quake-devel — Main discussion forum. Watch the fur fly.

Re: Huzzah fore ye Softwaere Renderyre

[Bob <ma...@mb...>]

>>	Yes I know software rendering is what folks might call "legacy" support and a good argument could be made for improving it as little as necessary, but I am curious to find out if anyone thinks it would be worthwhile to implement some of the basic software rendering enhancements found in Quake2 in the Quake1 based QuakeForge engine.

>>	Specifically I am thinking of transparent water and skyboxes, since these being OpenGL-only features complicates the art path of future games using the Quake engine (which is what I'm working on.)

>>	I can't dismiss the need for software rendering because I just happen to have one of those graphics cards which does not have freeware or open source 3d acceleration drivers (I think commercial drivers are more expensive than the card itself now and I don't know how well they work).

>>	Would the Quake2 source have any useful code for implementing these features, or are the engines just too different?



It's not just drivers but that so many Open Source GL projects (nobody
here though, surely... ahem) use methods that aren't supported by (and
can even cause serious problems with) many consumer level graphics
cards. So if you are sticking with the Q1 codebase (instead of
hybridizing the two engines or migrating your enhancements to the Q2
source foundation), I think it would be a pretty good idea to look at
the Q2 software renderer.

Merging features of Q2 is apparrently possible, since I see hybids like
MHQuake <http://mhquake.quakesrc.org> popping up already. I wouldn't
expect it to be a simple task, though, as both engines' software
rendering are primarily hand-optimized Assembly.

--

[patch] quake2 cross-platform fixes

[Jamie W. <ja...@sp...>]

This patch against the current quake2 head removes architecture specific
code, allowing one to build on things like mipsel and arm.

commit log entry:
=3D=3D=3D=3D=3D=3D
- removes architecture specific code from Makefile, linux/sys_linux.c
- fixes build errors on alpha machines with -Werror, cleaned up some
  typecasting
- typo correction in ctf/g_save.c
- Used README.sdl from relnev's 0.0.8 code (to match the rest of the sync)
- SVGALib code doesn't use <asm/io.h>
=3D=3D=3D=3D=3D=3D

