This has been obsoleted by C++ thread_local for a long time.
As far as I know, Xcode was the last supported toolchain to add
support for C++ thread_local in 2016.
As a precaution, use LLVM_THREAD_LOCAL which provides even greater
backwards compatibility, allowing this to function even pre-C++11
versions of GCC.
Reviewed By: majnemer
Differential Revision: https://reviews.llvm.org/D141347
Conditions for dominating branches and llvm.assumes are already
collected. This also adds conditions from guards.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D141243
Using the option `-print-sorted-by=.` cause to core dump, so change to a legal value.
Reviewed By: maksfb
Differential Revision: https://reviews.llvm.org/D140847
See https://reviews.llvm.org/D140867.
The signature of the constructor/destrucor of the global variables is
different in different targets. And the CHECKs in the test is
unavailable in other targets (e.g., AIX). So the patch tries to avoid
the test failure by requiring the target type.
`h` was the prefix of multi-letter extension name, but it become a
extension name in later RISC-V isa spec.
Fortunately we don't have any extension really defined is prefixed
with `h`, so we can just change that.
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D136817
C++ constexpr allows a non-constant-expresssion lvalue to be used in a constant expression if it's not subject to lvalue-to-rvalue conversion. Subtly, this means you can make a constant-expression copy of a non-constant-expression object of empty type since the copy constructor doesn't perform lvalue-to-rvalue conversion. MSVC has had bugs with this usage forever, which will hopefully finally be mashed implementing C++23's relaxation on the use of pointers and references in constant expressions.
There's no need for this particular test to use this particular constexpr feature, we can simply make the predicates constant expressions.
Differential Revision: https://reviews.llvm.org/D141336
The target node being removed did not have side effects; the generic one does. Nothing in the code indicates the lack was intentional, so this is probably fixing a bug.
We use a special TIED instructions for vwadd(u).wv to avoid an
earlyclobber constraint preventing the first source and the destination
from being the same register.
This prevents our normal post process for forming TU instructions.
Add manual isel pattern instead. This matches what we do for FMA
for example.
These tests show failure to form vwadd(u).wv instructions.
One of the operands of the add is already wide and has an
additional user. We are unnecessary enforcing that operand to
have a single use.
PartSize for a part in dx container is only the size of the content of the part, not include size of the PartHeader.
Differential Revision: https://reviews.llvm.org/D141207
Recent commit introducing AA for getting underluying objects of a
pointer created an uninitialized boolean, which causes tests to fail
when built unter asan/ubsan. This initialized that variable.
Limit trusting the arange accelerator tables (8b259fe573e1) to dSYMs
only, and not any debug info object file.
Differential revision: https://reviews.llvm.org/D141330
Allow running llvm/utils/update_any_test_checks.py from working
directories other llvm/utils/.
Fixes:
$ ./llvm/utils/update_any_test_checks.py path/to/foo.ll
utils/update_llc_test_checks.py: not found (used in path/to/foo.ll)
Reviewed By: arichardson
Differential Revision: https://reviews.llvm.org/D140586
Every other subclass of SelectionDAGISel calls this pass "<arch>-isel".
No existing tests refer to ppc-codegen so this is purely a cosmetic
change to bring the pass name in line with other architecture's
SelectionDAGISel subclasses.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D140497
We have multiple targets which have defined custom instructions and sdag nodes to represent a compiler memory barrier. This patch consolidates the sdag node definition into common code.
This is a companion to D92842, but a bit different in focus. This change consolidates the existing sdag node definitions; that patch skipped defining a sdag node by instead going straight to a target node. That patch is also not NFC - as being so is quite hard for commoning up the instruction definitions.
I started with two backends to ensure the new common code was reusable while not having a massive diff. Once this lands, I'll submit a series of NFCs for backends where the changes are obvious, or reviews if more discussion is needed.
Differential Revision: https://reviews.llvm.org/D141317
Implicit def operands are needed when we spill partially undef super registers by each individual subregister. The implicit-def operands will allow us to lower spills without the verifier complaining. Currently, we are overzeously attaching implicit operands, when we really only need them on the first sub reg spill op. By more selectively attached the implicit ops, we will free up some unneeded dependencies for the post-ra scheduler.
Moreover, this enables a previously incorrect optimization / resolves a correctness issue in indirectCopyToAGPR. When lowering AGPR copies on GFX908, we can improve CodeGen by reusing accvgpr_writes. However, we could not reliably determine which agprs accvgpr_writes actually define due to implicit-defs.
Differential Revision: https://reviews.llvm.org/D141101
This patch implements the enhancement proposed by
https://github.com/llvm/llvm-project/issues/59312.
Suppose we have following code
v0 = load %addr
br %LoadBB
LoadBB:
v1 = load %addr
...
PredBB:
...
br %cond, label %LoadBB, label %SuccBB
SuccBB:
v2 = load %addr
...
Instruction v1 in LoadBB is partially redundant, edge (PredBB, LoadBB) is a
critical edge. SuccBB is another successor of PredBB, it contains another load
v2 which is identical to v1. Current GVN splits the critical edge
(PredBB, LoadBB) and inserts a new load in it. A better method is move the load
of v2 into PredBB, then v1 can be changed to a PHI instruction.
If there are two or more similar predecessors, like the test case in the bug
entry, current GVN simply gives up because otherwise it needs to split multiple
critical edges. But we can move all loads in successor blocks into predecessors.
Differential Revision: https://reviews.llvm.org/D139582
Our static analyzer likes to warn when loop bodies are never executed, which is true for `make_string<T>("")`. Build the result with `basic_string`'s iterator-pair constructor instead, which is simpler (one liner), faster (single pass), and doesn't trigger the warning.
Differential Revision: https://reviews.llvm.org/D141263
The current interface theoretically could lead to a use-after-free
when a client holds on to the returned pointer. Fix this by returning
a shared_ptr to the scratch typesystem.
rdar://103619233
Differential Revision: https://reviews.llvm.org/D141100
* Using one-or-two letter names for globals is asking for shadowing warnings.
* MSVCSTL's container proxy allocations strike again
* MSVCSTL's `<string>` doesn't define `std::out_of_range`
* `basic_string::substr` takes two arguments of type `size_type`. Let's use that type instead of `size_t` and `ptrdiff_t` to avoid narrowing warnings.
Differential Revision: https://reviews.llvm.org/D141253
Previously, we use a temporary tensor with identity ordering. We now use a
temporary tensor with the destination dimension ordering, to enable the use of
sort_coo for sorting the tensor.
Reviewed By: Peiming
Differential Revision: https://reviews.llvm.org/D141295
