Instead of going through all those trailing blocks, just count the
number and increase the counter at once.

Reviewed By: cferris

Differential Revision: https://reviews.llvm.org/D145419

This only covers the common load/store case. There will be further patches required for masked load/store and some of the fast-path optimization cases.

Differential Revision: https://reviews.llvm.org/D145175

Fix the last runtime issue as some sequent comparisons need be spilted.
For the origin equal comparisons chain, the new spilted Icmp chain will
still be end with equal, while for the new not-equal comparisons chain,
the new spilted Icmp chain will still be end with equal, so should address
this carefully, see detail wih case partial_sequent_ne.

Thanks for @glandium and @ayzhao report the runtime issue and carefully
examine.
Fix https://github.com/llvm/llvm-project/issues/59740.

Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D141188

We are not mutating the descriptors here.

Only apply the change made in D145604 for derived-type.
zero-sized character for example are supposed to give a false result

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D145675

Allocatable intent(out) are deallocated at the beginning of a function/subroutine.
For polyrmophic entities, the dynamic type need to be reseted to the declared
type. This patch makes sure this is done when the dummy argument is optional and
present.

Depends on D145674

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D145679

Restore the behavior changed in D145384 and add proper
unit tests.

Unallocated unlimited poymorphic allocatable and disassociated
unlimited polymorphic pointer should return false.

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D145674

Ignoring freeze(undef) if it has multiple uses in LowerAVXCONCAT_VECTORS
causes the custom INSERT_SUBVECTOR for vector widening to be ignored.

Differential Revision: https://reviews.llvm.org/D144903

This silences a GCC conversion diagnostic about assigning `1` to a
1-bit signed bit-field changing the value from `1` to `-1`.

Co-authored-by: Igor Kushnir <igorkuo@gmail.com>

The function makes liveness tests for the entire live register set for every instruction it passes by.
This becomes very slow on high RP regions such as ASAN enabled code.

Instead only uses of last tracked instruction should be tested and this greatly improves compilation time.

This patch revealed few bugs in SIFormMemoryClauses and PreRARematStage::sinkTriviallyRematInsts which should
be fixed first.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D136267

The outer context private check for reduction variables was firing
for all constructs. This check is not applicable to non-worksharing
constructs.

OpenMP 5.2: Section 5.5.8
A list item that appears in a reduction clause on a worksharing construct
must be shared in the parallel region to which a correspodning worksharing
region binds.

Reviewed By: peixin

Differential Revision: https://reviews.llvm.org/D144824

We were failing tests where an ALLOCATE statement that allocated an
array had a non-character scalar MOLD argument.

I fixed this by merging the code for ALLOCATE statements with MOLD and
SOURCE arguments.

Differential Revision: https://reviews.llvm.org/D145418

The patch adds the lowering from Flang parse-tree to FIR+OpenMP. The
conversion code is also added in MLIR.

Reviewed By: kiranchandramohan

Differential Revision: https://reviews.llvm.org/D133442



Co-authored-by: Kiran Chandramohan <kiran.chandramohan@arm.com>

Summary:
We passed `-fopenmp-target=` when we compiled the bitcode, which isn't
necessary since the 15 release. Also adjust an error message.

After commit 0a4aa8a1 we see failures
like:
optional:722:7: error: missing '#include <__type_traits/conjunction.h>';
'_And' must be declared before it is used
and:
optional:683:46: error: missing '#include
<__type_traits/disjunction.h>'; '_Or' must be declared before it is used
      using __check_constructible_from_opt = _Or<

Adding these here fixes that.
I'm not familiar with the libcxx codebase, will ask the author to take a
look too.

This reverts commit 24b82355.

These changes broke the PPC build bot.

This reverts commit d4fcc692.

(This change removed all test coverage from a previous change, also to be reverted.)

Summray:
Forgot to do this in the last patch

This function mimics the std::atomic_thread_fence function from
<atomic>. This has no uses in source currently, but this will be used by
the proposed RPC client for the GPU mode support. There is varying
support for direct memory ordering for the GPU atomics on shared memory
resources. So the implementation will use relaxed atomics and explicit
memory fences.

Some additional work may need to be done to map this to `NVPTX` system
level fences.

Reviewed By: sivachandra

Differential Revision: https://reviews.llvm.org/D145608

This really should be in shuffle combining, but we're still struggling to handle mismatched vector sizes

This revision adds:
- New test allocator, which cleans memory during allocation and deallocation,
- tests using that allocator to vector.

This patch is part of our efforts to add support for ASan annotations with every
allocator.

This commit adds a new allocator for testing purposes only. The safe allocator
ensures that memory is cleand (zeroed) during allocation and deallocation, and
is intendted to test ASan annotations for every allocator in std::vector.
Check: D136765

Those tests should work correctly, even if support for every allocator in std::vector
is not yet available.

