llvm-project
llvm-project copied to clipboard
The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.
Building https://github.com/ColinIanKing/stress-ng with clang and `-flto` results in an assertion failure during the LTO step: ```console ld.gold: /home/npopov/repos/llvm-project/llvm/lib/Bitcode/Reader/BitcodeReader.cpp:6597: std::pair {anonymous}::ModuleSummaryIndexBitcodeReader::getValueInfoFromValueId(unsigned int): Assertion `VGI.first' failed. ``` The problematic input appears to...
There is some inconsistency in how the clang-cl driver handles `-Wno-unknown-pragmas`: ```console $ echo "#pragma deprecated(foo)" > foo.h $ clang-cl -W3 foo.h foo.h(1,9): warning: unknown pragma ignored [-Wunknown-pragmas] #pragma deprecated(foo)...
If one passes `-verify=foo`, the error message when no `foo-*` directives are found is: ```console error: no expected directives found: consider use of 'expected-no-diagnostics' ``` however, what's meant here is...
MLIR currently supports forward dataflow (`Analysis/Dataflow/{Dense,Sparse}Analysis.h`), but not backward dataflow. The latter would e.g. be needed for removing unused arguments and results from functions. For example, it would allow us...
When using Ubuntu 21.10, Ubuntu 22.04, Ubuntu 22.10 Mesa bug (I don't think that this is Mesa bug) - https://gitlab.freedesktop.org/mesa/mesa/-/issues/5699#note_1586842 ```console terminate called after throwing an instance of 'std::logic_error' what():...
| | | | --- | --- | | Bugzilla Link | [52016](https://llvm.org/bz52016) | | Version | trunk | | OS | All | | Reporter | LLVM Bugzilla Contributor...
Clang 10 and newer produce sub-optimal code when doing a signed floor division for the same LLVM IR. I have 3 versions of `floor_div` that preform division with rounding towards...
These docs look stale: https://clang.llvm.org/docs/UsersManual.html#cmdoption-fuse-ctor-homing "This flag is not enabled by default, and needs to be used with -cc1 or -Xclang." I believe it is enabled by default at this...
This seems to see a ~100x slowdown in qemu in scalar vs vector mode. Reduced test case from XNNPACK: ``` #include #include #include #include bool z{true}; size_t a{1}; uint32_t b{1};...
log: https://llvm-jenkins.debian.net/job/llvm-toolchain-binaries/architecture=i386,distribution=unstable,label=i386/680/console ```console Testing: 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50...