1. Disable hugepages (not supported).
2. Force W^X (required).
3. When allocating JIT memory, PROT_EXEC must be reserved
in order to set the pages executable later.
Adds more parallelizm into AES loop so modern CPUs can take advantage of it. Also, scratchpad data moves between L1 and L3 caches only one time which saves time and energy per hash.
Since OpenBSD 6.0 W^X is enforced.
Added `RANDOMX_FLAG_SECURE` in tests and benchmarks.
Updated comment.
Excluded `cpu_set_t` since it is not defined on OpenBSD.
* code buffer size is calculated based on RandomX parameters
* added a maximum value constraint for program size and superscalar latency
* reduced the x86 code size of memory instructions by 1 byte
* disclaimer note in configuration documentation
* Blake2Generator::getInt32 renamed to getUInt32 to avoid confusion
* isPowerOf2 renamed to isZeroOrPowerOf2 to avoid confusion
* added asserts to validate the input/output size of AES functions
* fixed possible overflow in JitCompilerX86::getCodeSize (unused function)
* this fixes identical sequences of columns 0/2 and 1/3 if their states are the same
* added TestU01 results for AesGenerator1R and AesGenerator4R
* added a note about the reversibility of AesHash1R
* instruction decode/execute separated into class BytecodeMachine
* added randomx-tests project
* removed the use of non-portable __COUNTER__ macro
* removed the use of unsupported FENV_ACCESS pragma