Returning failure when tile sizes are all zero prevents the change in
the marker. This makes pattern rewriter run the pattern multiple times
only to exit when it hits a limit. Instead just clone the operation
(since tiling is essentially cloning in this case). Then the
transformation filter kicks in to avoid the pattern rewriter to be
invoked many times.
Differential Revision: https://reviews.llvm.org/D113949
In 'STOP bye bye', do not print 'Fortran STOP:` before 'bye bye' when
NO_STOP_MESSAGE environment variable is set at runtime.
Also only exit with code 1 in StopStatementText if this is an ERROR STOP.
This matches other compiler behaviors.
Move STOP related unit tests in their own test file and add new tests to
cover this change.
Differential Revision: https://reviews.llvm.org/D114152
We won't talk about how long it took me to figure this out. The difference in batch vs interactive on the "did you mean to print bitcode" message made this confusing.
Use the factory class in the FIRBuilder.
Add unit tests for the factory class function and the convert function
of the Complex class.
Reviewed By: clementval, rovka
Differential Revision: https://reviews.llvm.org/D114125
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
checkOrderedReductions looks for Phi nodes which can be classified as in-order,
meaning they can be vectorised without unsafe math. In order to vectorise the
reduction it should also be classified as in-loop by getReductionOpChain, which
checks that the reduction has two uses.
In this patch, a similar check is added to checkOrderedReductions so that we
now return false if there are more than two uses of the FAdd instruction.
This fixes PR52515.
Reviewed By: fhahn, david-arm
Differential Revision: https://reviews.llvm.org/D114002
This patch extends the `FIRToLLVMLowering` pass in Flang by adding a
hook to transform `fir.field_index` to a sequence of LLVM MLIR
instructions.
This is part of the upstreaming effort from the `fir-dev` branch in [1].
[1] https://github.com/flang-compiler/f18-llvm-project
Differential Revision: https://reviews.llvm.org/D113988
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Compiling code for AMD GPUs requires knowledge of which chipset is
being targeted, especially if the code uses chipset-specific
intrinsics (which is the case in a downstream convolution generator).
This commit adds `target`, `chipset` and `features` arguments to the
SerializeToHsaco constructor to enable passing in this required
information.
It also amends the ROCm integration tests to pass in the target
chipset, which is set to the chipset of the first GPU on the system
executing the tests.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D114107
The libgcc runtime library provides __register_frame and
__deregister_frame functions, which can be used by dynamic code
generators to register an .eh_frame section, which contains one or
more Call Frame Information records, each consisting of a Common
Information Entry record followed by one or more Frame Description
Entry records. This libunwind library also provides __register_frame
and __deregister_frame functions, but they are aliases for
__unw_add_dynamic_fde and __unw_remove_dynamic_fde and thus can only
take a single FDE.
This patch adds __unw_add_dynamic_eh_frame_section and
__unw_remove_dynamic_eh_frame_section functions which explicitly use
the .eh_frame format. Clients such as the ORCv2 platform and runtime
can check for these functions and use them if unwinding is being
provided by libunwind, or fall back to __register_frame and
__deregister_frame if unwinding is provided by libgcc.
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D111863
We might emit functions that are private and never called. The coro
split pass only processes functions that might be called. Remove
intrinsics that we can't generate code for.
rdar://84619859
Differential Revision: https://reviews.llvm.org/D114021
Summary:
This patch marks libunwind test case signal_frame.pass.cpp as UNSUPPORTED on AIX because the AIX assembler does not support CFI directives.
Reviewed by: danielkiss, MaskRay, ldionne, libunwind
Differential Revision: https://reviews.llvm.org/D113607
This covers both C-style variadic functions and template variadic w/
parameter packs.
Previously we would return no signatures when working with template
variadic functions once activeParameter reached the position of the
parameter pack (except when it was the only param, then we'd still
show it when no arguments were given). With this commit, we now show
signathure help correctly.
Additionally, this commit fixes the activeParameter value in LSP output
of clangd in the presence of variadic functions (both kinds). LSP does
not allow the activeParamter to be higher than the number of parameters
in the active signature. With "..." or parameter pack being just one
argument, for all but first argument passed to "..." we'd report
incorrect activeParameter value. Clients such as VSCode would then treat
it as 0, as suggested in the spec) and highlight the wrong parameter.
In the future, we should add support for per-signature activeParamter
value, which exists in LSP since 3.16.0. This is not part of this
commit.
Differential Revision: https://reviews.llvm.org/D111318
This patch extends the FIRToLLVMLowering pass in Flang by adding a hook
to transform fir.end. This is just a placeholder for now as fir.end is
not required yet.
This is part of the upstreaming effort from the `fir-dev` branch in [1].
[1] https://github.com/flang-compiler/f18-llvm-project
Patch originally written by:
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Differential Revision: https://reviews.llvm.org/D113295
The StringPrinter class was using a Process instance to read memory.
This automatically prevented it from working before starting the
program.
This patch changes the class to use the Target object for reading
memory, as targets are always available. This required moving
ReadStringFromMemory from Process to Target.