diff -u -r1.6 .cvsignore
--- .cvsignore	2002/01/07 05:02:03	1.6
+++ .cvsignore	2002/01/25 05:31:48
@@ -3,6 +3,6 @@
 *.opt
 *.plg
 .vimrc
-debugi386
+debug*
 quake2
-releasei386
+release*
diff -u -r1.6 Makefile
--- Makefile	2002/01/12 02:14:07	1.6
+++ Makefile	2002/01/25 05:31:50
@@ -11,7 +11,7 @@
 # Here are your build options, no more will be added!
 # (Note: not all options are available for all platforms).
 # quake2 (uses OSS for sound, cdrom ioctls for cd audio) is automatically =
built.
-# game{i386,axp,ppc}.so is automatically built.
+# game$(ARCH).so is automatically built.
 BUILD_SDLQUAKE2=3DYES	# sdlquake2 executable (uses SDL for cdrom and sound)
 BUILD_SVGA=3DYES		# SVGAlib driver. Seems to work fine.
 BUILD_X11=3DYES		# X11 software driver. Works somewhat ok.
@@ -22,48 +22,46 @@
 BUILD_CTFDLL=3DYES		# gamei386.so for ctf
 # i can add support for building xatrix and rogue libs if needed
=20
+# unportable
 # Check OS type.
-OSTYPE :=3D $(shell uname -s)
+#OSTYPE :=3D $(shell uname -s)
+#
+#ifneq ($(OSTYPE),Linux)
+#$(error OS $(OSTYPE) is currently not supported)
+#endif
=20
-ifneq ($(OSTYPE),Linux)
-$(error OS $(OSTYPE) is currently not supported)
-endif
-
 # this nice line comes from the linux kernel makefile
-ARCH :=3D $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc64/ -e s/=
arm.*/arm/ -e s/sa110/arm/ -e s/alpha/axp/)
+ARCH :=3D $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc/ -e s/sp=
arc64/sparc/ -e s/arm.*/arm/ -e s/sa110/arm/ -e s/alpha/axp/)
+#ARCH :=3D $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc64/ -e s=
/arm.*/arm/ -e s/sa110/arm/ -e s/alpha/axp/)
=20
-ifneq ($(ARCH),i386)
-ifneq ($(ARCH),axp)
-ifneq ($(ARCH),ppc)
-ifneq ($(ARCH),sparc)
-$(error arch $(ARCH) is currently not supported)
-endif
-endif
-endif
-endif
+# This is preventing builds on sparc and ia64, etc
+#ifneq ($(ARCH),i386)
+#ifneq ($(ARCH),axp)
+#ifneq ($(ARCH),ppc)
+#$(error arch $(ARCH) is currently not supported)
+#endif
+#endif
+#endif
=20
 CC=3Dgcc
=20
-ifeq ($(ARCH),axp)
-RELEASE_CFLAGS=3D$(BASE_CFLAGS) -O2 -ffast-math -funroll-loops \
+# make this more port friendly
+#ifeq ($(ARCH),axp)
+RELEASE_CFLAGS=3D$(BASE_CFLAGS) -ffast-math -funroll-loops \
 	-fomit-frame-pointer -fexpensive-optimizations
-endif
+#endif
+#
+#ifeq ($(ARCH),ppc)
+#RELEASE_CFLAGS=3D$(BASE_CFLAGS) -ffast-math -funroll-loops \
+#	-fomit-frame-pointer -fexpensive-optimizations
+#endif
=20
-ifeq ($(ARCH),ppc)
-RELEASE_CFLAGS=3D$(BASE_CFLAGS) -O2 -ffast-math -funroll-loops \
-	-fomit-frame-pointer -fexpensive-optimizations
-endif
-
-ifeq ($(ARCH),sparc)
-RELEASE_CFLAGS=3D$(BASE_CFLAGS) -O2 -ffast-math -funroll-loops \
-	-fomit-frame-pointer -fexpensive-optimizations
-endif
-
 ifeq ($(ARCH),i386)
-RELEASE_CFLAGS=3D$(BASE_CFLAGS) -O2 -ffast-math -funroll-loops -malign-loo=
ps=3D2 \
-	-malign-jumps=3D2 -malign-functions=3D2
+#RELEASE_CFLAGS=3D$(BASE_CFLAGS) -O2 -ffast-math -funroll-loops -malign-lo=
ops=3D2 \
+#	-malign-jumps=3D2 -malign-functions=3D2 -g
+RELEASE_CFLAGS+=3D-O2 -malign-loops=3D2 -malign-jumps=3D2 -malign-function=
s=3D2 -g
 # compiler bugs with gcc 2.96 and 3.0.1 can cause bad builds with heavy op=
ts.
-#RELEASE_CFLAGS=3D$(BASE_CFLAGS) -O6 -march=3Di686 -ffast-math -funroll-lo=
ops \
+#RELEASE_CFLAGS=3D$(BASE_CFLAGS) -O6 -m486 -ffast-math -funroll-loops \
 #	-fomit-frame-pointer -fexpensive-optimizations -malign-loops=3D2 \
 #	-malign-jumps=3D2 -malign-functions=3D2
 endif
@@ -85,8 +83,12 @@
 CTF_DIR=3D$(MOUNT_DIR)/ctf
 XATRIX_DIR=3D$(MOUNT_DIR)/xatrix
=20
-BASE_CFLAGS=3D-Dstricmp=3Dstrcasecmp -Wall -Werror
+BASE_CFLAGS=3D-Dstricmp=3Dstrcasecmp -Wall -Werror -pipe
=20
+ifneq ($(ARCH),i386)
+  BASE_CFLAGS+=3D-DC_ONLY
+endif
+
 DEBUG_CFLAGS=3D$(BASE_CFLAGS) -g
=20
 LDFLAGS=3D-lm -ldl
@@ -114,11 +116,12 @@
 SHLIBCFLAGS=3D-fPIC
