…in the first function-attrs pass

Fix lit test too

Summary:
  Argument attributes like NoAlias and ReadOnly could affect memoryssa and thus earlyCSE in the function simplification pipeline.
https://reviews.llvm.org/D145210 adjusted PostOrderFunctionAttrs placement and caused the argument attributes not referred for the use
in the pipeline. This work (initiated by @nikic) unconditionally performs argument attribute inference in the first function-attrs pass.

Reviewers:
  aeubanks and nikic

Differential Revision:
  https://reviews.llvm.org/D156397

Change-Id: If9d1a1b165b708dddc03dfb4d33de2ee48e42844

Replace the ASO variant with upstream Debug.cpp

Change-Id: I5a1b0ae8d49a9d8d7ab49ce7a37eb46bde9d8c1b

Change-Id: I98f9dbf2b938cfe1774bc25b22bb543f46027f6d

Change-Id: I63bbe1eb7b76198ea4d1663bf3be6041f8d3db40

…ssert messages

Change-Id: Ia59e5bd3f9645213a15e68a959757749bef564aa

Change-Id: Id991d428db0c107790505e631257c3122e33281c

unxfails 3 tests
xfails: clang/test/CodeGen/X86/sm3-error.c

Change-Id: If8017b561cc0534a1c717119ecf861f8d6288d5a

Change-Id: If4d53d1e9fe69bd8d21eb155c0552acb03706d55

…upstream no longer uses named variables for master-worker handshaking in generic kernels

Change-Id: I50ca6c010f4c0b7c58b706385ceb71cea32f7c28

Change-Id: I415b21f1ce259e2ec018765ed97748d715e83de6

Change-Id: Id4842d20619920792cfb76b939695427857bf139

… target launch and data transfer operations

Implemented RAII objects, initialized at target entry points, that
invoke tool-supplied callbacks. Updated status of target callbacks as
implemented.

Depends on D127365

Patch from John Mellor-Crummey <[email protected]>
With contributions from:
Dhruva Chakrabarti <[email protected]>
Jan-Patrick Lehr <[email protected]>

Reviewed By: jdoerfert, dhruvachak, jplehr

Differential Revision: https://reviews.llvm.org/D127367

Change-Id: Ic6fcee7059aa4e6237e81e2a702adca6a26bcdc6

Change-Id: I1a79f3e6361fece67bee02fb55a62849d650b15d

Change-Id: Ib53689bd8e7eed86af1cff01a97eb3d6592fe9e4

The attributes changes were left out of Clang 17.
Attributes that used to take a string literal now accept an unevaluated
string literal instead, which means they reject numeric escape sequences
and strings literal with an encoding prefix - but the later was already
ill-formed in most cases.

We need to know that we are going to parse an unevaluated string literal
before we do - so we can reject numeric escape sequence,
so we derive from Attrs.td which attributes parameters are expected
to be string literals.

Reviewed By: aaron.ballman

Differential Revision: https://reviews.llvm.org/D156237

Change-Id: I8dcf4c4de75a3f7b089d04cf25b9a20682fa72ff

…source_symbol' and 'uuid' attributes as unevaluated

This is a complementary to D156237.
These attributes have custom parsing logic.

Reviewed By: cor3ntin

Differential Revision: https://reviews.llvm.org/D159024

Change-Id: Icb6d3e0f9ea02b4058a567e5c998be71a3aea7c2

Change-Id: Ie972cb4507ff7b20727e7d1ce7275b18786f2efb

Change-Id: Ia0a6077c7a6eaef44d8a915123e11b4cf24489b0

Change-Id: I3c8e3576cef0a1cbf77bc7e76c0af527938128de

Change-Id: I17ec52cc181708d1134949c66781d2ae0be8ba9f

Change-Id: Iad8bd30ae767ce015d6d39987c7e8e4b07d186df

Change-Id: Id69edef44c1f1fb3eefe9a71413a00c831ba10ee

Revert "[OpenMPOpt] Allow indirect calls in AAKernelInfoCallSite (#65836)"

Change-Id: I790c81ab7d92e0f828e81535cb131c6c45248138

Only a subset of the fields of DbgVariable are meaningful at any time,
and some fields are re-used for multiple purposes (for example
FrameIndexExprs is used with a throw-away frame-index of 0 to hold a
single DIExpression without needing to add another member). The exact
invariants must be reverse-engineered by inspecting the actual use of
the class, its imprecise/outdated doc-comment, and some asserts.

