I don't think this is actually defined for mesa, but this is what we
were doing on the DAG path.

Users can define aliases for long symbols using import aliases:

    import X = A.B.C;

Previously, these were unhandled and would terminate import sorting.
With this change, aliases sort as their own group, coming last after all
other imports.

Aliases are not sorted within their group, as they may reference each
other, so order is significant.

Revision URI: https://reviews.llvm.org/D118361

[X86][SSE] Add combineBitOpWithShift for BITOP(SHIFT(X,Z),SHIFT(Y,Z)) -> SHIFT(BITOP(X,Y),Z) vector folds

InstCombine performs this more generally with SimplifyUsingDistributiveLaws, but we don't need anything that complex here - this is mainly to fix up cases where logic ops get created late on during lowering, often in conjunction with sext/zext ops for type legalization.

https://alive2.llvm.org/ce/z/gGpY5v

Swizzled accesses are not merged, but there is no particular reason not
to merge two instructions if any of the intervening instructions happens
to be a swizzled access.

This moves the check for swizzled accesses out of checkAndPrepareMerge
into collectMergeableInsts where I think it makes more sense.

Differential Revision: https://reviews.llvm.org/D118267

Precommit some additional tests for D118256.

This test is completely nondeterministic, environment-dependent and does
not test what it was supposed to test (reverting the associated patch
does not make it fail).

I tried to figure out what the patch was meant to fix to see if I can
create a better test with the current tools, but I was not able to
understand the problem (it sounds like it has something to do with local
classes, but I don't understand the details).

Use isOpaqueOrPointeeTypeEquals() instead.

When checking x == 0, checking x u<= 0 is sufficient and simpler than
x u>= 0 && x u<= 0.

https://alive2.llvm.org/ce/z/btM7d3

    ----------------------------------------
    define i1 @src(i4 %a) {
    %0:
      %c = icmp eq i4 %a, 0
      ret i1 %c
    }
    =>
    define i1 @tgt(i4 %a) {
    %0:
       %c = icmp ule i4 %a, 0
       ret i1 %c
    }
    Transformation seems to be correct!

Helps fix a number of poor codegen cases for allof(cmp()) with 256-bit vectors on AVX1

Reviewed By: david-arm

Differential Revision: https://reviews.llvm.org/D118146

This has the additional advantage of supporting more types.

Differential Revision: https://reviews.llvm.org/D118348

This fixes a regression from 480cd4cb

Differential Revision: https://reviews.llvm.org/D118347

This patch adds a struct to manage a list of constraints. It simplifies
a follow-up change, that adds pre-conditions that must hold before a
list of constraints can be used.

Add Thread::GetSiginfo() and SBThread::GetSiginfo() methods to retrieve
the siginfo value from server.

Differential Revision: https://reviews.llvm.org/D118055

Support synthesizing the siginfo_t type from the Platform plugin.
This type is going to be used by LLDB client to process the raw siginfo
data received from lldb-server without the necessity of relying
on target's debug info being present.

Differential Revision: https://reviews.llvm.org/D117707

This results in excessive memory usage and eats a lot of screen estate.
Especially in the cases with lots of nested macro calls.

This patch tries to remedy it before the release cut by suppressing the
initializers. For better UX we should probably update the expression printer to
truncate those (behind some policy).

Fixes https://github.com/clangd/clangd/issues/917

Differential Revision: https://reviews.llvm.org/D118260

Relax the type assertion for opaque pointers, and enumerate the
value type in TypeFinder and ValueEnumerator.

Following up on commit 177176f7, if we
failed to setDeleteDisposition(true) during TempFile creation, then
don't try to setDeleteDisposition(false) during TempFile::keep, since it
will likely fail as well.

Instead of letting TempFile::keep just fail, we should let it go ahead
and try renaming the file.

This fixes an issue we are seeing when running clang-cl.exe through the
Incredibuild distributed build system.  We're seeing that renaming
temporary object files would fail here:
https://github.com/llvm/llvm-project/blob/5c1f7b296ac0dddeca02891976e6ab5cfc006719/clang/lib/Frontend/CompilerInstance.cpp#L789

Reviewed By: mstorsjo

Differential Revision: https://reviews.llvm.org/D118212

This reverts commit 1043107c.

This commit caused a breakage in `finalizing-bufferize.mlir`.

Reviewed By: david-arm

Differential Revision: https://reviews.llvm.org/D118145

It appears that this mangling type is currently unused. Make it
compatible with opaque pointers in case it becomes used again...

This code replaces the address space of the pointers while keeping
the element type. Use the appropriate helpers to make this work
with opaque pointers.

This feature requires support of __opencl_c_generic_address_space and
__opencl_c_program_scope_global_variables so diagnostics for that is provided as well.

Reviewed By: Anastasia

Differential Revision: https://reviews.llvm.org/D115640

Differential Revision: https://reviews.llvm.org/D118235

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D118057

We already perform some basic folds (add/sub with zero etc.) on scalar types, this patch adds some basic support for constant splats as well in a few cases (we can add more with future test coverage).

In the cases I've enabled, we can handle buildvector implicit truncation as we're not creating new constant nodes from the vector types - we're just returning existing nodes. This allows us to get a number of extra cases in the aarch64 tests.

I haven't enabled support for undefs in buildvector splats, as we're often checking for zero/allones patterns that return the original constant and we shouldn't be returning undef elements in some of these cases - we can enable this later if we're OK with creating new constants.

Differential Revision: https://reviews.llvm.org/D118264

Once again, this fold is meaningless with opaque pointers, as there
is no pointer element type to canonicalize. At some point, we may
want to do GEP type canonicalizations.

This is in preparation of switching `-tensor-constant-bufferize` and `-arith-bufferize` to BufferizableOpInterface-based implementations.

Differential Revision: https://reviews.llvm.org/D118324

Also make the function static to make it more obvious that it is
only used in the one place.

The get_image_num_mip_levels overloads that take a read_write image
parameter were missing the __opencl_c_read_write_images guard.

This commit switches the `tensor-bufferize` pass over to BufferizableOpInterface-based bufferization.

Differential Revision: https://reviews.llvm.org/D118246

The ISel patterns for PFALSE helps recognise the instructions as being
free of side-effects, which helps MachineCSE remove redundant
PFALSE instructions.

Reviewed By: dmgreen

Differential Revision: https://reviews.llvm.org/D118054

x86_64 isn't supported with the recent Xcode SDK.