This reverts commit d9333e36.

Ported from old amdgcn intrinsic which will soon be deleted.

https://reviews.llvm.org/D149587

In Orc runtime, we use `dlopen(nullptr, ...)` to open current executable and use `dlsym` to find addresses of symbols, this requires `-rdynamic` flag.

As `llvm/CMakeLists.txt` suggests

```
# Make sure we don't get -rdynamic in every binary. For those that need it,
# use export_executable_symbols(target).
```
This patch exports symbols in `ClangReplInterpreterExceptionTests`. This also fixes `ClangReplInterpreterExceptionTests` is skipped on ppc64 when jitlink is used.

Reviewed By: v.g.vassilev

Differential Revision: https://reviews.llvm.org/D159167

[D156301](https://reviews.llvm.org/D156301

) introduced atomic
optimizations for FAdd/FSub. For FSub, reduction/scan needs to be
performed using add operation (`not sub`) and memory location will be
updated by reduced value using atomic sub later by only one lane.

---------

Authored-by: Pravin Jagtap <Pravin.Jagtap@amd.com>

While creating a temporary alloca for a box in OpenMp region, the
insertion point should be the OpenMP region block instead of the
function entry block.

Updates GetInstructionSize to account for arm64 instruction sizes.

ARM64 instruction are always 4 bytes long but GetInstructionSize in
interception_win.cpp assumes x86_64 which has mixed sizes.

Fix is for: https://github.com/llvm/llvm-project/issues/64319

Before the changeclang_rt.asan_dynamic-aarch64.dll would crash at:
OverrideFunction -> OverrideFunctionWithHotPatch -> GetInstructionSize:825

After the change:
dllthunkintercept -> dllthunkgetrealaddressordie -> InternalGetProcAddress

This reverts commit 7fda1b74.

This fixes a build error introduced by https://reviews.llvm.org/D153587
when using an old version of GCC. See
https://reviews.llvm.org/D153587#4644735 for details.

Since we use C++20 to build the dylib, we can use a lambda to do the
first-time initialization instead of emulating std::bind. This should
not change the behavior of the code at all, it merely simplifies it.

This removes a symbol from the dylib, however that symbol was only ever
used inside the dylib so it shouldn't break the ABI for anyone. I
confirmed that by searching for that symbol on the ABI boundary of a
large number of programs and couldn't find any references to that
function.

This reverts commit e831a32c.

Actually pass along the depth parameter of getKnownBits to computeKnownBitsImpl.

Reviewed By: aemerson

Differential Revision: https://reviews.llvm.org/D159321

Instead of creating a copy of the vector, we should just pass a
reference along. The only method that calls this Ctor also holds onto a
non-mutable reference to the vector of strings so a copy should be
unnecessary.

- Remove usages of the non-existent `ignore-forks` field, conditions in
jobs already exist to prevent the jobs from running in forks.
- Don't use variables in the `printf` format string. Use `printf
"..%s.." "$foo"`. ([SC2059](https://www.shellcheck.net/wiki/SC2059))
- Double quote variable expansion to prevent globbing and word
splitting. ([SC2086](https://www.shellcheck.net/wiki/SC2086))
- Prefer `[ p ] || [ q ]` as `[ p -o q ]` is not well defined.
([SC2166](https://www.shellcheck.net/wiki/SC2166))
- Consider `{ cmd1; cmd2; } >> file` instead of individual redirects.
([SC2129](https://www.shellcheck.net/wiki/SC2129))
- Use `$(...)` notation instead of legacy notation `...`.
([SC2006](https://www.shellcheck.net/wiki/SC2006))
- Use `./*glob*` or `-- *glob*` so names with dashes won't become
options. ([SC2035](https://www.shellcheck.net/wiki/SC2035))
- Refactor JavaScript code in certain workflows.
- Change workflow variable substitution style of some wor...

This "bug" was probably not noticed because it doesn't affect any integer
type we currently support. It requires integers with more than 2x the
size of `unsigned long long`. However, with such types, the algorithm
used to break down the large integer into groups of size `unsigned long long`
didn't work because we rotated in the wrong direction.

