This should have happened a long time ago, now that openmp.llvm.org
redirects to openmp.llvm.org/docs we completely switched over to the
sphinx documentation page instead.
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D108588
Since these assumptions are coming from OpenMP it makes sense to mark
them as such in the generic IR encoding. Standardized assumptions will
be named
omp_ASSUMPTION_NAME
and extensions will be named
ompx_ASSUMPTION_NAME
which is the OpenMP 5.2 syntax for "extensions" of any kind.
This also matches what the OpenMP-Opt pass expects.
Summarized,
#pragma omp [...] assume[s] no_parallelism
now generates the same IR assumption annotation as
__attribute__((assume("omp_no_parallelism")))
Reviewed By: jhuber6
Differential Revision: https://reviews.llvm.org/D105937
Not all address spaces support initializers for globals and we can
therefore not set them without checking if they are allowed. This
patch adds a hook into TTI to check if an AS allows non-undef
initializers. We disable it for all but address space 0 by default,
NVPTX and AMDGPU targets allow all but address space 3.
Reviewed By: tra
Differential Revision: https://reviews.llvm.org/D109337
When we guard side-effects as part of SPMDzation we do it for
consecutive instructions that need guarding. This patch will try to
reorder guarded side-effects in a block to decrease the number of
guarded regions we need. It does not use any smarts, e.g., alias
analysis, to move side-effects over non-interfering reads. Instead,
it only moves side-effects downwards to the next guarded side-effect
if there was nothing in between that could have possibly be affected.
Reviewed By: ggeorgakoudis
Differential Revision: https://reviews.llvm.org/D109070
Always use the byval/inalloca/preallocated type (which is required
nowadays), don't fall back on the pointer element type.
This requires adding Function::getParamPreallocatedType() to
mirror the CallBase API, so that the templated code can work with
both.
LICM may have pulled out a splat, but with .vx instructions we
can fold it into an operation.
This patch enables CGP to reverse the LICM transform and move the
splat back into the loop.
I've started with the commutable integer operations and shifts, but we can
extend this with more operations in future patches.
Reviewed By: frasercrmck
Differential Revision: https://reviews.llvm.org/D109394
Remove some unnecessary logging from wasm-ld when running under
`--verbose`. Unlike `-debug` this logging is available in release
builds. This change makes it little more minimal/readable.
Also, avoid compiling the `debugWrite` function in releaase builds
where it does nothing. This should remove a lot debug strings from
the binary, and avoid having to construct unused debug strings at
runtime.
Differential Revision: https://reviews.llvm.org/D109583
The re-use of this struct across iterations of the loop was causing
fields (specifically Name) to be incorrectly shared between multiple
sections.
Differential Revision: https://reviews.llvm.org/D108984
Test using debug-only=CodeObjectComaptibility was failing in
non-assert builds, so it has been moved to a different file which
requires assert.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D109592
Currently, opaque pointers are supported in two forms: The
-force-opaque-pointers mode, where all pointers are opaque and
typed pointers do not exist. And as a simple ptr type that can
coexist with typed pointers.
This patch removes support for the mixed mode. You either get
typed pointers, or you get opaque pointers, but not both. In the
(current) default mode, using ptr is forbidden. In -opaque-pointers
mode, all pointers are opaque.
The motivation here is that the mixed mode introduces additional
issues that don't exist in fully opaque mode. D105155 is an example
of a design problem. Looking at D109259, it would probably need
additional work to support mixed mode (e.g. to generate GEPs for
typed base but opaque result). Mixed mode will also end up
inserting many casts between i8* and ptr, which would require
significant additional work to consistently avoid.
I don't think the mixed mode is particularly valuable, as it
doesn't align with our end goal. The only thing I've found it to
be moderately useful for is adding some opaque pointer tests in
between typed pointer tests, but I think we can live without that.
Differential Revision: https://reviews.llvm.org/D109290
This patch tests code in D108976. This split is necessary to avoid temporary regression.
Depends on D108974,
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D109158
During dependency scanning, we generally want to suppress -Werror. Apply the same logic to the DiagnosticOptions instance used for command-line parsing.
This fixes a test failure on the PS4 bot, where the system header directory could not be found, which was reported due to -Werror being on the command line and not being sanitized.
This patch implements legalization of EXTRACT_SUBVECTOR for the case
where the result needs promoting, and the input type requires widening.
Reviewed By: frasercrmck
Differential Revision: https://reviews.llvm.org/D109509
This patch implements legalization of EXTRACT_SUBVECTOR for the case
where the result needs promoting, and the input type is either legal
or requires splitting.
The idea is that the operation is broken down into simpler steps,
by first extracting a smaller subvector until the input vector
becomes legal or requires promotion.
Reviewed By: CarolineConcatto
Differential Revision: https://reviews.llvm.org/D109313
GDB uses normalized errno values for vFile errors. Implement
the translation between them and system errno values in the gdb-remote
plugin.
Differential Revision: https://reviews.llvm.org/D108148
Fall back to QEnvironmentHexEncoded if QEnvironment is not supported.
The latter packet is an LLDB extension, while the former is universally
supported.
Add tests for both QEnvironment and QEnvironmentHexEncoded packets,
including both use due to characters that need escaping and fallback
when QEnvironment is not supported.
