On 2012-04-13 00:31, Joerg Sonnenberger wrote:
On Fri, Apr 13, 2012 at 12:21:59AM +0200, Johnny Billquist wrote:On 2012-04-12 23:58, Matthew Mondor wrote:On Thu, 12 Apr 2012 23:04:33 +0200 Johnny Billquist<bqt%softjar.se@localhost> wrote:Funny you should say that, while a CVS checkout of all of the NetBSD source tree have grown in time from about 20 minutes to about 5 hours on a VAX-8650 as we move from NetBSD 2 to NetBSD 5. Did we really improve performance, or did we only just get faster computers, while the code itself got slower?Faster hardware, but also software that adapts better to that hardware: Most of the performance improvements I was refering to happened between netbsd-4 and netbsd-5, when Andrew Doran worked full-time on performance and SMP: http://blog.netbsd.org/tnf/entry/netbsd_5_0_an_overview This included changing a number of operations that were previously executing in O(n) time to now execute in O(1) time, making MP-safe some critical sections, not using the big lock anymore for MP-safe sections, scheduler and interrupt processing changes, a new 1:1 POSIX threading system that can more easy scale with SMP, etc.And yet, things are slower now... The only reason it appears faster is that the hardware is so much faster. But I'm sure some improvements have been done (I'm sure of it, it would be horrific otherwise), it's just that we've lost even more performance at other corners. Net result is a loss.Posts like this completely disqualify your opinion on more technical issues. Please upgrade to RetroBSD. It will allow the discussion about this to turn to someelse than code historic reasoning.
I already gave up on actually getting through to you anyway. Keep believing in what you do. Few others do...
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt%softjar.se@localhost || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol