ulfis@nada.kth.se (Anders Ulfheden) (03/06/91)
When playing around with Speedometer for measuring system performance I have noticed that turning the cache memory ON, reduces the speed in some cases with a factor 2! Can anyone enlighten me in how this can happen and how the setting of cache memory affect the system performance. This behaviour is also noticed together with screensavers like After Dark. Thanks! Ulfis ulfis@nada.kth.se -- +------------------------------------------------------------------------------ | Anders Ulfheden | USENET: ulfis@nada.kth.se | Royal Institute of Technology
Lawson.English@p88.f15.n300.z1.fidonet.org (Lawson English) (03/12/91)
Anders Ulfheden writes in a message to All AU> When playing around with Speedometer for measuring system performance AU> I have noticed that turning the cache memory ON, reduces the AU> speed in some cases with a factor 2! Just a guess, assuming that ON and OFF aren't mixed up: perhaps there is an overhead to "load the cache" that is noticable when a long *sequence* of instructions is executing instead of a (short) loop. Which CPU are you running? Lawson -- Uucp: ...{gatech,ames,rutgers}!ncar!asuvax!stjhmc!300!15.88!Lawson.English Internet: Lawson.English@p88.f15.n300.z1.fidonet.org
ulfis@cyklop.nada.kth.se (Anders Ulfheden) (03/14/91)
In article <4745.27DDB399@stjhmc.fidonet.org> Lawson.English@p88.f15.n300.z1.fidonet.org (Lawson English) writes: > Anders Ulfheden writes in a message to All > > AU> When playing around with Speedometer for measuring system performance > AU> I have noticed that turning the cache memory ON, reduces the > AU> speed in some cases with a factor 2! > > Just a guess, assuming that ON and OFF aren't mixed up: perhaps there is an > overhead to "load the cache" that is noticable when a long *sequence* of instructions > is executing instead of a (short) loop. Which CPU are you running? This test has been done on SI and LC, both with and without coprocessor. The same behavior. So, answer to your question: 68020 and 030. Ulfis -- +------------------------------------------------------------------------------ | Anders Ulfheden | USENET: ulfis@nada.kth.se | Royal Institute of Technology | Stockholm, Sweden