 SHLIBLDFLAGS=3D-shared
=20
-DO_CC=3D$(CC) $(CFLAGS) -o $@ -c $<
-DO_SHLIB_CC=3D$(CC) $(CFLAGS) $(SHLIBCFLAGS) -o $@ -c $<
-DO_GL_SHLIB_CC=3D$(CC) $(CFLAGS) $(SHLIBCFLAGS) $(GLCFLAGS) -o $@ -c $<
-DO_AS=3D$(CC) $(CFLAGS) -DELF -x assembler-with-cpp -o $@ -c $<
-DO_SHLIB_AS=3D$(CC) $(CFLAGS) $(SHLIBCFLAGS) -DELF -x assembler-with-cpp -=
o $@ -c $<
+# added $(ARCH) here as make would munge the quoting in BASE_CFLAGS above
+DO_CC=3D$(CC) -DARCH=3D\""$(ARCH)"\" $(CFLAGS) -o $@ -c $<
+DO_SHLIB_CC=3D$(CC) -DARCH=3D\""$(ARCH)"\" $(CFLAGS) $(SHLIBCFLAGS) -o $@ =
-c $<
+DO_GL_SHLIB_CC=3D$(CC) -DARCH=3D\""$(ARCH)"\" $(CFLAGS) $(SHLIBCFLAGS) $(G=
LCFLAGS) -o $@ -c $<
+DO_AS=3D$(CC) -DARCH=3D\""$(ARCH)"\" $(CFLAGS) -DELF -x assembler-with-cpp=
 -o $@ -c $<
+DO_SHLIB_AS=3D$(CC) -DARCH=3D\""$(ARCH)"\" $(CFLAGS) $(SHLIBCFLAGS) -DELF =
-x assembler-with-cpp -o $@ -c $<
=20
 ##########################################################################=
###
 # SETUP AND BUILD
@@ -132,74 +135,16 @@
  TARGETS +=3D $(BUILDDIR)/ctf/game$(ARCH).$(SHLIBEXT)
 endif
=20
-ifeq ($(ARCH),axp)
  ifeq ($(strip $(BUILD_SDLQUAKE2)),YES)
   TARGETS +=3D $(BUILDDIR)/sdlquake2
  endif
-
- ifeq ($(strip $(BUILD_SVGA)),YES)
-  $(warning Warning: SVGAlib support not supported for $(ARCH))
- endif
-
- ifeq ($(strip $(BUILD_X11)),YES)
-  $(warning Warning: X11 support not supported for $(ARCH))
- endif
-
- ifeq ($(strip $(BUILD_GLX)),YES)
-  $(warning Warning: support not supported for $(ARCH))
- endif
-
- ifeq ($(strip $(BUILD_FXGL)),YES)
-  $(warning Warning: FXGL support not supported for $(ARCH))
- endif
=20
- ifeq ($(strip $(BUILD_SDL)),YES)
-  $(warning Warning: SDL support not supported for $(ARCH))
- endif
-
- ifeq ($(strip $(BUILD_SDLGL)),YES)
-  $(warning Warning: SDLGL support not supported for $(ARCH))
- endif
-endif # ARCH axp
-
-ifeq ($(ARCH),ppc)
- ifeq ($(strip $(BUILD_SDLQUAKE2)),YES)
-  $(warning Warning: SDLQuake2 not supported for $(ARCH))
- endif
-=20
- ifeq ($(strip $(BUILD_SVGA)),YES)
-  $(warning Warning: SVGAlib support not supported for $(ARCH))
- endif
-
- ifeq ($(strip $(BUILD_X11)),YES)
-  $(warning Warning: X11 support not supported for $(ARCH))
- endif
-
- ifeq ($(strip $(BUILD_GLX)),YES)
-  $(warning Warning: GLX support not supported for $(ARCH))
- endif
-
- ifeq ($(strip $(BUILD_FXGL)),YES)
-  $(warning Warning: FXGL support not supported for $(ARCH))
- endif
-
- ifeq ($(strip $(BUILD_SDL)),YES)
-  $(warning Warning: SDL support not supported for $(ARCH))
- endif
-
- ifeq ($(strip $(BUILD_SDLGL)),YES)
-  TARGETS +=3D $(BUILDDIR)/ref_sdlgl.$(SHLIBEXT)
- endif
-endif # ARCH ppc
-
+# svgalib just doesn't work with non-i386 at the moment, so don't bother
 ifeq ($(ARCH),i386)
- ifeq ($(strip $(BUILD_SDLQUAKE2)),YES)
-  TARGETS +=3D $(BUILDDIR)/sdlquake2
- endif
-
  ifeq ($(strip $(BUILD_SVGA)),YES)
   TARGETS +=3D $(BUILDDIR)/ref_soft.$(SHLIBEXT)
  endif
+endif
=20
  ifeq ($(strip $(BUILD_X11)),YES)
   TARGETS +=3D $(BUILDDIR)/ref_softx.$(SHLIBEXT)
@@ -220,7 +165,6 @@
  ifeq ($(strip $(BUILD_SDLGL)),YES)
   TARGETS +=3D $(BUILDDIR)/ref_sdlgl.$(SHLIBEXT)
  endif
-endif # ARCH i386
=20
 all: build_debug build_release
=20
@@ -307,11 +251,11 @@
 	$(BUILDDIR)/client/cd_sdl.o \
 	$(BUILDDIR)/client/snd_sdl.o
=20
-ifeq ($(ARCH),axp)
-QUAKE2_AS_OBJS =3D  #blank
+# more i386 asm
+ifeq ($(ARCH),i386)
+QUAKE2_AS_OBJS =3D $(BUILDDIR)/client/snd_mixa.o
 else
-QUAKE2_AS_OBJS =3D \
-	$(BUILDDIR)/client/snd_mixa.o
+QUAKE2_AS_OBJS =3D  #blank
 endif
=20
 $(BUILDDIR)/quake2 : $(QUAKE2_OBJS) $(QUAKE2_LNX_OBJS) $(QUAKE2_AS_OBJS)
@@ -401,7 +345,7 @@
 $(BUILDDIR)/client/files.o :      $(COMMON_DIR)/files.c
 	$(DO_CC)
=20
-$(BUILDDIR)/client/mdfour.o :     $(COMMON_DIR)/mdfour.c
+$(BUILDDIR)/client/mdfour.o :        $(COMMON_DIR)/mdfour.c
 	$(DO_CC)
=20
