[SWDEV-544298] AINIC implementation (#2798)

[oosman330]

Motivation

Show AINIC information from amd-smi command line

Technical Details

JIRA ID

SWDEV-544298

Test Plan

Test Result

Pass on device with and without AINIC

Commands:

amd-smi list --nic 0

amd-smi static --nic 0

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

[marifamd]

This re-reverted #2798 that was causing failures on the nightly builds: ROCm/TheRock/actions/runs/22168787472

Originally reported in #3500
Now this unblocks rocProf team so they can start AINIC integration

[stellaraccident]

FYI - this causes a build failure with gcc14. Attempting to fix forward as this is a local, non-CI compiler suspected false positive:

● The build error is in projects/amdsmi/src/nic/ai-nic/amdsmi_unified/src/smi_sysfs.cpp in SmiSysfsReader::readAll(). GCC 14 with -Werror flags two lines:

• Line 54: content.push_back(hex_value);
• Line 56: content.push_back(std::stoi(token));

Both push an int into a std::vector<std::variant<int, std::string>>. GCC's inliner traces through the variant move constructor and concludes the string length member _M_string_length may be used uninitialized — a known GCC false positive with
std::variant containing std::string (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106621 and related).

The error is:
error: '...std::__cxx11::basic_string::_M_string_length' may be used
uninitialized [-Werror=maybe-uninitialized]

Introduced by 410afdc — the AINIC re-land.

[stellaraccident]

FYI - this causes a build failure with gcc14. Attempting to fix forward as this is a local, non-CI compiler suspected false positive:

● The build error is in projects/amdsmi/src/nic/ai-nic/amdsmi_unified/src/smi_sysfs.cpp in SmiSysfsReader::readAll(). GCC 14 with -Werror flags two lines:

• Line 54: content.push_back(hex_value);
• Line 56: content.push_back(std::stoi(token));

Both push an int into a std::vector<std::variant<int, std::string>>. GCC's inliner traces through the variant move constructor and concludes the string length member _M_string_length may be used uninitialized — a known GCC false positive with std::variant containing std::string (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106621 and related).

The error is: error: '...std::__cxx11::basic_string::_M_string_length' may be used uninitialized [-Werror=maybe-uninitialized]

Introduced by 410afdc — the AINIC re-land.

Fixed here: #3633