Base compiler and amdfam10 - anybody/anything?
Hi,
Are there any attempts to bring to -CURRENT newer AMD chips support?
Personally, I've just tried to apply the patches from openSUSE's gcc-4.2.1
SRPM. With slight adaptation they've applied and gave rather significant
boost in resulting code speed. At least, testfcpy by Alexander Konovalenko
(http://daemon.safety.sci.kth.se/~kono/testfcpu) gave me ~20% (!) speedup
with -march=amdfam10 compared to our -march=athlon64-sse3 on Phenom II
970.
Unfortunately, the patched compiler with -march=amdfam10 fails in
buildworld ("internal compiler error"'s while compiling clang). The
buildworld was successful with patched compiler and -march=athlon64-sse3
but since this is my main working system... Well, I had to come back to
our unpatched compiler :-(
If anyone is interested, the patches were taken from
gcc42-4.2.1_20070724-17.src.rpm (actually, I applied all the patches
marked as AMD stuff), the resulting patches towards our src/contrib/gcc
and share/mk/bsd.cpu.mk are attached (or I can send them by email), and I
am quite ready to test what comes out of it.
WBR,
Vladimir
gcc-amdfam10.diff.gz
Description: Binary data
bsd.cpu.mk.amdfam10.gz
Description: Binary data
___
freebsd-hackers@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
To unsubscribe, send any mail to "freebsd-hackers-unsubscr...@freebsd.org"
Re: The optimization of malloc(3): FreeBSD vs GNU libc
Sorry for intrusion. On Mon, 14 Aug 2006, Brooks Davis wrote: On Tue, Aug 15, 2006 at 07:10:47AM +0800, Intron wrote: One day, a friend told me that his program was 3 times slower under FreeBSD 6.1 than under GNU/Linux (from Redhat 7.2 to Fedora Core 5). I was astonished by the real repeatable performance difference on AMD Athlon XP 2500+ (1.8GHz, 512KB L2 Cache). After hacking, I found that the problem is nested in malloc(3) of FreeBSD libc. Download the testing program: http://ftp.intron.ac/tmp/fdtd.tar.bz2 You may try to compile the program WITHOUT the macro "MY_MALLOC" defined (in Makefile) to use malloc(3) provided by FreeBSD 6.1. Then, time the running of the binary (on Athlon XP 2500+): #/usr/bin/time ./fdtd.FreeBSD 500 500 1000 ... 165.24 real 164.19 user 0.02 sys Please try to recompile the program (Remember to "make clean") WITH the macro "MY_MALLOC" defined (in Makefile) to use my own simple implementation of malloc(3) (i.e. my_malloc() in cal.c). And time the running again: #/usr/bin/time ./fdtd.FreeBSD 500 500 1000 ... 50.41 real49.95 user 0.04 sys You may repeat this testing again and again. I guess this kind of performance difference comes from: 1. His program uses malloc(3) to obtain so many small memory blocks. 2. In this case, FreeBSD malloc(3) obtains small memory blocks from kernel and pass them to application. But malloc(3) of GNU libc obtains large memory blocks from kernel and splits & reallocates them in small blocks to application. You may verify my judgement with truss(1). 3. The way of FreeBSD malloc(3) makes VM page mapping too chaotic, which reduces the efficiency of CPU L2 Cache. In contrast, my my_malloc() simulates the behavior of GNU libc malloc(3) partially and avoids the over-chaos. Callgrind is broken under FreeBSD, or I will verify my guess with it. I have also verified the program on Intel Pentium 4 511 (2.8GHz, 1MB L2 cache, running FreeBSD 6.1 i386 though this CPU supports EM64T) /usr/bin/time ./fdtd.FreeBSD 500 500 1000 ... 185.30 real 184.28 user 0.02 sys /usr/bin/time ./fdtd.FreeBSD 500 500 1000 ... 36.31 real35.94 user 0.03 sys NOTE: you probably cannot see the performance difference on CPU with small L2 cache such as Intel Celeron 1.7GHz with 128 KB L2 Cache. In CURRENT we've replaced phkmalloc with jemalloc. It would be useful to see how this benchmark performs with that. I believe it does similar things. -- Brooke On -CURENT amd64 (Athlon64 3000+, 512k L2 cache): With jemalloc (without MY_MALLOS): ~/fdtd> /usr/bin/time ./fdtd.FreeBSD 500 500 1000 ... 116.34 real 113.69 user 0.00 sys With MY_MALLOC: ~/fdtd> /usr/bin/time ./fdtd.FreeBSD 500 500 1000 ... 45.30 real44.29 user 0.00 sys Regards, Vladimir ___ freebsd-hackers@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: ATI drivers for FreeBSD - anything?
