Re: K6-x optimized .debs for 4.0.1-1
I have been told that -march=k6 may not be beneficial for athlons. But some compilers may be able to use -march=k7. See previous emails on this subject. Thanks. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown **
Re: K6-x optimized .debs for 4.0.1-1
I have been told that -march=k6 may not be beneficial for athlons. But some compilers may be able to use -march=k7. See previous emails on this subject. Thanks. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown ** -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1
On Mon, Nov 06, 2000 at 12:26:14PM -0600, [EMAIL PROTECTED] wrote: > Hi, all. I finally succeeded in compiling K6-optimized X binaries for > XF4.0.1. Are these optimizations also benificial for the K7 (Athlon) line of CPU's? <>
Re: K6-x optimized .debs for 4.0.1-1
On Mon, Nov 06, 2000 at 12:26:14PM -0600, [EMAIL PROTECTED] wrote: > Hi, all. I finally succeeded in compiling K6-optimized X binaries for > XF4.0.1. Are these optimizations also benificial for the K7 (Athlon) line of CPU's? Navidad.exe
Re: K6-x optimized .debs for 4.0.1-1
sure. not a problem. I don't reboot my box very often (or don't plan to). Don't be surprised if it slows down a bit during that time. this is my personal box. :-) I won't be recompiling X on this box very often. But, i'll mail the list when i do. thanks. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown **
Re: K6-x optimized .debs for 4.0.1-1
sure. not a problem. I don't reboot my box very often (or don't plan to). Don't be surprised if it slows down a bit during that time. this is my personal box. :-) I won't be recompiling X on this box very often. But, i'll mail the list when i do. thanks. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown ** -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1
I will mirror (daily, sometime between 0600-1100 CET) your site to: ftp://kalle.csb.ki.se/pub/linux/debian-x-k6 if you don't mind. Regards, /Karl --- Karl HammarAspö Data [EMAIL PROTECTED] Lilla Aspö 2340 +46 173 140 57Networks S-742 94 Östhammar +46 70 511 97 84 Computers Sweden Consulting --- From: [EMAIL PROTECTED] Subject: K6-x optimized .debs for 4.0.1-1 Date: Mon, 6 Nov 2000 12:26:14 -0600 (CST) > Hi, all. I finally succeeded in compiling K6-optimized X binaries for > XF4.0.1. So far, it's quite stable... I'm still having troubles with my > s3virge server hard-locking my box me when running moire2 > screensaver...cheap card perhaps? If people would like to try these out, I > put the .debs up on ftp via > ftp://fh-wrstlr-198035.flinthills.com/pub/debian/. Please be aware that > my DSL uplink is only 112Kbps. > > I also included the patches for both the s3virge/gx2 chip and the k6 > optimizations. Of course, I make no guarantees (proceed at your own > peril?) about these. Just put these into the debian/patches directory > before building X. > > > -- > ** Derek J Witt** > * Email: mailto:[EMAIL PROTECTED] * > * Home Page: http://www.flinthills.com/~djw/ * > *** "Houston, the Eagle has landed and laid an egg!" -- Unknown **
Re: K6-x optimized .debs for 4.0.1-1
I will mirror (daily, sometime between 0600-1100 CET) your site to: ftp://kalle.csb.ki.se/pub/linux/debian-x-k6 if you don't mind. Regards, /Karl --- Karl HammarAspö Data [EMAIL PROTECTED] Lilla Aspö 2340 +46 173 140 57Networks S-742 94 Östhammar +46 70 511 97 84 Computers Sweden Consulting --- From: [EMAIL PROTECTED] Subject: K6-x optimized .debs for 4.0.1-1 Date: Mon, 6 Nov 2000 12:26:14 -0600 (CST) > Hi, all. I finally succeeded in compiling K6-optimized X binaries for > XF4.0.1. So far, it's quite stable... I'm still having troubles with my > s3virge server hard-locking my box me when running moire2 > screensaver...cheap card perhaps? If people would like to try these out, I > put the .debs up on ftp via > ftp://fh-wrstlr-198035.flinthills.com/pub/debian/. Please be aware that > my DSL uplink is only 112Kbps. > > I also included the patches for both the s3virge/gx2 chip and the k6 > optimizations. Of course, I make no guarantees (proceed at your own > peril?) about these. Just put these into the debian/patches directory > before building X. > > > -- > ** Derek J Witt** > * Email: mailto:[EMAIL PROTECTED] * > * Home Page: http://www.flinthills.com/~djw/ * > *** "Houston, the Eagle has landed and laid an egg!" -- Unknown ** -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1 -- OT
