nx_simd_sse_rot_gate.nx

nx_simd_sse_rot_gate.nx -- SOVEREIGN 128-bit SSE2 packed-shift/bitwise run-KAT (rung R1 of the sovereign VECTOR backend). Proves the packed ROTATE primitive -- the inner loop of every ARX cipher (ChaCha/BLAKE/Salsa) -- on REAL x86 silicon, NO gcc/qemu/binutils. AUTHOR=ORGAN, no-false-green. rotl32(x,n) = (pslld x,n) OR (psrld x,32-n). With x=0x80000001, n=1: pslld $1 -> 0x00000002 (top bit shifted out of the 32-bit lane) psrld $31 -> 0x00000001 (top bit becomes bit 0) por -> 0x00000003 (both halves load-bearing => por is real) POS exit = 3. NEG swaps por->pand: 0x2 & 0x1 = 0 -> exit 0 (distinct => real silicon read). GREEN iff pos==3 AND neg==0 AND pos!=neg. Run-proves: pslld, psrld, por (POS), pand (NEG), movdqa reg-reg copy, + movdqu load/store again. The padd{b,w,q}/ psub{b,w,d,q} twins share the proven x86_sse_rr path (different opcode byte only). Appends knowledge/status/simd_vec_kat.log. license_tier: ORIGINAL