Am Donnerstag, 6. Dezember 2001 13:50 schrieb Keith Whitwell:
> Dieter Nützel wrote:
> > Oh,
> >
> > I've forgotten to send you a little screenshot.
> > It shows some texture/clipping errors, too.
>
> I don't have this, and I haven't been able to see anything similar in other
> programs. Can you retest?
>
> > > Q3A 1.30 (final)
> > > is running at the same speed as with the trunk.
> > > But some texture problems?
> > > The Intro screen and all "Cinematics" are playing but do not show up.
>
> Fixed.
Yes. I get 30 fps with in fullscreen, now.
Hope, we get S3TC sometime...
...maybe with SLI... --- 60 fps were GREAT...
System here:
1 GHz Athlon II SlotA (0,18µm, mmx fxsr syscall mmxext 3dnowext 3dnow)
MSI MS-6167 Rev 1.0B (AMD Irongate C4 without bypass)
640 MB PC100-2-2-2 SDRAM
AHA-2940 UW
IBM UW/U160 disks
Voodoo5 5500 AGP
> > > PARSEC V0.99 build 0196
> > > is running but some texture are missing (on the space ships and some of
> > > the cockpit environment.
>
> Again, I don't have this - can you retest?
Not, tried with PARSEC V0.99 build 0196 (shipped with SuSE 7.3) and 0197
update from November. Trunk worked with 0196.
> > > Now I see a slowdown with ipers, gloss (teapot mode) and with VTK's
> > > Simple Sphere Benchmark V
> > > 2.1/2.3 (yes, V 2.1 is faster, it use a bigger sphere per default).
> > >
> > > trunk 3-5 4-0
> > > -----------------------------------------------------------------------
> > >---- -------------- gears (fps) 770 766 770
> > >
> > > ipers (fps) 15-16 not tested 11-12 :-(
> > > texture disabled, no fog, but I see no difference if I swith fog on or
> > > off 80.000-100.000 fewer Poly/sec
1 fps faster
> > >
> > > gloss (fps) 57 not tested 50
> > > teapot
2 frames/s faster, now
> > >
> > > sph V 2.1 1670 1500 1580
> > > sph V 2.3 1370 1250 1340
> > > kpolys/s
little faster, too 1613 kpolys/s
VTK/Local> vtk sphere-bench.tcl-2.1 &
[1] 18330
VTK/Local> ***** XMesaCloseFullScreen *****
[1] Fertig vtk sphere-bench.tcl-2.1
graphics/examplesTcl> vtk TestLargeTextures.tcl
Mesa implementation error: tdfx driver: extreme texmem fragmentation
Please report to the DRI bug database at dri.sourceforge.net
tdfxTMAllocTexMem returned NULL! unit=0 size=16777216
> > >
> > > Is there a texture rendering slowdown?
>
> These don't seem too bad except for ipers, which is probably a Mesa issue.
> I'll look at these after the merge to the trunk.
Some more polys, now (30.000).
******************************************************
Now it is time for optimization.
Let's start with fixing the remaining Glide3 texture bug when 3DNow! is
enabled. When ever I try to compile a Glide3 lib with debug enabled it
sigfault. Can I reach Daryll anywhere? Any advice?
Mesa/demos> texdown
GL_VENDOR = VA Linux Systems, Inc.
GL_VERSION = 1.2 Mesa 4.0.1
GL_RENDERER = Mesa DRI 20010624 Voodoo4 x86/MMX/3DNow!
Speicherschutzverletzung (core dumped)
[-]
Reading symbols from /usr/X11R6/lib/libXxf86dga.so.1...done.
Loaded symbols for /usr/X11R6/lib/libXxf86dga.so.1
Reading symbols from /usr/X11R6/lib/libXxf86vm.so.1...done.
Loaded symbols for /usr/X11R6/lib/libXxf86vm.so.1
#0 0x466adac6 in _trisetup_3DNow_win_nocull_valid ()
from /usr/lib/libglide3.so
(gdb) bt
#0 0x466adac6 in _trisetup_3DNow_win_nocull_valid ()
from /usr/lib/libglide3.so
#1 0x00000003 in ?? ()
Error accessing memory address 0x3: No such process.
(gdb) info registers
eax 0x466adac0 1181407936
ecx 0x0 0
edx 0x40 64
ebx 0x468f5060 1183797344
esp 0xbfffeb68 0xbfffeb68
ebp 0x3 0x3
esi 0x468f5020 1183797280
edi 0x40 64
eip 0x466adac6 0x466adac6
eflags 0x210202 2163202
cs 0x23 35
ss 0x2b 43
ds 0x2b 43
es 0x2b 43
fs 0x2b 43
gs 0x2b 43
fctrl 0x7f 127
fstat 0x0 0
ftag 0x0 0
fiseg 0x0 0
fioff 0x0 0
foseg 0x0 0
fooff 0x0 0
fop 0x0 0
xmm0 0xffffffffffffffffffffffffffffffff
xmm1 0xffffffffffffffffffffffffffffffff
xmm2 0xffffffffffffffffffffffffffffffff
xmm3 0xffffffffffffffffffffffffffffffff
xmm4 0xffffffffffffffffffffffffffffffff
xmm5 0xffffffffffffffffffffffffffffffff
xmm6 0xffffffffffffffffffffffffffffffff
xmm7 0xffffffffffffffffffffffffffffffff
mxcsr 0x1f80 8064
******************************************************
Second:
You can not build Glide3 with 3DNow! acceleration the "normal" way:
./chores.3dfx --clean --generate --configure=--enable-amd3d --build
'cause something in the build system (configuration) is broken. If you do it
this way you will get a much bigger lib in which one token is undefined.
SunWave1>cd build/h5/glide3/src/.libs/
Directory: /tmp/INSTALL/SOURCE/Glide3/build/h5/glide3/src/.libs
SunWave1>l
insgesamt 1071
drwxr-xr-x 2 nuetzel users 248 Okt 12 23:57 .
drwxr-xr-x 4 nuetzel users 1712 Okt 12 23:57 ..
-rw-r--r-- 1 nuetzel users 583204 Okt 12 23:57 libglide3.a
lrwxrwxrwx 1 nuetzel users 15 Okt 12 23:57 libglide3.la ->
../libglide3.la
-rw-r--r-- 1 nuetzel users 706 Okt 12 23:57 libglide3.lai
lrwxrwxrwx 1 nuetzel users 19 Okt 12 23:57 libglide3.so ->
libglide3.so.3.10.0lrwxrwxrwx 1 nuetzel users 19 Okt 12 23:57
libglide3.so.3 -> libglide3.so.3.10.0
-rwxr-xr-x 1 nuetzel users 501449 Okt 12 23:57 libglide3.so.3.10.0
SunWave1>nm libglide3.so.3.10.0 |grep 3DNow
00032a40 T _grDrawTriangles_3DNow
0003353d T _grDrawVertexList_3DNow_Clip
00033180 T _grDrawVertexList_3DNow_Window
U _grTexDownload_3DNow_MMX
00032040 T _trisetup_3DNow_clip_cull_invalid
00032040 T _trisetup_3DNow_clip_cull_valid
00032040 T _trisetup_3DNow_clip_nocull_invalid
00032040 T _trisetup_3DNow_clip_nocull_valid
00032280 T _trisetup_3DNow_win_cull_invalid
00032540 T _trisetup_3DNow_win_cull_valid
00032040 T _trisetup_3DNow_win_nocull_invalid
00032800 T _trisetup_3DNow_win_nocull_valid
You have to do it the "old" way:
setenv FX_GLIDE_HW h5
setenv GL_AMD3D 1
make -f makefile.linux
SunWave1>l /usr/lib/libglide3.so*
lrwxrwxrwx 1 root root 16 Okt 12 04:19 /usr/lib/libglide3.so
-> libglide3.so.3.1
lrwxrwxrwx 1 root root 16 Okt 12 04:20
/usr/lib/libglide3.so.3
-> libglide3.so.3.1
-rwxr-xr-x 1 root root 354355 Sep 27 08:53
/usr/lib/libglide3.so.3.1
SunWave1>nm /usr/lib/libglide3.so.3.1|grep 3DNow
00035d00 T _grDrawTriangles_3DNow
000367fd T _grDrawVertexList_3DNow_Clip
00036440 T _grDrawVertexList_3DNow_Window
000351c0 T _grTexDownload_3DNow_MMX
00035300 T _trisetup_3DNow_clip_cull_invalid
00035300 T _trisetup_3DNow_clip_cull_valid
00035300 T _trisetup_3DNow_clip_nocull_invalid
00035300 T _trisetup_3DNow_clip_nocull_valid
00035540 T _trisetup_3DNow_win_cull_invalid
00035800 T _trisetup_3DNow_win_cull_valid
00035300 T _trisetup_3DNow_win_nocull_invalid
00035ac0 T _trisetup_3DNow_win_nocull_valid
******************************************************
Third:
In summer 2000/2001 (???) before Joseph Kain fixed the "pixel smearing bug" I
had a Glide3 which gave nearly 4 times (!!!) for fill rate with "tenmillion"
(courtusy of SGI). See below.