Refactor DbgVariable into a sum type by inheriting from std::variant.
This makes the active fields for any given state explicit and removes
the need to re-use fields in disparate contexts. As a bonus, it seems to
reduce the size on my x86_64 linux box from 144 bytes to 96 bytes.

There is some potential cost to `std::get` as it must check the active
alternative even when context or an assert obviates it. To try to help
ensure the compiler can optimize out the checks the patch also adds a
helper `get` method which uses the noexcept `std::get_if`.

Some of the extra cost would also be avoided more cleanly with a
refactor that exposes the alternative types in the public interface,
which will come in another patch.

Differential Revision: https://reviews.llvm.org/D158675

[NFC][AsmPrinter] Remove dead multi-MMI handling from DwarfFile::addScopeVariable

Differential Revision: https://reviews.llvm.org/D158676

[NFC][AsmPrinter] Expose std::variant-ness of DbgVariable

Differential Revision: https://reviews.llvm.org/D158677

[NFC][AsmPrinter] Use std::visit in constructVariableDIEImpl

This potentially has a slightly positive performance impact, as
std::visit can be implemented as a `switch`-like jump rather than
a series of `if`s.

More importantly, the reader can be confident is no overlap between the
cases.

Differential Revision: https://reviews.llvm.org/D158678

Change-Id: Ie5b1fead7b4a4407f73b295530e46e5ce37f638e

This change follows from https://reviews.llvm.org/D156416 We include
`llvm-remark-size-diff` as a part of `llvm-remarkutil` under a
subcommand `size-diff`.

Differential Revision: https://reviews.llvm.org/D156515

Change-Id: I1d9ae39265a31d55e56930f910bf5d3a6d5f235b

Change-Id: I0e7d3045cb8d9e0256f22a0e96bb19d7cc6185a9

Change-Id: Id23677ba24e8ac05d55a269a29abd499c8c0d1d2

Pass a default DW_MSPACE_LLVM_none to satisfy new API

Change-Id: I50df461f00b5510a715f55f61107122318102d22

Change-Id: I11367080ee9fc14a8692fc63f58a31f84623d83d

Change-Id: I8efbea3c5f26b760629b912cf11519af1b9b5f97

Change-Id: I32fde7c902bf410c6c8208292b4f475f243635d4

Change-Id: I5d333e8d2a166874dabd6ad82e4376fedfd77733

Introduces an initial implementation of an OMPT-based unit testing
framework inspired by GoogleTest. In a unit test, defined via the
OMPTTESTCASE(TestSuite, TestCase) macro, a sequence of OMPT events can
be asserted to happen.
As of now, it needs to be done by hand via the `insert` method of the
SequencedAsserter. Eventually this will be hidden behind a macro.

The file openmp/libomptarget/test/ompTest/src/main.cpp serves as a demo
for both what is currently available and an intial sketch of how the
macros may look like.

As of now, the OMPT events are checked when they happen. This means that
the events
- Device Init
- Device Load
do currently not work for assertions as they happen before the Asserter
object is created. This is likely going to be addressed in future
additions, as it seems to be a critical part.

Change-Id: Id178c5f2b378a3a6ebbecf7deea3bf7f0a52f6cb

Revert "clang/OpenCL: Add inline implementations of sqrt in builtin header"
  breaks libdevice builds

Change-Id: Ib56f8fcb1924a2e1a504c511756974b2a8d5474a

This reverts commit dcd907b.