Well, k6 will work just fine for athlon. Looks like k7 is either removed or buggy. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown **
Re: K6-x optimized .debs for 4.0.1-1 -- OT
* Michael Meding <[EMAIL PROTECTED]> [001107 19:50]: [..athlon..] > Anybody any ideas ? Chances are good there will be very few ideas coming from debian-x. Try google, usenet, gcc mailling lists, perhaps debian-user. (Sadly, the gcc manpage didn't seem to shout the right answer to me. It did mention loads of other CPU types though...) -- ``Oh Lord; Ooh you are so big; So absolutely huge; Gosh we're all really impressed down here, I can tell you.''
Re: K6-x optimized .debs for 4.0.1-1 -- OT
Hi Seth, Seth Arnold wrote: > > * Michael Meding <[EMAIL PROTECTED]> [001107 19:50]: > [..athlon..] > > Anybody any ideas ? > > Chances are good there will be very few ideas coming from debian-x. Try > google, usenet, gcc mailling lists, perhaps debian-user. (Sadly, the gcc > manpage didn't seem to shout the right answer to me. It did mention > loads of other CPU types though...) Sure that may be the case. But maybe there is somebody also trying to figure out which compiler flags to use to get the most optimized *.deb files for XF and maybe also doing benchmarks on it. With best regards Michael Meding
Re: K6-x optimized .debs for 4.0.1-1 -- OT
Well, k6 will work just fine for athlon. Looks like k7 is either removed or buggy. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown ** -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1 -- OT
Hi there, [EMAIL PROTECTED] wrote: > > there's a different flag for athlon, it's -march=k7 (vs. -march=k6 for ^^^ guess that is not going to work with gcc-2.95.2. So what is best for athlon (classic, duron, thunderbird) then ? Took some from www.athlonlinux.org but I am still not quite sure how to produce fastest code with gcc-2.95.2 for athlons. Anybody any ideas ? Thanks in advance. Greetings Michael Meding
Re: K6-x optimized .debs for 4.0.1-1 -- OT
Hi Seth, Seth Arnold wrote: > > * Michael Meding <[EMAIL PROTECTED]> [001107 19:50]: > [..athlon..] > > Anybody any ideas ? > > Chances are good there will be very few ideas coming from debian-x. Try > google, usenet, gcc mailling lists, perhaps debian-user. (Sadly, the gcc > manpage didn't seem to shout the right answer to me. It did mention > loads of other CPU types though...) Sure that may be the case. But maybe there is somebody also trying to figure out which compiler flags to use to get the most optimized *.deb files for XF and maybe also doing benchmarks on it. With best regards Michael Meding -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1 -- OT
* Michael Meding <[EMAIL PROTECTED]> [001107 19:50]: [..athlon..] > Anybody any ideas ? Chances are good there will be very few ideas coming from debian-x. Try google, usenet, gcc mailling lists, perhaps debian-user. (Sadly, the gcc manpage didn't seem to shout the right answer to me. It did mention loads of other CPU types though...) -- ``Oh Lord; Ooh you are so big; So absolutely huge; Gosh we're all really impressed down here, I can tell you.'' -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1 -- OT
