This allows us to store more info about where we're emitting the remarks
without cluttering LLVMContext. This is needed for future support for
the remark section.
Differential Revision: https://reviews.llvm.org/D58996
Original llvm-svn: 355507
llvm-svn: 355514
This allows us to store more info about where we're emitting the remarks
without cluttering LLVMContext. This is needed for future support for
the remark section.
Differential Revision: https://reviews.llvm.org/D58996
llvm-svn: 355507
Summary:
The changes to disable LTO unit splitting by default (r350949) and
detect inconsistently split LTO units (r350948) are causing some crashes
when the inconsistency is detected in multiple threads simultaneously.
Fix that by having the code always look for the inconsistently split
LTO units during the thin link, by checking for the presence of type
tests recorded in the summaries.
Modify test added in r350948 to remove single threading required to fix
a bot failure due to this issue (and some debugging options added in the
process of diagnosing it).
Reviewers: pcc
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D57561
llvm-svn: 354062
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:
Records in the module summary index whether the bitcode was compiled
with the option necessary to enable splitting the LTO unit
(e.g. -fsanitize=cfi, -fwhole-program-vtables, or -fsplit-lto-unit).
The information is passed down to the ModuleSummaryIndex builder via a
new module flag "EnableSplitLTOUnit", which is propagated onto a flag
on the summary index.
This is then used during the LTO link to check whether all linked
summaries were built with the same value of this flag. If not, an error
is issued when we detect a situation requiring whole program visibility
of the class hierarchy. This is the case when both of the following
conditions are met:
1) We are performing LowerTypeTests or Whole Program Devirtualization.
2) There are type tests or type checked loads in the code.
Note I have also changed the ThinLTOBitcodeWriter to also gate the
module splitting on the value of this flag.
Reviewers: pcc
Subscribers: ormris, mehdi_amini, Prazek, inglorion, eraman, steven_wu, dexonsmith, arphaman, dang, llvm-commits
Differential Revision: https://reviews.llvm.org/D53890
llvm-svn: 350948
Summary:
Instead of using two separate callbacks to return the entry count and the
relative block frequency, use a single callback to return callsite
count. This would allow better supporting hybrid mode in the future as
the count of callsite need not always be derived from entry count (as in
sample PGO).
Reviewers: davidxl
Subscribers: mehdi_amini, steven_wu, dexonsmith, dang, llvm-commits
Differential Revision: https://reviews.llvm.org/D56464
llvm-svn: 350755
Summary:
The LTO/ThinLTO driver currently creates invalid bitcode by setting
symbols marked dllimport as dso_local. The compiler often has access
to the definition (often dllexport) and the declaration (often
dllimport) of an object at link-time, leading to a conflicting
declaration. This patch resolves the inconsistency by removing the
dllimport attribute.
Reviewers: tejohnson, pcc, rnk, echristo
Reviewed By: rnk
Subscribers: dmikulin, wristow, mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, dang, llvm-commits
Differential Revision: https://reviews.llvm.org/D55627
llvm-svn: 349667
Summary:
This patch computes the synthetic function entry count on the whole
program callgraph (based on module summary) and writes the entry counts
to the summary. After function importing, this count gets attached to
the IR as metadata. Since it adds a new field to the summary, this bumps
up the version.
Reviewers: tejohnson
Subscribers: mehdi_amini, inglorion, llvm-commits
Differential Revision: https://reviews.llvm.org/D43521
llvm-svn: 349076
This breaks C and C++ semantics because it can cause the address
of the global inside the module to differ from the address outside
of the module.
Differential Revision: https://reviews.llvm.org/D55237
llvm-svn: 348321
Without this, we don't consider types used by aliasees in our cache key.
This caused issues when using the same cache for thin-linking the same
TU with different sets of virtual call candidates for a virtual call
inside of a constructor. That's sort of a mouthful. :)
Differential Revision: https://reviews.llvm.org/D55060
llvm-svn: 348216
Summary:
We can sometimes end up with multiple copies of a local variable that
have the same GUID in the index. This happens when there are local
variables with the same name that are in different source files having the
same name/path at compile time (but compiled into different bitcode objects).
In this case make sure we import the copy in the caller's module.
This enables importing both of the variables having the same GUID
(but which will have different promoted names since the module paths,
and therefore the module hashes, will be distinct).
Importing the wrong copy is particularly problematic for read only
variables, since we must import them as a local copy whenever
referenced. Otherwise we get undefs at link time.
Note that the llvm-lto.cpp and ThinLTOCodeGenerator changes are needed
for testing the distributed index case via clang, which will be sent as
a separate clang-side patch shortly. We were previously not doing the
dead code/read only computation before computing imports when testing
distributed index generation (like it was for testing importing and
other ThinLTO mechanisms alone).
Reviewers: evgeny777
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, dang, llvm-commits
Differential Revision: https://reviews.llvm.org/D55047
llvm-svn: 347886
Summary:
The old legacy LTO API had a separate cache key computation, which was
a subset of the cache key computation in the new LTO API (from what I
can tell this is largely just because certain features such as CFI,
dsoLocal, etc are only utilized via the new LTO API). However, having
separate computations is unnecessary (much of the code is duplicated),
and can lead to bugs when adding new optimizations if both cache
computation algorithms aren't updated properly - it's much easier to
maintain if we have a single facility.
This patch refactors the old LTO API code to use the cache key
computation from the new LTO API. To do this, we set up an lto::Config
object and fill in the fields that the old LTO was hashing (the others
will just use the defaults).
There are two notable changes:
- I added a Freestanding flag to the LTO Config. Currently this is only
used by the legacy LTO API. In the patch that added it (D30791) I had
asked about adding it to the new LTO API, but it looks like that was not
addressed. This should probably be discussed as a follow up to this
change, as it is orthogonal.
- The legacy LTO API had some code that was hashing the GUID of all
preserved symbols defined in the module. I looked back at the history of
this (which was added with the original hashing in the legacy LTO API in
D18494), and there is a comment in the review thread that it was added
in preparation for future internalization. We now do the internalization
of course, and that is handled in the new LTO API cache key computation
by hashing the recorded linkage type of all defined globals. Therefore I
didn't try to move over and keep the preserved symbols handling.
Reviewers: steven_wu, pcc
Subscribers: mehdi_amini, inglorion, eraman, dexonsmith, dang, llvm-commits
Differential Revision: https://reviews.llvm.org/D54635
llvm-svn: 347592
An attempt to recommit r346584 after failure on OSX build bot.
Fixed cache key computation in ThinLTOCodeGenerator and added
test case
llvm-svn: 347033
In a lot of places an empty string was passed as the ErrorBanner to
logAllUnhandledErrors. This patch makes that argument optional to
simplify the call sites.
llvm-svn: 346604
This patch allows internalising globals if all accesses to them
(from live functions) are from non-volatile load instructions
Differential revision: https://reviews.llvm.org/D49362
llvm-svn: 346584
Summary:
This fixes PR 37422
In ELF, non-weak symbols can also be non-prevailing. In this particular
PR, the __llvm_profile_* symbols are non-prevailing but weren't getting
dropped - causing multiply-defined errors with lld.
Also add a test, strong_non_prevailing.ll, to ensure that multiple
copies of a strong symbol are dropped.
To fix the test regressions exposed by this fix,
- do not mark prevailing copies for symbols with 'appending' linkage.
There's no one prevailing copy for such symbols.
- fix the prevailing version in dead-strip-fulllto.ll
- explicitly pass exported symbols to llvm-lto in fumcimport.ll and
funcimport_var.ll
Reviewers: tejohnson, pcc
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith,
dang, srhines, llvm-commits
Differential Revision: https://reviews.llvm.org/D54125
llvm-svn: 346436
Summary:
LTO and ThinLTO optimizes the IR differently.
One source of differences is the amount of internalizations that
can happen.
Add an option to enable/disable internalization so that other
differences can be studied in isolation. e.g. inlining.
There are other things lto and thinlto do differently, I will add
flags to enable/disable them as needed.
Reviewers: tejohnson, pcc, steven_wu
Subscribers: mehdi_amini, inglorion, steven_wu, dexonsmith, dang, llvm-commits
Differential Revision: https://reviews.llvm.org/D53294
llvm-svn: 346140
All the PassBuilder::parse interfaces now return descriptive StringError
instead of a plain bool. It allows to make -passes/aa-pipeline parsing
errors context-specific and thus less confusing.
TODO: ideally we should also make suggestions for misspelled pass names,
but that requires some extensions to PassBuilder.
Reviewed By: philip.pfaffe, chandlerc
Differential Revision: https://reviews.llvm.org/D53246
llvm-svn: 344685
Summary:
All the PassBuilder::parse interfaces now return descriptive StringError
instead of a plain bool. It allows to make -passes/aa-pipeline parsing
errors context-specific and thus less confusing.
TODO: ideally we should also make suggestions for misspelled pass names,
but that requires some extensions to PassBuilder.
Reviewed By: philip.pfaffe, chandlerc
Differential Revision: https://reviews.llvm.org/D53246
llvm-svn: 344519
This can be used to preserve profiling information across codebase
changes that have widespread impact on mangled names, but across which
most profiling data should still be usable. For example, when switching
from libstdc++ to libc++, or from the old libstdc++ ABI to the new ABI,
or even from a 32-bit to a 64-bit build.
The user can provide a remapping file specifying parts of mangled names
that should be treated as equivalent (eg, std::__1 should be treated as
equivalent to std::__cxx11), and profile data will be treated as
applying to a particular function if its name is equivalent to the name
of a function in the profile data under the provided equivalences. See
the documentation change for a description of how this is configured.
Remapping is supported for both sample-based profiling and instruction
profiling. We do not support remapping indirect branch target
information, but all other profile data should be remapped
appropriately.
Support is only added for the new pass manager. If someone wants to also
add support for this for the old pass manager, doing so should be
straightforward.
This is the LLVM side of Clang r344199.
Reviewers: davidxl, tejohnson, dlj, erik.pilkington
Subscribers: mehdi_amini, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D51249
llvm-svn: 344200
Given a library call that is represented as an llvm intrinsic call, but
later transformed to an actual call, if an overriding definition of that
library routine is provided indirectly via an alias, prevent LTO from
eliminating the definition.
This is a fix for PR38547.
Differential Revision: https://reviews.llvm.org/D52836
llvm-svn: 344198
During D51276 discussion it was decided that legacy PassTimingInfo
interface can not be reused for new pass manager's implementation
of -time-passes.
This is a cleanup in preparation for D51276 to make legacy interface
as concise as possible, moving the PassTimingInfo from the header
into the anonymous legacy namespace in .cpp.
It is rather close to a revert of rL340872 in a sense that it hides
the interface and gets rid of templates. However as compared to
a complete revert it resides in a different translation unit and has
an additional pass-instance counting funcitonality (PassIDCountMap).
Reviewers: philip.pfaffe
Differential Revision: https://reviews.llvm.org/D52356
llvm-svn: 343104
Summary:
In D49565/r337503, the type id record writing was fixed so that only
referenced type ids were emitted into each per-module index for ThinLTO
distributed builds. However, this still left an efficiency issue: each
per-module index checked all type ids for membership in the referenced
set, yielding O(M*N) performance (M indexes and N type ids).
Change the TypeIdMap in the summary to be indexed by GUID, to facilitate
correlating with type identifier GUIDs referenced in the function
summary TypeIdInfo structures. This allowed simplifying other
places where a map from type id GUID to type id map entry was previously
being used to aid this correlation.
Also fix AsmWriter code to handle the rare case of type id GUID
collision.
For a large internal application, this reduced the thin link time by
almost 15%.
Reviewers: pcc, vitalybuka
Subscribers: mehdi_amini, inglorion, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D51330
llvm-svn: 343021
Currently the code-model does not get saved in the module IR,
so if a code model is specified when compiling with LTO,
it gets lost and is not propagated properly to LTO. This patch,
along with one for the front end, fixes that.
Differential Revision: https://reviews.llvm.org/D52322
llvm-svn: 342760
Summary:
During threaded thinLTO, it is possible that the entry for current
module doesn't exist in StringMaps (like ExportLists, ResolvedODR,
etc.). Using operator[] might trigger a rehash for the StringMap, which
might happen on multiple threads at the same time.
rdar://problem/43846199
Reviewers: tejohnson, mehdi_amini, kromanova, pcc
Reviewed By: tejohnson
Subscribers: dang, inglorion, eraman, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D52049
llvm-svn: 342263
Summary:
Issue occurs when doing ThinLTO with CodeGenOnly flag.
TMBuilder.TheTriple is assigned to by multiple threads in an unsafe way resulting in double-free of std::string memory.
Pseudocode:
if (CodeGenOnly) {
// Perform only parallel codegen and return.
ThreadPool Pool;
int count = 0;
for (auto &ModuleBuffer : Modules) {
Pool.async([&](int count) {
...
/// Now call OutputBuffer = codegen(*TheModule);
/// Which turns into initTMBuilder(moduleTMBuilder, Triple(TheModule.getTargetTriple()));
/// Which turns into
TMBuilder.TheTriple = std::move(TheTriple); // std::string = "....."
/// So, basically std::string assignment to same string on multiple threads = memory corruption
}
return;
}
Patch by Alex Borcan
Reviewers: llvm-commits, steven_wu
Reviewed By: steven_wu
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D51651
llvm-svn: 341422
Summary:
Some links were failing with "Global is external, but doesn't have
external or weak linkage!" in ThinLTO builds with debug
information. This happened when we elide the body of a global that is
referenced by debug info. This results in a declaration, which we
would then internalize - but declarations cannot be internal. This
change avoids the problem by not internalizing these declarations.
Fixes PR38046.
Reviewers: pcc, tejohnson
Subscribers: mehdi_amini, aprantl, hiraditya, JDevlieghere, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D49777
llvm-svn: 338100
Summary:
Similar to what lld already does for dllimport symbols which are
prefaced with __imp_ (see lld patch r240620), strip off the __imp_
prefix in LTO. Otherwise we can get 2 separate GlobalResolution for
a single symbol, the dllimport declaration, and the definition, which
leads to incorrect LTO handling.
Fixes PR38105.
Reviewers: pcc
Subscribers: mehdi_amini, inglorion, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D49138
llvm-svn: 337762
Summary:
Enable these passes for CFI and WPD in ThinLTO and LTO with the new pass
manager. Add a couple of tests for both PMs based on the clang tests
tools/clang/test/CodeGen/thinlto-distributed-cfi*.ll, but just test
through llvm-lto2 and not with distributed ThinLTO.
Reviewers: pcc
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D49429
llvm-svn: 337461
This reverts commit r337081, therefore restoring r337050 (and fix in
r337059), with test fix for bot failure described after the original
description below.
In order to always import the same copy of a linkonce function,
even when encountering it with different thresholds (a higher one then a
lower one), keep track of the summary we decided to import.
This ensures that the backend only gets a single definition to import
for each GUID, so that it doesn't need to choose one.
Move the largest threshold the GUID was considered for import into the
current module out of the ImportMap (which is part of a larger map
maintained across the whole index), and into a new map just maintained
for the current module we are computing imports for. This saves some
memory since we no longer have the thresholds maintained across the
whole index (and throughout the in-process backends when doing a normal
non-distributed ThinLTO build), at the cost of some additional
information being maintained for each invocation of ComputeImportForModule
(the selected summary pointer for each import).
There is an additional map lookup for each callee being considered for
importing, however, this was able to subsume a map lookup in the
Worklist iteration that invokes computeImportForFunction. We also are
able to avoid calling selectCallee if we already failed to import at the
same or higher threshold.
I compared the run time and peak memory for the SPEC2006 471.omnetpp
benchmark (running in-process ThinLTO backends), as well as for a large
internal benchmark with a distributed ThinLTO build (so just looking at
the thin link time/memory). Across a number of runs with and without
this change there was no significant change in the time and memory.
(I tried a few other variations of the change but they also didn't
improve time or peak memory).
The new commit removes a test that no longer makes sense
(Transforms/FunctionImport/hotness_based_import2.ll), as exposed by the
reverse-iteration bot. The test depends on the order of processing the
summary call edges, and actually depended on the old problematic
behavior of selecting more than one summary for a given GUID when
encountered with different thresholds. There was no guarantee even
before that we would eventually pick the linkonce copy with the hottest
call edges, it just happened to work with the test and the old code, and
there was no guarantee that we would end up importing the selected
version of the copy that had the hottest call edges (since the backend
would effectively import only one of the selected copies).
Reviewers: davidxl
Subscribers: mehdi_amini, inglorion, llvm-commits
Differential Revision: https://reviews.llvm.org/D48670
llvm-svn: 337184
In order to always import the same copy of a linkonce function,
even when encountering it with different thresholds (a higher one then a
lower one), keep track of the summary we decided to import.
This ensures that the backend only gets a single definition to import
for each GUID, so that it doesn't need to choose one.
Move the largest threshold the GUID was considered for import into the
current module out of the ImportMap (which is part of a larger map
maintained across the whole index), and into a new map just maintained
for the current module we are computing imports for. This saves some
memory since we no longer have the thresholds maintained across the
whole index (and throughout the in-process backends when doing a normal
non-distributed ThinLTO build), at the cost of some additional
information being maintained for each invocation of ComputeImportForModule
(the selected summary pointer for each import).
There is an additional map lookup for each callee being considered for
importing, however, this was able to subsume a map lookup in the
Worklist iteration that invokes computeImportForFunction. We also are
able to avoid calling selectCallee if we already failed to import at the
same or higher threshold.
I compared the run time and peak memory for the SPEC2006 471.omnetpp
benchmark (running in-process ThinLTO backends), as well as for a large
internal benchmark with a distributed ThinLTO build (so just looking at
the thin link time/memory). Across a number of runs with and without
this change there was no significant change in the time and memory.
(I tried a few other variations of the change but they also didn't
improve time or peak memory).
