Commit Graph

260 Commits

Author SHA1 Message Date
Rui Ueyama e28c146423 Avoid unnecessary buffer allocation and memcpy for compressed sections.
Previously, we uncompress all compressed sections before doing anything.
That works, and that is conceptually simple, but that could results in
a waste of CPU time and memory if uncompressed sections are then
discarded or just copied to the output buffer.

In particular, if .debug_gnu_pub{names,types} are compressed and if no
-gdb-index option is given, we wasted CPU and memory because we
uncompress them into newly allocated bufers and then memcpy the buffers
to the output buffer. That temporary buffer was redundant.

This patch changes how to uncompress sections. Now, compressed sections
are uncompressed lazily. To do that, `Data` member of `InputSectionBase`
is now hidden from outside, and `data()` accessor automatically expands
an compressed buffer if necessary.

If no one calls `data()`, then `writeTo()` directly uncompresses
compressed data into the output buffer. That eliminates the redundant
memory allocation and redundant memcpy.

This patch significantly reduces memory consumption (20 GiB max RSS to
15 Gib) for an executable whose .debug_gnu_pub{names,types} are in total
5 GiB in an uncompressed form.

Differential Revision: https://reviews.llvm.org/D52917

llvm-svn: 343979
2018-10-08 16:58:59 +00:00
Rui Ueyama e7688e6663 Revert r342297: Discard uncompressed buffer after creating .gdb_index contents.
Looks like it broke some local builds that use -gdb-index.

llvm-svn: 342298
2018-09-14 23:28:13 +00:00
Rui Ueyama 751dfbe39b Discard uncompressed buffer after creating .gdb_index contents.
Once we create .gdb_index contents, .zdebug_gnu_pub{names,types}
are useless, so there's no need to keep their uncompressed data
in memory.

I observed that for a test case in which lld creates a 3GB .gdb_index
section, the maximum resident set size reduced from 43GB to 29GB after
this patch.

Differential Revision: https://reviews.llvm.org/D52126

llvm-svn: 342297
2018-09-14 22:57:39 +00:00
Rui Ueyama 5cd9c6bcd8 Support RISC-V
Patch by PkmX.

This patch makes lld recognize RISC-V target and implements basic
relocation for RV32/RV64 (and RVC). This should be necessary for static
linking ELF applications.

The ABI documentation for RISC-V can be found at:
https://github.com/riscv/riscv-elf-psabi-doc/blob/master/riscv-elf.md.
Note that the documentation is far from complete so we had to figure out
some details from bfd.

The patch should be pretty straightforward. Some highlights:

 - A new relocation Expr R_RISCV_PC_INDIRECT is added. This is needed as
   the low part of a PC-relative relocation is linked to the corresponding
   high part (auipc), see:
   https://github.com/riscv/riscv-elf-psabi-doc/blob/master/riscv-elf.md#pc-relative-symbol-addresses

 - LLVM's MC support for RISC-V is very incomplete (we are working on
   this), so tests are given in objectyaml format with the original
   assembly included in the comments. Once we have complete support for
   RISC-V in MC, we can switch to llvm-as/llvm-objdump.

 - We don't support linker relaxation for now as it requires greater
   changes to lld that is beyond the scope of this patch. Once this is
   accepted we can start to work on adding relaxation to lld.

Differential Revision: https://reviews.llvm.org/D39322

llvm-svn: 339364
2018-08-09 17:59:56 +00:00
Sterling Augustine 4fd84c18df Implement framework for linking split-stack object files, and x86_64 support.
llvm-svn: 337332
2018-07-17 23:16:02 +00:00
Peter Collingbourne 11dc7fcae2 ELF: Do not ICF two sections with different output sections.
Note that this doesn't do the right thing in the case where there is
a linker script. We probably need to move output section assignment
before ICF to get the correct behaviour here.

Differential Revision: https://reviews.llvm.org/D47241

llvm-svn: 333052
2018-05-23 01:58:43 +00:00
Peter Smith dbef8cc67c [ELF] Implement --keep-unique option
The --keep-unique <symbol> option is taken from gold. The intention is that
<symbol> will be prevented from being folded by ICF. Although not
specifically mentioned in the documentation <symbol> only matches
global symbols, with a warning if the symbol is not found.

