While dense tensors support random accesses, it is critical to visit them in a row-major order for better cache locality. However, we previously consider dense inputs and outputs together when computing constraints for building iteration graph, it could lead us to less efficient iteration graphs.

This patch adds a new `SortMask::kIncludeDenseInput` to treat dense inputs/outputs separately when building iteration graph, thus increasing the chance for use to construct a better iteration graph.

A more fine-grained approach is to treat each input separately.

Note, related to:
 https://github.com/llvm/llvm-project/issues/51651

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D144932

Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>

This is in preparation for the transition to a solution to make libc tests
hermetic with respect to their use of errno. The implementation of strdup
has been switched over to libc_errno as an example of what the code looks
like in the new way.

See #61037 for more information.

Reviewed By: lntue

Differential Revision: https://reviews.llvm.org/D144928

If use_device_ptr/use_device_addr clauses are used on target data
directive and no device was specified during the compilation, only host
part should be emitted. But it still required to emit captured decls for
partially mapped data fields.

Differential Revision: https://reviews.llvm.org/D144993

Otherwise clang may invoke ld. If ld is a shell script using `~`, the
command will fail since `HOME` is changed.

Follow-up to D144914.
-mcpu=help seems fine as a Clang extension not in GCC, because llc supports -mcpu=help.
-mcpu=? is a bad choice as ? may be expanded by the shell.

Reviewed By: sameerds

Differential Revision: https://reviews.llvm.org/D144162

This is UB, but we shouldn't crash the compiler either.

Fixes #61056

Reviewed By: jpienaar

Differential Revision: https://reviews.llvm.org/D144978

In order to run a {break,watch}point command, lldb can resolve to the
script interpreter to run an arbitrary piece of code or call into a
user-provided function. To do so, we will generate a wrapping function,
where we first copy lldb's internal dictionary keys into the
interpreter's global dictionary, copied inline the user code before
resetting the global dictionary to its previous state.

However, {break,watch}point commands can optionally return a value that
would tell lldb whether we should stop or not. This feature was
only implemented for breakpoint commands and since we inlined the user
code directly into the wrapping function, introducing an early return,
that caused lldb to let the interpreter global dictionary tinted with the
internal dictionary keys.

This patch fixes that issue while also adding the stopping behaviour to
watchpoint commands.

To do so, this patch refactors the {break,watch}point command creation
method, to let the lldb wrapper function generator know if the user code is
a function call or a arbitrary expression.

Then the wrapper generator, if the user input was a function call, the
wrapper function will call the user function and save the return value into
a variable. If the user input was an arbitrary expression, the wrapper  will
inline it into a nested function, call the nested function and save the
return value into the same variable. After resetting the interpreter global
dictionary to its previous state, the generated wrapper function will return
the varible containing the return value.

rdar://105461140

Differential Revision: https://reviews.llvm.org/D144688



Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>

Removed unnecessary warning workaround.

Differential Revision: https://reviews.llvm.org/D144992

(C - X) + Y --> (Y - X) + C

Moving the constant operand to an 'add' gives more
flexibility to subsequent reassociation patterns,
and it may be better for codegen on targets that
don't have subtract-from-immediate instructions.

Even within this file, the usual spelling is 'Opcode',
so make it consistent.

This is too conservative as noted in the TODO comment.

It appears that ICK_Zero_Queue_Conversion was inserted into the ICK
enum without updating this table. Easy to do since the table size
was set to ICK_Num_Conversion_Kinds.

I've used ICR_Exact_Match to match what was previously done for
ICK_Zero_Event_Conversion that last time someone noticed this had happened.

To prevent this from happening again, I've removed the explicit size
and used a static_assert to check the size against ICK_Num_Conversion_Kinds.

Differential Revision: https://reviews.llvm.org/D144990

This fixes the Clang modular CI, but breaks other CIs.

This reverts commit 2ae39902.

Differential Revision: https://reviews.llvm.org/D144991

When building the set of referenced fields for the `DataflowAnalysisContext`,
include fields referenced only in default member initializers. These
initializers are visited in the CFGs of constructors and so the fields must be
included when analysing constructor bodies.

Differential Revision: https://reviews.llvm.org/D144987

The internal implementation of the string to float function previously
used pointer arguments for returning several values. Additionally it
set errno in several unexpected places. Now all of that goes through
return structs. For readability I also moved the function away from raw
pointer arithmetic towards proper indexing. I also added support for
rounding modes.

