This patch implements a new design for the symbol table that stores
SymbolBodies within a memory region of the Symbol object. Symbols are mutated
by constructing SymbolBodies in place over existing SymbolBodies, rather
than by mutating pointers. As mentioned in the initial proposal [1], this
memory layout helps reduce the cache miss rate by improving memory locality.
Performance numbers:
old(s) new(s)
Without debug info:
chrome 7.178 6.432 (-11.5%)
LLVMgold.so 0.505 0.502 (-0.5%)
clang 0.954 0.827 (-15.4%)
llvm-as 0.052 0.045 (-15.5%)
With debug info:
scylla 5.695 5.613 (-1.5%)
clang 14.396 14.143 (-1.8%)
Performance counter results show that the fewer required indirections is
indeed the cause of the improved performance. For example, when linking
chrome, stalled cycles decreases from 14,556,444,002 to 12,959,238,310, and
instructions per cycle increases from 0.78 to 0.83. We are also executing
many fewer instructions (15,516,401,933 down to 15,002,434,310), probably
because we spend less time allocating SymbolBodies.
The new mechanism by which symbols are added to the symbol table is by calling
add* functions on the SymbolTable.
In this patch, I handle local symbols by storing them inside "unparented"
SymbolBodies. This is suboptimal, but if we do want to try to avoid allocating
these SymbolBodies, we can probably do that separately.
I also removed a few members from the SymbolBody class that were only being
used to pass information from the input file to the symbol table.
This patch implements the new design for the ELF linker only. I intend to
prepare a similar patch for the COFF linker.
[1] http://lists.llvm.org/pipermail/llvm-dev/2016-April/098832.html
Differential Revision: http://reviews.llvm.org/D19752
llvm-svn: 268178
This patch redefines the default optimization level as 1 and adds
new level 0. In the command line, it is -O0. The flag disables
costly but optional features so that the linker produces semantically
correct but larger output quickly. Currently it only disables
section merging.
This flag is not intended to be used for final production linking.
It is intended to be used in compile-link-test cycle.
Time to link clang with debug info is about 2x faster with the flag.
Head:
13.24 seconds
Output size: 1227189664 bytes
With this patch:
7.41 seconds
Output size: 2490281784 bytes
Differential Revision: http://reviews.llvm.org/D19705
llvm-svn: 268056
Using multiple context used to be a really big memory saving because we
could free memory from each file while the linker proceeded with the
symbol resolution. We are getting lazier about reading data from the
bitcode, so I was curious if this was still a good tradeoff.
One thing that is a bit annoying is that we still have to copy the
symbol names. The problem is that the names are stored in the Module and
get freed when we move the module bits during linking.
Long term I think the solution is to add a symbol table to the bitcode.
That way IRObject file will not need to use a Module or a Context and we
can drop it while still keeping a StringRef to the names.
This patch is still be an interesting medium term improvement.
When linking llvm-as without debug info this patch is a small speedup:
master: 29.861877513 seconds
patch: 29.814533787 seconds
With debug info the numbers are
master: 34.765181469 seconds
patch: 34.563351584 seconds
The peak memory usage when linking llvm-as with debug info was
master: 599.10MB
patch: 600.13MB
llvm-svn: 267921
There seems to be no reason to keep st_size of undefined symbols.
This patch removes the member for it. This patch will change outputs
in cases that undefined symbols are copied to output, but I think
this is unimportant.
Differential Revision: http://reviews.llvm.org/D19574
llvm-svn: 267826
It turns out that this will read data from the section to properly
handle Elf_Rel implicit addends.
Sorry for the noise.
Original messages:
Try to fix Windows lld build.
Move getRelocTarget to ObjectFile.
It doesn't use anything from the InputSection.
llvm-svn: 267163
Since there is a copy in every translation unit that uses them, they can
be omitted from the symbol table if the address is not significant.
This still doesn't catch as many cases as the gold plugin. The
difference is that we check canBeOmittedFromSymbolTable in each file and
use lazy loading which limits what it can do. Gold checks it in the merged file.
I think the correct way of getting the same results as gold is just to
cache in the IR the result of canBeOmittedFromSymbolTable.
llvm-svn: 267063
Previously, Lazy symbols were created for undefined symbols even though
such symbols cannot be resolved by loading object files. This patch
fixes that bug.
llvm-svn: 265847
start-lib and end-lib are options to link object files in the same
semantics as archive files. If an object is in start-lib and end-lib,
the object is linked only when the file is needed to resolve
undefined symbols. That means, if an object is in start-lib and end-lib,
it behaves as if it were in an archive file.
