The size of this class actually matters because this is the
most popular class among all classes. We create a Defined symbol
for each defined symbol in a symbol table. That can be millions
for a large program. For example, linking LLD instantiates this
class millions times.
llvm-svn: 241025
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
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
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
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
Most build system depends on existence or time stamp of a file.
This patch is to create an empty file for /pdb:<filename> option
just to satisfy some build rules.
llvm-svn: 240948
The previous logic to find default entry name or subsystem does not
seem correct (i.e. was not compatible with MSVC linker). Previously,
default entry name was inferred from CRT functions and user-defined
entry functions. Subsystem was inferred from CRT functions.
Default entry name and subsystem are now inferred based on the
following table. Note that we no longer use CRT functions to infer
them.
Entry name Subsystem
main mainCRTStartup console
wmain wmainCRTStartup console
WinMain WinMainCRTStartup windows
wWinMain wWinMainCRTStartup windows
llvm-svn: 240922
Usually dllexported symbols are defined with 'extern "C"',
so identifying them is easy. We can just do hash table lookup
to look up exported symbols.
However, C++ non-member functions are also allowed to be exported,
and they can be specified with unmangled name. So, if /export:foo
is given, we need to look up not only "foo" but also its all
mangled names. In MSVC mangling scheme, that means that we need to
look up any symbol which starts with "?foo@@Y".
In this patch, we scan the entire symbol table to search for
a mangled symbol. The symbol table is a DenseMap, and that doesn't
support table lookup by string prefix. This is of course very
inefficient. But that should be probably OK because the user
should always add 'extern "C"' to dllexported symbols.
llvm-svn: 240919
This option is to ignore remaining undefined symbols and force
the linker to create an output file anyways.
The existing code assumes that there's no undefined symbol after
reportRemainingUndefines(). That assumption is legitimate.
I also don't want to mess up the existing code for this minor feature.
In order to keep it as is, remaining undefined symbols are replaced
with dummy defined symbols.
llvm-svn: 240913
When comparing two COMDAT sections, we need to take section values
and associative sections into account. This patch fixes that bug.
It fixes a crash bug of llvm-tblgen when linked with /opt:lldicf.
One thing I don't understand yet is that this logic seems to be
too strict. MSVC linker is able to create more compact executables
(which of course work correctly). With this ICF algorithm, LLD is
able to make executable smaller, but the outputs are larger than
MSVC's. There must be something I'm missing here.
llvm-svn: 240897
This function is actually *very* hot. It is hard to see currently
because the call graph is very recursive, but I'm working to remove that
and when I do this function becomes significantly higher on the profile
(up to 5%!) and so worth avoiding the call overhead.
No specific perf gain I can measure yet (below the noise), but likely to
have more impact as we stop cluttering the call graph.
Differential Revision: http://reviews.llvm.org/D10788
llvm-svn: 240873
StringRefs. This uses the LLVM hashing rather than the standard library
and a closed addressed hash table rather than chaining.
This improves the Windows self-link of LLD by 4.4% (averaged over 10
runs, with well under 1% of variance on each).
There is still some room to improve here. Two things I clearly see in
the profile:
1) This is one of the biggest stress tests for the LLVM hashing code. It
actually consumes something like 3-4% of the link time after the
change.
2) The way that StringRef keys are handled in the DenseMap interface is
pretty suboptimal. We pay the price of checking for empty and
tombstone keys when we could only possibly be looking for a normal
key. But fixing this requires invasive API changes.
So there is still some headroom here.
Differential Revision: http://reviews.llvm.org/D10684
llvm-svn: 240871
There were a few issues with the previous delay-import tables.
- "Attribute" field should have been 1 instead of 0.
(I don't know the meaning of this field, though.)
- LEA and CALL operands had wrong addresses.
- Address tables are in .didat (which is read-only).
They should have been in .data.
llvm-svn: 240837
This flag can be used to produce a map file, which is essentially a list
of objects linked into the final output file together with the RVAs of
their symbols. Because our format differs from MSVC's we expose it as a
separate flag.
Differential Revision: http://reviews.llvm.org/D10773
llvm-svn: 240812
We were resolving entry symbols and /include'd symbols after all other
symbols are resolved. But looks like it's too late. I found that it
causes some program to fail to link.
Let's say we have an object file A which defines symbols X and Y in an
archive. We also have another file B after A which defines X, Y and
_DLLMainCRTStartup in another archive. They conflict each other, so
either A or B can be linked.
If we have _DLLMainCRTStartup as an undefined symbol, file B is always
chosen. If not, there's a chance that A is chosen. If the linker
find it needs _DllMainCRTStartup after that, it's too late.
This patch adds undefined symbols to the symbol table as soon as
possible to fix the issue.
llvm-svn: 240757
Absolute symbols were always handled as external symbols, so if two
or more object files define the same absolute symbol, they would
conflict even if the symbol is private to each file.
This patch fixes that bug.
llvm-svn: 240756
Peter Collingbourne