For example, the 256 bit number (1 << 255) would yield the wrong answer
when used with the algorithm before this patch.

In particular, note that the current rotation happens to work for 128 bit
integers because it just swaps the halves in this case.

Differential Revision: https://reviews.llvm.org/D134625


Co-authored-by: Louis Dionne <ldionne.2@gmail.com>

Adopt the new markup overload, introduced in 77d10325, in the ARM
backend. This commit completes the migration and removes the old
overload.

Summary:
This patch improves the implementation of the standard `rand()` function
by implementing it in terms of the xorshift64star pRNG as described in
https://en.wikipedia.org/wiki/Xorshift#xorshift*. This is a good,
general purpose random number generator that is sufficient for most
applications that do not require an extremely long period. This patch
also correctly initializes the seed to be `1` as described by the
standard. We also increase the `RAND_MAX` value to be `INT_MAX` as the
standard only specifies that it can be larger than 32768.

This commit removes an optimization that skips the initialization of the
reduction struct if the number of threads in a team is 1. This
optimization
caused a bug with Hidden Helper Threads. When the task group is
initially
initialized by the master thread but a Hidden Helper Thread executes a
target
nowait region, it requires the reduction struct initialization to
properly
accumulate the data.

This commit also adds a LIT test for issue #57522 to ensure that the
issue is
properly addressed and that the optimization removal does not introduce
any
regressions.

Fixes: #57522

- Check `MakePointer*` load/store attribute values.
- Support coop matrix types in `MatrixTimesScalar` verification.
- Add test cases for all the remaining ops that accept coop matrix types.
- Split NV and KHR tests.

This patch fixes:

  llvm/lib/Target/AMDGPU/AMDGPU.h:297:18: error: private field 'TM' is
  not used [-Werror,-Wunused-private-field]

This addresses issues introduced by efe4a548

AutoFDO profile has no leading 0x in hex dumps.

Reviewed By: #bolt, rafauler

Differential Revision: https://reviews.llvm.org/D159507

Add const qualifier to avoid warning "_cast from 'const char *' to 'char*' drops const qualifier [-Werror,-Wcast-qual]_"

This will allow enable LLVM_ENABLE_WERROR when build with clang-cl on Windows.
This is imported from https://github.com/openbsd/src/commit/b81763002452802e4f7304ea60f121253bd94

Also removed the gcc change for the cast-qual warning which is not needed with const qualifier added.

This patch is to add the support of declaring a Cray pointer in a
module.

Propagate "branch_weights" metadata whe turning a select into a
conditional branch in tryToUnfoldSelectInCurrBB

Fixes https://github.com/llvm/llvm-project/issues/57544

This patch ports the AMDGPURewriteUndefForPHI pass to the new pass
manager. With this, the pass is supported under both the legacy and the
new pass managers.

---------

Co-authored-by: Jun Wang <jun.wang7@amd.com>

It's weird that we're specifying this in both the dxil and the LLVM IR
way, but if we are we should at least be consistent about it.

These tests are still relatively flaky on certain buildbots and on some
developer machines. This patch hopefully reduces the noise produced by
these tests in those environments by setting the retry count to 2 so
that they will hopefully flaky pass rather than just fail while I
perform more investigation.

https://reviews.llvm.org/D158130

https://reviews.llvm.org/D158131

We want the !fpmath metadata to be attached to the sqrt intrinsic to
make it to the backend lowering. Emit an available_externally
definition which uses the builtin, which emits the !fpmath.

Fixes #64264

https://reviews.llvm.org/D156743

Make codegen emit correctly rounded sqrt by default.

Emit the fast but only kind of fast expansion in AMDGPUCodeGenPrepare
based on !fpmath, like the fdiv case. Hack around visitation ordering
problems from AMDGPUCodeGenPrepare using forward iteration instead of
a well behaved combiner.

https://reviews.llvm.org/D158129

This will reduce the number of notifications created when a pull request
label is added. Each team will only get a notification when their team's
label is added and not when other teams' labels are added.

This loop is wrong, most of targets are not defined yet.
Also if we build with LLVM_ENABLE_RUNTIMES, these deps
are irrelevant.