Hi there, [EMAIL PROTECTED] wrote: > > there's a different flag for athlon, it's -march=k7 (vs. -march=k6 for ^^^ guess that is not going to work with gcc-2.95.2. So what is best for athlon (classic, duron, thunderbird) then ? Took some from www.athlonlinux.org but I am still not quite sure how to produce fastest code with gcc-2.95.2 for athlons. Anybody any ideas ? Thanks in advance. Greetings Michael Meding -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1
there's a different flag for athlon, it's -march=k7 (vs. -march=k6 for K6-x). -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown **
Re: K6-x optimized .debs for 4.0.1-1
Sure, I just used -march=k6 -funrolled-loops -finline-functions in hosts.def. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown **
Re: K6-x optimized .debs for 4.0.1-1
there's a different flag for athlon, it's -march=k7 (vs. -march=k6 for K6-x). -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown ** -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1
Sure, I just used -march=k6 -funrolled-loops -finline-functions in hosts.def. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown ** -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1
Hi there, what have you done to get k6 optimized binaries ? Which flags have you used if I may ask ? Greetings Michael Meding
Re: K6-x optimized .debs for 4.0.1-1
Hi there, what have you done to get k6 optimized binaries ? Which flags have you used if I may ask ? Greetings Michael Meding -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: K6-x optimized .debs for 4.0.1-1
On Mon, Nov 06, 2000 at 12:26:14PM -0600, [EMAIL PROTECTED] wrote: > Hi, all. I finally succeeded in compiling K6-optimized X binaries for > XF4.0.1. Are these optimizations also benificial for the K7 (Athlon) line of CPU's? pgpK2Ow4AXj7y.pgp Description: PGP signature
Re: K6-x optimized .debs for 4.0.1-1
On Mon, Nov 06, 2000 at 12:26:14PM -0600, [EMAIL PROTECTED] wrote: > Hi, all. I finally succeeded in compiling K6-optimized X binaries for > XF4.0.1. Are these optimizations also benificial for the K7 (Athlon) line of CPU's? PGP signature
Re: K6-x optimized .debs for 4.0.1-1
BTW, the patched xserver-xfree86 s3virge support will only work for the virge/gx2 chip. If you don't have the gx2, but do have an s3 chip, I recommend not getting my binaries,but instead just not getting the s3virge/gx2 patch. Cheers. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown **
K6-x optimized .debs for 4.0.1-1
Hi, all. I finally succeeded in compiling K6-optimized X binaries for XF4.0.1. So far, it's quite stable... I'm still having troubles with my s3virge server hard-locking my box me when running moire2 screensaver...cheap card perhaps? If people would like to try these out, I put the .debs up on ftp via ftp://fh-wrstlr-198035.flinthills.com/pub/debian/. Please be aware that my DSL uplink is only 112Kbps. I also included the patches for both the s3virge/gx2 chip and the k6 optimizations. Of course, I make no guarantees (proceed at your own peril?) about these. Just put these into the debian/patches directory before building X. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown ** --- xc/programs/Xserver/hw/xfree86/drivers/s3virge/s3v_macros.h Fri Mar 31 14:13:33 2000 +++ xc/programs/Xserver/hw/xfree86/drivers/s3virge/s3v_macros2.hFri Nov 3 01:59:40 2000 @@ -97,7 +97,7 @@ #define WAITFIFO(n) if(ps3v->NoPCIRetry) \ -while(((INREG(SUBSYS_STAT_REG) >> 8) & 0x1f) < n){} +while(((INREG(SUBSYS_STAT_REG) >> 9) & 0x60) < n){} #define WAITIDLE()\ do { int loop=0; mem_barrier(); \ --- /dev/null Wed Jun 7 10:48:23 2000 +++ xc/config/cf/host.def Thu Jul 6 19:05:20 2000 @@ -0,0 +1,710 @@ +XCOMM $XFree86: xc/config/cf/xf86site.def,v 3.162 2000/05/31 07:14:52 eich Exp $ +/**/ +/* + * This file is to provide a quick method for most people to change the + * behaviour of their XFree86 installation without having to fully + * understand the