• fix type mismatch for inline_memset

• rebase and squash

• rebase

Format commit message

• Use enum class for Temporal and Aligned attributes

• Fix documentation for address.h

In D124958#3492446, @sivachandra wrote:

I have a few questions about this patch:

1. Do you want all libc functions to be aligned to 64 bytes, or all public functions to be aligned to 64 bytes? Or, do you just want the memory functions to be aligned to 64 bytes?

@sivachandra at first I thought I should use LLVM_LIBC_FUNCTION_ATTR but it does not work. Should it be moved to the definition instead of the declaration?

In D124214#3483666, @michaelrj wrote:

The problem is that there are certain parts of libc++ that the tests include and depend on that aren't available. Most noticeably is iostream, which has #include "locale.h", which doesn't work since locales are disabled. The other problem is the memory functions, which have references to posix_memalign, which doesn't appear to be provided by your environment.

@avieira can you approve this change?

In D41895#3429454, @RKSimon wrote:

I raised this on D111029 - its still not confirmed which avx512 capable targets work better with 512-bit vectors (other than KNL of course).

LLVM currently prefers 256-bit operations even on recent Intel processors.
https://godbolt.org/z/dffKPxo53

@lgxbslgx can you abandon this patch?

rebase with all the diffs

rebase

Thx for challenging my patch @lgxbslgx.

• Guard libc benchmarks behind a flag
• Use ExternalProject as the benchmark target is not always available

@lgxbslgx I'vre created D122481 that should fix the issue.

I believe the benchmark CMake target is gated on some variable being defined.
Let me have a look, I'll reply ASAP.

Also it compiles just fine without the patch.
Can you share the error message you're getting in the first place?

This patch fails locally for me in the same way it fails on the build bot.

• rebase

• Try to use prebuilt tools

• Update bazel files to get closer to internal files

Thx!

• Fix bazel files after cpp_standalone splitting

• rebase

• rebase and test CI now that https://github.com/google/llvm-premerge-checks/commit/86e2347137812b329d0ed603140737ceebaa1e9f is submitted.

Much better, thank you.

Thx for the patch, please address the comment before submitting.

• Use git_repository with tags, re-enable tool registration to check if CI works (m4 is missing), apply .bazelrc flags more specifically

Another hack is to modify the .bazelrc file to pass -Wall -Werror only to cxxopt, instead of copt. Since make, gmp and mpfr are C and not C++ it works. But yeah, quite a hacky solution as well,

@GMNGeoffrey I'm having an issue that you may know how to solve.
.bazelrc automatically appends -Wall -Werror to everything that is built.

rebase and squash

• Add documentation

• Reverting print function, this will go in a different patch.

rebase and squash

• Use tuple ordering to simplify sorting

rebase and squash commits

• Cleanup StringSet angle brackets, they are not needed in C++17
• Rename flag so the intent is clearer

rebase

• Add documentation on why -Wno-error is needed to compile gmp and mpfr.

We'd need to have autotools installed on the CI machine though.

gmp and mpfr are built from source. Right now this is quite slow as Bazel does not parallelize the job yet (https://github.com/bazelbuild/rules_foreign_cc/issues/329).
Once the build is cached this is fast though.

This patch has been broken down and submitted:

• D119142: [libc][NFC] moving template specialization outside class declaration
• D119143: [libc] Don't use Clang flags on other compilers
• D119145: [libc] Replace type punning with bit_cast.
• D119242: [libc] undefined reference in LibcTest.cpp.

In D119145#3305009, @sivachandra wrote:

I am not an expert, but normally the word "illegal" in code/commit logs is not preferred as it can imply non-technical errors. So, may be replace it with "incorrect".

rebase and change commit message

• Add missing change

FYI I also had the following constness warning

/redacted/git/llvm-project/libc/src/fenv/fesetexceptflag.cpp: In function ‘int __llvm_libc::__fesetexceptflag_impl__(const fexcept_t*, int)’:
/redacted/git/llvm-project/libc/src/fenv/fesetexceptflag.cpp:23:24: warning: type qualifiers ignored on cast result type [-Wignored-qualifiers]
   23 |   int excepts_to_set = static_cast<const int>(*flagp) & excepts;
      |                        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

• Non recursive Repeated::move implementation