The implementation finds the Section defining <symbol> and removes it from
the set of sections considered for ICF.

Differential Revision: https://reviews.llvm.org/D46755

llvm-svn: 332332
2018-05-15 08:57:21 +00:00
Rafael Espindola 9bf1006278 Split merge sections early.
Now that getSectionPiece is fast (uses a hash) it is probably OK to
split merge sections early.

The reason I want to do this is to split eh_frame sections in the same
place.

This does mean that we have to decompress early. Given that the only
compressed sections are debug info, I don't think we are missing much.

It is a small improvement: 0.5% on the geometric mean.

llvm-svn: 331058
2018-04-27 16:29:57 +00:00
Rafael Espindola 4809e2c11d Define InputSection::getOffset inline.
This is much simpler than the other section types and there are many
places where the section type is statically know.

llvm-svn: 330350
2018-04-19 18:00:46 +00:00
Rafael Espindola 6275a7aa39 Rename MergeInputSection::getOffset.
Unlike the getOffset in the base class, this one computes the offset
in the parent synthetic section, not the final output section.

llvm-svn: 330339
2018-04-19 16:05:07 +00:00
Rafael Espindola ea2c78369c Reduce code duplication.
getVA was already implemented in the base class.

llvm-svn: 330036
2018-04-13 16:07:27 +00:00
Rafael Espindola 5f8e77afb5 Initialize OutputOff to zero.
We have a dedicated Live bit, so we don't need a special value and we
were not accounting for in at least one place.

llvm-svn: 329307
2018-04-05 15:56:04 +00:00
Rafael Espindola 6cd7af51e1 Inline initOffsetMap.
In the lld perf builder r328686 had a negative impact in
stalled-cycles-frontend. Somehow that stat is not showing on my
machine, but the attached patch shows an improvement on cache-misses,
which is probably a reasonable proxy.

My working theory is that given a large input the pieces vector is out
of cache by the time initOffsetMap runs.

Both finalizeContents implementation have a convenient location for
initializing the OffsetMap, so this seems the best solution.

llvm-svn: 329117
2018-04-03 21:38:18 +00:00
Rafael Espindola 816127ea17 Initialize OffsetMap in a known location.
This is a small optimization and avoids the need to use call_once.

llvm-svn: 328686
2018-03-28 03:20:18 +00:00
Rafael Espindola 92eba0e14a Define a trivial method inline.
llvm-svn: 328685
2018-03-28 03:14:11 +00:00
Rafael Espindola 5a7ca96e2d Store live offsets as uint32_t.
We don't support input merge sections larger than 4gb, so these can be
uint32_t.

llvm-svn: 328684
2018-03-28 02:32:31 +00:00
Rafael Espindola 4f058a2c6b Add a SectionBase::getVA helper. NFC.
There were a few too many places duplicating this.

llvm-svn: 328402
2018-03-24 00:35:11 +00:00
Rui Ueyama ac114d27ae s/uncompress/decompress/g.
In lld, we use both "uncompress" and "decompress" which is confusing.
Since LLVM uses "decompress", we should use the same term.

llvm-svn: 324944
2018-02-12 21:56:14 +00:00
Rafael Espindola c7945c827d Move function to the file where it is used.
llvm-svn: 323780
2018-01-30 16:24:04 +00:00
Rafael Espindola 9a84f6b954 Detemplate reportDuplicate.
We normally avoid "switch (Config->EKind)", but in this case I think
it is worth it.

It is only executed when there is an error and it allows detemplating
a lot of code.

llvm-svn: 321404
2017-12-23 17:21:39 +00:00
Rafael Espindola ce3b52c186 Pass an InputFile to the InputSection constructor.
This simplifies toRegularSection and reduces the noise in a followup
patch.

llvm-svn: 321240
2017-12-21 02:11:51 +00:00
Rafael Espindola 604032729c Convert a few more InputFiles to references.
We use null files in sections to represent linker created sections,
so ObjFile<ELFT> is never null.

llvm-svn: 321238
2017-12-21 02:03:39 +00:00
Rafael Espindola 5c73c49c9f Detemplate createCommentSection.
It was only templated so it could create a dummy section header that
was immediately parsed back.

llvm-svn: 321235
2017-12-21 01:21:59 +00:00
Rafael Espindola f4fb5fd752 Move Repl to SectionBase.
It is currently in InputSectionBase. Only InputSections are used in
ICF, so Repl should be move to InputSection to clear the class
hierarchy or, like this patch does, to SectionBase for convenience.

