/*` exclusion to the more specific `projects/*/`.
llvm-svn: 240973
Remove use of assignment to multiple struct fields using .fieldname syntax.
This doesn't work with gcc 4.8 and earlier. Replace with elementwise field assignments.
Patch by John Mellor-Crummey
Differential Revision: http://reviews.llvm.org/D10798
llvm-svn: 240972
Several tests wouldn't pass when executed on an armv7a_pc_linux triple
due to the non-default arm_aapcs calling convention produced on the
function definitions in the IR output. Account for this with the
application of a little regex.
Patch by Ying Yi.
llvm-svn: 240971
Fix OMPT support for barriers so that state changes occur even if OMPT_TRACE turned off.
These state changes are needed by performance tools that use callbacks for either
ompt_event_wait_barrier_begin or ompt_event_wait_barrier_end. Change ifdef flag to OMPT_BLAME
for callbacks ompt_event_wait_barrier_begin or ompt_event_wait_barrier_end rather than
OMPT_TRACE -- they were misclassified. Without this patch, when the runtime is compiled with
LIBOMP_OMPT_SUPPORT=true, LIBOMP_OMPT_BLAME=true, and LIBOMP_OMPT_TRACE=false, and a callback
is registered for either ompt_event_wait_barrier_begin or ompt_event_wait_barrier_end, then an
assertion will trip. Fix the scoping of one OMPT_TRACE ifdef, which should not have surrounded
an update of an OMPT state. Add a missing initialization of an OMPT task id for an implicit task.
Patch by John Mellor-Crummey
Differential Revision: http://reviews.llvm.org/D10759
llvm-svn: 240970
Make sure to remove the unique lock file, which is what the .lock
symlink points to, if there is a signal while the lock is held. This
will release the lock, since the symlink will point to nothing (already
tested in unit tests). For good measure, also clean up the unique lock
file if there is an error or signal before the lock is acquired.
I will add a clang test.
rdar://problem/21512307
llvm-svn: 240967
This commit implements serialization of the register mask machine
operands. This commit serializes only the call preserved register
masks that are defined by a target, it doesn't serialize arbitrary
register masks.
This commit also extends the TargetRegisterInfo class and TableGen so that
the users of TRI can get the list of all the call preserved register masks and
their names.
Reviewers: Duncan P. N. Exon Smith
Differential Revision: http://reviews.llvm.org/D10673
llvm-svn: 240966
struct sigaction was not initialized. As the result if SA_RESETHAND is set in sa_flags, then the handler is reset after first invocation leading to crash.
Initialize struct sigaction to zero.
Reviewed in http://reviews.llvm.org/D10803
llvm-svn: 240965
which caches the executable name upon the first invocation.
This is necessary because Google Chrome (and potentially other programs)
restrict the access to /proc/self/exe on linux.
This change should fix https://code.google.com/p/chromium/issues/detail?id=502974
llvm-svn: 240960
The random module in Python 2.7.8 and later leaks /dev/[u]random into
children. The test is expected to fail, not be flakey.
This will be fixed in Python 2.7.10 for some operating systems, but not
all e.g. OS X 10.4.
llvm.org/pr23983
bugs.freebsd.org/197376
bugs.python.org/issue23458
llvm-svn: 240958
Summary: This makes the Mozilla style defaults more compliant with the Mozilla style guide. A few options were removed in order to use the LLVM style defaults.
Differential Revision: http://reviews.llvm.org/D10771
llvm-svn: 240957
This is very preliminary support, but it seems to work for the most common case.
When observing more/different test cases, we can work on generalizing this.
llvm-svn: 240955
Compilers recognize "main" function and don't mangle its name.
But if you use a different function as a user-defined entry name,
and if you didn't define that function with extern C, your entry
point function name is mangled. And the linker has to be able to
find that. This is relatively rare but can happen.
llvm-svn: 240953
The expressions we delinearize do not necessarily have to have a SCEVAddRecExpr
at the outermost level. At this moment, the additional flexibility is not
exploited in LLVM itself, but in Polly we will soon soonish use this
functionality. For LLVM, this change should not affect existing functionality
(which is covered by test/Analysis/Delinearization/)
llvm-svn: 240952
Chandler Carruth