* Refactor compression namespaces across the project, making way for a possible
introduction of alternatives to zlib compression.
Changes are as follows:
* Relocate the `llvm::zlib` namespace to `llvm::compression::zlib`.
Reviewed By: MaskRay, leonardchan, phosek
Differential Revision: https://reviews.llvm.org/D128953
Update the references to the old Mailman mailing lists to point to Discourse forums.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D128766
The existing provision is not sufficient, it did not allow for the cases
where an implementation partition includes the primary module interface,
or for the case that an exported interface partition is contains a decl
that is then implemented in a regular implementation unit.
It is somewhat unfortunate that we have to compare top level module names
to achieve this, since built modules are not necessarily available.
TODO: It might be useful to cache a hash of the primary module name if
this test proves to be a significant load.
Differential Revision: https://reviews.llvm.org/D127624
The motivation here is to a) bring us closer into alignment with AArch64 under the assumption that codepath is better tested, and b) simplify pattern matching in an upcoming change.
The immediate impact is a significant IR reduction but a fairly minimal change in the generated assembly. Due to a difference in expansion behavior we get a saturating add vs an unsaturating one for the old code, but that's about it. This difference comes down to different handling of overflow, which doesn't seem to be possible here anyways, so the assembly codegen is arguably a minor regression. I don't expect that to matter in practice.
Differential Revision: https://reviews.llvm.org/D129221
This allows fixed length vectors involving splats on the LHS to commute into the _vx form of the instruction. Oddly, the generic canonicalization rules appear to catch the scalable vector cases. I haven't fully dug in to understand why, but I suspect it's because of a difference in how we represent splats (splat_vector vs build_vector).
Differential Revision: https://reviews.llvm.org/D129302
Originally encountered with RUST, but also there are cases with distributed LTO
where debug info dwo units contain structurally the same debug information, with
difference in DW_AT_linkage_name. This causes collision on DWO ID.
Differential Revision: https://reviews.llvm.org/D129317
`SymbolUserMap` relied on `try_emplace` and `std::move` to relocate an entry to another key. However, if this triggered the resizing of the `DenseMap`, the value was destroyed before it could be moved to the new storage location, leading to a dangling `users` reference to be inserted into the map. On destruction, since a new entry was created from one that was already freed, a double-free error occurred.
Fixed issue by re-fetching the iterator after the mutation of the container.
Differential Revision: https://reviews.llvm.org/D129345
Array-value-copy fails to generate a temporary array for case like this:
subroutine bug(b)
real, allocatable :: b(:)
b = b(2:1:-1)
end subroutine
Since LHS may need to be reallocated, lowering produces the following FIR:
%rhs_load = fir.array_load %b %slice
%lhs_mem = fir.if %b_is_allocated_with_right_shape {
fir.result %b
} else {
%new_storage = fir.allocmem %rhs_shape
fir.result %new_storage
}
%lhs = fir.array_load %lhs_mem
%loop = fir.do_loop {
....
}
fir.array_merge_store %lhs, %loop to %lhs_mem
// deallocate old storage if reallocation occured,
// and update b descriptor if needed.
Since %b in array_load and %lhs_mem in array_merge_store are not the same SSA
values, array-value-copy does not detect the conflict and does not produce
a temporary array. This causes incorrect result in runtime.
The suggested change in lowering is to generate this:
%rhs_load = fir.array_load %b %slice
%lhs_mem = fir.if %b_is_allocated_with_right_shape {
%lhs = fir.array_load %b
%loop = fir.do_loop {
....
}
fir.array_merge_store %lhs, %loop to %b
fir.result %b
} else {
%new_storage = fir.allocmem %rhs_shape
%lhs = fir.array_load %new_storage
%loop = fir.do_loop {
....
}
fir.array_merge_store %lhs, %loop to %new_storage
fir.result %new_storage
}
// deallocate old storage if reallocation occured,
// and update b descriptor if needed.
Note that there are actually 3 branches in FIR, so the assignment loops
are currently produced in three copies, which is a code-size issue.
It is possible to generate just two branches with two copies of the loops,
but it is not addressed in this change-set.
Differential Revision: https://reviews.llvm.org/D129314
As far as I can tell treating s1 values as legal makes no sense. There
are no allocatable 1-bit registers. SelectionDAG legalizes the usual
set of boolean operations to 32-bits, and this should do the
same. This avoids some special case handling in the selector of s1
values, and some extra code to look through truncates.
This makes some code worse at -O0, since nothing cleans up the and 1
the artifact combiner inserts. We could probably add some
non-essential combines or teach the artifact combiner to elide
intermediates betweeen boolean uses and defs.
SelectionDAG has a target hook, getExtendForAtomicOps, which it uses
in the computeKnownBits implementation for ATOMIC_LOAD. This is pretty
ugly (as is having a separate load opcode for atomics), so instead
allow making use of atomic zextload. Enable this for AArch64 since the
DAG path defaults in to the zext behavior.