The convenience of having it on the base class is that we can just
access the replacement without having to first check if it is an
InputSection. It is a bit less code and a bit faster as some of this
code is very hot.

I got up to 1.77% improvement in clang-gdb-index and no regressions
according to lnt.

llvm-svn: 320654
2017-12-13 22:59:23 +00:00
Rafael Espindola b01cd86458 Fix the type of the Discared section.
It is constructed with a kind of Regular and will dyn_cast to
InputSection, but is declared to be an InputSectionBase.

llvm-svn: 320539
2017-12-13 01:39:35 +00:00
Rafael Espindola 10bcc1cf90 Fix line endings. NFC.
llvm-svn: 320502
2017-12-12 17:37:01 +00:00
James Henderson 8d0efdd5db [ELF] Reset OutputSection size prior to processing linker script commands
The size of an OutputSection is calculated early, to aid handling of compressed
debug sections. However, subsequent to this point, unused synthetic sections are
removed. In the event that an OutputSection, from which such an InputSection is
removed, is still required (e.g. because it has a symbol assignment), and no longer
has any InputSections, dot assignments, or BYTE()-family directives, the size
member is never updated when processing the commands. If the removed InputSection
had a non-zero size (such as a .got.plt section), the section ends up with the
wrong size in the output.

The fix is to reset the OutputSection size prior to processing the linker script
commands relating to that OutputSection. This ensures that the size is correct even
in the above situation.

Additionally, to reduce the risk of developers misusing OutputSection Size and
InputSection OutSecOff, they are set to simply the number of InputSections in an
OutputSection, and the corresponding index respectively. We cannot completely
stop using them, due to SHF_LINK_ORDER sections requiring them.

Compressed debug sections also require the full size. This is now calculated in
maybeCompress for these kinds of sections.

Reviewers: ruiu, rafael

Differential Revision: https://reviews.llvm.org/D38361

llvm-svn: 320472
2017-12-12 11:51:13 +00:00
Rafael Espindola bdcfb178b5 Delete dead code.
llvm-svn: 319403
2017-11-30 05:52:42 +00:00
Peter Collingbourne e9a9e0a1e7 ELF: Merge DefinedRegular and Defined.
Now that DefinedRegular is the only remaining derived class of
Defined, we can merge the two classes.

Differential Revision: https://reviews.llvm.org/D39667

llvm-svn: 317448
2017-11-06 04:35:31 +00:00
Peter Collingbourne 6c55a70838 ELF: Remove DefinedCommon.
Common symbols are now represented with a DefinedRegular that points
to a BssSection, even during symbol resolution.

Differential Revision: https://reviews.llvm.org/D39666

llvm-svn: 317447
2017-11-06 04:33:58 +00:00
Rui Ueyama f52496e1e0 Rename SymbolBody -> Symbol
Now that we have only SymbolBody as the symbol class. So, "SymbolBody"
is a bit strange name now. This is a mechanical change generated by

  perl -i -pe s/SymbolBody/Symbol/g $(git grep -l SymbolBody lld/ELF lld/COFF)

nd clang-format-diff.

Differential Revision: https://reviews.llvm.org/D39459

llvm-svn: 317370
2017-11-03 21:21:47 +00:00
George Rimar 82f0c42dad [ELF] - Teach LLD to report line numbers for data symbols.
This is PR34826.

Currently LLD is unable to report line number when reporting
duplicate declaration of some variable.

That happens because for extracting line information we always use
.debug_line section content which describes mapping from machine
instructions to source file locations, what does not help for
variables as does not describe them.

In this patch I am taking the approproate information about
variables locations from the .debug_info section.

