This add constant folding for all the integer vector reduce intrinsics,
providing that the argument is a constant vector. zeroinitializer always
produces 0 for all intrinsics, and other values can be handled with
APInt operators.
Differential Revision: https://reviews.llvm.org/D80516
Don't run tests that use undefined behavior sanitizer inside an
address-sanitized LLDB. The tests don't support that
configuration. Incidentally they were skipped on green dragon for a
different reason, so this hasn't come up there before.
Don't run tests that use thread sanitizer inside an address-sanitized
LLDB. The tests don't support that configuration. Incidentally they
were skipped on green dragon for a different reason, so this hasn't
come up there before.
optimizations
As discussed in the thread http://lists.llvm.org/pipermail/llvm-dev/2020-May/141838.html,
some bit field access width can be reduced by ReduceLoadOpStoreWidth, some
can't. If two accesses are very close, and the first access width is reduced,
the second is not. Then the wide load of second access will be stalled for long
time.
This patch add command line options to guard ReduceLoadOpStoreWidth and
ShrinkLoadReplaceStoreWithStore, so users can use them to disable these
store width reduction optimizations.
Differential Revision: https://reviews.llvm.org/D80745
Summary:
In this patch I am trying to get rid of the `Irrelevant` types from the
signatures of the functions from the standard C library. For that I've
introduced `lookupType()` to be able to lookup arbitrary types in the global
scope. This makes it possible to define the signatures precisely.
Note 1) `fread`'s signature is now fixed to have the proper `FILE *restrict`
type when C99 is the language.
Note 2) There are still existing `Irrelevant` types, but they are all from
POSIX. I am planning to address those together with the missing POSIX functions
(in D79433).
Reviewers: xazax.hun, NoQ, Szelethus, balazske
Subscribers: whisperity, baloghadamsoftware, szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy, dkrupp, gamesh411, Charusso, steakhal, ASDenysPetrov, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D80016
One can have multiple simulator runtimes installed, supporting
various generations of OSs. The logic in TestAppleSimulatorOSType
might select a rnutime older than the one targeted by the current
tools, preventing the executable from running. This commit changes
the test to look for the most recent runtime available instead.
The ARM specific code was trying to determine endianness using the
`__LITTLE_ENDIAN__` macro which is not guaranteed to be defined.
When not defined, it makes libunwind to build the big-endian code even
when the compiler builds for a little-endian target.
This change allows building libunwind with the `musl-gcc` toolchain
which does not define `__LITTLE_ENDIAN__`. Use `__BYTE_ORDER__`
instead.
Patch by Idan Freiberg!
These patterns for i8 and i16 VMLA's were missing. They end up from
legalized vector.reduce.add.v8i16 and vector.reduce.add.v16i8, and
although the instruction works differently (the mul and add are
performed in a higher precision), I believe it is OK because only an
i8/i16 are demanded from them, and so the results will be the same. At
least, they pass any testing I can think to run on them.
There are some tests that end up looking worse, but are quite artificial
due to passing half vector types through a call boundary. I would not
expect the vmull to realistically come up like that, and a vmlava is
likely better a lot of the time.
Differential Revision: https://reviews.llvm.org/D80524
SUMMARY:
when there are two symbol has the same address. llvm-objdump -D -symbol-description will select symbol based on the following rule:
1. using Label first if there is a Label symbol.
2. If there is not Label, using a symbol which has Storage Mapping class.
3. if more than one symbol has storage mapping class, put the TC0 has the low priority, for other storage mapping class , compare based on the value.
Reviewers: James Henderson ,hubert.reinterpretcast,
Differential Revision: https://reviews.llvm.org/D78387
As LLVM has moved from SVN to git, there is no need to
keep SVN related code. Also, this code piece was never used.
Differential Revision: https://reviews.llvm.org/D79400
Summary:
Diagnostic is emitted if some declaration of unsupported type
declaration is used inside device code.
Memcpy operations for structs containing member with unsupported type
are allowed. Fixed crash on attempt to emit diagnostic outside of the
functions.
The approach is generalized between SYCL and OpenMP.
CUDA/OMP deferred diagnostic interface is going to be used for SYCL device.
