Re: Recent -current Performance Drop?
In message <[EMAIL PROTECTED]>, "Thomas D. Dean" writes: >I turned off the malloc AJ flags, via malloc.conf. It improved 'make >world' by something like 17% == mean_aj/mean_AJ. Make sense, make world is dominated by gcc/cc1 which is doing a lot of malloc/free operations. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 [EMAIL PROTECTED] | TCP/IP since RFC 956 FreeBSD coreteam member | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Recent -current Performance Drop?
I turned off the malloc AJ flags, via malloc.conf. It improved 'make world' by something like 17% == mean_aj/mean_AJ. Make World Statistics -current SMP, 2xP133, 96MB RAM, IBM Superstor 9G disk. Lines is the number of lines of output produced by 'make world'. Date Lines Make Time Lines/Minute - - 19991018 35281 6h40m16s 88.20 <== malloc aj? 2401 39490 7h56m18s 82.96 . 2402 39427 7h28m41s 88.01 . 2412 39404 7h10m21s 91.64 . 2621 38679 7h20m20s 87.91 . 2630 39530 7h56m20s 83.05 <== malloc aj? 2709 39679 8h40m28s 76.31 <== malloc AJ? 2710 39568 9h13m46s 71.55 <== malloc AJ? 2711 41178 9h22m05s 73.27 <== malloc AJ? 2712 39577 8h03m04s 81.94 <== malloc aj tomdean To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Recent -current Performance Drop?
On Tue, 11 Jul 2000, Poul-Henning Kamp wrote: > >I seem to see somewhat of a performance drop in the past week. > > You should have read the commit messages :-) > > I enabled malloc flags AJ by default, this has a performance > cost. It will be turned off for releases of course. > > It has already exposed on bug (see peters commit). ^- Multiple bugs, thankyouverymuch :) > You can disable it if you want to run benchmarks: If you run a desktop system (need good response) and aren't willing to take the large performance hit, too. Note it's a large performance hit when you tend to run a LOT of stuff and always dig into swap very quickly. I imagine for most people, the performance drop isn't nearly as high, so they can live with it :) > ln -sf aj /etc/malloc.conf > > -- > Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 > [EMAIL PROTECTED] | TCP/IP since RFC 956 > FreeBSD coreteam member | BSD since 4.3-tahoe > Never attribute to malice what can adequately be explained by incompetence. -- Brian Fundakowski Feldman \ FreeBSD: The Power to Serve! / [EMAIL PROTECTED]`--' To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Recent -current Performance Drop?
Actually in the past week or two I have seen a make buildworld drop from 205-210 minutes to 180-190 minutes. But I would be more inclined to attribute that to changes in the source tree itself, not performance. On Tue, 11 Jul 2000, Thomas D. Dean wrote: > I don't know if this belongs in -current or -smp, so here goes. > > I am running -current SMP, cvsup, etc. late last night PDT. > > # uname -a > FreeBSD celebris 5.0-CURRENT FreeBSD 5.0-CURRENT #0: \ > Tue Jul 11 09:53:15 PDT 2000 \ > tomdean@celebris:/usr/src/sys/compile/CELEBRIS-SMP i386 > > I have been using 'make world' as a stability/performance indicator. > > The stability has been good. I have had no recent panics, etc., > during 'make world'. I choose the times to update carefully. > > Good work core team. > > I seem to see somewhat of a performance drop in the past week. > > I save the output of 'make world' I know this is not a benchmark nor a > very accurate indicator. However, I believe 'make world' is a pretty > good torque tool for a lightly loaded, interactive, 'play' system. > And, disk I/O, compiling and executing user code is what interactive > users do. > > If 'make world' puts out approximately the same number of lines, I > think it is doing about the same amount of work. > > I see an increase in 'make world' time over the past week, or so. The > number of lines of output has gone up slightly. The time elapsed has > gone up more. > > # world_time.sh > Make World Statistics > -current SMP, 2xP133, 96MB RAM, IBM Superstor 9G disk. > Lines is the number of lines of output produced by 'make world'. > Date Lines Make Time Lines/Minute > - - > 19991018 35281 6h40m16s 88.20 > 2401 39490 7h56m18s 82.96 > 2402 39427 7h28m41s 88.01 > 2412 39404 7h10m21s 91.64 > 2621 38679 7h20m20s 87.91 > 2630 39530 7h56m20s 83.05 > 2709 39679 8h40m28s 76.31 > 2710 39568 9h13m46s 71.55 > 2711 41178 9h22m05s 73.27 > > tomdean > > > To Unsubscribe: send mail to [EMAIL PROTECTED] > with "unsubscribe freebsd-current" in the body of the message > -- _ __ ___ ___ ___ ___ Wesley N Morgan _ __ ___ | _ ) __| \ [EMAIL PROTECTED] _ __ | _ \._ \ |) | FreeBSD: The Power To Serve _ |___/___/___/ Hi! I'm a .signature virus! Copy me into your ~/.signature to help me spread! To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Recent -current Performance Drop?
>I seem to see somewhat of a performance drop in the past week. You should have read the commit messages :-) I enabled malloc flags AJ by default, this has a performance cost. It will be turned off for releases of course. It has already exposed on bug (see peters commit). You can disable it if you want to run benchmarks: ln -sf aj /etc/malloc.conf -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 [EMAIL PROTECTED] | TCP/IP since RFC 956 FreeBSD coreteam member | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Recent -current Performance Drop?
phk turned on some extra malloc options by default that a) catch more bugs b) make it slower. I hope you are reading cvs-all when you are using -current. On Tue, 11 Jul 2000, Thomas D. Dean wrote: >I don't know if this belongs in -current or -smp, so here goes. > >I am running -current SMP, cvsup, etc. late last night PDT. > ># uname -a >FreeBSD celebris 5.0-CURRENT FreeBSD 5.0-CURRENT #0: \ > Tue Jul 11 09:53:15 PDT 2000 \ > tomdean@celebris:/usr/src/sys/compile/CELEBRIS-SMP i386 > >I have been using 'make world' as a stability/performance indicator. > >The stability has been good. I have had no recent panics, etc., >during 'make world'. I choose the times to update carefully. > >Good work core team. > >I seem to see somewhat of a performance drop in the past week. > >I save the output of 'make world' I know this is not a benchmark nor a >very accurate indicator. However, I believe 'make world' is a pretty >good torque tool for a lightly loaded, interactive, 'play' system. >And, disk I/O, compiling and executing user code is what interactive >users do. > >If 'make world' puts out approximately the same number of lines, I >think it is doing about the same amount of work. > >I see an increase in 'make world' time over the past week, or so. The >number of lines of output has gone up slightly. The time elapsed has >gone up more. > ># world_time.sh >Make World Statistics >-current SMP, 2xP133, 96MB RAM, IBM Superstor 9G disk. >Lines is the number of lines of output produced by 'make world'. >Date Lines Make Time Lines/Minute > - - >19991018 35281 6h40m16s 88.20 >2401 39490 7h56m18s 82.96 >2402 39427 7h28m41s 88.01 >2412 39404 7h10m21s 91.64 >2621 38679 7h20m20s 87.91 >2630 39530 7h56m20s 83.05 >2709 39679 8h40m28s 76.31 >2710 39568 9h13m46s 71.55 >2711 41178 9h22m05s 73.27 > >tomdean > > >To Unsubscribe: send mail to [EMAIL PROTECTED] >with "unsubscribe freebsd-smp" in the body of the message > To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message