This is sufficient to make frame/target variable work, but further
changes are necessary for the expression evaluator. Preliminary analysis
indicates the failures are due to the expression result ValueObjects
failing to provide an address, presumably because we're operating on
file addresses before starting. I haven't looked into what would it take
to make that work.
Differential Revision: https://reviews.llvm.org/D113098
This removes the ability to disable roundtripping in assert builds.
(Roundtripping happens by default in assert builds both before and after
this patch.)
The CLANG_ROUND_TRIP_CC1_ARGS was added as an escape hatch 9 months ago
in https://reviews.llvm.org/D97462, with a FIXME to remove it eventually.
It's probably time to remove it.
Differential Revision: https://reviews.llvm.org/D114120
It turns out that mlir::ComplexType is translated the same way with or
without a conversion in TypeConverter.h. Add tests to make sure this
doesn't change in the future.
This patch is part of the upstreaming effort from fir-dev.
Differential Revision: https://reviews.llvm.org/D113883
Add the FIR to LLVM conversion patterns for the BoxProcHostOp, EmboxProcOp,
and UnboxProcOp ops and the boxproc type. These are currently unimplemented.
Implementation will come at a later time when support for Fortran 2003
procedure pointer feature is added.
Reviewed By: clementval, rovka
Differential Revision: https://reviews.llvm.org/D113879
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
This patch adds a reduced version of the test case from PR52024.
Together with 764d9aa979 the test causes a crash, because LV expands a
SCEV expression during code generation, when the dominator tree is not
up-to-date.
These tests were introduced in D109809 which I pushed on behalf of
@tangxingxin1008. I must have not understood the correct arcanist
workflow for this and as such may have locally tested a stale build.
This patch fixes the issue by re-running update_llc_test_checks.py on
all four tests.
This patch refactors the code that checks whether a file has just been included for the first time.
The `HeaderSearch::FirstTimeLexingFile` function is removed and the information is threaded to the original call site from `HeaderSearch::ShouldEnterIncludeFile`. This will make it possible to avoid tracking the number of includes in a follow up patch.
Depends on D114092.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D114093
This patch removes `HeaderFileInfo::isNonDefault`, which is not being used anywhere.
Reviewed By: dexonsmith, vsapsai
Differential Revision: https://reviews.llvm.org/D114092
During explicit modules build, when all modules are provided via `-fmodule-file=<path>` and implicit modules and implicit module maps are disabled (`-fno-implicit-modules`, `-fno-implicit-module-maps`), we don't need to load the original module map files at all. This patch stops emitting the `-fmodule-map-file=` arguments we don't need, saving some compilation time due to avoiding parsing such module maps and making the command line shorter.
Reviewed By: bnbarham
Differential Revision: https://reviews.llvm.org/D113473
Fixed the vector type issue that where we used getVectorNumElements()
should be replaced by getVectorElementCount() when lowering these
intrinsics.
This is similar to D94149
Signed-off-by: Eric Tang <tangxingxin1008@gmail.com>
Reviewed By: craig.topper, frasercrmck
Differential Revision: https://reviews.llvm.org/D109809
Problem:
PCM file includes references to all module maps used in compilation which created PCM. This problem leads to PCM-rebuilds in distributed compilations as some module maps could be missing in isolated compilation. (For example in our distributed build system we create a temp folder for every compilation with only modules and headers that are needed for that particular command).
Solution:
Add only affecting module map files to a PCM-file.
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D106876
Convert a `fir.embox` operation to LLVM IR dialect.
A `fir.embox` is converted to a sequence of operation that
create, allocate if needed, and populate a descriptor.
Current limitiation: alignment is set by default but should be retrieved in the specific target.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: kiranchandramohan, awarzynski
Differential Revision: https://reviews.llvm.org/D113756
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
We were missing patterns for vector_reverse of unpacked FP vector
types, as well as all the supported bfloat vectors.
Tests added here:
CodeGen/AArch64/named-vector-shuffle-reverse-sve.ll
Differential Revision: https://reviews.llvm.org/D114089
The initial two cases require a SCEVConstant as RHS. Pull up the condition
to check and swap SCEVConstants from below. Also remove a redundant
check & swap if RHS is SCEVUnknown.
This is a cleanup of the only llvm-prefer-isa-or-dyn-cast-in-conditionals finding in the clangd code base. This patch was created by automatically applying the fixes from clang-tidy.
Differential Revision: https://reviews.llvm.org/D113899
Add the codegen for fir.constc.
This patch is part of the upstreaming effort from fir-dev.
Differential Revision: https://reviews.llvm.org/D114063
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Previously, in case there was only one `Optional` operand/result within
the list, we would always return `None` from the accessor, e.g., for a
single optional result we would generate:
```
return self.operation.results[0] if len(self.operation.results) > 1 else None
```
But what we really want is to return `None` only if the length of
`results` is smaller than the total number of element groups (i.e.,
the optional operand/result is in fact missing).
This commit also renames a few local variables in the generator to make
the distinction between `isVariadic()` and `isVariableLength()` a bit
more clear.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D113855
MaheshRavishankar