Reviewers: rsmith, rjmccall, ABataev, erichkeane, bader, jdoerfert, aaron.ballman
Reviewed By: jdoerfert
Subscribers: guansong, sstefan1, yaxunl, mgorny, bader, ebevhan, Anastasia, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D74387
Summary:
unittests/AST/Language.h defines some helpers that we would like to
reuse in other tests, for example, in tests for syntax trees.
Reviewers: sammccall
Reviewed By: sammccall
Subscribers: mgorny, martong, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D80792
Summary:
Once we found a matching FunctionDecl for the given summary then we
validate the given constraints against that FunctionDecl. E.g. we
validate that a NotNull constraint is applied only on arguments that
have pointer types.
This is needed because when we matched the signature of the summary we
were working with incomplete function types, i.e. some intricate type
could have been marked as `Irrelevant` in the signature.
Reviewers: NoQ, Szelethus, balazske
Subscribers: whisperity, xazax.hun, baloghadamsoftware, szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy, dkrupp, gamesh411, Charusso, steakhal, ASDenysPetrov, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D77658
cmake configure fails when it tries to setup target for llvm_vcsrevision_h
This happens only when source is checked out using repo in a read
only filesystem, because cmake tries to create `.git/logs/HEAD` file.
This patch:
1. Recovers from failure gracefully.
2. Ensures that VCSRevision.h is successfully created and updated
in above scenarios.
Differential Revision: https://reviews.llvm.org/D79400
Replace calls to getNumElements() with getElementCount() in order
to avoid warnings for scalable vectors. The warnings were discovered
by this existing test:
test/CodeGen/AArch64/sve-gep.ll
Differential revision: https://reviews.llvm.org/D80782
Summary:
Further develop the buffer size argumentum constraint so it can handle sizes
that we can get by multiplying two variables.
Reviewers: Szelethus, NoQ, steakhal
Subscribers: whisperity, xazax.hun, baloghadamsoftware, szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy, dkrupp, gamesh411, Charusso, ASDenysPetrov, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D77148
When creating a new vector type based on another vector type we
should pass in the element count instead of the number of elements
and scalable flag separately.
I encountered this warning whilst compiling this test:
CodeGen/AArch64/sve-intrinsics-int-compares.ll
Differential revision: https://reviews.llvm.org/D80621
Summary:
Introducing a new argument constraint to confine buffer sizes. It is typical in
C APIs that a parameter represents a buffer and another param holds the size of
the buffer (or the size of the data we want to handle from the buffer).
Reviewers: NoQ, Szelethus, Charusso, steakhal
Subscribers: whisperity, xazax.hun, baloghadamsoftware, szepet, rnkovacs, a.sidorin, mikhail.ramalho, donat.nagy, dkrupp, gamesh411, ASDenysPetrov, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D77066
It would be nice to switch to `reportUniqueWarnings` from
`reportError` in this class, but first of all it needs a cleanup.
This patch:
1) Eliminates autos.
2) Removes code duplication.
3) Changes how the code works with `Expected<>`.
4) Introduces 2 new `using`s to make the code a bit shorter.
Differential revision: https://reviews.llvm.org/D80726
When a group member is removed, the corresponding record in the
SHT_GROUP section has to be deleted.
This fixes PR46064.
Differential Revision: https://reviews.llvm.org/D80568
When a SHT_GROUP section is removed, but other sections of the group are
kept, the SHF_GROUP flag of these sections should be dropped, otherwise
the resulting ELF file will be malformed.
Differential Revision: https://reviews.llvm.org/D80511
Summary:
Declaring these helpers in the ast_matcher namespace in the clangAST
unit test seems inappropriate -- neither these helpers, nor clangAST have
anything to do with AST matchers. Therefore, I moved these helpers to
the clang namespace.
Declaring another typedef called "ArgVector" is not a good idea -- we
already have both "ArgVector", "ArgsVector", and "ArgList". I expanded
it into the underlying type.
Declaring another enum called "Language" is not a good idea because we
arleady have the "clang::Language" enum. I renamed it to
"TestLanguage".
Similarly, I renamed "getBasicRunOptionsForLanguage" to
"getCommandLineArgsForTesting" to explain the semantics better (what are
"run options"?) and not repeat types in the function name
("ForLanguage").
Reviewers: shafik, rengolin, sammccall
Reviewed By: sammccall
Subscribers: gribozavr2, sammccall, martong, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D80786
These are mapped in MachO::getMachOArchName already, but were missing
in ToolChain::getDefaultUniversalArchName.
