In D148797#4299288, @sivachandra wrote:

This change LGTM in general. But, I would like to have instructions on how to test this before I accept it. Couple of reasons:

1. We want to have an estimate for the public CI.

• Added comments about sys_*64 syscalls
• Added assertion about offset size in sys_*64 syscalls
• Fixed wait4 implementation
• Rewrote wait and waitpid to use wait4

hey @lntue, can you replace __int128_t with

-nostdlib++ is a clang-only flag, gcc fails with:

This is required because libc's UInt<T> can't be used with libc's optional, when trying to get libc built for riscv32, we get the following errors:

/home/mgadelha/tools/llvm-project/libc/src/__support/UInt.h:347:12: note: non-literal type 'optional<UInt<128U>>' cannot be used in a constant expression
    return remainder;
           ^
/home/mgadelha/tools/llvm-project/libc/src/__support/UInt.h:352:12: note: in call to '&result->div(COUNT)'
    result.div(other);
           ^
/home/mgadelha/tools/llvm-project/libc/test/src/math/CopySignTest.h:37:44: note: in call to '&UIntType(-1)->operator/(COUNT)'
    constexpr UIntType STEP = UIntType(-1) / COUNT;
                                           ^
1 error generated.

Addressed comments

• Implement get_val() for x86_64 long double class
• Use dyadic float in clinger_fast_path to fix conversion to long double

Fixed compilation with clang < 16, where there is no add_lvalue_reference and add_rvalue_reference

Removed wrong conversion from uint<T> to long double

Changed some defines to LIBC_TARGET_ARCH_IS_ANY_RISCV

In D150211#4330579, @michaelrj wrote:

Before I review this patch, I want to confirm that this is the right strategy. If the main goal is to support Uint<> in optional is there a way to do that without a complete replacement of optional?

addressed comments

ping

• Fixed wrong define
• Rebased

• make struct linux_dirent a private header
• rebased

• Use static_cast<foo> instead of foo(...)
• Changed operator++ and operator-- to use operator+ and operator-, respectively

@sivachandra does my last chance address your comment?

Apply clang-format to syscall_numbers.h.inc

Moved linux specific implementation to linux_dir.cpp

Fixed strtol tests to pass in riscv32

@sivachandra can you clarify what you mean by "This file is supposed to contain the platform independent implementation. Can we improve the platform abstraction to push variance within Linux into the Linux implementation?"?

• Added details about the current status of the riscv32 port in the commit message
• Added LIBC_TARGET_ARCH_IS_ANY_RISCV and fix a syscall infinite recursion

• added siginfo_t dependency to wait.h
• added FUTEX_SYSCALL_ID
• changed SYS_sched_rr_get_interval_time64 to use struct __kernel_timespec instead of a long arr[2]
• improve wait* documentation

Are you able to validate your guess with testing?

In D148797#4292915, @michaelrj wrote:

Overall LGTM, though are we sure that having the riscv 64 bit and 32 bit targets is the correct path? On other platforms so far we've separated the 32 bit and 64 bit targets.

ping

ping

Added SYS_sched_rr_get_interval_time64 support to fix sched compilation in riscv32

• Changed ifdef unix to linux
• Changed strlen, memset and memmove to use libc's internal implementation

rebased

Renamed dirs riscv64 to riscv

Addressed comments

Added static assertions

@sivachandra while we don't need this anymore, I think it's a good addition to the codebase.

Add linux_dirent to dirent.h only if unix is defined

• Added new struct linux_dirent to llvm-libc-types
• Removed both ino64_t and off64_t

In D148739#4282973, @sivachandra wrote:

While this change is OK, we do want the UInt class to be trivially copyable. Do you know why it is not trivially copyable even when defaulted?

Thanks, it fixed the problem with sub_with_borrow and add_with_carry in riscv32 for me!

Remove C++ standard library in the runtime components of the libc

In D147961#4275174, @sivachandra wrote:

This is the failure the bot is still seeing: https://lab.llvm.org/buildbot/#/builders/256/builds/1030

In D147961#4274137, @sivachandra wrote:

The last time I checked, log10 had failures. Was it fixed with the FMA change?

Changed ino_t and ino64_t type to UINT64_TYPE

rebased

Added SYS_futex_time64 to syscall_numbers.h.inc

• Added comment about string lenght
• changed memcopy to memmove to prevent UB

Update waitid syscall

Fix error message

ping

Fix variable name

Added a variable to count how many times longjmp was called

• Fixed error message
• Fixed function not being called in release mode

Update return

In D146253#4201018, @lntue wrote:

@mikhail.ramalho : why do you need to force the FMA_OPT and ROUND_OPT for RISC-V target? These are to expand to build and test both with and without those instructions automatically. Do you plan the same thing for RISC-V? Without this, the current setup should test the trig functions with FMA and rounding instructions by default, and not testing trig functions without those instructions.

With this change, on AArch64, AArch32, RISC-V, and GPU, each math function will have 2-4 targets, with and without FMA_OPT and ROUND_OPT flags, and since the build system does not do anything, add any special compile options or flags, those targets are all identical. So I don't think that's your intention?

Do you want to test both non-FMA and FMA math implementations on RISC-V64 automatically similar to x86-64 instead?

Address comments

@sivachandra I only tested this in x86_64 and riscv 64, since I don't have aarch64 or armv7 machines available for testing.

We submitted D146245, which should enable llvm to generate better code for setjmp/longjmp.

1. Added -fomit-frame-pointer
2. Removed attribute((naked)) from setjmp/longjmp
3. Because of (2) we don't need the __NOEXCEPT anymore
4. Rewrote the riscv functions to use the function arguments directly, instead of fixed offset
5. Split the platform specific implementations into different directories/files

Thanks for the review, Siva, I think I've managed to address all your comments thanks to your code snippet (tests are running now, I'll update here if they pass).

Fixed header include

Small change to the cmake file: we only need to set -fno-omit-frame-pointer it's not riscv