divappr tests

[bodrato at mail.dm.unipi.it]

Niels,

> I was about to check in a hacked version of mpn_sbpi1_divappr_q that
> uses the new udiv_qr_3by2 macro, but I hesitate because this code seems
> to not be exercised by the testsuite. I attach the patch.

I finally did commit the mpn_invert code, that uses divappr.
It passes the test "try mpn_invert", so the current divappr code seems
correct (but may also introduce small errors, compensated by the Newton
iteration).

I tried your patch, but "make tune" aborted with:
#define DC_DIVAPPR_Q_THRESHOLD        sbpi1_divappr_q.c:181: GNU MP
assertion failed: np[1] == n1
*** Signal 6

Regards,
Marco

-- 
http://bodrato.it/papers/