NewGVN will find operator from other context. ValueTracking currently doesn't have a way to run completely without context instruction.
So it will use operator itself as conext instruction.
If the operator in another branch will never be executed but it has an assume, it may caused value tracking use the assume to do wrong simpilfy.
It would be better to make these simplification queries not use context at all, but that would require some API changes.
For now we just use the orignial instruction as context instruction to fix the issue.
Fix#56039
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D127942
createInductionResumeValues creates a phi node placeholder
without filling incoming values. Then it generates the incoming values.
It includes triggering of SCEV expander which may invoke SSAUpdater.
SSAUpdater has an optimization to detect number of predecessors
basing on incoming values if there is phi node.
In case phi node is not filled with incoming values - the number of predecessors
is detected as 0 and this leads to segmentation fault.
In other words SSAUpdater expects that phi is in good shape while
LoopVectorizer breaks this requirement.
The fix is just prepare all incoming values first and then build a phi node.
Reviewed By: fhahn
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D128033
Summary:
This patch removes a duplicated exit from the OpenMP data envrionment.
We already have an RAII method that guards this environment so it is
unnecessary.
This avoid creating empty bins in AAPointerInfo which can lead to
segfaults. Also ensure we do not try to translate from callee to caller
except if we really take the argument state and move it to the call site
argument state.
Fixes: https://github.com/llvm/llvm-project/issues/55726
When determining liveness via Attributor::isAssumedDead(...) we might
end up without a liveness AA or with one pointing into another function.
Neither is helpful and we will avoid both from now on.
Reapplied after fixing the ASAN error which caused the revert:
db68a25ca9
Add a setting to configure how LLDB parses dynamic Objective-C class
metadata. By default LLDB will choose the most appropriate method for
the target OS.
Differential revision: https://reviews.llvm.org/D128312
Recent revisions of the Objective-C runtime changed
objc_debug_class_getNameRaw() in a way that no longer triggers lazy
names to be instantiated. This has the unintended side-effect of making
generic bridged Swift classes, such as _SwiftDeferredNSDictionary<U,V>
to become invisible to the Objective-C runtime.
This patch detects this situation and forces the names to be
instantiated by calling class_getName() and discarding the result before
calling objc_debug_class_getNameRaw() again.
Many thanks to Mike Ash for outlining the solution and Adrian for
authoring the downstream patch.
rdar://95245318
Differential revision: https://reviews.llvm.org/D128306
erf and round op are able to lowered to libm supporting vector type as other math operations.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D127934
According to ministat, this is a small but measurable speedup
(using the repro in PR56121):
N Min Max Median Avg Stddev
x 10 3.7439518 3.7783802 3.7730219 3.7655502 0.012375226
+ 10 3.6149218 3.692198 3.6519327 3.6502951 0.025905601
Difference at 95.0% confidence
-0.115255 +/- 0.0190746
-3.06078% +/- 0.506554%
(Student's t, pooled s = 0.0203008)
(Without 858e8b17f7, this change here to use parallelFor is an 18% speedup,
and doing 858e8b17f7 on top of this change is just a 2.55% +/- 0.58% win.
Doing both results in a total speedup of 20.85% +/- 0.44%.)
Differential Revision: https://reviews.llvm.org/D128298
During the reordering transformation we should try to avoid reordering bundles
like fadd,fsub because this may block them being matched into a single vector
instruction in x86.
We do this by checking if a TreeEntry is such a pattern and adding it to the
list of TreeEntries with orders that need to be considered.
Differential Revision: https://reviews.llvm.org/D125712
This enables an existing transformation that when combined with an
add will emit saba/uaba instructions.
Differential Revision: https://reviews.llvm.org/D128198
The error used to look like this:
ld64.lld: error: undefined symbol: _foo
>>> referenced by /path/to/bar.o:(symbol _baz+0x4)
If DWARF line information is available, we now show where in the source
the references are coming from:
ld64.lld: error: unreferenced symbol: _foo
>>> referenced by: bar.cpp:42 (/path/to/bar.cpp:42)
>>> /path/to/bar.o:(symbol _baz+0x4)
The reland is identical to the first time this landed. The fix was in D128294.
This reverts commit 0cc7ad4175.
Differential Revision: https://reviews.llvm.org/D128184
Marking bufferization allocation operation as invalid
during sparse lowering is too strict, since dense and
sparse allocation can co-exist. This revision refines
the lowering with a dynamic type check.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D128305
[JITLink][Orc] Add MemoryMapper interface with InProcess implementation
MemoryMapper class takes care of cross-process and in-process address space
reservation, mapping, transferring content and applying protections.
Implementations of this class can support different ways to do this such
as using shared memory, transferring memory contents over EPC or just
mapping memory in the same process (InProcessMemoryMapper).
The original patch landed with commit 6ede652050
It was reverted temporarily in commit 6a4056ab2a
Reviewed By: sgraenitz, lhames
Differential Revision: https://reviews.llvm.org/D127491
As noted by @nikic, D126061 causes a compile time regression of about
0.5% on -O0 builds:
http://llvm-compile-time-tracker.com/compare.php?from=7acc88be0312c721bc082ed9934e381d297f4707&to=8c7b64b5ae2a09027c38db969a04fc9ddd0cd6bb&stat=instructions
This happens because, in a number of places, D126061 creates an
additional local variable of type `ParsedAttributes`. In the large
majority of cases, no attributes are added to this `ParsedAttributes`,
but it turns out that creating an empty `ParsedAttributes`, then
destroying it is a relatively expensive operation.
The reason for this is because `AttributePool` uses a `TinyPtrVector` as
its underlying vector class, and the technique that `TinyPtrVector`
employs to achieve its extreme memory frugality makes the `begin()` and
`end()` member functions relatively slow. The `ParsedAttributes`
destructor iterates over the attributes in its `AttributePool`, and this
is a relatively expensive operation because `TinyPtrVector`'s `begin()` and
`end()` are relatively slow.
The fix for this is to replace `TinyPtrVector` in `ParsedAttributes` and
`AttributePool` with `SmallVector`. `ParsedAttributes` and
`AttributePool` objects are only ever allocated on the stack (they're
not part of the AST), and only a small number of these objects are live
at any given time, so they don't need the extreme memory frugality of
`TinyPtrVector`.
I've confirmed with valgrind that this patch does not increase heap
memory usage, and it actually makes compiles slightly faster than they
were before D126061.
Here are instruction count measurements (obtained with callgrind)
running `clang -c MultiSource/Applications/JM/lencod/parsetcommon.c`
(a file from llvm-test-suite that exhibited a particularly large
compile-time regression):
7acc88be03
(baseline one commit before D126061 landed)
102,280,068 instructions
8c7b64b5ae
(the patch that landed D126061)
103,289,454 instructions
(+0.99% relative to baseline)
This patch applied onto
8c7b64b5ae
101,117,584 instructions
(-1.14% relative to baseline)
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D128097
Put it before the VL instead of as the first operand. I want to add
passthru to more operands, but the commutable ones like VADD_VL
require the commutable operands to be operand 0 and 1. So we can't
have the passthru as operand 0 for those.
This just adds some very basic vectorizer testing with both fixed and scalable vectorization enabled. For context, I just yesterday fixed a crash in costing of the splat_ptr example - see bbf3fd.
We're slowly removing SelectionDAG::GetDemandedBits and replacing it with SimplifyMultipleUseDemandedBits, we no longer have any uses for the vector demanded elt variant.
Summary:
This patch adds _LIBCXXABI_FUNC_VIS to the definitions of the personality and helper routines for the state table based EH, now that the support of the visibility attribute is being added to AIX Clang. Currently an export list is generated in the absence of the visibility attribute support downstream.
Reviewed by: MaskRay, daltenty
Differential Revision: https://reviews.llvm.org/D126915
This reverts commit 90c9d41c8a.
Keeping arm* LLVM_ENABLE_PER_TARGET_RUNTIME_DIR=on is sufficient to work around
LLVM_ENABLE_RUNTIMES not working well with arm. It is more important for
LLVM_ENABLE_PROJECTS and LLVM_ENABLE_RUNTIMES to match in the file hierarchy.
Michał's change in https://reviews.llvm.org/D127193 did a search &
replace for a pattern that also appears in debugserver, but it
shouldn't be done there.
In some cases, a recurrence splice instructions needs to be inserted
between to regions, for example if the regions get re-arranged during
sinking.
Fixes#56146.
If ld64.lld was supplied an object file that had a `__debug_abbrev` or
`__debug_str` section, but didn't have any compile unit DIEs in
`__debug_info`, it would dereference an iterator pointing to the empty
array of DIEs. This underlying issue started causing segmentation faults
when parsing for `__debug_info` was addded in D128184. That commit was
reverted, and this one fixes the invalid dereference to allow relanding
it.
This commit adds an assertion to `filter_iterator_base`'s dereference
operators to catch bugs like this one.
Ran check-llvm, check-clang and check-lld.
Differential Revision: https://reviews.llvm.org/D128294
This created a weird loop making the tested registers live out of the
block, which I don't think is relevant to the purpose of the
tests. This caused regressions when the validity queries are changed
to use tests based whether the use instruction was a kill. If the
register was live out for the loop, it was still live.
I guess we could still do this in a narrow case where the value loops
back, but that's most a pointlessly complex case to handle.
It's in libSystem, so it doesn't bring in any new deps, and it's
currently much faster than LLVM's current SHA256 implementation.
Makes linking (arm64) Chromium Framework with ld64.lld 17% faster.
See also PR56121.
No behavior change.
Differential Revision: https://reviews.llvm.org/D128290
For non-mem-intrinsic and non-lifetime `CallBase`s, the current
`isRemovable` function only checks if the `CallBase` 1. has no uses 2.
will return 3. does not throw:
80fb782336/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp (L1017)
But we should also exclude invokes even in case they don't throw,
because they are terminators and thus cannot be removed. While it
doesn't seem to make much sense for `invoke`s to have an `nounwind`
target, this kind of code can be generated and is also valid bitcode.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D128224
Add more diagnostics to fir.coordinate_of to provide better checking
that the IR is sane.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D128255
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Tom Stellard