Basecase assembly optimisation project

[Torbjorn Granlund]

nisse at lysator.liu.se (Niels Möller) writes:

  Another feature, which I looked into while ago without getting very far
  with the loopmixer, is to make it understand associativity. I.e, try
  reordering certain instructions with the same destination register, like
  
    xor	%r8, %rax
    xor	%r9, %rax
  
  or
  
    add	%r8, %rax
    add	%r9, %rax
    
  This is possible if nothing else depends on the intermediate results
  (including no dependencies on status flags).
  
  Probably more useful for cryptographic loops than for arithmetic.
  
I think this couldd be very useful also for the critical multiply-
accumulate loops of GMP.  The schema loops I create for the loopmixer
try many accumulation strategies.  This is a lot of work, but I only in
practice try regular variants, while an automatic random-driven
re-association feature could try beneficial irregular variants too.

-- 
Torbjörn