workings of site.def and all the various '.cf' files. + * + * These are the most common settings you would choose for compiling and + * installing XFree86 on the systems supported by it. + * + * A good way to use this file is to copy it to host.def, and make the + * changes there. That way, future patches to this file won't fail. + * The host.def file will never be patched. + * + * The distributed version of this file should contain no uncommented + * definitions. Such default definitions belong in xfree86.cf. + */ +/**/ + +/* + * If you have build-specific modifications in your host.def file, but + * want an empty host.def file installed when doing 'make install', + * uncomment the following + * + */ +#define InstallEmptyHostDef + +/* + * If you want to enable some developer settings, like more verbose + * compiler warnings, uncomment this. + * +*/ +#define XFree86Devel YES + +/* + * If using GCC 2.x on a system where it isn't the default, uncomment + * the following + * +#define HasGcc2YES +#define HasGcc YES + */ + +/* + * If using GCC 2.x with C++ on a system where it isn't the default, uncomment + * the following. + * +#define HasGcc2ForCplusplusYES + */ + +/* + * The default optimisation flags for GCC 2.x. -fno-strength-reduce is + * here to work around a bug in -O2 for GCC 2.x on i386 platforms. + * If you are using a version that doesn't have this bug, you can + * uncomment the following line, and remove '-fno-strength-reduce' + * If you are building binaries for a 486, it may be beneficial to add + * -m486 + */ +#define DefaultGcc2i386Opt -O3 -fno-strength-reduce -march=k6 -funroll-loops -finline-functions + + +/* + * This allows the GCC warning flags to be set. The default is shown here. + * +#define GccWarningOptions -Wall -Wpointer-arith -Wstrict-prototypes \ + -Wmissing-prototypes -Wmissing-declarations \ + -Wredundant-decls -Wnested-externs + */ + +/* + * Sun Compiler stuff... + * +#define HasSunC YES +#define HasSunCplusplus YES +#define CplusplusCompilerMajorVersion 5 +#define CplusplusCompilerMinorVersion 0 +#define CCompilerMajorVersion 5 +#define CCompilerMinorVersion 0 + */ + +/* + * Optimized Sun Compiler Build. + * +#define DefaultCDebugFlags -xO4 -xtarget=pentium_pro +#define OptimizedCDebugFlags-xO4 -xtarget=pentium_pro + */ + +/* + * Debuggable Sun Compiler Build. + * Note: This builds _EVERYTHING_ as debuggable + * +#define DefaultCDebugFlags -g -xs +#define OptimizedCDebugFlags-g -xs + */ + +/* + * For Linux, this should match the Binutils version you have. This example + * is for 2.6.0.7. See linux.cf for the default setting. + * + * This should automatically get set correctly by imake. + * +#define LinuxBinUtilsMajorVersion 26 + */ + +/* + * For Linux, these should match the libc version you have. This example + * is for libc.5.4.x. See linux.cf for the default setting. + * + * This should automatically get set correctly by imake. + * +#def
Re: K6-x optimized .debs for 4.0.1-1
BTW, the patched xserver-xfree86 s3virge support will only work for the virge/gx2 chip. If you don't have the gx2, but do have an s3 chip, I recommend not getting my binaries,but instead just not getting the s3virge/gx2 patch. Cheers. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown ** -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
K6-x optimized .debs for 4.0.1-1
Hi, all. I finally succeeded in compiling K6-optimized X binaries for XF4.0.1. So far, it's quite stable... I'm still having troubles with my s3virge server hard-locking my box me when running moire2 screensaver...cheap card perhaps? If people would like to try these out, I put the .debs up on ftp via ftp://fh-wrstlr-198035.flinthills.com/pub/debian/. Please be aware that my DSL uplink is only 112Kbps. I also included the patches for both the s3virge/gx2 chip and the k6 optimizations. Of course, I make no guarantees (proceed at your own peril?) about these. Just put these into the debian/patches directory before building X. -- ** Derek J Witt** * Email: mailto:[EMAIL PROTECTED] * * Home Page: http://www.flinthills.com/~djw/ * *** "Houston, the Eagle has landed and laid an egg!" -- Unknown ** --- xc/programs/Xserver/hw/xfree86/drivers/s3virge/s3v_macros.h Fri Mar 31 14:13:33 2000 +++ xc/programs/Xserver/hw/xfree86/drivers/s3virge/s3v_macros2.hFri Nov 3 +01:59:40 2000 @@ -97,7 +97,7 @@ #define WAITFIFO(n) if(ps3v->NoPCIRetry) \ -while(((INREG(SUBSYS_STAT_REG) >> 8) & 0x1f) < n){} +while(((INREG(SUBSYS_STAT_REG) >> 9) & 0x60) < n){} #define WAITIDLE()\ do { int loop=0; mem_barrier(); \ --- /dev/null Wed Jun 7 10:48:23 2000 +++ xc/config/cf/host.def Thu Jul 6 19:05:20 2000 @@ -0,0 +1,710 @@ +XCOMM $XFree86: xc/config/cf/xf86site.def,v 3.162 2000/05/31 07:14:52 eich Exp $ +/**/ +/* + * This file is to provide a quick method for most people to change the + * behaviour of their XFree86 installation without having to fully + * understand the workings of site.def and all the various '.cf' files. + * + * These are the most common settings you would choose for compiling and + * installing XFree86 on the systems supported by it. + * + * A good way to use this file is to copy it to host.def, and make the + * changes there. That way, future patches to this file won't fail. + * The host.def file will never be patched. + * + * The distributed version of this file should contain no uncommented + * definitions. Such default definitions belong in xfree86.cf. + */ +/**/ + +/* + * If you have build-specific modifications in your host.def file, but + * want an empty host.def file installed when doing 'make install', + * uncomment the following + * + */ +#define InstallEmptyHostDef + +/* + * If you want to enable some developer settings, like more verbose + * compiler warnings, uncomment this. + * +*/ +#define XFree86Devel YES + +/* + * If using GCC 2.x on a system where it isn't the default, uncomment + * the following + * +#define HasGcc2YES +#define HasGcc YES + */ + +/* + * If using GCC 2.x with C++ on a system where it isn't the default, uncomment + * the following. + * +#define HasGcc2ForCplusplusYES + */ + +/* + * The default optimisation flags for GCC 2.x. -fno-strength-reduce is + * here to work around a bug in -O2 for GCC 2.x on i386 platforms. + * If you are using a version that doesn't have this bug, you can + * uncomment the following line, and remove '-fno-strength-reduce' + * If you are building binaries for a 486, it may be beneficial to add + * -m486 + */ +#define DefaultGcc2i386Opt -O3 -fno-strength-reduce -march=k6 -funroll-loops +-finline-functions + + +/* + * This allows the GCC warning flags to be set. The default is shown here. + * +#define GccWarningOptions -Wall -Wpointer-arith -Wstrict-prototypes \ + -Wmissing-prototypes -Wmissing-declarations \ + -Wredundant-decls -Wnested-externs + */ + +/* + * Sun Compiler stuff... + * +#define HasSunC YES +#define HasSunCplusplus YES +#define CplusplusCompilerMajorVersion 5 +#define CplusplusCompilerMinorVersion 0 +#define CCompilerMajorVersion 5 +#define CCompilerMinorVersion 0 + */ + +/* + * Optimized Sun Compiler Build. + * +#define DefaultCDebugFlags -xO4 -xtarget=pentium_pro +#define OptimizedCDebugFlags-xO4 -xtarget=pentium_pro + */ + +/* + * Debuggable Sun Compiler Build. + * Note: This builds _EVERYTHING_ as debuggable + * +#define DefaultCDebugFlags -g -xs +#define OptimizedCDebugFlags-g -xs + */ + +/* + * For Linux, this should match the Binutils version you have. This example + * is for 2.6.0.7. See linux.cf for the default setting. + * + * This should automatically get set correctly by imake. + * +#define LinuxBinUtilsMajorVersion 26 + */ + +/* + * For Linux, these should match the libc version you have. This example + * is f