On Mon, 21 Nov 2005, Maxime Henrion wrote: Albert Vest wrote: On Sat, 19 Nov 2005 01:58:25 +0200 (EET) Vladimir Kushnir <[EMAIL PROTECTED]> wrote: Hi all, Is there any project on FreeBSD wrapper for ATI Linux drivers (like nVidia's used to be)? If so - I'd be more than happy to test (sorry I can hardly write it myself). Regards, Vladimir I too would welcome this, mainly for ATI sound but also for video. I see the x11/nvidia-driver port can still be built with LINUX compatibility turned on; maybe if we "make extract" with LINUX=yes, the source code will contain some hints to how it can be done? The Linux compatibility in the nvidia-driver has nothing to do with a wrapper to run Linux drivers; nVidia releases a build of this driver for FreeBSD. The Linux compatibility option is here to install nVidia's Linux OpenGL libraries so that Linux binaries can run with FreeBSD's driver. This is possible because nVidia's OpenGL libraries communicate with the driver by using /dev/nvidia and the FreeBSD driver offers the same interface. In short, there is no easy way to use the ATI drivers for Linux under FreeBSD. True, but before they've started releasing them there was a wrapper in ports which used the Linux driver (and I think that was what motivated nVidia to release FreeBSD drivers). I just thought might be somebody who knows better than myself woud feel interested enough to write a similar wrapper for ATI drivers (which could eventualy stir them towards the same direction as nVidia). It looks I was wrong. A pity. Regards, Vladimir ___ freebsd-hackers@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to "[EMAIL PROTECTED]"
ATI drivers for FreeBSD - anything?
Hi all, Is there any project on FreeBSD wrapper for ATI Linux drivers (like nVidia's used to be)? If so - I'd be more than happy to test (sorry I can hardly write it myself). Regards, Vladimir ___ freebsd-hackers@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: libmap.conf: mapping directories?
On Saturday 03 September 2005 01:13, Pascal Hofstee wrote: [snip] > There is actually an effort underway right now to make happen exactly > what you suggested: making a /compat/ia32 available with a > freebsd32-syscall table similarly to how we treat Linux. > > The last word i got from the main person involved here is they > discovered a few bugs they're trying to iron out at the moment. > > The person who can tell you more about that and i am sure would be > thrilled to receive additional testing support goes by the name of > Willow` on ##FreeBSD on the FreeNode IRC-network. Feel free to drop by > sometime. Thanks a bunch, I'd be happy to participate. Regards, Vladimir ___ freebsd-hackers@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to "[EMAIL PROTECTED]"
libmap.conf: mapping directories?
Hi all,
Is it doable to teach rtld (via libmap.conf) to map directories rather
than libraries one-by-one? My problem is: I'm trying to use i386-built
packages on my amd64 box. Most of them don't work 'cause of RPATH set to
/usr/{X11R6,local}/lib where (obviously) 64-bit libs reside. And this
is one of the reasons ia32 compatibility under amd64 arch is almost
useless.
One way: to completely remove RPATH (with chrpath, for example - BTW, this
is nice enough utility but to make it work with 32-bit objects one has to
use some workarounds). It's not always convenient, though. Much, much better
it would be to place ALL of ia32 compat stuff into something like
/compat/freebsd32 like we do it for Linux stuff, but somehow nobody seems
to be even remotely interested. And the last way I see (a workaround as
well but hey, it's better than nothing at all) would be $SUBJECT. So my
question is: is it possible to map, say, /usr/local/lib to
/usr/local/lib32 and if yes how do I do it?
TIA,
Vladimir
___
freebsd-hackers@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Are there any on-going projects on v4l porting?
Hello all, As subj. said - does anybody work on porting v4l & (especially!) drivers for non- bt8x based cards? Specifically saa7134 based (got one and would rather not have to reboot to Linux to watch TV :-) Yes, I know, the simplest answer would be "you're interested - you do" but that'd be quite beyond my skills. Still I'd happily help with testing/debugging/whatever else. Regards, Vladimir To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: cpu class & features
On Thu, 7 Nov 2002, Sam wrote: > On Thursday 07 November 2002 06:37 pm, Terry Lambert wrote: > > Sam wrote: > > > I'm writing an application that needs info on the machine cpu > > > architecture, the cpu class, and the cpu features. > > > > Is this a status display (e.g. "About This Computer...") for a human > > to read? If not, the entire point of an OS is to hide that information > > from you, so that you can write code that runs on the OS, instead of > > writing code that runs only on particular hardware. > > No. If you really need to know, it is for implementing a Windows API call > on FreeBSD for Wine. > > I'm not interested in getting into a flame war on what a point of an OS > is, what should be allowed and what should not. But there are _many_ > other legitimate reasons to need that info other than "about this > computer" display. > > > -- Terry > > Sam > Take a look at MPlayer sources (ports/graphic/mplayer). They got what you need in TOOLS/cpuinfo.c Regards, Vladimir -- Vladimir Kushnir - [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: -STABLE NVidia drivers
Hi, Well, since nobody seems to be interested (I can't for the life of mine understand why), thanks and a couple of (perhaps silly) questions. On Thu, 15 Nov 2001, David Rufino wrote: > Hi, > > Just thought I would mention my -STABLE NVidia drivers have been working for > the past week, as far as 2D and XVideo goes. Available on CVS at > http://sourceforge.net/projects/nv-bsd. > > David > i) any plans on porting to -CURRENT? On GLX stuff? And does it work with DEVFS? ii) any problems that should be addressed? iii) (sorry, this goes before reading your code) is your port very different from Matthew Dodd's (which I run right now)? iv) any help you need? v) thanks again, this kind of work is a miracle which no-one seems to have noticed - usage for alien _kernel_ modules is not something one can see very often. Regards, Vladimir -- Vladimir Kushnir - [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: Framebuffer device under FreeBSD?
On Fri, 9 Nov 2001, Kevin D.Wooten wrote: > > > > i. Is it feasible to port Linux fbdev modules to FreeBSD (as a modules, > > again)? > > > I was under the impression that the subsystem is there ( /usr/src/sys/dev/fb > ? ). Maybe not the specific accelerated drivers, but the basics look there. > I will say the video mode selection looks weak though, Linux does have a much > more flexible mode selection. > > P.S. - I have been trying to get NVidia docs ( I have to sign an NDA ) to > build a GeForce3 framebuffer. > Actually, rivafb under Linux provides some HW acceleration (based on utah-glx code, I think). There's no GeForce3 there, though. -- Vladimir Kushnir - [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Framebuffer device under FreeBSD?
Hi all, A couple of questions here. i. Is it feasible to port Linux fbdev modules to FreeBSD (as a modules, again)? They work quite nice under Linux for multimedia apps, and it's a pity we haven't got anything of this kind (TNT2 based v/cards with HW accelleration, for one). Did anybody try to do that? If so, I would happily join this project (no money for XFree86-supported card like Matrox or Radeon :-) ii. If not, is there any chance to port Linux's vm86 - based video module (see MPlayer CVS version, VESA video-out) to our i386_vm86? It alse work nice under Linux (no HW accelleration but very low overhead instead). Any hints would be greatly appreciated. TIA, Vladimir -- Vladimir Kushnir - [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: Screen saver module
Hi, On Thu, 8 Nov 2001, Jonathan Belson wrote: > Hiya > > > I've written a screen saver module for syscons which uses > VESA modes (where available) for higher resolutions. > > It runs quite happily on my GeForce 2 but I found it > crashes almost straight away on a Matrox Millenium > when I try to run it in M_VESA_CG800x600 - I changed > the code to use 640x480 and it seemed fine. > It runs just as happily on Vanta here (FreeBSD-CURRENT). Thanks Jonathan. > > The source is at http://www.witchspace.com//kix.tgz > Extract it in /usr/src/sys/modules/syscons, cd to > 'kix' and type 'make && make install'. > > Cheers, > > > --Jon > Would somebody commit it (with some warning message, perhaps)? Regards, Vladimir -- Vladimir Kushnir - [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
Re: Cdrom and russian sysmbols
Hello, First - thanks a bunch for your patch. The only problem here is that usually in Russia (and here in Ukraine too) they burn CDs unduer DOS/Windows, and so the usual encoding is CP866 rather than KOI8-R. Here's a patch to your patch with this encoding. Besides, there still some additional editing may be needed. And the last. If one compiles cd866 as a module ane needs to compile kernel+modules with "make -DCHARSET_" (in Eugene's case probably "make -DCHARSET_CP866"). On Thu, 30 Nov 2000, Motomichi Matsuzaki wrote: > > At Wed, 29 Nov 2000 15:02:43 +, > Kaltashkin Eugene <[EMAIL PROTECTED]> wrote: > > How i can mount cdrom drive created on Windows machine with specific russian > > symbols in filename ? > > Try my patch: > > http://triaez.kaisei.org/~mzaki/joliet/ > > This was originally for 3.1-stable, > but will be applied for 4-stable or -current with a few obvious fix. > Regards, Vladimir -- ===|=== Vladimir Kushnir | [EMAIL PROTECTED] |Powered by FreeBSD joluni.diff.gz