Reviewers: davidxl
Subscribers: mehdi_amini, inglorion, llvm-commits
Differential Revision: https://reviews.llvm.org/D48670
llvm-svn: 337050
Summary:
I noticed that the .imports files emitted for distributed ThinLTO
backends do not have consistent ordering. This is because StringMap
iteration order is not guaranteed to be deterministic. Since we already
have a std::map with this information, used when emitting the individual
index files (ModuleToSummariesForIndex), use it for the imports files as
well.
This issue is likely causing some unnecessary rebuilds of the ThinLTO
backends in our distributed build system as the imports files are inputs
to those backends.
Reviewers: pcc, steven_wu, mehdi_amini
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D48783
llvm-svn: 336721
On Windows we've observed that if you open a file, write to it, map it into
memory and close the file handle, the contents of the memory mapping can
sometimes be incorrect. That was what we did when adding an entry to the
ThinLTO cache using the TempFile and MemoryBuffer classes, and it was causing
intermittent build failures on Chromium's ThinLTO bots on Windows. More
details are in the associated Chromium bug (crbug.com/786127).
We can prevent this from happening by keeping a handle to the file open while
the mapping is active. So this patch changes the mapped_file_region class to
duplicate the file handle when mapping the file and close it upon unmapping it.
One gotcha is that the file handle that we keep open must not have been
created with FILE_FLAG_DELETE_ON_CLOSE, as otherwise the operating system
will prevent other processes from opening the file. We can achieve this
by avoiding the use of FILE_FLAG_DELETE_ON_CLOSE altogether. Instead,
we use SetFileInformationByHandle with FileDispositionInfo to manage the
delete-on-close bit. This lets us remove the hack that we used to use to
clear the delete-on-close bit on a file opened with FILE_FLAG_DELETE_ON_CLOSE.
A downside of using SetFileInformationByHandle/FileDispositionInfo as
opposed to FILE_FLAG_DELETE_ON_CLOSE is that it prevents us from using
CreateFile to open the file while the flag is set, even within the same
process. This doesn't seem to matter for almost every client of TempFile,
except for LockFileManager, which calls sys::fs::create_link to create a
hard link from the lock file, and in the process of doing so tries to open
the file. To prevent this change from breaking LockFileManager I changed it
to stop using TempFile by effectively reverting r318550.
Differential Revision: https://reviews.llvm.org/D48051
llvm-svn: 334630
With the upcoming patch to add summary parsing support, IsAnalysis would
be true in contexts where we are not performing module summary analysis.
Rename to the more specific and approprate HaveGVs, which is essentially
what this flag is indicating.
llvm-svn: 334140
The DEBUG() macro is very generic so it might clash with other projects.
The renaming was done as follows:
- git grep -l 'DEBUG' | xargs sed -i 's/\bDEBUG\s\?(/LLVM_DEBUG(/g'
- git diff -U0 master | ../clang/tools/clang-format/clang-format-diff.py -i -p1 -style LLVM
- Manual change to APInt
- Manually chage DOCS as regex doesn't match it.
In the transition period the DEBUG() macro is still present and aliased
to the LLVM_DEBUG() one.
Differential Revision: https://reviews.llvm.org/D43624
llvm-svn: 332240
Summary:
This change is necessary for D46464, which will pass -1 as the Task
ID for distributed backends, so that the save temps files don't end
up with "4294967295" in their path. For distributed back ends, when -1
is passed, don't append any Task ID.
An existing test (tools/clang/test/CodeGen/thinlto_backend.ll) will
fail without this change after D46464.
Reviewers: pcc
Subscribers: mehdi_amini, inglorion, llvm-commits
Differential Revision: https://reviews.llvm.org/D46488
llvm-svn: 331591
Summary:
Set setDiagnosticsHotnessRequested before the early exit check for a
diagnostic output file, so that pass remarks with hotness works when
emitting pass remarks to stderr (e.g. via -pass-remarks=.).
Also fix the llvm-lto2 diagnistic handler so that it only calls exit(1)
when the diagnistic is an error type. Otherwise the new test invocation
of llvm-lto2 with -pass-remarks causes it to fail. The new code is
consistent with the diagnostic handler elsewhere (e.g. on the
LLVMContext).
Reviewers: pcc, davide
Subscribers: fhahn, mehdi_amini, llvm-commits, inglorion
Differential Revision: https://reviews.llvm.org/D46387
llvm-svn: 331569
Summary:
Support was added to the regular LTO backend, but not thinBackend.
This patch adds that support.
Reviewers: pcc, davide
Subscribers: mehdi_amini, inglorion, llvm-commits
Differential Revision: https://reviews.llvm.org/D46376
llvm-svn: 331481
Francis Visoiu Mistrih