Support in ASan API was added here: rGdd1b7b79

Reviewed By: philnik, #libc

Spies: libcxx-commits

Differential Revision: https://reviews.llvm.org/D145597

fir.if currently isn't treated as a 'proper' conditional, so passes are unable to determine which regions are executed at times.

This patch gives fir.if this interface, which shouldn't do too much on its own but should allow future changes to take advantage
for various purposes

Reviewed By: vzakhari

Differential Revision: https://reviews.llvm.org/D145165

[X86] Add test case showing failure to fold concat_v32i8(trunc_v16i16(x),trunc_v16i16(y)) -> trunc_v32i16(concat_v16i16(x,y))

We have similar tests but nothing that should only fold on BWI targets

This canonicalization just makes the implicit sext/trunc explicit,
and does not affect the inbounds-ness of the GEP.

This file used CRLF newlines.

This reverts commit e2817933.

This reverts commit d14e7ee3.

Makes clang_format_diff.py always fail with:

    Traceback (most recent call last):
      File "/home/npopov/repos/llvm-project/clang/tools/clang-format/clang-format-diff.py", line 177, in <module>
        main()
      File "/home/npopov/repos/llvm-project/clang/tools/clang-format/clang-format-diff.py", line 174, in main
        process_subprocess_result(proc, args)
      File "/home/npopov/repos/llvm-project/clang/tools/clang-format/clang-format-diff.py", line 42, in process_subprocess_result
        with open(filename) as f:
                  ^^^^^^^^
    NameError: name 'filename' is not defined

When linking a shared library with Flang on MinGW, the functions from the
Flang runtime are exported from the shared library. When trying to link an
executable to that library using Flang, the linker errors out because the
functions from the runtime conflict with the functions exported from the
shared library.

Add the Flang runtime libraries to the list of libraries for which no
symbols are exported.

Reapplying the patch with the git author name corrected.

Reviewed By: mstorsjo

Differential Revision: https://reviews.llvm.org/D145389

This reverts commit 04fc67bf.

The patch was applied with the wrong git author name (forgot to
amend in the right user name after applying the patch); I'll
reapply it with the right author name.

When linking a shared library with Flang on MinGW, the functions from the
Flang runtime are exported from the shared library. When trying to link an
executable to that library using Flang, the linker errors out because the
functions from the runtime conflict with the functions exported from the
shared library.

Add the Flang runtime libraries to the list of libraries for which no
symbols are exported.

Reviewed By: mstorsjo

Differential Revision: https://reviews.llvm.org/D145389

Ignoring freeze(undef) if it has multiple uses in LowerAVXCONCAT_VECTORS
causes the custom INSERT_SUBVECTOR for vector widening to be ignored.

Differential Revision: https://reviews.llvm.org/D144903

The code was difficult to maintain (big internal class definitions
with long inline functions, other functions of the same class at
different location far away, irregular ordering of classes and
function definitions). It is now improved to some extent.
New functions are added to RangeConstraint to remove code repetition,
these are useful for planned new features too.
Comments are improved.

Reviewed By: Szelethus

Differential Revision: https://reviews.llvm.org/D143751

Add an emitter to produce something similar to opencl-c.h from the
OpenCL builtin descriptions in OpenCLBuiltins.td

This only adds the emitter, without any direct use of it.  This allows
opencl-c.h additions to be generated from the builtin descriptions by
manually invoking `clang-tblgen -gen-clang-opencl-builtin-header`.

Differential Revision: https://reviews.llvm.org/D104040

Values in SGPR and VGPR register are treated as unsigned by hardware.

When value in 32-bit SGPR or VGPR base can be negative calculate offset
using 32-bit add instructions, otherwise use
sgpr(unsigned) + vgpr(unsigned) + offset.

LoopStrengthReduce.cpp changes offsets to negative and in some
iterations value in SGPR or VGPR register could be negative.

Differential Revision: https://reviews.llvm.org/D144957

Values in VGPR register are treated as unsigned by hardware.

When value in 32-bit VGPR base can be negative calculate offset using
32-bit add instruction, otherwise use vgpr base(unsigned) + offset.
Does not affect case where whole offset comes from VGPR register
(immediate offset is 0).

LoopStrengthReduce.cpp changes offsets to negative and in some
iterations value in VGPR register could be negative.

Differential Revision: https://reviews.llvm.org/D144956

Values in SGPR register are treated as unsigned by hardware.

When value in 32-bit SGPR base can be negative calculate offset using
32-bit add instruction, otherwise use sgpr base(unsigned) + offset.
Does not affect case where whole offset comes from SGPR register
(immediate offset is 0).

LoopStrengthReduce.cpp changes offsets to negative and in some
iterations value in SGPR register could be negative.

Differential Revision: https://reviews.llvm.org/D144955