Improved binvert algorithm

[Torbjörn Granlund]

Niels Möller <nisse at lysator.liu.se> writes:

  E.g., the decade old Intel Atom processors seem to lack mul pipelining
  (if I read https://gmplib.org/~tege/x86-timing.pdf correctly, 18
  cycles). Looks like there's at least one atom machine, "glm", in the gmp
  farm that could be used to test performance. Any other examples?

According to https://gmplib.org/devel/asm, glm has good multiply
performance.  The older slm is worse.  But the original atom is the
worst, but "gege" is not connected at this time.

Many ARM CPUs have poor multiply performance and there are plenty of
those in the GMP test setup.

It is a shame we no longer have any SPARC hardware...the first few
generations of 64-bit CPUs need some 18 cycles in mpn_mul_1, more modern
CPUs like the T1 need 68 cycles...


-- 
Torbjörn
Please encrypt, key id 0xC8601622