 $(BUILDDIR)/client/net_chan.o :   $(COMMON_DIR)/net_chan.c
@@ -1020,7 +964,12 @@
 	$(BUILDDIR)/ref_soft/r_scan.o \
 	$(BUILDDIR)/ref_soft/r_sprite.o \
 	$(BUILDDIR)/ref_soft/r_surf.o \
-	\
+	$(BUILDDIR)/ref_soft/q_shared.o \
+	$(BUILDDIR)/ref_soft/q_shlinux.o \
+	$(BUILDDIR)/ref_soft/glob.o
+
+ifeq ($(ARCH),i386)
+REF_SOFT_OBJS +=3D \
 	$(BUILDDIR)/ref_soft/r_aclipa.o \
 	$(BUILDDIR)/ref_soft/r_draw16.o \
 	$(BUILDDIR)/ref_soft/r_drawa.o \
@@ -1031,11 +980,8 @@
 	$(BUILDDIR)/ref_soft/math.o \
 	$(BUILDDIR)/ref_soft/d_polysa.o \
 	$(BUILDDIR)/ref_soft/r_varsa.o \
-	$(BUILDDIR)/ref_soft/sys_dosa.o \
-	\
-	$(BUILDDIR)/ref_soft/q_shared.o \
-	$(BUILDDIR)/ref_soft/q_shlinux.o \
-	$(BUILDDIR)/ref_soft/glob.o
+	$(BUILDDIR)/ref_soft/sys_dosa.o
+endif
=20
 REF_SOFT_SVGA_OBJS =3D \
 	$(BUILDDIR)/ref_soft/rw_svgalib.o \
@@ -1283,6 +1229,6 @@
 	$(REF_GL_OBJS)
=20
 distclean:
-	@-rm -rf $(BUILD_DEBUG_DIR) $(BUILD_RELEASE_DIR)
-	@-rm -f `find . \( -not -type d \) -and \
+	-rm -rf $(BUILD_DEBUG_DIR) $(BUILD_RELEASE_DIR)
+	-rm -f `find . \( -not -type d \) -and \
 		\( -name '*~' \) -type f -print`
diff -u -r1.4 cl_cin.c
--- client/cl_cin.c	2002/01/07 04:54:10	1.4
+++ client/cl_cin.c	2002/01/25 05:31:52
@@ -412,7 +412,7 @@
=20
 	if (input - in.data !=3D in.count && input - in.data !=3D in.count+1)
 	{
-		Com_Printf ("Decompression overread by %i", (input - in.data) - in.count=
);
+		Com_Printf ("Decompression overread by %i", (int) (input - in.data) - in=
.count);
 	}
 	out.count =3D out_p - out.data;
=20
diff -u -r1.4 cl_scrn.c
--- client/cl_scrn.c	2002/01/12 02:14:08	1.4
+++ client/cl_scrn.c	2002/01/25 05:31:54
@@ -614,11 +614,11 @@
 	*/
 	if ( a->model =3D=3D b->model )
 	{
-		return ( ( int ) a->skin - ( int ) b->skin );
+		return (long int) a->skin - (long int) b->skin;
 	}
 	else
 	{
-		return ( ( int ) a->model - ( int ) b->model );
+		return (long int) a->model - (long int) b->model;
 	}
 }
=20
diff -u -r1.3 snd_mem.c
--- client/snd_mem.c	2002/01/12 02:14:08	1.3
+++ client/snd_mem.c	2002/01/25 05:31:55
@@ -253,7 +253,7 @@
 		memcpy (str, data_p, 4);
 		data_p +=3D 4;
 		iff_chunk_len =3D GetLittleLong();
-		Com_Printf ("0x%x : %s (%d)\n", (int)(data_p - 4), str, iff_chunk_len);
+		Com_Printf ("0x%p : %s (%d)\n", data_p - 4, str, iff_chunk_len);
 		data_p +=3D (iff_chunk_len + 1) & ~1;
 	} while (data_p < iff_end);
 }
diff -u -r1.8 g_save.c
--- ctf/g_save.c	2002/01/12 03:23:32	1.8
+++ ctf/g_save.c	2002/01/25 05:31:57
@@ -175,7 +175,7 @@
 //ZOID
 //This game.dll only supports deathmatch
 	if (!deathmatch->value) {
-		gi.dprintf("Forcing deathmatch.");
+		gi.dprintf("Forcing deathmatch.\n");
 		gi.cvar_set("deathmatch", "1");
 	}
 	//force coop off
diff -u -r1.1 README.sdl
--- docs/README.sdl	2002/01/12 02:14:08	1.1
+++ docs/README.sdl	2002/01/25 05:31:58
@@ -11,14 +11,34 @@
 Be sure to install SDL 1.2 (http://www.libsdl.org) if you want to use the
 softsdl or sdlgl drivers, or the sdlquake2 binary.
=20
-You can change what drivers you wish to build by editing the Makefile and
-changing the BUILD_ lines at the very top.
-
 'make' will, by default, build both the debug and release files.
 To build fully optimized binaries: make build_release
 The resulting binaries are then put in releasei386.
=20
=20
+Makefile options:
+-----------------
+(quake2 and gamei386.so are always built, but the following options can be
+ changed by editing the Makefile)
+BUILD_SDLQUAKE2		Build sdlquake2, a quake2 binary that uses SDL for
+                        CD audio and sound access (default =3D YES).
+BUILD_SVGA              Build ref_soft.so, a quake2 video driver that uses
+                        SVGAlib (default =3D NO).
+BUILD_X11               Build ref_softx.so, a quake2 video driver that uses
+                        X11 (default =3D YES).
+BUILD_GLX               Build ref_glx.so, a quake2 video driver that uses
+                        X11's GLX (default =3D YES).
+BUILD_FXGL              Build ref_gl.so [might be renamed to fxgl later],=
=20
+                        a quake2 video driver that uses fxMesa (default =3D
+                        NO).  This option is currently untested because I =
do=20
+                        not have a Voodoo 1 or 2.
+BUILD_SDL               Build ref_softsdl.so, a quake2 video driver that
+                        uses SDL (default =3D YES).
+BUILD_SDLGL             Build ref_sdlgl.so, a quake2 video driver that uses
+                        OpenGL with SDL (default =3D YES).
+BUILD_CTFDLL            Build the Threewave CTF gamei386.so (default =3D N=
O).
+
+
 To install the Quake2 gamedata:
 -------------------------------
 (installdir is wherever you want to install quake2, and cdromdir is wherev=
er
@@ -59,11 +79,20 @@
 Configuration files and such are saved in ~/.quake2/, so <installdir> can =
be
 made read-only or whatever.
=20
-WARNING: Please do not make quake2 or any of the libraries suid root!
+WARNING: Please do not make quake2 or any of the libraries suid root!  Doi=
ng
+so is at your own risk.
=20
-NOTE: Save games will most likely not work across different versions or
-builds (this is due to how savegames were stored).
+NOTE: Save games will not work across different versions or builds, because
+of the way they are stored.
=20
+Dedicated Server:
+-----------------
+If there is a demand for it, I can add support for an explicit q2ded binar=
y.
+Else, using +set dedicated 1 should be fine.
+
+Joystick Support:
+-----------------
+None yet.
=20
 Commonly used commands:
 -----------------------
@@ -80,28 +109,63 @@
 snd_restart             // restart sound driver
 basedir <dir>           // point quake2 to where the data is
 gl_driver <libGL.so>    // point quake2 to your libGL
+dedicated 1             // run quake2 as a dedicated server
+
+When using these commands on the quake2 command line, use +set to cause the
+variables be set before the config files are loaded (important for
+gl_driver). e.g.
+./quake2 +set vid_ref glx +set gl_driver /usr/lib/libGL.so.1
+
+If quake2 crashes when trying to load an OpenGL based driver (glx, sdlgl),
+make sure its not loading the wrong libGL.
=20
+Have a NVIDIA card and it _still_ crashes? Try=20
+export LD_PRELOAD=3D/usr/lib/libGL.so, and run quake2 again.
=20
+Is lighting slow in OpenGL (while firing, explosions, etc.)? Disable
+multitexturing (gl_ext_multitexture 0; vid_restart).
+
+
 Website:
 --------
-I'll post any updates I make at http://www.icculus.org/quake/=20
+I'll post any updates I make at http://www.icculus.org/quake2/=20
 (which currently redirects to http://www.icculus.org/~relnev/)
=20
+Mailing List:
+-------------
+to subscribe: send a blank email to qua...@ic...
+to post: send email to qu...@ic...
+
 Anonymous CVS access:
 ---------------------
 cvs -d:pserver:ano...@ic...:/cvs/cvsroot login
       (password is "anonymous" without the quotes.)
 cvs -z3 -d:pserver:ano...@ic...:/cvs/cvsroot co quake2
=20
-Questions:
-----------
-What's the best way of handling international keyboards with SDL?
+Bugzilla:
+---------
+https://bugzilla.icculus.org
=20
 TODO:
 -----
+Try out RCG's key idea.
 Fix save games.
+Verify FXGL works.
+Joystick support.
+Fullscreen/DGA support in X11 driver.
+Fully switch to glext.h.
 Suggestions, anyone?
=20
+v0.0.8: [01/04/02]
+-------
++ Fixed C-only ref_soft building.
++ SDL CD audio looping fix (Robert B=E4uml)
++ ~/.quake2/<game> added to the search path for mods. (Ludwig Nussel)
++ Minor change to fix compilation with OpenGL 1.3 headers.
++ Fixed changing video drivers using the menu.
++ Fixed autoexec.cfg on startup.
++ Sparc Linux support (Vincent Cojot)
+
 v0.0.7: [12/28/01]
 -------
 + Merged in Quake2 3.21 source.
@@ -150,7 +214,10 @@
 Stephen Anthony
 William Aoki
 Robert B=E4uml
+Vincent Cojot
+Michel D=E4nzer
 Ryan C. Gordon
 Ludwig Nussel
+Peter van Paassen
 Zachary 'zakk' Slater
 Matti Valtonen
diff -u -r1.4 rw_in_svgalib.c
--- linux/rw_in_svgalib.c	2002/01/03 04:22:46	1.4
+++ linux/rw_in_svgalib.c	2002/01/25 05:32:00
@@ -26,8 +26,6 @@
 #include <signal.h>
 #include <sys/mman.h>
=20
-#include <asm/io.h>
-
 #include "vga.h"
 #include "vgakeyboard.h"
 #include "vgamouse.h"
diff -u -r1.4 rw_svgalib.c
--- linux/rw_svgalib.c	2002/01/12 02:14:08	1.4
+++ linux/rw_svgalib.c	2002/01/25 05:32:00
@@ -42,8 +42,6 @@
 #include <unistd.h>
 #include <sys/mman.h>
=20
-#include <asm/io.h>
-
 #include "vga.h"
 #include "vgakeyboard.h"
 #include "vgamouse.h"
diff -u -r1.7 sys_linux.c
--- linux/sys_linux.c	2002/01/12 02:14:08	1.7
+++ linux/sys_linux.c	2002/01/25 05:32:01
@@ -40,7 +40,7 @@
 #include <dlfcn.h>
=20
 #include "../qcommon/qcommon.h"
-
+#include "../game/game.h"
 #include "../linux/rw_linux.h"
=20
 cvar_t *nostdout;
@@ -214,22 +214,13 @@
 */
 void *Sys_GetGameAPI (void *parms)
 {
-	void	*(*GetGameAPI) (void *);
+	game_export_t *(*GetGameAPI) (game_import_t *);
=20
 	char	name[MAX_OSPATH];
 	char	*path;
 	char	*str_p;
-#if defined __i386__
-	const char *gamename =3D "gamei386.so";
-#elif defined __alpha__
-	const char *gamename =3D "gameaxp.so";
-#elif defined __powerpc__
-	const char *gamename =3D "gameppc.so";
-#elif defined __sparc__
-	const char *gamename =3D "gamesparc.so";
-#else
-#error Unknown arch
-#endif
+
+	const char *gamename =3D "game"ARCH".so";
=20
 	setreuid(getuid(), getuid());
 	setegid(getgid());
@@ -265,7 +256,7 @@
 		}
 	}
=20
-	GetGameAPI =3D (void *)dlsym (game_library, "GetGameAPI");
+	GetGameAPI =3D (game_export_t * (*)(game_import_t *)) dlsym(game_library,=
 "GetGameAPI");
=20
 	if (!GetGameAPI)
 	{
diff -u -r1.9 common.c
--- qcommon/common.c	2002/01/12 02:14:09	1.9
+++ qcommon/common.c	2002/01/25 05:32:05
@@ -69,7 +69,7 @@
 static int	rd_buffersize;
 static void	(*rd_flush)(int target, char *buffer);
=20
-void Com_BeginRedirect (int target, char *buffer, int buffersize, void (*f=
lush))
+void Com_BeginRedirect (int target, char *buffer, int buffersize, void (*f=
lush)(int, char*))
 {
 	if (!target || !buffer || !buffersize || !flush)
 		return;
diff -u -r1.8 qcommon.h
--- qcommon/qcommon.h	2002/01/12 03:23:32	1.8
+++ qcommon/qcommon.h	2002/01/25 05:32:08
@@ -735,7 +735,7 @@
 #define	PRINT_ALL		0
 #define PRINT_DEVELOPER	1	// only print when "developer 1"
=20
-void		Com_BeginRedirect (int target, char *buffer, int buffersize, void (*=
flush));
+void		Com_BeginRedirect (int target, char *buffer, int buffersize, void (*=
flush)(int, char*));
 void		Com_EndRedirect (void);
 void 		Com_Printf (char *fmt, ...) __attribute__((format(printf,1,2)));
 void 		Com_DPrintf (char *fmt, ...) __attribute__((format(printf,1,2)));
diff -u -r1.2 r_edge.c
--- ref_soft/r_edge.c	2001/12/22 21:49:58	1.2
+++ ref_soft/r_edge.c	2002/01/25 05:32:11
@@ -1074,7 +1074,7 @@
=20
 		// make a stable color for each surface by taking the low
 		// bits of the msurface pointer
-		D_FlatFillSurface (s, (int)s->msurf & 0xFF);
+		D_FlatFillSurface (s, (long int) s->msurf & 0xFF);
 		D_DrawZSpans (s->spans);
 	}
 }
diff -u -r1.2 r_surf.c
--- ref_soft/r_surf.c	2001/12/22 21:49:58	1.2
+++ ref_soft/r_surf.c	2002/01/25 05:32:12
@@ -450,7 +450,7 @@
 	if ((size <=3D 0) || (size > 0x10000))
 		ri.Sys_Error (ERR_FATAL,"D_SCAlloc: bad cache size %d\n", size);
 =09
-	size =3D (int)&((surfcache_t *)0)->data[size];
+	size =3D (long int)&((surfcache_t *)0)->data[size];
 	size =3D (size + 3) & ~3;
 	if (size > sc_size)
 		ri.Sys_Error (ERR_FATAL,"D_SCAlloc: %i > cache size of %i",size, sc_size=
);

--=20
ja...@sp...                           http://spacepants.org/jaq.gpg
=20
When C++ is your hammer, everything looks like a thumb
        -- Latest seen from Steven M. Haflich, in c.l.l

Hello Coders!

[<42...@in...>]

NEWS FROM THE OPENSOURCE HIDDEN FACE OF MOON

Hello Friends, Girlfriends and Coders.

I am Tei, "Oscar Vives" at berlios.de and in Real Life[tm]. 

My engine modder plattform project as moved to CVS here:
  http://cvs.berlios.de/cgi-bin/cvsweb.cgi/src/tj19/?cvsroot=telejano#dirlist

You can download all the code, and specific "WFX Tech Test" here:
  ftp://ftp.berlios.de/pub/telejano/

The main services (mail list, bugs, report, expect features) will be here:
 https://developer.berlios.de/projects/telejano/

 - - -

My target its:
- WfX - Wheater FX 
- .jpg loading 
- .png loading
- engine text editor (builtin )
- more gui features ( client side gui, scripted gui events )
- conversational engine ( .cml files )
- rpg games support features ( groupware for quake players, persistent profiles)
- rts games support features ( more sst mods, chess engine builtin )
- fps games support features ( more qrally mods )
- racer games support features ( patches, terrain levels )
- others modders features...

This prj target is to be a good mod platform, with exciting features from the 
latest engines, like Unreal2, Quake3, Operation Flashpoint, etc..

Theoricaly, you can join our efforts:
 CVS, coding.
Mail list, ideas. 
 FTP/Forums, playing betas, and sending feedback.
Bugs
 Sending feedback.

If you need more about "the big picture", get you browser point and fire to 
 qbism.telefragged.com/tei
 qbism.telefragged.com/tei/airshots1.shtml
 qbism.telefragged.com/tei/airshots2.shtml
 qbism.telefragged.com/tei/airshots3.shtml
 qbism.telefragged.com/tei/fxshots1.shtml

Download, play, compile, code :D  as you wish ;)

1 saludo
Tei

---------------------------------------------
This message was sent using Endymion MailMan.
http://www.endymion.com/products/mailman/

Re: Quakeforge Homepage?

[Bill C. <bi...@ta...>]

On Tue, Jan 22, 2002 at 07:55:38PM -0500, BL...@ao... wrote:
> Has anyone noticed the news has dissapeared, and the most recent news is from 
> Nov 2000?

Hmm, if I had a buck for every time that was asked, beer would be on me (ok, not
a lot of beer, probably a sip, but still:)

The news is back now. We had some sql issues (the news was actually
deliberately broken). A security problem was pointed out to us and the website
was promptly denied access to the database till things could be worked around.

Bill
-- 
Leave others their otherness. -- Aratak

Quakeforge Homepage?

[<BL...@ao...>]

Has anyone noticed the news has dissapeared, and the most recent news is from 
Nov 2000?

