Reviewed By: cota

Differential Revision: https://reviews.llvm.org/D119959

Since the code for apply data relocations can sometimes use
the values stored in he globals, they need to be relocated
before the data relocations can be run.

Fixes: https://github.com/emscripten-core/emscripten/issues/13398

Differential Revision: https://reviews.llvm.org/D119666

When we move an allocation from the heap to the stack we need to
allocate it in the alloca AS and then cast the result. This also
prevents us from inserting the alloca after the allocation call but
rather right before.

Fixes https://github.com/llvm/llvm-project/issues/53858

When we use liveness for edges during the `genericValueTraversal` we
need to make sure to use the AAIsDead of the correct function. This
patch adds the proper logic and some simple caching scheme. We also
add an assertion to the `isEdgeDead` call to make sure future misuse
is detected earlier.

Fixes https://github.com/llvm/llvm-project/issues/53872

llvm-config wants all libraries referenced in
llvm/lib/CMakeLists.txt to exist on disk.

But WindowsDriver is only referenced in clang and lld and hence
wasn't built as a dependency of check-llvm.

Add it as an explicit dependency to make llvm-config happy.

This patch adds requirement for the `scripted_crashlog_json` test to
make sure it only runs on apple silicon systems.

This should fix the following green dragon failure:
https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake/41454



Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>

This patch adds Win32 to the list of supported triples in
`fir::CodeGenSpecifics`. This change means that we can use the "native"
triple, even when running tests on Windows. Currently this affects only
1 test, but it will change once we start adding more tests for lowering
and code-generation.

Differential Revision: https://reviews.llvm.org/D119332

This patch passes in the AMDPGU math libraries to the linker wrapper.
The wrapper already handles linking OpenMP bitcode libraries via the
`--target-library` option. This should be sufficient to link in math
libraries for the accompanying architecture.

Fixes #53526.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D119841

Summary:
We were not previously saving strings when saving symbol names during
LTO symbol resolution. This caused a crash inside the dense set when
some of the strings would rarely be moved internally by the object file
class.

Fixes https://github.com/llvm/llvm-project/issues/24781.
Fixes https://github.com/llvm/llvm-project/issues/38160.

This patch splits `TT_RecordLBrace` for classes/enums/structs/unions (and other records, e.g. interfaces) and uses the brace type to avoid the error-prone scanning for record token.

The mentioned bugs were provoked by the scanning being too limited (and so not considering `const` or `constexpr`, or other qualifiers, on an anonymous struct variable declaration).

Moreover, the proposed solution is more efficient as we parse tokens once only (scanning being parsing too).

Reviewed By: MyDeveloperDay, HazardyKnusperkeks

Differential Revision: https://reviews.llvm.org/D119785

I added posix_madvise out of order before.

This patch upstreams some changes we've made internally to how we're
building the libc++ dylib on Apple platforms. The goal is still to
eventually get rid of `apple-install-libcxx.sh` entirely and have a
proper way to mirror what we do internally with just the normal CMake
configuration.

Differential Revision: https://reviews.llvm.org/D118912

Differential Revision: https://reviews.llvm.org/D119947

A significant number of our tests in C accidentally use functions
without prototypes. This patch converts the function signatures to have
a prototype for the situations where the test is not specific to K&R C
declarations. e.g.,

  void func();

becomes

  void func(void);

This is the twelfth batch of tests being updated (the end may be in
sight soon though).

We now use posix_madvise, so we need to allow this symbol.

Differential Revision: https://reviews.llvm.org/D119970

The `IsSPMD` global can only be read by threads other than the main
thread *after* initialization is complete. To allow usage of
`mapping::getBlockSize` before initialization is done, we can pass the
`IsSPMD` state explicitly. This is similar to other APIs that take
`IsSPMD` explicitly to avoid such a race, e.g.,
`mapping::isInitialThreadInLevel0(IsSPMD)`

Fixes https://github.com/llvm/llvm-project/issues/53857

`UsedAssumedInformation` is a return argument utilized to determine what
information is known. Most APIs used it already but
`genericValueTraversal` did not. This adds it to `genericValueTraversal`
and replaces `AllCallSitesKnown` of `checkForAllCallSites` with the
commonly used `UsedAssumedInformation`.

This was supposed to be a NFC commit, then the test change appeared.
Turns out, we had one user of `AllCallSitesKnown` (AANoReturn) and the
way we set `AllCallSitesKnown` was wrong as we ignored the fact some
call sites were optimistically assumed dead. Included a dedicated test
for this as well now.

Fixes https://github.com/llvm/llvm-project/issues/53884

I missed it in my grep. Fixes broken buildbot.`

Identifiers with __ anywhere are reserved. I picked this up via the
bugprone-reserved-identifier clang-tidy check but -Wreserved-identifier will
also flag these uses as well.

