mirror of
https://github.com/HeyPuter/gex.git
synced 2026-11-13 20:12:43 -10:00
velzieandClaude Opus 5.5
a007e903e3
--simd draws candidates from the Advanced SIMD data-processing groups (vector and scalar) instead of the whole space, and the harness now seeds and compares v0-v31 as well as the general registers. Only the integer and bitwise members are drawn by default: gex's floating-point vector lowerings knowingly differ from hardware on NaN payloads, fused multiply-add rounding and the reciprocal estimates, so --simd-fp includes them only on request. An encoding that decodes but lifts to a SIGILL is counted and listed as unimplemented rather than reported as a difference. --inst WORD (repeatable) tests only the given encodings against fresh seeds, the regression check for one fix. --prefix names the cross toolchain's tool prefix (aarch64-linux-gnu- works as well as the musl one). smulh/umulh with Ra != 31 are no longer drawn: the field is should-be-one, and QEMU and gex make different, equally valid choices. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01FFqjijkLKTTuaQutt3168C