Having these reverse mapped here fixes weird inconsistencies like
-dumpmachine showing a target triple like "aarch64-apple-darwin",
while "clang -target aarch64-apple-darwin" didn't use to work (ended
up mapped as unknown-apple-ios).
Differential Revision: https://reviews.llvm.org/D79117
Currently combineInsertEltToShuffle turns insert_vector_elt into a
vector_shuffle, even if the inserted element is a vector with a single
element. In this case, it should be unlikely that the additional shuffle
would be more efficient than a insert_vector_elt.
Additionally, this fixes a infinite cycle in DAGCombine, where
combineInsertEltToShuffle turns a insert_vector_elt into a shuffle,
which gets turned back into a insert_vector_elt/extract_vector_elt by
a custom AArch64 lowering (in visitVECTOR_SHUFFLE).
Such insert_vector_elt and extract_vector_elt combinations can be
lowered efficiently using mov on AArch64.
There are 2 test changes in arm64-neon-copy.ll: we now use one or two
mov instructions instead of a single zip1. The reason that we need a
second mov in ins1f2 is that we have to move the result to the result
register and is not really related to the DAGCombine fold I think.
But in any case, on most uarchs, mov should be cheaper than zip1. On a
Cortex-A75 for example, zip1 is twice as expensive as mov
(https://developer.arm.com/docs/101398/latest/arm-cortex-a75-software-optimization-guide-v20)
Reviewers: spatel, efriedma, dmgreen, RKSimon
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D80710
This revision adds custom rewrites for patterns that arise during linalg structured
ops vectorization. These patterns allow the composition of linalg promotion,
vectorization and removal of redundant copies.
The patterns are voluntarily limited and restrictive atm.
More robust behavior will be implemented once more powerful side effect modeling and analyses are available on view/subview.
On the transfer_read side, the following pattern is rewritten:
```
%alloc = ...
[optional] %view = std.view %alloc ...
%subView = subview %allocOrView ...
[optional] linalg.fill(%allocOrView, %cst) ...
...
linalg.copy(%in, %subView) ...
vector.transfer_read %allocOrView[...], %cst ...
```
into
```
[unchanged] %alloc = ...
[unchanged] [optional] %view = std.view %alloc ...
[unchanged] [unchanged] %subView = subview %allocOrView ...
...
vector.transfer_read %in[...], %cst ...
```
On the transfer_write side, the following pattern is rewriten:
```
%alloc = ...
[optional] %view = std.view %alloc ...
%subView = subview %allocOrView...
...
vector.transfer_write %..., %allocOrView[...]
linalg.copy(%subView, %out)
```
Differential Revision: https://reviews.llvm.org/D80728
This patch helps make the `PubSection` optional in the DWARF structure.
Reviewed By: jhenderson, aprantl
Differential Revision: https://reviews.llvm.org/D80722
This utility factors out the machinery required to add iterArgs and yield values to an scf.ForOp.
Differential Revision: https://reviews.llvm.org/D80656
AddressingModeMatcher::matchScaledValue was calling getSExtValue for a constant before ensuring that we can actually represent the value as int64_t
Fixes OSSFuzz#22723 which is a followup to rGc479052a74b2 (PR46004 / OSSFuzz#22357)
Summary: Depends on D80198.
This patch implies ASTs might be built with stale preambles without
blocking for a fresh one. It also drops any guarantees on every preamble
version being built. In case of multiple preamble build requests, in
addition to being debounced.
Any preamble requested with a WantDiags::Yes will always be built, this
is ensured by blocking enqueueing of any subsequent reqest.
AST worker will still block for initial preamble to reduce duplicate
work.
Subscribers: ilya-biryukov, javed.absar, MaskRay, jkorous, arphaman, jfb, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D80293
Summary:
The description of EXTACT_SUBVECTOR and INSERT_SUBVECTOR has been
changed to accommodate scalable vectors (see ISDOpcodes.h). This
patch updates the asserts used to verify these requirements when
using SelectionDAG's getNode interface.
This patch introduces the MVT function getVectorMinNumElements
that can be used against fixed-length and scalable vectors when
only the known minimum vector length is required.
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80709
Christopher Tetreault