Fast MUL in ASM?

Comments

• Ale Posts: 2,363

  2007-11-29 11:59 edited 2007-11-29 11:59

  There is a document by Parallax with code for mul. If you unroll it, it is quite fast
  
  

  ' 
  ' Multiply x[noparse][[/noparse]15..0] by y[noparse][[/noparse]15..0] (y[noparse][[/noparse]31..16] must be 0) 
  ' on exit, product in y[noparse][[/noparse]31..0] 
  ' 
  multiply        shl     x,#16           'get multiplicand into x[noparse][[/noparse]31..16] 
                  mov     t,#16           'ready for 16 multiplier bits 
                  shr     y,#1    wc      'get initial multiplier bit into c 
   
  :loop   if_c    add     y,x     wc      'if c set, add multiplicand into product 
                  rcr     y,#1    wc      'get next multiplier bit into c, shift product 
                  djnz    t,#:loop        'loop until done 
   
  multiply_ret    ret                     'return with product in y[noparse][[/noparse]31..0] 
  
  

  
  
  just unrool and enjoy
  
  It is available in "Propeller Guts"
• Direct Digital Labs Posts: 27

  2007-12-14 18:35 edited 2007-12-14 18:35

  Now if we could get a 16x16 signed multiply example......
  
  
  -Dan
  
  Post Edited (Direct Digital Labs) : 12/14/2007 6:40:30 PM GMT
• Mike Green Posts: 23,101

  2007-12-14 18:41 edited 2007-12-14 18:41

  It's not hard to do yourself. Use ABS to fetch the X and Y values and do the multiply. Then XOR the original values and shift the sign bit into the carry flag. NEGC will negate the result if the carry flag is true. There you are.
• deSilva Posts: 2,967

  2007-12-14 19:43 edited 2007-12-14 19:43

  As the MSB will be never set, you can also reduce the loopcount...