This patch adds builtins for:
- svadrb, svadrh, svadrw, svadrd

This patch adds builtins for:
- svrdffr, svrdffr_z
- svsetffr
- svwrffr

This patch adds builtins for saturating add/sub instructions:
- svqadd, svqadd_n
- svqsub, svqsub_n

and builtins for dot product instructions:
- svdot, svdot_lane

Remove unnecessary EncoderMethod and DecoderMethod which cause errors in
supporting MC layer.

Differential Revision: https://reviews.llvm.org/D79544

Summary:
Correct instruction bitfield addresses to generate machine code correctly.  Also
add some variables to represent all instructions correctly and change default
values to use registers by default.

Differential Revision: https://reviews.llvm.org/D79539

Summary:
Conflicting types for the same section name defined in clang section
pragmas and GNU-style section attributes were not properly captured by
Clang's Sema. The lack of diagnostics was caused by the fact the section
specification coming from attributes was handled by Sema as implicit,
even though explicitly defined by the user.

This patch enables the diagnostics for section type conflicts between
those specifications by making sure sections defined in section
attributes are correctly handled as explicit.

Reviewers: hans, rnk, javed.absar

Reviewed By: rnk

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D78573

Summary:
Section names used in clang section pragmas were not validated against
previously defined sections, causing section type conflicts to be
ignored by Sema.

This patch enables Clang to capture these section type conflicts by
using the existing Sema's UnifySection method to validate section names
from clang section pragmas.

Reviewers: hans, rnk, javed.absar

Reviewed By: rnk

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D78572

This patch adds builtins for:
- sveorbt
- sveortb
- svpmul
- svpmullb, svpmullb_pair
- svpmullt, svpmullt_pair

The svpmullb and svpmullt builtins are expressed using the svpmullb_pair
and svpmullt_pair LLVM IR intrinsics, respectively.

Reviewers: SjoerdMeijer, efriedma, rengolin

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D79480

With 3 llvm regr tests fixed/updated that I had missed.

Summary:
We used getEnclosingNamespaceContext(), which calls getParent() rather
than getLexicalParent(), so we would end up adding the "using" line in
places that do not affect the cursor location, or just return an error
when declaration was in another file.

Patch by Adam Czachorowski!

Reviewers: hokein

Reviewed By: hokein

Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, kadircet, usaxena95, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D79496

Summary:
Adds AArch64ISD nodes for:
 - SHL_PRED (logical shift left)
 - SHR_PRED (logical shift right)
 - SRA_PRED (arithmetic shift right)

Existing patterns for unpredicated left shift by immediate
have also been moved into the appropriate multiclasses
in SVEInstrFormats.td.

Reviewers: sdesmalen, efriedma, ctetreau, huihuiz, rengolin

Reviewed By: efriedma

Subscribers: huihuiz, tschuett, kristof.beyls, hiraditya, rkruppe, psnobl, cfe-commits, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D79478

This functionality was introduced in a87db48e
but only only tested indirectly though Linalg tests. Add direct tests.

Adds a test exercising the rewriting pattern in the test dialect that calls
OperationFolder.create.

Originally, these operations were folded only if all expressions in their
affine maps could be folded to a constant expression that can be then subject
to numeric min/max computation. This introduces a more advanced version that
partially folds the affine map by lifting individual constant expression in it
even if some of the expressions remain variable. The folding can update the
operation in place to use a simpler map. Note that this is not as powerful as
canonicalization, in particular this does not remove dimensions or symbols that
became useless. This allows for better composition of Linalg tiling and
promotion transformation, where the latter can handle some canonical forms of
affine.min that the folding can now produce.

Differential Revision: https://reviews.llvm.org/D79502

Summary: Depends on D78740.

Reviewers: sammccall

Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D78743

Summary:
Add DAG scheduling mutation to cluster export instructions.
This avoids unnecessary waitcnts being added when computation
ends up interspersed with exports.

Reviewers: foad, arsenm, rampitec, nhaehnle

Reviewed By: foad

Subscribers: kzhuravl, jvesely, wdng, mgorny, yaxunl, dstuttard, tpr, t-tye, hiraditya, kerbowa, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D79481

Test shows unnecessary s_waitcnt between shader exports.

ROCLD -> ROCDL.

Differential Revision: https://reviews.llvm.org/D79441

In the Vector to LLVM conversion, the `replaceTransferOp` function calls
into a type converter that may fail and suppresses the status. Change
the function to return the failure status instead, Since it is called
from a pattern, the failure can be readily propagated to the rest of
infrastructure.

When emitting masked load / store, set alignment from data layout.

Differential Revision: https://reviews.llvm.org/D79246

Summary:
This patch handles illegal scalable types when lowering IR operations,
addressing several places where the value of isScalableVector() is
ignored.

For types such as <vscale x 8 x i32>, this means splitting the
operations. In this example, we would split it into two
operations of type <vscale x 4 x i32> for the low and high halves.

In cases such as <vscale x 2 x i32>, the elements in the vector
will be promoted. In this case they will be promoted to
i64 (with a vector of type <vscale x 2 x i64>)

Reviewers: sdesmalen, efriedma, huntergr

Reviewed By: efriedma

Subscribers: david-arm, tschuett, hiraditya, rkruppe, psnobl, cfe-commits, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D78812

Splitting critical edges for indirect branches
the SplitIndirectBrCriticalEdges() function may break branch
probabilities if target basic block happens to have unset
a probability for any of its successors. That is because in
such cases the getEdgeProbability(Target) function returns
probability 1/NumOfSuccessors and it is called after Target
was split (thus Target has a single successor). As the result
the correspondent successor of the split block gets
probability 100% but 1/NumOfSuccessors is expected (or better
be left unset).

Reviewers: yamauchi
Differential Revision: https://reviews.llvm.org/D78806

This reverts commit 617aa64c.

while I investigate buildbot failures.

If tail-folding of the scalar remainder loop is applied, the primary induction
variable is splat to a vector and used by the masked load/store vector
instructions, thus the IV does not remain scalar. Because we now mark
that the IV does not remain scalar for these cases, we don't emit the vector IV
if it is not used. Thus, the vectoriser produces less dead code.

Thanks to Ayal Zaks for the direction how to fix this.

Differential Revision: https://reviews.llvm.org/D78911

Reviewers: gribozavr2

Subscribers: xazax.hun, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D79494

When calculating the natural alignment for scalable vectors it
is acceptable to calculate an allocation size based on the minimum
number of elements in the vector.

This code path is exercised by an existing test:

  CodeGen/AArch64/sve-intrinsics-int-arith.ll

Differential Revision: https://reviews.llvm.org/D79475

We're truncating so the extra bits will be discarded.

[X86] Add test cases for missed opportunity to match pmulh from multiplies with elements larger than i32.

We currently look for vXi32 sext/zext to match PMULH, but it
doesn't matter how many extra bits above i32 there are.

The list of destination load ops while evaluating producer-consumer
fusion wasn't being maintained as a set, and as such, duplicate load ops
were being added to it. Although this is harmless correctness-wise, it's
a killer efficiency-wise and it prevents interesting/useful fusions
(including for eg. reshapes into a matmul). The reason the latter
fusions would be missed is that a slice union would be unnecessarily
needed due to the duplicate load ops on a memref added to the 'dst
loads' list. Since slice union is unimplemented for the local var case,
a single destination load op that leads to local vars (like a floordiv /
mod producing fusion), a common case, would not get fused due to an
unnecessary union being tried with itself.  (The union would actually be
the same thing but we would bail out.)

Besides the above, this would also significantly speed up fusion as all
the unnecessary slice computations / unions, checks, etc. due to the
duplicates go away.

Differential Revision: https://reviews.llvm.org/D79547

The source file used to test Objective-C data-formatters didn't respected
any formatting (mixed tab and spaces, lines exceed column 80, etc...).

This patch reformats the file using clang-format to make it easier to
work with.

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

Fix DWARFLinker.cpp:2538:5: error: call to 'sort' is ambiguous.

I missed this case when I did the same for gather results and scatter
operands in c69a4d6b.

This patch adds statistics about the contribution of each object file to
the linked debug info. When --statistics is passed to dsymutil, it
prints a table after linking as illustrated below.

It lists the object file name, the size of the debug info in the object
file in bytes, and the absolute size contribution to the linked dSYM and
the percentage difference. The table is sorted by the output size, so
the object files contributing the most to the link are listed first.

.debug_info section size (in bytes)
-------------------------------------------------------------------------------
Filename                                           Object         dSYM   Change
-------------------------------------------------------------------------------
basic2.macho.x86_64.o                                210b         165b  -24.00%
basic3.macho.x86_64.o                                177b         150b  -16.51%
basic1.macho.x86_64.o                                125b         129b    3.15%
-------------------------------------------------------------------------------
Total                                                512b         444b  -14.23%
-------------------------------------------------------------------------------

Differential revision: https://reviews.llvm.org/D79513

f8990feb enabled installing PIC version of both libclang.a and
libclang.so when LIBCLANG_BUILD_STATIC is ON. But it broke the no-PIC
build when LLVM_ENABLE_PIC=OFF with the following error:

```
CMake Error at
/b/s/w/ir/cache/builder/src/third_party/llvm/clang/tools/libclang/CMakeLists.txt:123
(target_compile_definitions):
    target_compile_definitions called with non-compilable target type
```

This is because as the code loops through ${name} and ${name}_static, it
introduced a side effect, which is adding an empty libclang_static to
targets. Later target_compile_definitions is called on libclang_static.
That function requires that target must have been created by a command
such as add_executable() or add_library(), so it crashed.

The solution is to not naively loop through both libclang and
libclang_static, but only the ones that are actually added by
llvm_add_library(). Here's the library build type to library target name
mapping:

| SHARED only | libclang |
| STATIC only | libclang |
| SHARED and STATIC | libclang and libclang_static |

So only when SHARED and STATIC are both set should we loop through two
targets. Explicitly parse the STATIC argument and set the list
accordingly.

Reviewed By: smeenai

Differential Revision: https://reviews.llvm.org/D79059

When reporting diagnostics from ASan's (and other sanitizer's) signal
handlers we should strip the "invalid signature" bit before printing
addresses.  This makes the report less confusing and let's the user
focus on the real issue.

rdar://62615826

Reviewed By: kubamracek, delcypher

Differential Revision: https://reviews.llvm.org/D79132

Now using patterns, since there's a single-instruction lowering. (We
could convert to VSELECT and pattern-match that, but there doesn't seem
to be much point.)

I think this might be the first instruction to use nested multiclasses
this way? It seems like a good way to reduce duplication between
different integer widths. Let me know if it seems like an improvement.

Also, while I'm here, fix the return type of SETCC so we don't try to
merge a sign-extend with a SETCC.

Differential Revision: https://reviews.llvm.org/D79193