Fix bignum.cc build failure on GCC 8 (x86_64)

[akhanzode]

Description

The _addcarry_u64 and _subborrow_u64 intrinsics in bignum.cc require the ADX instruction set extension, which GCC < 9 does not enable by default on x86_64. The existing #ifdef __x86_64__ guard is too broad, causing build failures on GCC 8.5.0 (the default compiler on Rocky Linux 8 / RHEL 8).

bignum.cc:299:7: error: '_addcarry_u64' was not declared in this scope
bignum.cc:316:13: error: '_subborrow_u64' was not declared in this scope

Fix

Replace #ifdef __x86_64__ with a stricter check:

#if defined(__x86_64__) && (defined(__clang__) || (defined(__GNUC__) && __GNUC__ >= 9) || defined(__ADX__))

This allows the intrinsics on Clang (all versions), GCC >= 9, or any compiler with -madx. On older GCC, falls through to the existing portable absl::uint128 arithmetic path with no functional change.

Fixes #568

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[jmr]

#568 (comment)

[jmr]

You can replace __x86_64__ with __ADX__ if that's really all we need. If it happens to fix RHEL 8, great, but no guarantees with that.

#568 (comment)

[jmr]

I think __ADX__ is the wrong macro here. It indicates support for the ADCX/ADOX instructions (Intel ADX extension), but _addcarry_u64 and _subborrow_u64 map to plain ADC and SBB, which are base x86 instructions — no ADX required.