This is a simplified implementation of std::optional that can be used
in the offload builds for the device code. The methods are properly
marked with RT_API_ATTRS so that the device compilation succedes.
Reviewers: klausler, jeanPerier
Reviewed By: jeanPerier
Pull Request: https://github.com/llvm/llvm-project/pull/85177
Changed the lowering to call Norm2DimReal16 for REAL(16).
Added the corresponding entry point to FortranFloat128Math,
which required some restructuring in the related templates.
Detect NaN elements in data and handle them like gfortran does (at
runtime); namely, NaN can be returned if all the data are NaNs, but any
non-NaN value is preferable. Ensure that folding returns the same
results as runtime computation.
Fixes llvm-test-suite/Fortran/gfortran/regression/maxloc_2.f90 (and
probably others).
The implementation of MAXLOC/MINLOC without DIM=1 is wrong for CHARACTER
data -- change it to use the character comparator rather than the
numeric comparator.
Fixes llvm-test-suite/Fortran/gfortran/regression/maxloc_string_1.f90.
This update allows constant folding for many 128 bit floating point intrinsics
through the library quadmath, which is only available on some platforms.
Differential Revision: https://reviews.llvm.org/D156435
NORM2 is broken for arrays that start with a negative number
because it sets the initial running max_ value to that number
rather than to its absolute value.
Differential Revision: https://reviews.llvm.org/D155976
The implementation follows the pattern used in comparable intrinsics.
Change the runtime API for Norm2 so it does not expect a mask argument
since the Norm2 intrinsic does not accept a mask in Fortran.
Differential Revision: https://reviews.llvm.org/D138150
We used to have a big switch statement over the type categories and kinds
inside Minloc/Maxloc. After D133051 the switch grew bigger, and this
changed inlining decisions made by GCC (the build compiler). Some of the
simple methods stopped being inlined, and this caused slight performance
regression in Polyhedron/gas_dyn2. This change adds separate entries
for real/integer data types to let them be optimized separately.
Differential Revision: https://reviews.llvm.org/D135610
std::numeric_limits<__float128>::max/lowest return 0.0, so recreate
value of FLT128_MAX ourselves to avoid using quadmath.h's FLT128_MAX
that is currently causes warnings with GCC -Wpedantic.
Differential Revision: https://reviews.llvm.org/D134496
BytesFor() used to return KIND for the size, which is not always
correct, so I changed it to return the size of the actual CppType
corresponding to the given category and kind.
MinElemLen() used to calculate size incorrectly (e.g. CFI_type_extended_double
was sized 10, whereas it may occupy more bytes on a target), so I changed it
to call BytesFor().
Additional changes were needed to resolve new failures for transformational
intrinsics. These intrinsics used to work for not fully supported
data types (e.g. REAL(3)), but now stopped working because CppType
cannot be computed for those categories/kinds. The solution is to use
known element size from the source argument(s) for establishing
the destination descriptor - the element size is all that is needed
for transformational intrinsics to keep working.
Note that this does not help cases, where runtime still has to
compute the element size, e.g. when it creates descriptors for
components of derived types. If the component has unsupported
data type, BytesFor() will still fail. So these cases require
adding support for the missing types.
New regression unit test in Runtime/Transformational.cpp
demonstrates the case that will start working properly with
this commit.
The result of MINVAL over an empty default character array should
(per the standard) have a 127 in every character position, not just
the first.
Differential Revision: https://reviews.llvm.org/D132677
CharacterExtremumAccumulator::GetResult() needs to use byte counts, not wide
character counts, when calling memcpy() & memset().
Differential Revision: https://reviews.llvm.org/D132156
On targets with __float128 available and distinct from long double,
use it to support more kind=16 entry points. This affects mostly
x86-64 targets. This means that more runtime entry points are
defined for lowering to call.
Delete Common/long-double.h and its LONG_DOUBLE macro in favor of
testing the standard macro LDBL_MANT_DIG.
Differential Revision: https://reviews.llvm.org/D127025
When passing a scalar .FALSE. as the MASK argument to MAXLOC, we were getting
bad memory references. We were falling into the code intended when the MASK
argument was missing.
I fixed this by checking for a scalar MASK with a .FALSE. value and
setting the result to all zeroes in that case. I also added tests for
MAXLOC and MINLOC with scalar values of .TRUE. and .FALSE. for the MASK
argument.
I also special cased situations where the MASK argument is a scalar with
a .TRUE. value and passed along a nullptr in such cases.
Along the way, I eliminated the unused "chars" argument from the constructor
for ExtremumLocAccumulator.
Differential Revision: https://reviews.llvm.org/D124484
A template argument was hard-coded as the Integer type category
rather than properly forwarding the type category of the data for
type-specific instantiations of total (no DIM=) MAXLOC and MINLOC.
This broke total MAXLOC and MINLOC reductions for real and character
data.
Differential Revision: https://reviews.llvm.org/D124303
Where possible, I added additional information to the messages to help
programmers figure out what went wrong. I also removed all uses of the word
"bad" from the messages since (to me) that implies a moral judgement rather
than a programming error. I replaced it with either "invalid" or "unsupported"
where appropriate.
Differential Revision: https://reviews.llvm.org/D121493
Move the closure of the subset of flang/runtime/*.h header files that
are referenced by source files outside flang/runtime (apart from unit tests)
into a new directory (flang/include/flang/Runtime) so that relative
include paths into ../runtime need not be used.
flang/runtime/pgmath.h.inc is moved to flang/include/flang/Evaluate;
it's not used by the runtime.
Differential Revision: https://reviews.llvm.org/D109107
Implement the reduction transformational intrinsic function NORM2 in
the runtime, using infrastructure already in place for MAXVAL & al.
Differential Revision: https://reviews.llvm.org/D102024
The single source file reduction.cpp is a little large in
terms of both source lines and generated text bytes, so
split SUM, PRODUCT, FINDLOC, and MAXLOC/MAXVAL/MINLOC/MINVAL
off into their own C++ source files that share a set of
implementation function templates now in a common header.
Differential Revision: https://reviews.llvm.org/D101111
