Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Tue, 8 Nov 2011 23:04:02 +0400
From: Solar Designer <>
Subject: floating-point math after MMX

magnum -

I expect that
0022-j7-Fix-for-huge-figures-in-MPI-sum-avg-cps-64-bit-ov.patch and
0027-j7-Optionally-output-number-of-candidates-tried.patch would cause
misreporting when running MMX-using builds (and formats for which MMX
specifically is actually used) on certain CPUs.  I previously observed
this on Pentium 3, for which there's now a workaround in -jumbo's
status.c: status_get_ETA() and status_get_totalETA(), but not in other
functions.  Basically, we only issue the EMMS instruction right before
doing floating-point math.

Is this something you can test for (on a P3) and patch now or should I?



Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.