multibyte parallel mods, with zero byte detection

[dsouza123]

The SSE2, MMX, ALU fininshes off with a pair wrong instructions
movdqa (SSE2) mixed with mm0 (MMX)
should be

Code Select Expand


   movdqa   actarr+ 0, xmm0   ; save it/update the copy in memory
   movdqa   actarr+16, xmm2   ; save it/update the copy in memory
   movq     qword ptr actarr+32,  mm0  ; repaired line
   movq     qword ptr actarr+40,  mm2  ; repaired line
   mov      dword ptr actarr+48,  eax


A repaired version is attached.

[attachment deleted by admin]