From mboxrd@z Thu Jan 1 00:00:00 1970 From: Richard Palethorpe Date: Fri, 29 Sep 2017 11:44:43 +0200 Subject: [LTP] [PATCH v4 3/7] fzsync: Add long running thread support and deviation stats In-Reply-To: <20170925160441.GC31931@rei> References: <20170925095137.31491-1-rpalethorpe@suse.com> <20170925095137.31491-3-rpalethorpe@suse.com> <20170925160441.GC31931@rei> Message-ID: <87efqpx2ic.fsf@our.domain.is.not.set> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ltp@lists.linux.it Hello, Cyril Hrubis writes: > Hi! >> This does not replace the delay method of synchronisation which can be used in >> addition to the new method. Also the precision of the averages has been >> reduced because double precision appears to be unnecessary. > > Just a quick question, have you tested this on arm as well? The x86 FPU > computes with greater than float precission internally and I've seen a > cases where code would run perfectly fine on x86 with floats but failed > horrinbly on arm. Ah, that is interesting. However I tested it again on ARM and the figures appear to be the same. I think the average is fairly safe from accumulative rounding errors. -- Thank you, Richard.