That is very equal to results I get with the beta version of the latest
Accelerated-X server (with OpenGL) from Xig. Sadly they have dropped the
VSA-100 support from there final demo version, which you can grep from there
site.
There (final) Radeon numbers outperform Windows at a glance.
They have a different (OSS) AGPGART package which I've installed on my system.
I have taken it from there final release...;-)
SunWave1 xig/xsvc# ./xsvctest
xsvc: Version Major: 3 Minor: 0
bridge vendor: 0x1022 bridge device: 0x7006
agp status: 0xF000203 agp mode: 0xF000302
aperture base: 0xE8000000 aperture size: 0x2000000 (33554432)
pages bound: 0
xsvc: mmap'd 33554432 bytes at offset 0xE8000000, agp = 0x4014f000
xsvc: MTRR: Enabled Write Combining
xsvc: Starting bandwidth test, 3 tests, interval = 5 secs.
0: Wrote 667733196 (636MB) bytes/sec. (199 iterations of 16777216 bytes)
1: Wrote 667733196 (636MB) bytes/sec. (199 iterations of 16777216 bytes)
2: Wrote 667733196 (636MB) bytes/sec. (199 iterations of 16777216 bytes)
xsvc: Average: 667733196 (636MB) bytes/sec.
Not BAD, I think???
Now some benchmarks with there 3Daccelx_V2-2.0BETA-57 version.
Some are slower and had bugs (gears, isosurf, trispd, gloss and gltestperf)
but some are remarkably faster ("tenmillion" is nearly tenmillion and my VTK
benchmark had 2100 vs 1670, sorry only from memory).
SunWave1>./glinfo
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
SunWave1>./gears
2004 frames in 5.000 seconds = 400.800 FPS
2345 frames in 5.000 seconds = 469.000 FPS
2349 frames in 5.000 seconds = 469.800 FPS
2356 frames in 5.001 seconds = 471.106 FPS
SunWave1>./isosurf
7179 vertices, 7177 triangles
Vertex arrays not supported by this renderer
Nr unique vertex/normal pairs: 2723
num_tri_verts: 21531
primitive (0x10000): GL_TRIANGLE_STRIP,
render style (0x100): glVertex,
enabling normal arrays
new flags (0xa93169): glVertex, GL_TRIANGLE_STRIP, lit,
Benchmarking...
Result: triangles/sec: 732903 fps: 102.118
new flags (0x2a92169): glVertex, GL_TRIANGLE_STRIP, as a displaylist, lit,
rebuilding displaylist
Benchmarking...
Result: triangles/sec: 1.98058e+06 fps: 275.962
SunWave1>./trispd
For options: ./trispd -help
Dither: on
ShadeModel: smooth
DepthTest: off
Size: 50 pixels
RedBits: 8 GreenBits: 8 BlueBits: 8
Rate: 1557504 tri in 2.009s = 775263 tri/s 38763165 pixels/s
Rate: 3115008 tri in 3.383s = 920783 tri/s 46039136 pixels/s
Rate: 3115008 tri in 3.411s = 913224 tri/s 45661213 pixels/s
Rate: 3115008 tri in 3.228s = 964996 tri/s 48249814 pixels/s
Rate: 3115008 tri in 3.246s = 959645 tri/s 47982255 pixels/s
Rate: 3115008 tri in 3.292s = 946236 tri/s 47311786 pixels/s
Rate: 3115008 tri in 3.304s = 942799 tri/s 47139951 pixels/s
SunWave1>./gloss
482 frames in 5.007 seconds = 96.2652 FPS
573 frames in 5.003 seconds = 114.531 FPS
574 frames in 5.005 seconds = 114.685 FPS
575 frames in 5.006 seconds = 114.862 FPS
SunWave1>./tenmillion
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
So enough for the moment.
Yours,
Dieter
_______________________________________________
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel
