Hello, I got a helpfully mail pointing me to a post at glide.sourceforge.net on Thursday or Friday last week and bingo, I got it.
http://sourceforge.net/tracker/index.php?func=detail&aid=451696&group_id=369&atid=100369 So most kudos go to ??? --- I don't know. Both patches apply but the second (for h5) has a rejection. I found that the rejected part isn't needed with the current glide code base. [-] --- glide3x-20010309/h5/glide3/src/fifo.c.orig Tue Jul 31 12:38:36 2001 +++ glide3x-20010309/h5/glide3/src/fifo.c Tue Jul 31 15:28:31 2001 @@ -1181,7 +1181,7 @@ long -_griHwFifoPtr(FxBool ignored) +_grHwFifoPtr(FxBool ignored) { long rVal = 0; [-] I'll rediff it tomorrow and call for commit. I have a precompiled version for the Athlon ready (i686 optimizations, hand crafted) and ask everyone for testing. 'cause I have no space present I ask for hosting and/or forwarding it to the Glide/DRI site, too. The autoconf scripts aren't running completely so I call for help, here. They only build much bigger and slower (!!!) libs when I chmod 755 build.3dfx SOURCE/Glide3> ./chores.3dfx --clean --generate --configure=--enable-amd3d [-] config.status: creating swlibs/texus2/cmd/makefile.autoconf config.status: creating swlibs/include/config.h Done. SOURCE/Glide3> cd build Directory: /tmp/INSTALL/SOURCE/Glide3/build Glide3/build> l insgesamt 204 drwxr-xr-x 6 nuetzel users 304 Apr 9 22:52 . drwxr-xr-x 8 nuetzel users 928 Apr 9 22:52 .. -rw-r--r-- 1 nuetzel users 2391 Apr 9 22:52 build.3dfx -rw-r--r-- 1 nuetzel users 23069 Apr 9 22:52 config.log -rwxr-xr-x 1 nuetzel users 32270 Apr 9 22:52 config.status drwxr-xr-x 5 nuetzel users 160 Apr 9 22:52 cvg drwxr-xr-x 5 nuetzel users 160 Apr 9 22:52 h3 drwxr-xr-x 5 nuetzel users 160 Apr 9 22:52 h5 -rwxr-xr-x 1 nuetzel users 119610 Apr 9 22:52 libtool -rw-r--r-- 1 nuetzel users 17612 Apr 9 22:52 makefile.autoconf drwxr-xr-x 7 nuetzel users 208 Apr 9 22:52 swlibs Glide3/build> chmod 755 build.3dfx Glide3/build> ./build.3dfx Dateisystem 1k-Bl�cke Benutzt Verf�gbar Ben% montiert auf /dev/sda2 1518088 274728 1243360 19% /tmp Making all in swlibs make[1]: Entering directory `/tmp/INSTALL/SOURCE/Glide3/build/swlibs' Making all in fxmisc make[2]: Entering directory `/tmp/INSTALL/SOURCE/Glide3/build/swlibs/fxmisc' /bin/sh ../../libtool --mode=compile gcc -DX11 -fomit-frame-pointer -funroll-loops -fexpensive-optimizations -ffast-math -DBIG_OPT -Wall -I/tmp/INSTALL/SOURCE/Glide3/h5/glide3/src -I../../h5/incsrc -I/tmp/INSTALL/SOURCE/Glide3/h5/incsrc -I/tmp/INSTALL/SOURCE/Glide3/h5/minihwc -I. -I/tmp/INSTALL/SOURCE/Glide3/swlibs/fxmemmap -I/tmp/INSTALL/SOURCE/Glide3/swlibs/fxmisc -I/tmp/INSTALL/SOURCE/Glide3/swlibs/newpci/pcilib -I/tmp/INSTALL/SOURCE/Glide3/swlibs/texus2/lib -O6 -m486 -c /tmp/INSTALL/SOURCE/Glide3/swlibs/fxmisc/fx64.c libtool: ltconfig version `' does not match ltmain.sh version `1.3.5' Fatal configuration error. See the libtool docs for more information. make[2]: *** [fx64.lo] Error 1 make[2]: Leaving directory `/tmp/INSTALL/SOURCE/Glide3/build/swlibs/fxmisc' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory `/tmp/INSTALL/SOURCE/Glide3/build/swlibs' make: *** [all-recursive] Error 1 Dateisystem 1k-Bl�cke Benutzt Verf�gbar Ben% montiert auf /dev/sda2 1518088 274728 1243360 19% /tmp Time Elapsed: 0 Years, 0 Days, 0 Hours, 0 Minutes, 0 Seconds expr: non-numeric argument ./build.3dfx: [: -lt: unary operator expected expr: Syntaxfehler expr: Syntaxfehler Disk usage: . Mb After I replace the generated libtool with my SuSE 7.3 /usr/bin/libtool version I get the big and slow libs. ar cru .libs/libglide3.a gsplash.o g3df.o gu.o gthread.o gpci.o diglide.o disst.o ditex.o gbanner.o gerror.o gaa.o gdraw.o gglide.o distate.o gstrip.o distrip.o diget.o glfb.o gsst.o gtex.o gtexdl.o fifo.o cpudtect.o xtexdl_def.o xdraw2_3dnow.o xdraw3_3dnow.o xtexdl_3dnow.o xdraw2_def.o xdraw3_def.o .libs/libglide3.lax/libfxmisc.al/fximg.lo .libs/libglide3.lax/libfxmisc.al/fx64.lo .libs/libglide3.lax/libfxmisc.al/linutil.lo .libs/libglide3.lax/libfxmisc.al/fxos.lo .libs/libglide3.lax/libfxpci.al/fxmsr.lo .libs/libglide3.lax/libfxpci.al/sst1_pci.lo .libs/libglide3.lax/libfxpci.al/fxpci.lo .libs/libglide3.lax/libfxpci.al/fxlinux.lo .libs/libglide3.lax/libfxpci.al/fxinfo.lo .libs/libglide3.lax/libtexus.al/nccnnet.lo .libs/libglide3.lax/libtexus.al/read.lo .libs/libglide3.lax/libtexus.al/pal256.lo .libs/libglide3.lax/libtexus.al/codec.lo .libs/libglide3.lax/libtexus.al/quantize.lo .libs/libglide3.lax/libtexus.al/ncc.lo .libs/libglide3.lax/libtexus.al/ppm.lo .libs/libglide3.lax/libtexus.al/rgt.lo .libs/libglide3.lax/libtexus.al/diffuse.lo .libs/libglide3.lax/libtexus.al/tga.lo .libs/libglide3.lax/libtexus.al/txs.lo .libs/libglide3.lax/libtexus.al/mipmap.lo .libs/libglide3.lax/libtexus.al/eigen.lo .libs/libglide3.lax/libtexus.al/view.lo .libs/libglide3.lax/libtexus.al/util.lo .libs/libglide3.lax/libtexus.al/resample.lo .libs/libglide3.lax/libtexus.al/3df.lo .libs/libglide3.lax/libtexus.al/pal6666.lo .libs/libglide3.lax/libtexus.al/clamp.lo .libs/libglide3.lax/libtexus.al/write.lo .libs/libglide3.lax/libtexus.al/texuslib.lo .libs/libglide3.lax/libtexus.al/bitcoder.lo .libs/libglide3.lax/libtexus.al/dequant.lo .libs/libglide3.lax/libminihwc.al/hwcio.lo .libs/libglide3.lax/libminihwc.al/gdebug.lo .libs/libglide3.lax/libminihwc.al/linhwc.lo ranlib .libs/libglide3.a rm -fr .libs/libglide3.lax creating libglide3.la (cd .libs && rm -f libglide3.la && ln -s ../libglide3.la libglide3.la) make[3]: Leaving directory `/tmp/INSTALL/SOURCE/Glide3/build/h5/glide3/src' make[3]: Entering directory `/tmp/INSTALL/SOURCE/Glide3/build/h5/glide3' make[3]: Nothing to be done for `all-am'. make[3]: Leaving directory `/tmp/INSTALL/SOURCE/Glide3/build/h5/glide3' make[2]: Leaving directory `/tmp/INSTALL/SOURCE/Glide3/build/h5/glide3' make[2]: Entering directory `/tmp/INSTALL/SOURCE/Glide3/build/h5' make[2]: Nothing to be done for `all-am'. make[2]: Leaving directory `/tmp/INSTALL/SOURCE/Glide3/build/h5' make[1]: Leaving directory `/tmp/INSTALL/SOURCE/Glide3/build/h5' make[1]: Entering directory `/tmp/INSTALL/SOURCE/Glide3/build' make[1]: Nothing to be done for `all-am'. make[1]: Leaving directory `/tmp/INSTALL/SOURCE/Glide3/build' Dateisystem 1k-Bl�cke Benutzt Verf�gbar Ben% montiert auf /dev/sda2 1518088 278252 1239836 19% /tmp Time Elapsed: 0 Years, 0 Days, 0 Hours, 2 Minutes, 6 Seconds expr: non-numeric argument ./build.3dfx: [: -lt: unary operator expected expr: Syntaxfehler expr: Syntaxfehler Disk usage: . Mb Glide3/build> l h5/glide3/src/.libs/ insgesamt 1107 drwxr-xr-x 2 nuetzel users 248 Apr 9 23:06 . drwxr-xr-x 4 nuetzel users 1904 Apr 9 23:06 .. -rw-r--r-- 1 nuetzel users 598428 Apr 9 23:06 libglide3.a lrwxrwxrwx 1 nuetzel users 15 Apr 9 23:06 libglide3.la -> ../libglide3.la -rw-r--r-- 1 nuetzel users 771 Apr 9 23:06 libglide3.lai lrwxrwxrwx 1 nuetzel users 19 Apr 9 23:06 libglide3.so -> libglide3.so.3.10.0 lrwxrwxrwx 1 nuetzel users 19 Apr 9 23:06 libglide3.so.3 -> libglide3.so.3.10.0 -rwxr-xr-x 1 nuetzel users 523720 Apr 9 23:06 libglide3.so.3.10.0 Apart from that I get a running texus2 tool! :-) Glide3/build> l swlibs/texus2/cmd/ insgesamt 218 drwxr-xr-x 4 nuetzel users 184 Apr 9 23:05 . drwxr-xr-x 4 nuetzel users 136 Apr 9 22:52 .. drwxr-xr-x 2 nuetzel users 72 Apr 9 23:05 .deps drwxr-xr-x 2 nuetzel users 48 Apr 9 23:05 .libs -rw-r--r-- 1 nuetzel users 13844 Apr 9 23:05 cmd.o -rw-r--r-- 1 nuetzel users 12884 Apr 9 22:52 makefile.autoconf -rwxr-xr-x 1 nuetzel users 186622 Apr 9 23:05 texus It would be really cool to get the Glide test progs running, too. I did my speedy lib like the old style: setenv FX_GLIDE_HW h5 setenv GL_AMD3D 1 make -f makefile.linux /home/nuetzel> l /usr/lib/libglide3.so.3.1 -rwxr-xr-x 1 root root 354937 Apr 7 07:55 /usr/lib/libglide3.so.3.1 Next thing: I will focus on speed improvements (prefetch) and former speed grade (summer 2000) as my results with latest Accelerated-X shows we can get more... Linux version 2.4.19-pre5 + -aa vm33 + preemption+lock-break + 00_k7-prefetch-1 + page_color Mesa/demos> ./tenmillion running on SunWave1 (Mesa DRI 20020221 Voodoo4 x86/MMX/3DNow!) Tue Apr 9 23:36:38 2002 visual: 0x23 50 pixel triangles, 90 triangles/strip, 1 strips vertex data: v3f flat shading, zbuffer off, light off, texture off, immediate off test time: 1.998 secs. geometry rate: 879496 triangles/sec fill rate: 43.975 Mpixels/sec draw_strips_count: 1 Mesa/demos> ./tenmillion smooth running on SunWave1 (Mesa DRI 20020221 Voodoo4 x86/MMX/3DNow!) Tue Apr 9 23:36:46 2002 visual: 0x23 50 pixel triangles, 90 triangles/strip, 1 strips vertex data: v3f smooth shading, zbuffer off, light off, texture off, immediate off test time: 1.579 secs. geometry rate: 929552 triangles/sec fill rate: 46.478 Mpixels/sec draw_strips_count: 1 Mesa/demos> ./tenmillion area=3 running on SunWave1 (Mesa DRI 20020221 Voodoo4 x86/MMX/3DNow!) Tue Apr 9 23:30:05 2002 visual: 0x23 3 pixel triangles, 90 triangles/strip, 4 strips vertex data: v3f flat shading, zbuffer off, light off, texture off, immediate off test time: 1.996 secs. geometry rate: 3883429 triangles/sec fill rate: 11.650 Mpixels/sec draw_strips_count: 1 Mesa/demos> ./tenmillion area=3 smooth running on SunWave1 (Mesa DRI 20020221 Voodoo4 x86/MMX/3DNow!) Tue Apr 9 23:30:14 2002 visual: 0x23 3 pixel triangles, 90 triangles/strip, 4 strips vertex data: v3f smooth shading, zbuffer off, light off, texture off, immediate off test time: 1.247 secs. geometry rate: 4551861 triangles/sec fill rate: 13.656 Mpixels/sec draw_strips_count: 1 With older kernel and Accelerated-X 2.0 beta I got: GL_VERSION: 1.2.1 GL_EXTENSIONS: GL_ARB_multitexture GL_EXT_abgr GL_EXT_blend_subtract GL_RENDERER: VSA-100/3DNow! GL_VENDOR: Xi Graphics, Inc. GLU_VERSION: 1.3 GLU_EXTENSIONS: GLU_EXT_nurbs_tessellator GLU_EXT_object_space_tess GLUT_API_VERSION: 3 GLUT_XLIB_IMPLEMENTATION: 15 running on SunWave1 (VSA-100/3DNow!) Thu Sep 27 05:25:10 2001 visual: 0x21 50 pixel triangles, 90 triangles/strip, 1 strips vertex data: v3f flat shading, zbuffer off, light off, texture off, immediate off test time: 1.997 secs. geometry rate: 3258835 triangles/sec fill rate: 162.942 Mpixels/sec draw_strips_count: 1 SunWave1>./tenmillion smooth running on SunWave1 (VSA-100/3DNow!) Thu Sep 27 05:25:17 2001 visual: 0x21 50 pixel triangles, 90 triangles/strip, 1 strips vertex data: v3f smooth shading, zbuffer off, light off, texture off, immediate off test time: 1.997 secs. geometry rate: 3259569 triangles/sec fill rate: 162.978 Mpixels/sec draw_strips_count: 1 SunWave1>./tenmillion area=3 running on SunWave1 (VSA-100/3DNow!) Thu Sep 27 05:25:28 2001 visual: 0x21 3 pixel triangles, 90 triangles/strip, 4 strips vertex data: v3f flat shading, zbuffer off, light off, texture off, immediate off test time: 1.996 secs. geometry rate: 6836102 triangles/sec fill rate: 20.508 Mpixels/sec draw_strips_count: 1 SunWave1>./tenmillion area=3 smooth running on SunWave1 (VSA-100/3DNow!) Thu Sep 27 05:25:39 2001 visual: 0x21 3 pixel triangles, 90 triangles/strip, 4 strips vertex data: v3f smooth shading, zbuffer off, light off, texture off, immediate off test time: 1.996 secs. geometry rate: 6836527 triangles/sec fill rate: 20.510 Mpixels/sec draw_strips_count: 1 They have a much better AGP driver I think. Look here, there 2.1 diver on my latest kernel (with page_color). Now, they have an updated 2.1 Accelerated-X server with 3.xx AGPGART version, but there precompiled kernel module do not work on my kernel. SunWave1 xig/xsvc# ./xsvctest xsvc: Version Major: 2 Minor: 1 bridge vendor: 0x1022 bridge device: 0x7006 agp status: 0xF000203 agp mode: 0x100 aperture base: 0xE8000000 aperture size: 0x2000000 (33554432) pages bound: 0 xsvc: mmap'd 33554432 bytes at offset 0xE8000000, agp = 0x4014a000 xsvc: MTRR type 0x1 set xsvc: Starting bandwidth test, 3 tests, interval = 5 secs. 0: Wrote 671088640 (640MB) bytes/sec. (200 iterations of 16777216 bytes) 1: Wrote 671088640 (640MB) bytes/sec. (200 iterations of 16777216 bytes) 2: Wrote 671088640 (640MB) bytes/sec. (200 iterations of 16777216 bytes) xsvc: Average: 671088640 (640MB) bytes/sec. These numbers with my "poor" system: 1 GHz Athlon II (0.18�m, L1 128K, L2 512K) MSI MS-6167 Rev 1.0B (AMD 750, Irongate C4, without bypass) 640 MB PC100-2-2-2 SDRAM Raw memory speed is like below: nuetzel/Entwicklung> ./athlon Athlon test program $Id: fast.c,v 1.6 2000/09/23 09:05:45 arjan Exp $ clear_page() tests clear_page function 'warm up run' took 23164 cycles per page clear_page function '2.4 non MMX' took 15233 cycles per page clear_page function '2.4 MMX fallback' took 15606 cycles per page clear_page function '2.4 MMX version' took 13053 cycles per page clear_page function 'faster_clear_page' took 6753 cycles per page clear_page function 'even_faster_clear' took 5488 cycles per page copy_page() tests copy_page function 'warm up run' took 33096 cycles per page copy_page function '2.4 non MMX' took 45402 cycles per page copy_page function '2.4 MMX fallback' took 48318 cycles per page copy_page function '2.4 MMX version' took 33428 cycles per page copy_page function 'faster_copy' took 13702 cycles per page copy_page function 'even_faster' took 14667 cycles per page nuetzel/Entwicklung> ./athlon2 998.074 MHz clear_page by 'normal_clear_page' took 13718 cycles (284.2 MB/s) clear_page by 'slow_zero_page' took 14503 cycles (268.8 MB/s) clear_page by 'fast_clear_page' took 11882 cycles (328.1 MB/s) clear_page by 'faster_clear_page' took 5408 cycles (720.9 MB/s) copy_page by 'normal_copy_page' took 22165 cycles (175.9 MB/s) copy_page by 'slow_copy_page' took 22942 cycles (169.9 MB/s) copy_page by 'fast_copy_page' took 13715 cycles (284.3 MB/s) copy_page by 'faster_copy' took 6008 cycles (648.9 MB/s) copy_page by 'even_faster' took 5995 cycles (650.2 MB/s) Last but not least: The noticed tdfx driver slowdown from 2.x over tdfx-3-5-branch to tdfx-4-0-branch (current trunk) is Mesa related like Keith told me. Regards, Dieter -- Dieter N�tzel Graduate Student, Computer Science University of Hamburg Department of Computer Science @home: [EMAIL PROTECTED] _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
