From: notsaying@nospicedham.invalid.org   
      
   "Kerr Mudd-John"  wrote in   
   news:XnsA7B1748CF8C1Cadmin127001@213.239.209.88:   
      
   > "wolfgang kern"  wrote in news:ok7d1h$1aud$1   
   > @gioia.aioe.org:   
   >   
   >>  Kerr Mudd-John wrote:   
   >> ...   
   >>> ll64 bytes 246 (sadly need to set cl=0 cost 3)   
   >>   
   >> push bp, pop cx ?   
   >   
   > Yup; push in plain, pop in 2:1; I can't do it earlier as I need   
   > ch=0x30 for skipping with the cmp [di+bp], ch.   
   >>   
   >>> ZRPRhp0XPR5p1PRRjL4hPa5LU1GJ1GLRX+G=1GeRSVG8+r6238wbt083t+RU=rY=   
   >>> 66Z2202Zffffff0222F6====MJ=20=17===eeee+++slOBI9UnEdoAasbaANE9E7   
   >>> jelperHQE9sAjesuH8lpssMIG7asA9babaaGv7h8A7IdH9RKi8A9EGI8sCoAlpin   
   >>> MIT9lpadITt7YoH9f7=tAm6CAHNIcNIZWxsbyB3b3JsZCEk=   
   >>   
   >>> 1622:0100 AC             lodsb   
   >>> 1622:0101 3C 2B          cmp          al,2B   
   >>> 1622:0103 72 25          jb           012A   
   >>>   
   >>> (getnum)   
   >>> 1622:0105 2C 41          sub          al,41   
   >>> 1622:0107 73 0A          jnb          0113   
   >>> 1622:0109 04 45          add          al,45   
   >>> 1622:010B 3C 33          cmp          al,33   
   >>> 1622:010D 77 0A          ja           0119   
   >>> 1622:010F B0 86          mov          al,86   
   >>> 1622:0111 1C 00          sbb          al,00   
   >>> 1622:0113 3C 1A          cmp          al,1A   
   >>> 1622:0115 72 02          jb           0119   
   >>> 1622:0117 2C 06          sub          al,06   
   >>>   
   >>>    ;put3:                 ; need cl=*0* on 1st run   
   >>> 41          inc cx   
   >>> 41          inc cx        ; need cl=8 for last; 'and' later   
   >>> 98          cbw           ; clr ah reqd   
   >>> D3E0        shl ax,cl     ; 2,4,6,8   
   >>> 0865FF      or [di-1],ah  ; 'or' on top bits to prev; OK as 1st is 0   
   >>> 80E106      and cl,0x06   ; 8->0. 2,4,6 ok   
   >   ; could cmp al,0 or cmp cl,8 here & put the "and" at top but this   
   >   costs 4 in 2:1 v 3 to clear cx.   
      
   cl=0 might even be free in SMC4:3   
      
   >>> 7401        jz nosave     ; jz not jcxz; saves fixup   
   >>> AA          stosb         ; store these bits   
   >>> nosave:   
   >>> (gnc)   
   >>   
   >> good, I see   
   >>   
   >> 0fb  AC      lodsb   
   >> 0fc  3C 3D   cmp al,3D   
   >> 0FE  75 D5   jnz  0d5   
   >>   
   >> you saved three (six 2:1) bytes in the decoder part.   
   >>   
   >> 9 instead of 11 modify needs: d3,98,08e0,80,ff,06e1,aa01,ac,d5   
   >> this are only 16 total instead of the 18 previous, and it make me   
   >> think of adopting your much shorter decoder and direct SMC it. :)   
   >>   
   You can reduce the fixup count further for 1 byte extra:   
      
   >>>    ;put3:                 ; need cl=*0* on 1st run   
   >>> 41          inc cx   
   >>> 41          inc cx        ; need cl=8 for last; 'and' later   
       4F          dec di   
   >>> 98          cbw           ; clr ah reqd   
   >>> D3E0        shl ax,cl     ; 2,4,6,8   
       08 25       or  [di],ah   
       47          inc di   
   >>> 80E106      and cl,0x06   ; 8->0. 2,4,6 ok   
      
   and 6 * dec ax instead of sub al,6 (cost 6-2,& saves 3 for fixup)   
      
   >> Using the command-line instead of 042 would save me on push_bx/ret   
   >> and one 3 byte modifier, but needs some more bytes to modify   
   >> 0c0..0ff.   
   > Yup.   
   >> Not all fit my current DX=0x7862 paired modify, but I'll check.   
   >> __   
   >> wolfgang   
   >>   
   >   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)