In this patch, I introduced a new notion, LazyObjectFile. That is
analogous to Archive file type, but that works for a single object
file instead of for an archive file.
http://reviews.llvm.org/D18814
llvm-svn: 265710
We have to differentiate undefined symbols from bitcode and undefined
symbols from other sources.
Undefined symbols from bitcode should not inhibit the symbol being
internalized. Undefined symbols from other sources should.
llvm-svn: 265536
Summary:
This bug was introduced by http://reviews.llvm.org/rL265059,
where InputSectionBase got Thunks field, which can do memory allocations.
Since InputSectionBase destructors were never called (I count it as another bug),
that caused a memory leak when 2 or more thunks are added to a section.
The fix to is properly call InputSectionBase destructors from ~ObjectFile.
Reviewers: atanasyan, ruiu, rafael
Subscribers: rafael, krasin, pcc
Differential Revision: http://reviews.llvm.org/D18809
llvm-svn: 265497
Our symbol representation was redundant, and some times would get out of
sync. It had an Elf_Sym, but some fields were copied to SymbolBody.
Different parts of the code were checking the bits in SymbolBody and
others were checking Elf_Sym.
There are two general approaches to fix this:
* Copy the required information and don't store and Elf_Sym.
* Don't copy the information and always use the Elf_Smy.
The second way sounds tempting, but has a big problem: we would have to
template SymbolBody. I started doing it, but it requires templeting
*everything* and creates a bit chicken and egg problem at the driver
where we have to find ELFT before we can create an ArchiveFile for
example.
As much as possible I compared the test differences with what gold and
bfd produce to make sure they are still valid. In most cases we are just
adding hidden visibility to a local symbol, which is harmless.
In most tests this is a small speedup. The only slowdown was scylla
(1.006X). The largest speedup was clang with no --build-id, -O3 or
--gc-sections (i.e.: focus on the relocations): 1.019X.
llvm-svn: 265293
For now just treat such sections as non-mergeable.
Resubmit r263660 with test fix.
Differential Revision: http://reviews.llvm.org/D18225
llvm-svn: 263664
which was reverted because included
unrelative changes by mistake.
Original commit message:
[ELF] - Change all messages to lowercase to be consistent.
That is directly opposite to http://reviews.llvm.org/D18045,
which was reverted.
This patch changes all messages to start from lowercase letter if
they were not before.
That is done to be consistent with clang.
Differential revision: http://reviews.llvm.org/D18085
llvm-svn: 263337
That is directly opposite to http://reviews.llvm.org/D18045,
which was reverted.
This patch changes all messages to start from lowercase letter if
they were not before.
That is done to be consistent with clang.
Differential revision: http://reviews.llvm.org/D18085
llvm-svn: 263252
In lld we usually avoid hash lookups. In addition to that, IR names are
not fully mangled, so it is best to avoid using them whenever possible.
llvm-svn: 263248
It was discussed to make all messages be
lowercase to be consistent with clang.
(also reverts the r263128 which fixed
build bot fail after r263125)
Original commit message:
[ELF] - Consistent spelling for error/warning messages
Previously error and warnings were not consistent in lld.
Some of them started from lowercase letter, others from
uppercase. Also there was one or two which had a dot at the end.
This patch changes all messages to start from uppercase letter if
they were not before.
Differential revision: http://reviews.llvm.org/D18045
llvm-svn: 263240
Summary:
More generally, appending linkage is a special case that we don't want
to create a SymbolBody for.
Reviewers: rafael, ruiu
Subscribers: Bigcheese, llvm-commits, joker.eph
Differential Revision: http://reviews.llvm.org/D18012
llvm-svn: 263179
Previously error and warnings were not consistent in lld.
Some of them started from lowercase letter, others from
uppercase. Also there was one or two which had a dot at the end.
This patch changes all messages to start from uppercase letter if
they were not before.
Differential revision: http://reviews.llvm.org/D18045
llvm-svn: 263125
Each object file compiled in split stack mode will have an empty
section with a special name: .note.GNU-split-stack
We don't support this in linker now, so patch just adds an error out for that.
Differential revision: http://reviews.llvm.org/D17918
llvm-svn: 263039
Peter Collingbourne