Differential Revision: https://reviews.llvm.org/D108018
Implement the simpler vRun packet and prefer it over the A packet.
Unlike the latter, it tranmits command-line arguments without redundant
indices and lengths. This also improves GDB compatibility since modern
versions of gdbserver do not implement the A packet at all.
Make qLaunchSuccess not obligatory when using vRun. It is not
implemented by gdbserver, and since vRun returns the stop reason,
we can assume it to be successful.
Differential Revision: https://reviews.llvm.org/D107931
Add a GDB-compatible fallback to vFile:fstat for vFile:mode, and to
vFile:open for vFile:exists. Note that this is only partial fallback,
as it fails if the file cannot be opened.
Differential Revision: https://reviews.llvm.org/D107811
Add two new commands 'platform get-file-permissions' and 'platform
file-exists' for the respective bits of LLDB protocol. Add tests for
them. Fix error handling in GetFilePermissions().
Differential Revision: https://reviews.llvm.org/D107809
Create a common GDBPlatformClientTestBase class and move the platform
select/connect logic there to reduce duplication.
Differential Revision: https://reviews.llvm.org/D109585
In this patch the dependency scanner starts using proper `DiagnosticOptions` parsed from the actual TU command-line in order to mimic what the actual compiler would do. The actual functionality will be enabled and tested in follow-up patches. (This split is necessary to avoid temporary regression.)
Depends on D108976.
Reviewed By: dexonsmith, arphaman
Differential Revision: https://reviews.llvm.org/D108982
The original version of the bufferization pattern for linalg.generic would
manually clone operations within the region to the bufferized clone of the
operation. This triggers legality requirements on those operations in the
conversion infra. Instead, this now uses the rewriter to inline the region
instead, avoiding those legality requirements.
Differential Revision: https://reviews.llvm.org/D109581
LoopFlatten wasn't triggering on this motivating case after IV widening:
void foo(int *A, int N, int M) {
for (int i = 0; i < N; ++i)
for (int j = 0; j < M; ++j)
f(A[i*M+j]);
}
The reason was that the old induction phi nodes were getting in the way. These
narrow and dead induction phis are not always trivially dead, and having both
the narrow and wide IVs confused the analysis and caused it to bail. This adds
some extra bookkeeping for these old phis, so we can filter them out when
checks on phi nodes are performed. Other clean up passes will get rid of these
old phis and increment instructions.
As this was one of the motivating examples from the beginning, it was
surprising this wasn't triggering from C/C++ code. It looks like the IR and CFG
is just slightly different.
Differential Revision: https://reviews.llvm.org/D109309
This patch allows the clients of `ToolInvocation` to provide custom diagnostic options to be used during driver -> cc1 command-line transformation and parsing.
Tests covering this functionality are in a follow-up commit. To make this testable, the `DiagnosticsEngine` needs to be properly initialized via `CompilerInstance::createDiagnostics`.
Reviewed By: dexonsmith, arphaman
Differential Revision: https://reviews.llvm.org/D108976
Update a section of OpenCLSupport page to reflect the latest
development in OpenCL 3.0 support for release 13.
Differential Revision: https://reviews.llvm.org/D109320
D101329 introduces the Process:SaveCore function returning a
`llvm::Expected<bool>`. That function causes that Clang with -fmodules crashes
while compiling LLDB's PythonDataObjects.cpp. With enabled asserts Clang fails
because of:
Assertion failed: (CachedFieldIndex && "failed to find field in parent")
Crash can be reproduced by building via -DLLVM_ENABLE_MODULES=On with Clang
12.0.1 and then building PythonDataObjects.cpp.o .
Clang bug is tracked at rdar://82901462
When an end of record is met in non advancing IO:
- Set IOSTAT if present according to 12.11.4 (5).
- Position the file to the next record (12.11.4 (4)).
The previous code was only signaling EOR for fixed record length IO.
Reading at 12.11.4, I do not find the rational for this condition, so I
removed it.
It also does not seem the presence of padding should prevent
the EOR signaling.
The positionning to the next record was block when EOR is signaling
in FinishReadingRecord because ErrorHandler.isError() is true in this
case.
EOR in input is not an error, but I am not confident to modify
ErrorHandler.isError() to cover that. However, In FinishReadingRecord,
the code should not bail if the error is simply an end of record.
I did not check the SIZE requirements here because GetSize runtime is
not yet implemented.
Differential Revision: https://reviews.llvm.org/D109505
We set breakpoint on child_func, so synchronization inside it is too
late to guarantee ordering between the parent output and child
breakpoint. Split the function in two, and perform synchronization
before the breakpoint.
Differential Revision: https://reviews.llvm.org/D109591
For SVE, when scalarising the PHI instruction the whole vector part is
generated as opposed to creating instructions for each lane for fixed-
width vectors. However, in some cases the lane values may be needed
later (e.g for a load instruction) so we still need to calculate
these values to avoid extractelement being called on the vector part.
Differential Revision: https://reviews.llvm.org/D109445
This fixes LanaiTTIImpl::getIntImmCost to return valid costs for i128
(and wider) values. Previously any immediate wider than
64 bits would cause Lanai llc to crash.
A regression test is also added that exercises this functionality.
Reviewed By: jpienaar
Differential Revision: https://reviews.llvm.org/D107091
Anton Afanasyev