[InstCombine] Signed saturation patterns
This adds an instcombine matcher for code that attempts to perform signed
saturating arithmetic by casting to a higher type. Unsigned cases are already
matched, this adds extra matches for the more complex signed cases, which
involves matching the min(max(add a b)) nodes with proper extends to ensure
Differential Revision: https://reviews.llvm.org/D68651
[libcxx] Remove shared_ptr::make_shared
Summary: This patch removes `shared_ptr::make_shared` as it is not part of the
standard. This patch also adds __create_with_cntrl_block, which is a help function that
can be used in std::allocate_shared and std::make_shared. This is the third patch (out of
4) from D66178.
Reviewers: EricWF, mclow.lists, ldionne
Subscribers: christof, dexonsmith, libcxx-commits
Differential Revision: https://reviews.llvm.org/D68805
[MIParser] Set RegClassOrRegBank during instruction parsing
RegClassOrRegBank to static_cast<RegisterBank *>(nullptr).
MIParser on the other hand doesn't. When we attempt to constrain
Register Class on such VReg, additional COPY is generated.
This way we avoid COPY instructions showing in test that have MIR
input while they are not present with llvm-ir input that was used
to create given MIR for a -run-pass test.
Differential Revision: https://reviews.llvm.org/D68946
[MIPS GlobalISel] Select MSA vector generic and builtin add
Select vector G_ADD for MIPS32 with MSA. We have to set bank
for vector operands to fprb and selectImpl will do the rest.
__builtin_msa_addv_<format> will be transformed into G_ADD
in legalizeIntrinsic and selected in the same way.
__builtin_msa_addvi_<format> will be directly selected into
ADDVI_<format> in legalizeIntrinsic. MIR tests for it have
unnecessary additional copies. Capture current state of tests
with run-pass=legalizer with a test in test/CodeGen/MIR/Mips.
Differential Revision: https://reviews.llvm.org/D68984
[PowerPC] Turn on CR-Logical reducer pass
This re-commits r375152 which was pulled in r375233 because it broke
the EXPENSIVE_CHECKS bot on Windows.
The reason for the failure was a bug in the pass that the commit turned
on by default. This patch fixes that bug and turns the pass back on.
This patch has been verified on the buildbot that originally failed
thanks to Simon Pilgrim.
Differential revision: https://reviews.llvm.org/D52431
[LLVMDebugInfoPDB] - Use cantFail() instead of assert().
Currently injected-sources-native.test fails with "Expected<T>
value was in success state.
(Note: Expected<T> values in success mode must still be checked
prior to being destroyed)"
when llvm is compiled with LLVM_ENABLE_ABI_BREAKING_CHECKS in Release.
The problem is that getStringForID returns Expected<StringRef>
and Expected value must always be checked, even if it is in success state.
Checking with assert only helps in Debug and is wrong.
Differential revision: https://reviews.llvm.org/D69251
[builtins][test] Avoid unportable mmap call in clear_cache_test.c
Within the last two weeks, the Builtins-*-sunos :: clear_cache_test.c started to FAIL
on Solaris. Running it under truss shows
mmap(0x00000000, 128, PROT_READ|PROT_WRITE|PROT_EXEC, MAP_PRIVATE|MAP_ANON, 0, 0) Err#22
While there are several possible reasons mmap can return EINVAL on Solaris, it turns
out it's this one (from mmap(2)):
MAP_ANON was specified, but the file descriptor was not
And indeed even the Linux mmap(2) documents this as unportable:
The mapping is not backed by any file; its contents are initial‐
ized to zero. The fd argument is ignored; however, some imple‐
mentations require fd to be -1 if MAP_ANONYMOUS (or MAP_ANON) is
specified, and portable applications should ensure this. The
This patch follows this advise. Tested on x86_64-pc-linux-gnu, amd64-pc-solaris2.11
[2 lines not shown]
Refactor DependencyScanningTool to its own file
There's no behavior change - just moving DependencyScanningTool to its own file
since this tool can be reused across both clang-scan-deps binary and an interface
exposed as part of libClang APIs.
Reviewers: arphaman, jkorous, Bigcheese, dexonsmith
Subscribers: mgorny, cfe-commits
Differential Revision: https://reviews.llvm.org/D69186
[SPECCPU2017] Check for 'all' input data directory before generating command
Summary: As of SPEC CPU2017 1.0.5, 526.blender_r/data does not contain an 'all'
subdirectory, which breaks the build.
Reviewers: Meinersbur, MatzeB
Subscribers: mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D68967
fix PythonDataObjectsTest.TestExceptions on windows
Looks like on windows googlemock regexes treat newlines differently
from on darwin. This patch fixes the regex in this test so it
will work on both.
remove multi-argument form of PythonObject::Reset()
With this patch, only the no-argument form of `Reset()` remains in
PythonDataObjects. It also deletes PythonExceptionState in favor of
PythonException, because the only call-site of PythonExceptionState was
also using Reset, so I cleaned up both while I was there.
Reviewers: JDevlieghere, clayborg, labath, jingham
Reviewed By: labath
Subscribers: mgorny, lldb-commits
Differential Revision: https://reviews.llvm.org/D69214
Fix lld detection in standalone compiler-rt.
Right now all hwasan tests on Android are silently disabled because they
require "has_lld" and standalone compiler-rt can not (and AFAIK was
never able to) set it.
Reviewers: pcc, dyung
Subscribers: dberris, mgorny, #sanitizers, llvm-commits
Tags: #sanitizers, #llvm
Differential Revision: https://reviews.llvm.org/D69196
[hwasan] Workaround unwinder issues in try-catch test.
Android links the unwinder library to every DSO. The problem is,
unwinder has global state, and hwasan implementation of personality
function wrapper happens to rub it the wrong way.
Switch the test to static libc++ as a temporary workaround.
[examples] Add a dependency on ExecutionEngine to LLJITWithObjectCache example.
ExecutionEngine.cpp contains the anchor() for the ObjectCache base class, so we
need an explicit dependency on it.
Patch by Stephen Neuendorffer. Thanks Stephen!