For x86_64 callable functions, ARM64EC requires an entry thunk generated
by the compiler. The linker interprets .hybmp sections to associate
function chunks with their entry points and writes an offset to thunks
preceding function section contents.
Additionally, ICF needs to be aware of entry thunks to not consider
chunks to be equal when they have different entry thunks, and GC needs
to mark entry thunks together with function chunks.
I used a new SectionChunkEC class instead of storing entry thunks in
SectionChunk, following the guideline to keep SectionChunk as compact as
possible. This way, there is no memory usage increase on non-EC targets.
Avoid replacing replacing a chunk with one from a different type. It's
mostly a concern for ARM64X, where we don't want to merge aarch64 and
arm64ec chunks, but it may also in theory happen between arm64ec and
x86_64 chunks.
This adds support for generating Chrome-tracing .json profile traces in
the LLD COFF driver.
Also add the necessary time scopes, so that the profile trace shows in
great detail which tasks are executed.
As an example, this is what we see when linking a Unreal Engine
executable:
This reverts commit 7370ff624d217b0f8f7512ca5b651a9b8095a411.
(and 47fb8ae2f9a4075de05433ef24f459b6befd1730).
This commit broke the symbol type in import libraries generated
for mingw autoexported symbols, when the source files were built
with LTO. I'll commit a testcase that showcases this issue after
the revert.
Patch created by running:
rg -l parallelForEachN | xargs sed -i '' -c 's/parallelForEachN/parallelFor/'
No behavior change.
Differential Revision: https://reviews.llvm.org/D128140
Original commit description:
[LLD] Remove global state in lld/COFF
This patch removes globals from the lldCOFF library, by moving globals
into a context class (COFFLinkingContext) and passing it around wherever
it's needed.
See https://lists.llvm.org/pipermail/llvm-dev/2021-June/151184.html for
context about removing globals from LLD.
I also haven't moved the `driver` or `config` variables yet.
Differential Revision: https://reviews.llvm.org/D109634
This reverts commit a2fd05ada9030eab2258fff25e77a05adccae128.
Original commits were b4fa71eed34d967195514fe9b0a5211fca2bc5bc
and e03c7e367adb8f228332e3c2ef8f45484597b719.
check for timer output"
Seems to be causing a number of asan test failures.
This reverts commit b4fa71eed34d967195514fe9b0a5211fca2bc5bc
and e03c7e367adb8f228332e3c2ef8f45484597b719.
This patch removes globals from the lldCOFF library, by moving globals
into a context class (COFFLinkingContext) and passing it around wherever
it's needed.
See https://lists.llvm.org/pipermail/llvm-dev/2021-June/151184.html for
context about removing globals from LLD.
I also haven't moved the `driver` or `config` variables yet.
Differential Revision: https://reviews.llvm.org/D109634
The only known reason why ICF should not merge otherwise identical
sections with differing associated sections has to do with exception
handling tables. It's not clear what ICF should do when there are other
kinds of associated sections. In every other case when this has come up,
debug info and CF guard metadata, we have opted to make ICF ignore the
associated sections.
For comparison, ELF doesn't do anything for comdat groups. Instead,
.eh_frame is parsed to figure out if a section has an LSDA, and if so,
ICF is disabled.
Another issue is that the order of associated sections is not defined.
We have had issues in the past (crbug.com/1144476) where changing the
order of the .xdata/.pdata sections in the object file lead to large ICF
slowdowns.
To address these issues, I decided it would be best to explicitly
consider only .pdata and .xdata sections during ICF. This makes it easy
to ignore the object file order, and I think it makes the intention of
the code clearer.
I've also made the children() accessor return an empty list for
associated sections. This mostly only affects ICF and GC. This was the
behavior before I made this a linked list, so the behavior change should
be good. This had positive effects on chrome.dll: more .xdata sections
were merged that previously could not be merged because they were
associated with distinct .pdata sections.
Reviewed By: mstorsjo
Differential Revision: https://reviews.llvm.org/D98993
This patch adds support for creating Guard Address-Taken IAT Entry Tables (.giats$y sections) in object files, matching the behavior of MSVC. These contain lists of address-taken imported functions, which are used by the linker to create the final GIATS table.
Additionally, if any DLLs are delay-loaded, the linker must look through the .giats tables and add the respective load thunks of address-taken imports to the GFIDS table, as these are also valid call targets.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D87544
This broke both Firefox and Chromium (PR47905) due to what seems like dllimport
function not being handled correctly.
