Goal is to remove use of isOpNewLike. I looked at a couple approaches to this, and this turned out to be the cheapest one. Just letting deref_or_null be generated causes a bunch of test diffs, and I couldn't convince myself there wasn't a real regression somewhere. A generic instcombine to convert deref_or_null + nonnull to deref is annoying complicated since you have to mix facts from callsite and declaration while manipulating only existing call site attributes. It just wasn't worth the code complexity.
Note that the change in new-delete-itanium.ll is a real regression. If you have a callsite which overrides the builtin status of a nobuiltin declaration, *and* you don't put the apppriate attributes on that callsite, you may lose the deref fact. I decided this didn't matter; if anyone disagrees, you can add this case to the generic non-null inference.
The reason for those nested namespaces is explained in D115315:
> AIUI, this keeps the CPO's own type from ADL'ing into the `std::ranges`
> namespace; e.g. `foobar(std::ranges::uninitialized_default_construct)`
> should not consider `std::ranges::foobar` a candidate, even if
> `std::ranges::foobar` is not a CPO itself. Also, of course, consistency
> (Chesterton's Fence, the economist's hundred-dollar bill): if it were
> safe to omit the namespace, we'd certainly want to do it everywhere,
> not just here.
This makes these three niebloids more consistent with the other Ranges
niebloids we've already implemented, such as the `ranges::begin` group
and the `ranges::uninitialized_default_construct` group.
FWIW, we still have three different indentation-and-comment styles
among these three groups.
Differential Revision: https://reviews.llvm.org/D116569
Ensure that errors in `frame variable` are reflected in result object.
The statistics for `frame variable` show invocations as being successful, even
when executing one of the error paths.
This change replaces `result.GetErrorStream()` with `result.AppendError()`,
which also sets the status to `eReturnStatusFailed`.
Differential Revision: https://reviews.llvm.org/D116788
To enable this on all targets there's still a number of regressions due to getSplatValue/getTargetVShiftNode but these don't really affect pre-AVX targets.
We're testing that the RegionLoop pointer is null in the first part of the check, so we need to check that its non-null before dereferencing it in a later part of the check.
This is very similar to the existing ROTL/ROTR support for scalar shifts in LowerRotate, I think as time goes on we should be able to share much of this code in helpers between Funnel Shift + Rotation lowering.
Fix crash in the presence of yield values. Multiple fixes to affine loop
tiling pre-condition checks and return status. Do not signal pass
failure on a failure to tile since the IR is still valid. Detect index
set computation failure in checkIfHyperrectangular and return failure.
Replace assertions with proper status return. Move checks to an
appropriate place earlier in the utility before mutation happens.
Differential Revision: https://reviews.llvm.org/D116738
Unsigned compares of the form <u 0 are always false. Do not create such
a redundant check in generateOverflowCheck.
The patch introduces a new lambda to create the check, so we can
exit early conveniently and skip creating some instructions feeding the
check.
I am planning to sink a few additional instructions as follow-ups, but I
would prefer to do this separately, to keep the changes and diff
smaller.
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D116811
This patch moves PresburgerSet to Presburger/ directory. This patch is purely
mechincal, it only moves and renames functionality and tests.
This patch is part of a series of patches to move presburger functionality to
Presburger/ directory.
Reviewed By: arjunp
Differential Revision: https://reviews.llvm.org/D116836
Change CombinerHelper::matchBitfieldExtractFromShrAnd to use
getPreferredShiftAmountTy for the shift-amount-like operands of G_UBFX
just like all the other G_[SU]BFX combines do. This better matches the
AMDGPU legality rules for these instructions.
Differential Revision: https://reviews.llvm.org/D116803
1. Fix CombinerHelper::matchBitfieldExtractFromAnd to check legality
with the correct types for the G_UBFX that it builds.
2. Fix AMDGPUTargetLowering::isConstantUnsignedBitfieldExtractLegal to
match the legality rules: result and first operand can be s32 or s64
but the "shift amount" operands are always s32.
3. Add AMDGPU tests where the post-legalizer combiner would create
illegal MIR without the above fixes.
Differential Revision: https://reviews.llvm.org/D116802
- Prevent symbols from being stripped so that it can run with 'RelWithDebInfo'.
- Adjust llvm-support CHECKs after code changes.
- Polish mlir-support CHECKs as suggested in https://reviews.llvm.org/D116646.
Differential Revision: https://reviews.llvm.org/D116837
Inspired by LLVM_DEBUG, but using environment variables rather than command line
options.
Code can use ORC_RT_DEBUG(...) (if ORC_RT_DEBUG_TYPE is set), or
ORC_RT_DEBUG_WITH_TYPE(<type>, ...) (if ORC_RT_DEBUG_TYPE is not set. E.g. in
headers).
Debug logging is enabled in the executor by setting the ORC_RT_DEBUG environment
variable. Debug logging can be restricted by type by setting the
ORC_RT_DEBUG_TYPES environment variable to a comma separated list of types,
e.g. ORC_RT_DEBUG_TYPES=macho_platform,sps.
Differential Revision: https://reviews.llvm.org/D116139
By default we return the width of an LMUL=1 register. We can enable
testing with larger LMUL values by returning a larger bit width.
This patch adds a RISCV specific option to provide a LMUL which will be
multiplied by the LMUL=1 bit width.
Reviewed By: kito-cheng
Differential Revision: https://reviews.llvm.org/D116339
getMinVectorRegisterBitWidth means what vector types is supported in
this target, and actually RISC-V support all fixed length vector types with
vector length less than `getMinRVVVectorSizeInBits`, so set it to 16,
means 2 x i8, that is minimal fixed length vector size in theory.
That also fixed one issue, some testcase migth become non-vectorizable
when `-riscv-v-vector-bits-min` set to larger value, because the vector size is
smaller than `-riscv-v-vector-bits-min`.
For example, following code can vectorize by SLP with
`-riscv-v-vector-bits-min=128` or `-riscv-v-vector-bits-min=256`, but
can't vectorize `-riscv-v-vector-bits-min=512` or larger:
```
void foo(double *da) {
da[0] = 0;
da[1] = 1;
da[2] = 2;
da[3] = 3;
}
```
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D116534
a32300a changed it to create a ThreadPool eagerly so that it gets reused
across buffers, however it also made it so that we create a ThreadPool
early even if we're not gonna use it later because of the command line
option `--mlir-disable-threading` is provided.
Fix#53056
Reland 45adf60802 after build fixes
Differential Revision: https://reviews.llvm.org/D116848
This reverts commit 3ca6928344.
492de35df4 relanded in feeff8a37c.
The reland only uses CMAKE_INSTALL_INCLUDEDIR and not CMAKE_INSTALL_BINDIR
in llvm-config, so this reland reflects that.
Currently, compiles with LLVM_ENABLE_THREADS=OFF fail due to this symbol missing. Add it but assert as calling code is (and should be) checking that threading is enabled.
Differential Revision: https://reviews.llvm.org/D116846
a32300a changed it to create a ThreadPool eagerly so that it gets reused
across buffers, however it also made it so that we create a ThreadPool
early even if we're not gonna use it later because of the command line
option `--mlir-disable-threading` is provided.
Fix#53056
Differential Revision: https://reviews.llvm.org/D116848
When building libcxx, libcxxabi, and libunwind the build environment may
specify any number of sanitizers. For some build feature tests these
sanitizers must be disabled to prevent spurious linking errors. With
-fsanitize= this is straight forward with -fno-sanitize=all. With
-fsanitize-coverage= there is no -fno-sanitize-coverage=all, but there
is the equivalent undocumented but tested -fsanitize-coverage=0.
The current build rules fail to disable 'trace-pc-guard'. By disabling
all sanitize-coverage flags, including 'trace-pc-guard', possible
spurious linker errors are prevented. In particular, this allows libcxx,
libcxxabi, and libunwind to be built with HonggFuzz.
CMAKE_REQUIRED_FLAGS is extra compile flags when running CMake build
configuration steps (like check_cxx_compiler_flag). It does not affect
the compile flags for the actual build of the project (unless of course
these flags change whether or not a given source compiles and links or
not). So libcxx, libcxxabi, and libunwind will still be built with any
specified sanitize-coverage as before. The build configuration steps
(which are mostly checking to see if certain compiler flags are
available) will not try to compile and link "int main() { return 0;}"
(or other specified source) with sanitize-coverage (which can fail to
link at this stage in building, since the final compile flags required
are yet to be determined).
The change to LIBFUZZER_CFLAGS was done to keep it consistent with the
obvious intention of disabling all sanitize-coverage. This appears to
be intentional, preventing the fuzzer driver itself from showing up in
any coverage calculations.
Reviewed By: #libunwind, #libc, #libc_abi, ldionne, phosek
Differential Revision: https://reviews.llvm.org/D116050
Even if findImplementors does not use
uninitialized parameter it's still UB and
it's going to be detected by msan with:
-Xclang -enable-noundef-analysis -mllvm -msan-eager-checks=1
Differential Revision: https://reviews.llvm.org/D116827
This is the patch for LLVM proper in my series for adding GNUInstallDirs support in all project.
Additionally:
Create a new `CACHE STRING` variable, `LLVM_EXAMPLES_INSTALL_DIR`, to control where the examples are installed on analogy with the other variables.
---
This patch supersedes D28234, which tried to do the same thing but hand-rolled without GNUInstallDirs.
This patch nearly reverts commit 3 0fc88bf1dc15a72e2d9809d28019d386b7a7cc0, which was a revert of a prior attempt."
(I had to add a space here or else Phabricator detects a reference cycle and won't let me do the form submit.)
Reviewed By: compnerd
Differential Revision: https://reviews.llvm.org/D100810
Users could pass flags by environment variables like CFLAGS/CXXFLAGS/LDFLAGS
or by using CMAKE_<LANG>_FLAGS_INIT/CMAKE_<t>_LINKER_FLAGS_INIT. So this
toolchain file should append to INIT flags instead. Otherwise, user
flags would be discarded here by assigning to CMAKE_<LANG>_FLAGS
directly.
Reviewed By: smeenai, mstorsjo
Differential Revision: https://reviews.llvm.org/D116709
The lldbconfig module was necessary to run the LLDB test suite against a
reproducer. Since this functionality has been removed, the module is no
longer necessary.
@beanz tells me it should in fact exists, and it would be bad to put a
list of dirs in `LLVM_LIBRARY_DIR`. I am therefore making some changes
to help avoid this an other mistakes:
- Use a separate `LLVM_CONFIG_LIBRARY_DIR` (no "S") variable so we don't
start putting a list in `LLVM_LIBRARY_DIR` by mistake.
- Define the individual dir variables first, and the define the list of
dirs variables programmatically.
Also, I rearranged the definitions of the "regular" single dirs below so
`LLVM_BINARY_DIR`, which is the unsuffixed one, comes first before the
suffixed ones.
Reviewed By: beanz
Differential Revision: https://reviews.llvm.org/D116497
Philip Reames