Niels Möller nisse at lysator.liu.se
Tue Sep 3 08:51:42 UTC 2019

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

> But... I get even better numbers if I keep the old code and just replace
> the div1 function with plain division q = a / b:

Attaching deletion patch. Tried it also on my broadwell machine.
Benchmarking is less reliable there, but seems to give about 10% gcd
speedup in the range 3-10 limbs.

> It should be fairly easy to find out, if we define a HGCD_DIV1_METHOD
> known to tuneup, to select between plain division and the div1 function.

Is there any even easier way to find out on which machines (if any) div1
improves performance? It's still a bit awkward to add tuning of
HGCD_DIV1_METHOD and wait for the nightly builds.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: delete-div1.patch
Type: text/x-diff
Size: 1939 bytes
Desc: not available
URL: <https://gmplib.org/list-archives/gmp-devel/attachments/20190903/48b2ddf4/attachment.bin>
-------------- next part --------------

Niels Möller. PGP-encrypted email is preferred. Keyid 368C6677.
Internet email is subject to wholesale government surveillance.

More information about the gmp-devel mailing list