Differential Revision: https://reviews.llvm.org/D119915

In some cases, it can happen that crashlogs don't have any signal in
the exception, which causes the parser to crash.

This fixes the parsing by checking if the `signal` field is in the
`exception` dictionary before trying to access it.

rdar://84552251

Differential Revision: https://reviews.llvm.org/D119504



Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>

This patch introduces a new type of ScriptedProcess: CrashLogScriptedProcess.
It takes advantage of lldb's crashlog parsers and Scripted Processes to
reconstruct a static debugging session with symbolicated stackframes, instead
of just dumping out everything in the user's terminal.

The crashlog command also has an interactive mode that only provide a
very limited experience. This is why this patch removes all the logic
for this interactive mode and creates CrashLogScriptedProcess instead.

This will fetch and load all the libraries that were used by the crashed
thread and re-create all the frames artificially.

rdar://88721117

Differential Revision: https://reviews.llvm.org/D119501



Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>

This patch adds the ability for the user to check if the command
interpreter's IOHandler is interactive.

Differential Revision: https://reviews.llvm.org/D119499



Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>

This patch adds the ability for ScriptedThread to load artificial stack
frames. To do so, the interpreter instance can create a list that will
contain the frame index and its pc address.

Then, when the Scripted Process plugin stops, it will refresh its
Scripted Threads state by invalidating their register context and load
to list from the interpreter object and reconstruct each frame.

This patch also removes all of the default implementation for
`get_stackframes` from the derived ScriptedThread classes, and add the
interface code for the Scripted Thread Interface.

rdar://88721095

Differential Revision: https://reviews.llvm.org/D119388



Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>

This patch removes the `my_scripted_process.py` blueprint since it's not
used anymore.

The patch also updates the base ScriptedProcess and ScriptedThread
initializers to automatically initialize convinience variables, to
access debugger from the ScriptedProcess, access the SBProcess and
ScriptedProcess object from a ScriptedThread instance.

Differential Revision: https://reviews.llvm.org/D119386



Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>

It's not particularly user-friendly to have to call `initLRU` everywhere. Also,
it wasn't particularly great that the LRU for registers used in a sequence was
also initialized by `initLRU`.

This patch hides this stuff behind some helper functions:

* `isAvailableAcrossAndOutOfSeq`
* `isAnyUnavailableAcrossOrOutOfSeq`
* `isAvailableInsideSeq`

This allows the user to avoid calling `initLRU` explicitly. Also, it allows
us to separate initializing the used-in-sequence LRU from the main LRU.

Since both ARM and AArch64 check LR liveness in `insertOutlinedCall`, this
refactor requires that we de-const the Candidate there.

Some other quality-of-code improvements:

* LRUs in outliner::Candidate now have more descriptive names
* Use `Register` instead of `unsigned` in some places
* Improve readability in some places by using ranges rather than `std::for_each`

This is a preparatory commit for a larger compile time related change for the
AArch64 outliner.

This patch adds infrsatrcutrue to be able to lower
arguments in functions and subroutines.

This patch is part of the upstreaming effort from fir-dev branch.

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D119957



Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>

zext(umin(x,y)) == umin(zext(x),zext(y))
zext(x) == 0  ->  x == 0

While it is not a very likely scenario, we probably should not expect
that instcombine already dropped such a redundant zext,
but handle directly. Moreover, perhaps there was no ZExtInst,
and SCEV somehow managed to  pull out said zext out of the SCEV expression.

zext(umin(x,y)) == umin(zext(x),zext(y))
zext(x) == 0  ->  x == 0

Extra leading zeros do not affect the result of comparison with zero,
nor do they matter for the unsigned min/max,
so we should not be dissuaded when we find a zero-extensions,
but instead we should just skip it.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D119912

Reviewed By: hctim

Differential Revision: https://reviews.llvm.org/D119961

Ensure that function definitions match their declrations in header
files, even if they have no effect on linking. This includes

 1. Both have the same __isl_* annotations

 2. Both use the same type alias

 3. Remove unused declarations that have no definition

 4. Use explicit polly namespace qualifier for definitions; generally,
    the .cpp file should use at most an anon namespace region since
    only symbols declared in the header file can be accessed from other
    translation units anyway. For defintions that have been declared in
    the header file, the explicit namespace qualifier ensures that both
    match.

This commit adds a pattern to wrap a tensor.pad op with
an scf.if op to separate the cases where we don't need padding
(all pad sizes are actually zeros) and where we indeed need
padding.

This pattern is meant to handle padding inside tiled loops.
Under such cases the padding sizes typically depend on the
loop induction variables. Splitting them would allow treating
perfect tiles and edge tiles separately.

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D117018