> This patch adds support for creating Guard Address-Taken IAT Entry Tables (.giats$y sections) in object files, matching the behavior of MSVC. These contain lists of address-taken imported functions, which are used by the linker to create the final GIATS table.
> Additionally, if any DLLs are delay-loaded, the linker must look through the .giats tables and add the respective load thunks of address-taken imports to the GFIDS table, as these are also valid call targets.
>
> Reviewed By: rnk
>
> Differential Revision: https://reviews.llvm.org/D87544
This reverts commit cfd8481da1adba1952e0f6ecd00440986e49a946.
This patch adds support for creating Guard Address-Taken IAT Entry Tables (.giats$y sections) in object files, matching the behavior of MSVC. These contain lists of address-taken imported functions, which are used by the linker to create the final GIATS table.
Additionally, if any DLLs are delay-loaded, the linker must look through the .giats tables and add the respective load thunks of address-taken imports to the GFIDS table, as these are also valid call targets.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D87544
This patch adds support for creating Guard Address-Taken IAT Entry Tables (.giats$y sections) in object files, matching the behavior of MSVC. These contain lists of address-taken imported functions, which are used by the linker to create the final GIATS table.
Additionally, if any DLLs are delay-loaded, the linker must look through the .giats tables and add the respective load thunks of address-taken imports to the GFIDS table, as these are also valid call targets.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D87544
Essentially takes the lld/Common/Threads.h wrappers and moves them to
the llvm/Support/Paralle.h algorithm header.
The changes are:
- Remove policy parameter, since all clients use `par`.
- Rename the methods to `parallelSort` etc to match LLVM style, since
they are no longer C++17 pstl compatible.
- Move algorithms from llvm::parallel:: to llvm::, since they have
"parallel" in the name and are no longer overloads of the regular
algorithms.
- Add range overloads
- Use the sequential algorithm directly when 1 thread is requested
(skips task grouping)
- Fix the index type of parallelForEachN to size_t. Nobody in LLVM was
using any other parameter, and it made overload resolution hard for
for_each_n(par, 0, foo.size(), ...) because 0 is int, not size_t.
Remove Threads.h and update LLD for that.
This is a prerequisite for parallel public symbol processing in the PDB
library, which is in LLVM.
Reviewed By: MaskRay, aganea
Differential Revision: https://reviews.llvm.org/D79390
Heap profiling with ETW shows that LLD performs 4,053,721 heap
allocations over its lifetime, and ~800,000 of them come from
assocEquals. These vectors are created just to do a comparison, so fuse
the comparison into the loop and avoid the allocation.
ICF is overall a small portion of the time spent linking, and I did not
measure overall throughput improvements from this change above the noise
threshold. However, these show up in the heap profiler, and the work is
done, so we might as well land it if the code is clear enough.
Reviewed By: hans
Differential Revision: https://reviews.llvm.org/D79297
This patch does the same thing as r365595 to other subdirectories,
which completes the naming style change for the entire lld directory.
With this, the naming style conversion is complete for lld.
Differential Revision: https://reviews.llvm.org/D64473
llvm-svn: 365730
The previous patch lost the call to PowerOf2Ceil, which causes LLD to
crash when handling common symbols with a non-power-of-2 size. I tweaked
the existing common.test to make the bsspad16 common symbol be 15 bytes
to add coverage for this case.
llvm-svn: 361426
Summary:
Valid section or chunk alignments are powers of 2 in the range [1,
8192]. These can be stored more canonically in log2 form to free up some
bits in Chunk. Combined with D61696, SectionChunk gets 8 bytes smaller.
Reviewers: ruiu, aganea
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D61698
llvm-svn: 361206
SectionChunk is one of the most frequently allocated data structures in
LLD, since there are about four per function when optimizations and
debug info are enabled (.text, .pdata, .xdata, .debug$S).
A PE COFF file cannot be larger than 2GB, so there is an inherent limit
on the length of the section name and the number of relocations.
Decompose the ArrayRef and StringRef into pointer and size, and put them
back together in the accessors for section name and relocation list.
I plan to gather complete performance numbers later by padding
SectionChunk with dead data and measuring performance after all the size
optimizations are done.
llvm-svn: 359923
Make some small adjustment while touching the code: make parameters
const, use less_first(), etc.