Re: "reload" crashes server :)

[Bill C. <bi...@ta...>]

On Sun, Jan 20, 2002 at 07:51:50PM -0600, Ryan Underwood wrote:
> Hi,
> 
> Using QF 0.5, bootstrapped last night, if a player does the action "reload"
> in a Teamfortress game (2.8 or 2.9), the server dies saying "Program error" or something
> to that effect.  Everything else seems to work okay.
> 
> A fix?

sv_old_entity_free 1

put that in your server.cfg and things will be fine

Bill
-- 
Leave others their otherness. -- Aratak

"reload" crashes server :)

[Ryan U. <ne...@ic...>]

Hi,

Using QF 0.5, bootstrapped last night, if a player does the action "reload"
in a Teamfortress game (2.8 or 2.9), the server dies saying "Program error" or something
to that effect.  Everything else seems to work okay.

A fix?

Thanks

-- 
Ryan Underwood, <nemesisATicequakeDOTnet>, jabber=same, icq=10317253
                                    http://www.icequake.net/~nemesis

 |=  icequake networks, ltd.  =|=  university of missouri rolla  =|
 |=   system administration   =|=       computer science         =|

Re: toggle console problem

[Bill C. <bi...@ta...>]

On Fri, Jan 18, 2002 at 07:33:38AM -0000, Colin Thompson wrote:
> Hi all,
> 
> I noticed a problem with binding a key to toggle the console. the
> 'traditional' key is usually tilde but it only brings the console down -
> after that the keypress is is captured by the console code and won't send it
> back up again....

This is actually by design.

> I'm not sure if this is classed as a bug or not, but I noticed that Taniwha
> is working with the console code so I thought this list would be more
> appropriate...

Yeah, I've been busy of late (finally got back into the swing of things:).

> Should the console only be capturing certain keys? or everything?

Well, the console /is/ capturing the keys. They're just not bound to any
action. qf 0.5 has a rather powerful key binding system (thanks to Mercury)
that can drasticly simplify config scripts. Rather than the one key binding
table you're used to, qf has eighteen: one for the console (IMT_CONSOLE) and
17 for in-game (IMT_0 to IMT_16). There is also IMT_DEFAULT, but that just
maps to IMT_0. The way to bind keys in this system is (eg): in_bind
imt_console k_backquote toggleconsole (there, even answered your main
question:). IMT -> input mapping table. Here's another, longer example:

-----8<-----

exec configs/common/alias/zoom.cfg
bind mwheeldown z_out
bind mwheelup z_in
init__zoom.cfg

in_bind imt_1 m_wheel_down "impulse 12"
in_bind imt_1 m_wheel_up "impulse 10"
in_bind imt_0 k_capslock +goto_imt_1
// unbound keys in imt_[1-16] fall through to imt_0
alias +goto_imt_1 "imt imt_1"
alias -goto_imt_1 "imt imt_0"

-----8<-----

Now, bind acts as a compatability layer taking the old bind syntax (and key
names), converting them to the new using the imt specified in the in_bind_imt
cvar (which defaults to imt_default (which points to imt_0: the main game
imt)). The imt command selects the current imt that the input engine is to
check for a binding.

What this setup does is give me variable zoom using my mouse wheel by default.
When I hold down the capslock key, the wheel cycles though my weapons.

As stated in the comment in the config, if the input engine doesn't find a
binding for a key in tables 1-16, it will look in table 0. If it still doesn't
find a binding, it drops the key.

imt_console is /totally/ independent of imt 0-16. /any/ time you leave the
game, imt_console is made the `default' and imt_0 to imt_16 are ignored. In
game, imt_0 is default with imt_1 to imt_16 being available.

Ways to leave the game: bring down the console (or have it forced down via
disconnect); go to the menu (they're not functional yet, but I'm making
progress); messagemode1 and messagemode2.

HTH
Bill

PS: incase nobodie's noticed: chat now has input history. just use the up/down
arrow keys.
-- 
Leave others their otherness. -- Aratak

toggle console problem

[Colin T. <col...@su...>]

Hi all,

I noticed a problem with binding a key to toggle the console. the
'traditional' key is usually tilde but it only brings the console down -
after that the keypress is is captured by the console code and won't send it
back up again....

I'm not sure if this is classed as a bug or not, but I noticed that Taniwha
is working with the console code so I thought this list would be more
appropriate...

Should the console only be capturing certain keys? or everything?

Colin

Re: use of engine

[Seth G. <sga...@li...>]

On Thu, 17 Jan 2002, adam petronaitis wrote:

> Hi; I would like to use a GL quake engine (win32) for a DOOM game; but
> there are many source ports to chose from. