In LTO, part of LLVM's middle-end runs after linking has finished. LTO's
semantics depend on the complete set of extracted bitcode files being
known at this time. If the middle-end inserts new calls to library
functions (libfuncs) that are implemented in bitcode, this could extract
new bitcode object files into the link. These cannot be compiled,
leading to undefined symbol references.
Additionally, the middle-end in LTO may reason that such library
functions have no references, and it may internalize them, then
manipulate their API or even delete them. Afterwards, it may emit a call
to them, again producing undefined symbol references.
This patch resolves the former issue by ensuring that the middle end
emits no new references to symbols defined in bitcode, and it resolves
the latter issue by ensuring that extracted bitcode for libfuncs is
considered external, since new calls may be emitted to them at any time.
The new semantics are not yet established for MachO LLD, which does not
yet appear to have any special handling for libcalls in LTO. It also
does not yet support distributed ThinLTO; doing so would require
additional (de)serialization work.
This is the patch referenced in @ilovepi's and my talk at the last LLVM
devmeeting: "LT-Uh-Oh"
Gemini 3.1 was used in porting to COFF and WASM LLDs.
Without this change, passing -fthinlto-index causes -fpass-plugin
arguments to be ignored. We want to be able to use plugins with
distributed thin-lto, so add support for this.
Only if module's name contains a string in -filter-save-modules, will
its temp BC files be saved. If -filter-save-modules= not set, all
modules' BC files will be saved. This feature is more useful for ThinLto
when huge numbers of modules are built separately. Using
-filter-save-modules= can reduce build time and size of generated files,
even avoid crash if some other unrelated files have issues during BC
files dumping.
This avoid pulling in the entire Passes library with all passes as
dependencies when just referring to PassPlugin, which is in fact
independent of the Passes themselves.
Pull Request: https://github.com/llvm/llvm-project/pull/173279
This avoid pulling in the entire Passes library with all passes as
dependencies when just referring to PassPlugin, which is in fact
independent of the Passes themselves.
Pull Request: https://github.com/llvm/llvm-project/pull/172478
Commit #167996 moved VecLib into TargetOptions and ensured clang
properly sets it. However, some LTO backend code paths were still
creating _TargetLibraryInfoImpl_ without passing the VecLib parameter
from `TargetMachine::Options`.
This PR completes the fix by ensuring that:
_LTOBackend.cpp, ThinLTOCodeGenerator.cpp, UpdateCompilerUsed.cpp_ all
pass `TM->Options.VecLib` when constructing _TargetLibraryInfoImpl_.
Without this fix, vector library information (e.g., -fveclib=ArmPL)
would not be properly recognized during LTO optimization and code
generation, potentially causing incorrect optimizations or linker errors
when vector library functions are referenced.
The libcall lowering decisions should be program dependent,
depending on the current module's RuntimeLibcallInfo. We need
another related analysis derived from that plus the current
function's subtarget to provide concrete lowering decisions.
This takes on a somewhat unusual form. It's a Module analysis,
with a lookup keyed on the subtarget. This is a separate module
analysis from RuntimeLibraryAnalysis to avoid that depending on
codegen. It's not a function pass to avoid depending on any
particular function, to avoid repeated subtarget map lookups in
most of the use passes, and to avoid any recomputation in the
common case of one subtarget (and keeps it reusable across
repeated compilations).
This also switches ExpandFp and PreISelIntrinsicLowering as
a sample function and module pass. Note this is not yet wired
up to SelectionDAG, which is still using the LibcallLoweringInfo
constructed inside of TargetLowering.
In preparation for a follow on change that will require checking every
time a new summary is added to the SummaryList for a GUID, make the
SummaryList private and require all accesses to go through one of two
new interfaces. Most changes are to access the list via the read only
getSummaryList() method, and the few that add new summaries (e.g. while
building the combined summary) use the new addSummary() method.
In preparation for a follow on fix that removes these attributes and
metadata in non-LTO pipelines, convert updateMemProfAttributes to a new
MemProfRemoveInfo pass that executes at the start of the LTO backend
pass pipelines when we don't have an index indicating that we linked
with a library support hot cold operator new.
This is largely NFC from an end user perspective but changes where the
removal can be observed, hence the test updates.
A follow on change will use the new pass for non-LTO pipelines (for
cases when the bitcode is initially matched with memprof data but we
decide to complete the compile without LTO).
Some LLVM passes need access to the filesystem to read configuration
files and similar. In some places, this is achieved by grabbing the VFS
from `PGOOptions`, but some passes don't have access to these and resort
to just calling `vfs::getRealFileSystem()`. This PR allows setting the
VFS directly on `PassBuilder` that's able to pass it down to all passes
that need it.
Currently there are two serialization modes for bitstream Remarks:
standalone and separate. The separate mode splits remark metadata (e.g.
the string table) from actual remark data. The metadata is written into
the object file by the AsmPrinter, while the remark data is stored in a
separate remarks file. This means we can't use bitstream remarks with
tools like opt that don't generate an object file. Also, it is confusing
to post-process bitstream remarks files, because only the standalone
files can be read by llvm-remarkutil. We always need to use dsymutil
to convert the separate files to standalone files, which only works for
MachO. It is not possible for clang/opt to directly emit bitstream
remark files in standalone mode, because the string table can only be
serialized after all remarks were emitted.
Therefore, this change completely removes the separate serialization
mode. Instead, the remark string table is now always written to the end
of the remarks file. This requires us to tell the serializer when to
finalize remark serialization. This automatically happens when the
serializer goes out of scope. However, often the remark file goes out of
scope before the serializer is destroyed. To diagnose this, I have added
an assert to alert users that they need to explicitly call
finalizeLLVMOptimizationRemarks.
This change paves the way for further improvements to the remark
infrastructure, including more tooling (e.g. #159784), size optimizations
for bitstream remarks, and more.
Pull Request: https://github.com/llvm/llvm-project/pull/156715
In order to better see what's going on during ThinLTO linking, this PR
adds more profile tags when using `--time-trace` on a `lld-link.exe`
invocation.
After PR, linking `clang.exe`:
<img width="3839" height="2026" alt="Capture d’écran 2025-09-02 082021"
src="https://github.com/user-attachments/assets/bf0c85ba-2f85-4bbf-a5c1-800039b56910"
/>
Linking a custom (Unreal Engine game) binary gives a completly
different picture, probably because of using Unity files, and the sheer
amount of input files (here, providing over 60 GB of .OBJs/.LIBs).
<img width="1940" height="1008" alt="Capture d’écran 2025-09-02 102048"
src="https://github.com/user-attachments/assets/60b28630-7995-45ce-9e8c-13f3cb5312e0"
/>
These are identified by misc-include-cleaner. I've filtered out those
that break builds. Also, I'm staying away from llvm-config.h,
config.h, and Compiler.h, which likely cause platform- or
compiler-specific build failures.
Usage errors in `LTOBackend.cpp` were previously, misleadingly, reported
as internal crashes.
This PR updates `LTOBackend.cpp` to use `reportFatalUsageError` for
reporting usage-related issues.
LLVM Issue: https://github.com/llvm/llvm-project/issues/140953
Internal Tracker: TOOLCHAIN-17744
The `CacheStream::commit()` function (defined in Caching.cpp) deletes
the underlying raw stream. Some output streamers may hold a pointer
to it, which then will outlive the stream object.
In particular, MCAsmStreamer keeps the pointer to the raw stream
though a separate `formatted_raw_stream` object, which buffers data and
there is no path to explicitly flush this data. Before this change,
the buffered data was flushed during the MCAsmStreamer destructor.
After #136121, this happened after the `commit()` function is called.
Therefore, it caused a crash because the `formatted_raw_stream` object
tries to write the buffered data into a deleted raw stream. Even if
we don't delete the stream to avoid the crash, it would be too late
as the output stream cannot accept data after commit().
Fixes: #138194.
This implements the result of the discussion at:
https://discourse.llvm.org/t/rfc-report-fatal-error-and-the-default-value-of-gencrashdialog/73587
There are two different use cases for report_fatal_error, so replace it
with two functions reportFatalInternalError() and
reportFatalUsageError(). The former indicates a bug in LLVM and
generates a crash dialog. The latter does not. The names have been
suggested by rnk and people seemed to like them.
This replaces a lot of the usages that passed an explicit value for
GenCrashDiag. I did not bulk replace remaining report_fatal_error usage
-- they probably require case by case review for which function to use.
…Stream.
CachedFileStream has previously performed the commit step in its
destructor, but this means its only recourse for error handling is
report_fatal_error. Modify this to add an explicit commit() method, and
call this in the appropriate places with appropriate error handling for
the location.
Currently the destructor of CacheStream gives an assert failure in Debug
builds if commit() was not called. This will help track down any
remaining uses of the API that assume the old destructior behaviour. In
Release builds we fall back to the previous behaviour and call
report_fatal_error if the commit fails.
This is version 2 of this PR, superseding reverted PR
https://github.com/llvm/llvm-project/pull/115331 . I have incorporated a
change to the testcase to make it more reliable on Windows, as well as
two follow-up changes
(df79000896
and
b0baa1d8bd)
that were also reverted when 115331 was reverted.
---------
Co-authored-by: Augie Fackler <augie@google.com>
Co-authored-by: Vitaly Buka <vitalybuka@google.com>
…Stream.
CachedFileStream has previously performed the commit step in its
destructor, but this means its only recourse for error handling is
report_fatal_error. Modify this to add an explicit commit() method, and
call this in the appropriate places with appropriate error handling for
the location.
Currently the destructor of CacheStream gives an assert failure in Debug
builds if commit() was not called. This will help track down any
remaining uses of the API that assume the old destructior behaviour. In
Release builds we fall back to the previous behaviour and call
report_fatal_error if the commit fails.
…argetMachine
RISC-V's data layout is determined by the ABI, not just the target
triple. However, the TargetMachine is created using the data layout from
the target triple, which is not always correct. This patch uses the
target ABI from the module and passes it to the TargetMachine, ensuring
that the data layout is set correctly according to the ABI.
The same problem will happen with other targets like MIPS, but
unfortunately, MIPS didn't emit the target-abi into the module flags, so
this patch only fixes the issue for RISC-V.
NOTE: MIPS with -mabi=n32 can trigger the same issue.
Another possible solution is add new parameter to the TargetMachine
constructor, but that would require changes in all the targets.
The module currently stores the target triple as a string. This means
that any code that wants to actually use the triple first has to
instantiate a Triple, which is somewhat expensive. The change in #121652
caused a moderate compile-time regression due to this. While it would be
easy enough to work around, I think that architecturally, it makes more
sense to store the parsed Triple in the module, so that it can always be
directly queried.
For this change, I've opted not to add any magic conversions between
std::string and Triple for backwards-compatibilty purses, and instead
write out needed Triple()s or str()s explicitly. This is because I think
a decent number of them should be changed to work on Triple as well, to
avoid unnecessary conversions back and forth.
The only interesting part in this patch is that the default triple is
Triple("") instead of Triple() to preserve existing behavior. The former
defaults to using the ELF object format instead of unknown object
format. We should fix that as well.
Follow up to PR118508, to avoid unnecessary compile time for an empty
combind regular LTO module if all modules end up being ThinLTO only.
This required minor changes to a few tests to ensure they weren't empty.
This feature is enabled by `-codegen-data-thinlto-two-rounds`, which
effectively runs the `-codegen-data-generate` and `-codegen-data-use` in
two rounds to enable global outlining with ThinLTO.
1. The first round: Run both optimization + codegen with a scratch
output.
Before running codegen, we serialize the optimized bitcode modules to a
temporary path.
2. From the scratch object files, we merge them into the codegen data.
3. The second round: Read the optimized bitcode modules and start the
codegen only this time.
Using the codegen data, the machine outliner effectively performs the
global outlining.
Depends on #90934, #110461 and #110463.
This is a patch for
https://discourse.llvm.org/t/rfc-enhanced-machine-outliner-part-2-thinlto-nolto/78753.
This patch turns type alias ImportMapTy into a proper class to provide
a more intuitive interface like:
ImportList.addDefinition(...)
as opposed to:
FunctionImporter::addDefinition(ImportList, ...)
Also, this patch requires all non-const accesses to go through
addDefinition, maybeAddDeclaration, and addGUID while providing const
accesses via:
const ImportMapTyImpl &getImportMap() const { return ImportMap; }
I realize ImportMapTy may not be the best name as a class (maybe OK as
a type alias). I am not renaming ImportMapTy in this patch at least
because there are 47 mentions of ImportMapTy under llvm/.
