From: krishna.myneni@ccreweb.org   
      
   On 3/23/24 21:18, dxf wrote:   
   > On 24/03/2024 12:47 pm, Krishna Myneni wrote:   
   >> On 3/23/24 14:16, Krishna Myneni wrote:   
   >>> On 3/23/24 14:13, Krishna Myneni wrote:   
   >>> ...   
   >>>> I had need for DLSHIFT and DRSHIFT recently, and this is what I came up   
   with.   
   >>>>   
   >>>> \ u is the number of bits to shift   
   >>>> 1 cells 8 * constant BITS_PER_CELL   
   >>>> 0 value ubits   
   >>>>   
   >>>> : DLSHIFT ( ud u -- ud2 ) BITS_PER_CELL min 0 ?DO D2* LOOP ;   
   >>>>       dup 0= IF drop EXIT THEN   
   >>>>       BITS_PER_CELL min to ubits   
   >>>>       ubits lshift swap   
   >>>>       dup >r ubits msbits or   
   >>>>       r> ubits lshift swap ;   
   >>>>   
   >>>> : DRSHIFT ( ud u -- ud2 )   
   >>>>       dup 0= IF drop EXIT THEN   
   >>>>       BITS_PER_CELL min to ubits   
   >>>>       swap ubits rshift   
   >>>>       swap dup >r ubits lsbits or   
   >>>>       r> ubits rshift ;   
   >>>>   
   >>>> With x86, it should be possible to write efficient versions of DLSHIFT   
   and DRSHIFT using SHLD and SHRD instructions.   
   >>>>   
   >>>   
   >>> I left out the definitions of MSBITS and LSBITS (shown below):   
   >>>   
   >>>    Return the u least significant bits of cell value u1   
   >>> \ as the most significant bits of u2   
   >>> : lsbits ( u1 u -- u2 )   
   >>>       BITS_PER_CELL min   
   >>>       BITS_PER_CELL - negate   
   >>>       lshift ;   
   >>>   
   >>> \ Return the u most significant bits of cell value u1   
   >>> \ as the least significant bits of u2   
   >>> : msbits ( u1 u -- u2 )   
   >>>       BITS_PER_CELL min   
   >>>       BITS_PER_CELL - negate   
   >>>       rshift ;   
   >>>   
   >>   
   >> Just realized that these versions of DRSHIFT and DLSHIFT are limited to   
   shift of 0 bits -- 1 cell width in bits, rather than the general shift count   
   of 0 bits -- 2 cells width in bits. They have to be modified for general use   
   on double length numbers.    
   Of course one can write the general shifts in terms of these by applying them   
   twice if needed.   
   >   
   > Indeed. I was expecting:   
   >   
   > : DLSHIFT ( ud u -- ud2 ) 0 ?DO D2* LOOP ;   
   >   
      
   I had considered coding D2* with a loop using D2* but thought it would   
   be faster to do it with LSHIFT and RSHIFT. In the end I will implement   
   it as an intrinsic word in assembly, using the SHLD instruction.   
      
   > DRSHIFT would require DU2/   
   >   
      
   Yes, but SHRD is more efficient, if it is available.   
      
   --   
   KM   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)