Change-Id: I25c29b6d0c40125fec0c22bd24fc57fb044804a4

Change-Id: Ifd6cb1c631bca497f029890a8ab4235fde5f6170

Change-Id: I5d4962df784c20863de48ccdee2db836faecaaba

wq
The Attributor has gained support for indirect calls but it is opt-in.
This patch makes AAKernelInfoCallSite able to handle multiple potential
callees.

[LLVM][OpenMPOpt] Fix a crash when associated function is nullptr (#66274)

The associated function can be a nullptr if it is an indirect call.
This causes a crash in `CheckCallee` which always assumes the callee
is a valid pointer.

Fix #66904.

Change-Id: Ief6ea28456dc5c01709049153fe7e30f6aa4cb29

Change-Id: I73de1015b02085ee43a40358fc198798dfe0ebfc

Change-Id: I069f2344d4a6e86c365e6cc692bfcca627178db8

Change-Id: I53c80d21eb059988b295a1abec9acdd1a3470bca

Change-Id: I308e2e5d3d4b237c2d7861095e2e438956cf496f

Change-Id: I024513a0810861b6aa8a9f16e14b6f6841e74468

Change-Id: I384198f35208b652012e094422eab882f2d5cbbf

Change-Id: Ic57706bf2d11a13a4105c471f4f69c603a295a3f

Change-Id: Ib913b4af93bbc62db992be9b3d6aba7058fddf79

Change-Id: Ib699a2fd9e9cf5f40f295163cd8f732fa85ad41b

Change-Id: I1c81a40a36ff47b36cafb37f7baa4e1aa66dbfc9

Change-Id: I27c20d24ef00ec92712f8144e3ac89fd7c2f4ca6

Change-Id: I4dc0838450583025dcf64438a1c7cdac32641621

Change-Id: I4494e70b6e969ccf480e5601adf11abe17740ee4

Change-Id: I84d1d117327870990b9f60ba039206d1c3e1e1e4

Change-Id: I7539b45f16e1a61f760dc47e03307a5f7d866b07

Change-Id: I2b265848a94f12db2472e91cd2c6192dfe1a4164

Change-Id: I45cbfc779e3da70bd601505d3fb0a2cc47cb8ad6

Change-Id: Ifeaa6e94db19208bb5ef31f24408bf1772c06f29

Change-Id: I8dc9993e558c98d1f3efdb1bc88dc5a964b6c5e3

Change-Id: I82d395afc76c019bea5708e97f39324edeceb726

Change-Id: Ic1074dc5a100be2684de61515e426064c364e439

Change-Id: If7f685e2073742656531f7d5542159d72a57351d

Change-Id: Iabd20a90fbf7e96b3f943072066e9ba2364222dd

Change-Id: I10c38e2210b0fd28fd08501164dd9cb98ecf95b5

… of using ROCt, the envvar HSA_XNACK and ther kernel parameter noretry is used. Removed dependencies on ROCt

Change-Id: I768d4566df664ff7fe45a8ed932d5905b6ae9b09

Check after linking bitcodes instead of before. Also fix issue with
wrong variable being printed after check

Change-Id: I14d219a2f5656c6789d5abbf164a04c27d4ab395

For tests that use stdout/stderr as part of the verification process,
sending logs to those output streams causes the test to fail. Previously,
we temporarily disabled the log redirect if set to stdout. With this patch,
we do the same for stderr

Change-Id: Ia0c2125d97fbe76094e8c62df6405e65f19c09df

Change-Id: Ia8742f8c50407ad2c728c708883692fdf5010d06

Change-Id: I497cbf9a073376975e41e9424277089cd9723f3c

Change-Id: I2971b6733b09d0285c2892d6832b0329aeeb20d2

Change-Id: I9085b3c3d606aac5e87456678804b3a94a6fa1ee

Fold constructVariableDIEImpl into constructVariableDIE,
simplify it and group related functions.

Pull out the previously inline lambdas for visiting the active
variant of the DbgVariable to add location and related attributes
as an overload set for a private method
applyConcreteDbgVariableAttributes.