Differential Revision: https://reviews.llvm.org/D60989
llvm-svn: 358943
Summary:
Reorder the fields in both to use padding more efficiently, and add more
comments on the purpose of the fields.
Replace `std::vector<SectionChunk*> AssociativeChildren` with a
singly-linked list. This avoids the separate vector allocation to list
associative children, and shrinks the 3 pointers used for the typically
empty vector down to 1.
In the end, this reduces the sum of heap allocations used to link
browser_tests.exe with NO PDB by 13.10%, going from 2,248,728 KB to
1,954,071 KB of heap. These numbers exclude memory mapped files, which
are of course a significant factor in LLD's memory usage.
Reviewers: ruiu, mstorsjo, aganea
Subscribers: jdoerfert, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D59797
llvm-svn: 357535
LLD's performance on PGO instrumented Windows binaries was still not
great even with the fix in D56955; out of the 2m41s linker runtime,
around 2 minutes were still being spent in ICF. I looked into this more
closely and discovered that the vast majority of the runtime was being
spent segregating .pdata sections with the following relocation chain:
.pdata -> identical .text -> unique PGO counter (not eligible for ICF)
This patch causes us to perform 2 rounds of relocation hash
propagation, which allows the hash for the .pdata sections to
incorporate the identifier from the PGO counter. With that, the amount
of time spent in ICF was reduced to about 2 seconds. I also found that
the same change led to a significant ICF performance improvement in a
regular release build of Chromium's chrome_child.dll, where ICF time
was reduced from around 1s to around 700ms.
With the same change applied to the ELF linker, median of 100 runs
for lld-speed-test/chrome reduced from 4.53s to 4.45s on my machine.
I also experimented with increasing the number of propagation rounds
further, but I did not observe any further significant performance
improvements linking Chromium or Firefox.
Differential Revision: https://reviews.llvm.org/D56986
llvm-svn: 351899
It turns out that sections in PGO instrumented object files on Windows
contain a large number of relocations pointing to themselves. With
r347429 this can cause many sections to receive the same hash (usually
zero) as a result of a section's hash being xor'ed with itself.
This patch causes the COFF and ELF linkers to avoid this problem
by adding the hash of the relocated section instead of xor'ing it.
On my machine this causes the regressing test case
provided by Mozilla to terminate in 2m41s.
Differential Revision: https://reviews.llvm.org/D56955
llvm-svn: 351898
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
Summary: They have an additional `ThreadsEnabled` check, which does not matter much.
Reviewers: pcc, ruiu, rnk
Reviewed By: ruiu
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D54812
llvm-svn: 347587
Previously we were taking over 13 minutes to link Firefox's xul.dll
on ARM64; this reduces link time to around 18s on my machine.
The root cause of the problem was that all of the input .pdata sections
had the same unrelocated section data and therefore the same hash,
which made segregation quadratic in the number of .pdata sections. The
reason why we weren't observing this on other architectures was that
ARM has a different .pdata format. On non-ARM the format is (start
address, end address, .xdata), which caused the size of the function
to appear in the unrelocated section data where the end address field
is. However, the ARM format omits the end address field.
Fixes PR39667.
Differential Revision: https://reviews.llvm.org/D54809
llvm-svn: 347429
After fixing up the runtime pseudo relocation, the .refptr.<var>
will be a plain pointer with the same value as the IAT entry itself.
To save a little binary size and reduce the number of runtime pseudo
relocations, redirect references to the IAT entry (via the __imp_<var>
symbol) itself and discard the .refptr.<var> chunk (as long as the
same section chunk doesn't contain anything else than the single
pointer).
As there are now cases for both setting the Live variable to true
and false externally, remove the accessors and setters and just make
the variable public instead.
Differential Revision: https://reviews.llvm.org/D51456
llvm-svn: 341175
This patch does the same thing as r338153 for COFF.
Note that this patch affects only the order of log messages.
The output file is already deterministic.
Differential Revision: https://reviews.llvm.org/D50023
llvm-svn: 338406
This is needed to avoid merging two functions with identical
instructions but different xdata. It also reduces binary size by
deduplicating identical pdata sections.
Fixes PR35337.
Differential Revision: https://reviews.llvm.org/D46672
llvm-svn: 332169
In an upcoming change I will need to make a distinction between section
type (code, data, bss) and permissions. The term that I use for both
of these things is "output characteristics".
Differential Revision: https://reviews.llvm.org/D45799
llvm-svn: 330361
