[thomasbuilds]

randomx_get_flags() reports RANDOMX_FLAG_HARD_AES only if HAVE_AES is 1, which it never was on RISC-V: rx_aesenc_vec_i128/rx_aesdec_vec_i128 had no RISC-V implementation. So callers that take their flags from it, such as Monero and randomx-benchmark --auto, used software AES even where hardware AES was detected.

This implements both functions with Zvkned in aes_hash_rv64_zvkned.cpp, the file already compiled with Zvkned, and sets HAVE_AES to 1 on RISC-V. They are only called from RANDOMX_FLAG_HARD_AES VMs, which VmBase::allocate() rejects unless cpu.hasAes(). In those VMs, the interpreter's v2 AES mix now uses them instead of software AES. platformError stays defined on RISC-V, where allocate() uses it, and is now a const pointer so GCC's -Wall doesn't flag it as unused. The "Hardware AES not supported" test still runs on RISC-V whenever cpu.hasAes() is false.

[hyc]

If the CPU has hardware AES but HAVE_AES is not being defined, then the right thing to do is to fix that, not muck around here.

[SChernykh]

Yes, HAVE_AES should be 1 on RISC-V because there are AES instruction extensions there.

[hyc]

I think this change is unnecessary. hard AES will always be checked at runtime on platforms that have hardware AES support.

The HAVE_AES define means this particular chip family has hardware AES support, but we always have to check at runtime whether the current chip has it or not.

[thomasbuilds]

This test covers that runtime check on RISC-V: allocate() rejects RANDOMX_FLAG_HARD_AES there unless cpu.hasAes() (#349). With HAVE_AES now 1 on RISC-V, the old condition would skip it, including in the riscv64 CI job, where it runs today. Other platforms don't check cpu.hasAes() in allocate() so a plain !cpu.hasAes() would fail the Win32, ppc64le and two arm64 CI jobs, where cpu.hasAes() is always false but the CPU has AES. #369 will replace __riscv here with HAVE_AES_CHECKER.

[hyc]

Likewise, I think this change is unnecessary.

[thomasbuilds]

Without it RISC-V doesn't build: allocate() throws platformError there (#349) and with HAVE_AES now 1, #ifndef HAVE_AES no longer defines it. SChernykh suggested this guard rather than changing that throw - #369 will replace defined(__riscv) with defined(HAVE_AES_CHECKER).