Differential revision: https://reviews.llvm.org/D38721

llvm-svn: 317080
2017-11-01 07:42:38 +00:00
Rui Ueyama 95c142e208 Revert r316305: Remove a fast lookup table from MergeInputSection.
This reverts commit r316305 because performance regression was observed.

llvm-svn: 317026
2017-10-31 19:14:06 +00:00
Rui Ueyama ea1398eeb3 Move "Assigned" bit from SectionBase to InputSectionBase.
This bit is to manage whether an input section has already been assigned
to some output section by linker scripts or not. So it logically belongs
to InputSectionBase. SectionBase is a common base class for input and
output sections, so that wasn't the right place to define the bit.

llvm-svn: 316879
2017-10-29 23:32:23 +00:00
Rui Ueyama da4d26dfa3 Initialize members not by assignment but by the member initializer list.
llvm-svn: 316876
2017-10-29 22:26:52 +00:00
Rui Ueyama d6b7a390d8 De-template elf::getObjMsg. NFC.
llvm-svn: 316732
2017-10-27 03:13:54 +00:00
Rui Ueyama 572247f22e De-template EhFrameSection. NFC.
llvm-svn: 316731
2017-10-27 03:13:39 +00:00
Rui Ueyama d5135b7e99 Remove obsolete comment.
llvm-svn: 316520
2017-10-24 21:44:43 +00:00
George Rimar f4ca4a6fa1 [ELF] - Do not collect SHT_REL[A] sections unconditionally when --gc-sections and --emit-relocs used together.
This is "Bug 34836 - --gc-sections remove relocations from --emit-relocs",

When --emit-relocs is used, LLD currently always drops SHT_REL[A] sections from
output if --gc-sections is present. Patch fixes the issue.

Differential revision: https://reviews.llvm.org/D38724

llvm-svn: 316418
2017-10-24 08:26:32 +00:00
Konstantin Zhuravlyov 70abb6e204 LLD: Fix large integer implicitly truncated to unsigned type gcc warning
This fixes gcc warning.

Change by Brian Sumner

llvm-svn: 316365
2017-10-23 19:31:31 +00:00
Rui Ueyama d96724db42 Remove a fast lookup table from MergeInputSection.
We used to have a map from section piece offsets to section pieces
as a cache for binary search. But I found that the map took quite a
large amount of memory and didn't make linking faster. So, in this
patch, I removed the map.

This patch saves 566 MiB of RAM (2.019 GiB -> 1.453 GiB) when linking
clang with debug info, and the link time is 4% faster in that test case.

Thanks for Sean Silva for pointing this out.

llvm-svn: 316305
2017-10-22 23:02:07 +00:00
Rui Ueyama 95bf509873 Assume that mergeable input sections are smaller than 4 GiB.
By assuming that mergeable input sections are smaller than 4 GiB,
lld's memory usage when linking clang with debug info drops from
2.788 GiB to 2.019 GiB (measured by valgrind, and that does not include
memory space for mmap'ed files). I think that's a reasonable assumption
given such a large RAM savings, so this patch.

According to valgrind, gold needs 3.54 GiB of RAM to do the same thing.

NB: This patch does not introduce a limitation on the size of
output sections. You can still create sections larger than 4 GiB.

llvm-svn: 316280
2017-10-21 23:20:13 +00:00
Rui Ueyama 0543343170 Use more precise type.
llvm-svn: 315426
2017-10-11 04:01:13 +00:00
Rui Ueyama ed94f36154 Remove an unused default argument.
llvm-svn: 315386
2017-10-10 23:09:23 +00:00
Rui Ueyama f987fe7869 Add more comments.
llvm-svn: 315267
2017-10-10 03:40:57 +00:00
Rui Ueyama 124bedbd50 Add comment to InputSectionBase::File.
This patch also moves declarations so that related declarations next
to each other.

llvm-svn: 315266
2017-10-10 03:22:29 +00:00
Rui Ueyama 339ddd1685 Remove dead function.
llvm-svn: 315265
2017-10-10 03:01:41 +00:00
Rui Ueyama e8936bca72 Add comment.
llvm-svn: 315163
2017-10-08 01:55:29 +00:00
Rui Ueyama 10f5867135 Remove redundant initialization code.
llvm-svn: 315158
2017-10-08 00:45:34 +00:00
Rui Ueyama e41f9e0651 Inline a small function. NFC.
llvm-svn: 315096
2017-10-06 20:08:51 +00:00