The tablegen changes are pretty ugly, but partially helps migrate
SelectionDAG from using ISD::ATOMIC_LOAD to regular ISD::LOAD with
atomic memory operands. For now the DAG emitter will emit matchers for
patterns which the DAG will not produce.
I'm still a bit confused by the intent of the isLoad/isStore/isAtomic
bits. The DAG implementation rejects trying to use any of these in
combination. For now I've opted to make the isLoad checks also check
isAtomic, although I think having isLoad and isAtomic set on these
makes most sense.
Check that the operation actually folded before trying to flush
denormals. A minor variation of the pr33453 test exposed this
with the FP binops marked as undesirable.
Summary:
The previous path reworked some handling of temporary files which
exposed some bugs related to capturing local state by reference in the
callback labmda. Squashing this by copying in everything instead. There
was also a problem where the argument name was changed for
`--bitcode-library=` but clang still used `--target-library=`.
Summary:
This patch reworks the command line argument handling in the linker
wrapper from using the LLVM `cl` interface to using the `Option`
interface with TableGen. This has several benefits compared to the old
method.
We use arguments from the linker arguments in the linker
wrapper, such as the libraries and input files, this allows us to
properly parse these. Additionally we can now easily set up aliases to
the linker wrapper arguments and pass them in the linker input directly.
That is, pass an option like `cuda-path=` as `--offload-arg=cuda-path=`
in the linker's inputs. This will allow us to handle offloading
compilation in the linker itself some day. Finally, this is also a much
cleaner interface for passing arguments to the individual device linking
jobs.
Avoid calling ConstantExpr::get() for associative/commutative
binops, call ConstantFoldBinaryOpOperands() instead. We only
want to perform the reassociation of the constants actually fold.
If all the demanded bits of the AND mask covering the inserted subvector 'X' are known to be one, then the mask isn't affecting the subvector at all.
In which case, if the base vector 'C' is undef/constant, then move the AND mask up to just (constant) fold it directly.
Addresses some of the regressions from D129150, particularly the cases where we're attempting to zero the upper elements of a widened vector.
Differential Revision: https://reviews.llvm.org/D129290
Replace ConstantExpr:getFAdd etc with call to
ConstantFoldBinaryOpOperands(). I'm using the constant folding API
rather than IRBuilder here to ensure that this does actually
constant fold. These transforms don't use m_ImmConstant(), so this
would not otherwise be guaranteed (and apparently, they can't use
m_ImmConstant because they want to handle scalable vector splats).
There is an opportunity here to further migrate these to the
ConstantFoldFPInstOperands() API, which would respect the denormal
mode. I've held off on doing so here, because some of this code
explicitly checks for denormal results, and I don't want to touch
it in a mostly NFC change.
Move the device_type parser to a separate parser AccDeviceTypeExprList. Preparatory work for D106968.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D106967
Set the isOptional flag for the self clause. Move the optional and parenthesis part of the parser. Update the rest of the code to deal with the optional value.
Preparatory work for D106968.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D106965
After D82916 `updateAllRanges()` started to fix holes in main range with
subranges but it fails on instructions with two subregs def which are parts of
one reg. The main range constructed with //all// subranges of subregs just after
processing the first operand. So the main range gets intervals from subranges
those are not updated yet.
The patch takes into account lane mask to update the main range.
Reviewed By: rampitec, arsenm
Differential Revision: https://reviews.llvm.org/D128553
This commit re-applies 9ee97ce3b8, which was reverted by 61d417ce
because it broke the LLDB data formatter tests. It also re-applies
6148c79a (the manual GN change associated to it).
Differential Revision: https://reviews.llvm.org/D127444
This is almost the same as the abandoned D48529, but it
allows splat vector constants too.
This replaces the x86-specific code that was added with
the alternate patch D48557 with the original generic
combine.
This transform is a less restricted form of an existing
InstCombine and the proposed SDAG equivalent for that
in D128080:
https://alive2.llvm.org/ce/z/OUm6N_
Differential Revision: https://reviews.llvm.org/D128123
See discussion here:
https://github.com/llvm/llvm-project/issues/55982
And the RFC here:
https://discourse.llvm.org/t/rfc-disable-clang-format-in-the-clang-test-tree/63498/2
We don't generally expect test files to be formatted according to the
style guide. Indeed, some tests may require specific formatting for the
purposes of the test.
When tests intentionally do not conform to the "correct" formatting,
this causes errors in the CI, which can drown out real errors and causes
people to stop trusting the CI over time.
From the history of the clang/test/.clang-format file, it looks as if
there have been attempts to make clang-format do a subset of formatting
that would be useful for tests. However, it looks as if it's hard to
make clang-format do exactly the right thing -- see the back-and-forth
between
13316a7
and
7b5bddf.
These changes disable the .clang-format file for clang/test, llvm/test,
and clang-tools-extra/test.
Fixes#55982
Differential Revision: https://reviews.llvm.org/D128706
Daniil Fukalov