Rename applyVariableAttribute to reflect what kinds of attributes it
applies, and to contrast it with the new
applyConcreteDbgVariableAttributes.

Move constructLabelDIE down in the implementation file, so all of
the constructVariableDIE-related function impls are adjacent.

Change-Id: I8c87340c3f859fed1a2cfb300b1d9e4c22eea93f

rocm-docs-core 0.24.0

Also adds amd/hipcc/.readthedocs.yaml configuration file for ReadtheDocs when building docs

Change-Id: Ie9c67d04b7a073fafddc80ccf5c283179791cb64

Change-Id: Ia8f490e0cb545168580e97866b903937fb7239be

Change-Id: I39edb8b5a926c15d53b164fcdcbe4fc7f05388f6

Fixes ROCm/ROCm#2058.

ompt_start/flush/stop_trace now accept a device and actually do
something with it. ompt_set_trace_ompt still ignores the device.

The runtime will now use the correct device-id while invoking buffer
request/complete callbacks.

For start_trace, this support enables tracing on selected devices.
Unique buffer request/complete callbacks can be attached to a given
device-id. All of this info will be tracked and the device-id
returned to the tool through the completion callbacks.

For flush_trace, the runtime will use the device-id to return buffers
only for the selected device to the tool.

For stop_trace, the runtime will first flush the buffers for the
selected device and stop tracing for that device only. If tracing has
been stopped for all devices, the helper threads will be shut down as
well.

Regardless the device-id used in the flush/stop APIs, buffers for all
devices will be automatically flushed as they become full, per the
OpenMP specification.

We don't do anything with the device id passed through
ompt_set_trace_ompt since using that information to filter out tracing
will introduce locking in the critical tracing path of OpenMP worker threads,
something we would like to avoid. If this feature is indeed very
important for the tool, we can introduce this feature later. The
downside of ignoring the device id is that a tool does not have the
ability to selectively control the events that get traced on a device
basis.

Added a related change to cache the thread_id in a thread local variable
to reduce locking.

Change-Id: Ida3401bb295158a0777e1c02c84dcf16bfa116c6

Change-Id: I989960f3b46412c70b377b97c57bb35c8ef17b64

Update to the commit landed in
http://gerrit-git.amd.com/c/lightning/ec/llvm-project/+/926948 to match
upstream review.

I had intended to wait until the upstream review
llvm/llvm-project#66435 was finished before
submitting the gerrit merge of it, but accidentally submitted to soon.

Change-Id: Ib244a34b662584b3314f70371e4527870ddea8e5

OpenCL allows a kernel function to be called from another kernel function,
and this behavior is featured in the primary OpenCL files used in the Comgr
tests. However, this behavior is atypical and rarely-used.

In this patch, we refactor this behavior into a separate test,
and restructure the primary tests to not call nested kernels. This will allow
us to more quickly identify issues specifically with nested kernels without
failing a large array of tests.

Change-Id: I1772db85b770977012798940dd2cba5fc2cdf8ac

Change-Id: I49a7fdd14470e7b1aee3bfacefa2d4246ddd67ae

Change-Id: I5592e87a1f83330264effd489c16fdcb186a8e8c

Change-Id: Ibc93f693bb08536769e4225d812aac68c904833a

Change-Id: Id00be9c00c2cf47db6a4e7de5b62464d347f97b9

Change-Id: Id3a9d34c7142e1ed7859592bcc4d793b95f203a1

Change-Id: Ia0d155f5580b450c5f224a1aabd02415cf869a16

Change-Id: I1d6dee4c7dd1b2af8e2c426bec8efac00e181884

Change-Id: Ia2e9ab7b4d063e9caf1f247b20dface4342a2cad

Change-Id: I04a2d4fba7be8dc01d677d7476c9d7ef88793c55

Change-Id: I5ef810f0892b20b70e78a27b290357d1e8ae4a0a

Change-Id: Ie6d1c30485f4a15f31ddce8658199929762a1517