Reviewed By: sivachandra

Differential Revision: https://reviews.llvm.org/D144597

Clang currently uses `-mcpu=?` and `-mtune=?`. The `?` causes errors on some
shells such as zsh since it is a special character. In order for it to work on
shells such as zsh, the option must be passed in quotes or escaped. This patch
adds `-mcpu=help` and `-mtune=help` as another alias for `--print-supported-cpus`.
In llc, `-mcpu=help` is an alias to print supported cpus.

Differential Revision: https://reviews.llvm.org/D144914

This commits breaks the libc++ modular build in the precommit CI.

This reverts commit 74565c3a.

Reviewed By: ldionne, Mordante, #libc

Spies: libcxx-commits, arichardson

Differential Revision: https://reviews.llvm.org/D143333

A block construct is an execution control construct that supports
declaration scopes contained within a parent subprogram scope or another
block scope. (blocks may be nested.) This is implemented by applying
basic scope processing to the block level.

Name uniquing/mangling is extended to support this. The term "block" is
heavily overloaded in Fortran standards. Prior name uniquing used tag `B`
for common block objects. Existing tag choices were modified to free up `B`
for block construct entities, and `C` for common blocks, and resolve
additional issues with other tags. The "old tag -> new tag" changes can
be summarized as:

     -> B  -- block construct -> new
  B  -> C  -- common block
  C  -> YI -- intrinsic type descriptor; not currently generated
  CT -> Y  -- nonintrinsic type descriptor; not currently generated
  G  -> N  -- namelist group
  L  ->    -- block data; not needed -> deleted

Existing name uniquing components consist of a tag followed by a name
from user source code, such as a module, subprogram, or variable name.
Block constructs are different in that they may be anonymous. (Like other
constructs, a block may have a `block-construct-name` that can be used
in exit statements, but this name is optional.) So blocks are given a
numeric compiler-generated preorder index starting with `B1`, `B2`,
and so on, on a per-procedure basis.

Name uniquing is also modified to include component names for all
containing procedures rather than for just the immediate host. This
fixes an existing name clash bug with same-named entities in same-named
host subprograms contained in different-named containing subprograms,
and variations of the bug involving modules and submodules.

F18 clause 9.7.3.1 (Deallocation of allocatable variables) paragraph 1
has a requirement that an allocated, unsaved allocatable local variable
must be deallocated on procedure exit. The following paragraph 2 states:

  When a BLOCK construct terminates, any unsaved allocated allocatable
  local variable of the construct is deallocated.

Similarly, F18 clause 7.5.6.3 (When finalization occurs) paragraph 3
has a requirement that a nonpointer, nonallocatable object must be
finalized on procedure exit. The following paragraph 4 states:

  A nonpointer nonallocatable local variable of a BLOCK construct
  is finalized immediately before it would become undefined due to
  termination of the BLOCK construct.

These deallocation and finalization requirements, along with stack
restoration requirements, require knowledge of block exits. In addition
to normal block termination at an end-block-stmt, a block may be
terminated by executing a branching statement that targets a statement
outside of the block. This includes

Single-target branch statements:
 - goto
 - exit
 - cycle
 - return

Bounded multiple-target branch statements:
 - arithmetic goto
 - IO statement with END, EOR, or ERR specifiers

Unbounded multiple-target branch statements:
 - call with alternate return specs
 - computed goto
 - assigned goto

Lowering code is extended to determine if one of these branches exits
one or more relevant blocks or other constructs, and adds a mechanism to
insert any necessary deallocation, finalization, or stack restoration
code at the source of the branch. For a single-target branch it suffices
to generate the exit code just prior to taking the indicated branch.
Each target of a multiple-target branch must be analyzed individually.
Where necessary, the code must first branch to an intermediate basic
block that contains exit code, followed by a branch to the original target
statement.

This patch implements an `activeConstructStack` construct exit mechanism
that queries a new `activeConstruct` PFT bit to insert stack restoration
code at block exits. It ties in to existing code in ConvertVariable.cpp
routine `instantiateLocal` which has code for finalization, making block
exit finalization on par with subprogram exit finalization. Deallocation
is as yet unimplemented for subprograms or blocks. This may result in
memory leaks for affected objects at either the subprogram or block level.
Deallocation cases can be addressed uniformly for both scopes in a future
patch, presumably with code insertion in routine `instantiateLocal`.

The exit code mechanism is not limited to block construct exits. It is
also available for use with other constructs. In particular, it is used
to replace custom deallocation code for a select case construct character
selector expression where applicable. This functionality is also added
to select type and associate constructs. It is available for use with
other constructs, such as select rank and image control constructs,
if that turns out to be necessary.

Overlapping nonfunctional changes include eliminating "FIR" from some
routine names and eliminating obsolete spaces in comments.

Currently SimplifyCFG hoists/sink common instructions in then/else basic blocks
when certain options are enabled, which is the case for default clang optimization
pipelines for -O3. It tries to hoist/sink convergent function calls in divergent
control flow, which causes incorrect ISA generated for GPU, e.g.
https://github.com/ROCm-Developer-Tools/HIP/issues/3172

This patch fixes that by conservatively disable hoisting/sinking common
convergent function calls in then/else blocks.

Reviewed by: Artem Belevich

Differential Revision: https://reviews.llvm.org/D144756

  3782. Should <math.h> declare ::lerp?

Libc++ doesn't declare ::lerp, adds tests to validate the requirement.

Reviewed By: #libc, philnik

Differential Revision: https://reviews.llvm.org/D142817

Using listremove makes it easier to confirm that the code matches the comments.  The only in tree users of these lists are not order sensative.

compile.fail.cpp tests are an anti-feature since they are too easy to
break when evolving code. This patch moves various allocator_mismatch
tests to .verify.cpp and normalizes the error messages from various
containers.

Differential Revision: https://reviews.llvm.org/D144913

This was discussed in https://llvm.org/PR61006.

Reviewed By: ChuanqiXu

Differential Revision: https://reviews.llvm.org/D144894

Previously the mask would be loaded as the appropriate integer type and cast to I1 to pass to
fir.if, however this truncates the integer and so would cast 6 to 0. By loading values as logicals
and casting to I1 this problem is avoided.

Reviewed By: Leporacanthicus

Differential Revision: https://reviews.llvm.org/D144974

Differential Revision: https://reviews.llvm.org/D144900

Factoring out a shared multiclass imply makes it easier to see that all of these do the same thing, just on different lists. It also makes it easier to see differences - such as we don't define read related pieces for FWRed.

Differential Revision: https://reviews.llvm.org/D144899

This commit adds support for atomic fmax/smax/umin support
for AMDGPU dialect and the dependent dialects to allow such
a lowering.

Reviewed By: krzysz00

Differential Revision: https://reviews.llvm.org/D144097

Fixed the bug in merging BatchGroups back to the FreeList. Added DCHECKs
to ensure the order of BatchGroups

This reverts commit 387452ec.

Reviewed By: cferris

Differential Revision: https://reviews.llvm.org/D144920

Try to fix flang-aarch64-sharedlibs build.

7344f8a8 introduced a use of
LLVMCreatePassBuilderOptions(), which is part of the Passes
library, while previously only InstCombine was linked.

Move plugin source and cmake files into separate directory.
Typically cmake targets in LLVM have a single target per directory.
This change brings this unittest more inline with that structure.

Reviewed By: thakis

Differential Revision: https://reviews.llvm.org/D140588

[mlir][sparse] Add runtime support for reading a COO tensor and writing the data to the given indices and values buffers.

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D143862

This allows `static_assert(false)` to not be ill-formed
in template definitions.

This change is applied as a DR in all C++ modes.

Of notes, a couple of tests were relying of the eager nature
of static_assert

* test/SemaTemplate/instantiation-dependence.cpp
* test/SemaTemplate/instantiate-var-template.cpp

I don't know if the changes to `static_assert`
still allow that sort of tests to be expressed.

Reviewed By: #clang-language-wg, erichkeane, aaron.ballman

Differential Revision: https://reviews.llvm.org/D144285

Invoke InstCombine via the new pass manager, instead of the
legacy pass manager.

When gathering the counter for the reused scalars, need to use reduced
value, not the original reduced value. Same values counter is gathered
for reduced values, not original ones.

When we allocate a variable using a MOLD argument, the function that
applies the type of the MOLD argument first checks to see if the
variable is already allocated by looking at its descriptor.  But in the
case of allocating a scalar, the descriptor was not yet been created and
the associated memory is uninitialized.  This change fixes that.

Differential Revision: https://reviews.llvm.org/D144761