The new helper functions make the intent clearer while hiding
implementation details, including how we handle previously added
entries. Note that:
- If we are adding a GUID as a GlobalValueSummary::Definition, then we
override a previously added GlobalValueSummary::Declaration entry
for the same GUID.
- If we are adding a GUID as a GlobalValueSummary::Declaration, then a
previously added GlobalValueSummary::Definition entry for the same
GUID takes precedence, and no change is made.
Without this patch, passing -load-pass-plugin=nonexistent.so to
llvm-lto2 produces a backtrace because LTOBackend.cpp does not handle
the error correctly:
```
Failed to load passes from 'nonexistant.so'. Request ignored.
Expected<T> must be checked before access or destruction.
Unchecked Expected<T> contained error:
Could not load library 'nonexistant.so': nonexistant.so: cannot open shared object file: No such file or directoryPLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.
```
Any tool using `lto::Config::PassPlugins` should suffer similarly.
Based on the message "Request ignored" and the continue statement, the
intention was apparently to continue on failure to load a plugin.
However, no one appears to rely on that behavior now given that it
crashes instead, and terminating is consistent with opt.
This reverts commit e33db249b53fb70dce62db3ebd82d42239bd1d9d.
The change from *set to *map increases memory usage, and caused indexing
OOM in some applications. Need to profile offline to bring the memory
usage down.
The base class llvm::ThreadPoolInterface will be renamed
llvm::ThreadPool in a subsequent commit.
This is a breaking change: clients who use to create a ThreadPool must
now create a DefaultThreadPool instead.
This option is not used. It was added in
[D122133](https://reviews.llvm.org/D122133), 5856f30b, with the only
usage in `ClangLinkerWrapper.cpp`, which was later updated in a1d57fc2,
and then finally removed in [D142650](https://reviews.llvm.org/D142650),
6185246f.
The performance of cold functions shouldn't matter too much, so if we
care about binary sizes, add an option to mark cold functions as
optsize/minsize for binary size, or optnone for compile times [1]. Clang
patch will be in a future patch.
This is intended to replace `shouldOptimizeForSize(Function&, ...)`.
We've seen multiple cases where calls to this expensive function, if not
careful, can blow up compile times. I will clean up users of that
function in a followup patch.
Initial version: https://reviews.llvm.org/D149800
[1]
https://discourse.llvm.org/t/rfc-new-feature-proposal-de-optimizing-cold-functions-using-pgo-info/56388
This allows us to not have to pass -mllvm flags to set the large data
threshold for (in-LLD/not-distributed) ThinLTO.
Follows https://reviews.llvm.org/D52322, which did the same for the code
model.
Since the large data threshold is tied to the code model and we disallow
mixing different code models, do the same for the large data threshold.
This restores commit b4a82b62258c5f650a1cccf5b179933e6bae4867, reverted
in 3ab7ef28eebf9019eb3d3c4efd7ebfd160106bb1 because it was thought to
cause a bot failure, which ended up being unrelated to this patch set.
Differential Revision: https://reviews.llvm.org/D154856
Previously the MemProf profile was expected to be in the same profile
file as a normal PGO profile, passed via the usual -fprofile-use=
option, and was matched in the same pass. To simplify profile
preparation, since the raw MemProf profile requires the binary for
symbolization and may be simpler to index separately from the raw PGO
profile, and also to enable providing a MemProf profile for a SamplePGO
build, separate out the MemProf feedback option and matching pass.
This patch adds the -fmemory-profile-use=${file} option, and the
provided file is passed down to LLVM and ultimately used in a new
MemProfUsePass which performs the matching of just the memory profile
contents of that file.
Note that a single profile file containing both normal PGO and MemProf
profile data is still supported, and the relevant profile data is
matched by the appropriate matching pass(es) based on which option(s)
the profile is provided with (the same profile file can be supplied to
both feedback options).
Differential Revision: https://reviews.llvm.org/D154856
Here's a high level summary of the changes in this patch. For more
information on rational, see the RFC.
(https://discourse.llvm.org/t/rfc-a-unified-lto-bitcode-frontend/61774).
- Add config parameter to LTO backend, specifying which LTO mode is
desired when using unified LTO.
- Add unified LTO flag to the summary index for efficiency. Unified
LTO modules can be detected without parsing the module.
- Make sure that the ModuleID is generated by incorporating more types
of symbols.
Differential Revision: https://reviews.llvm.org/D123803
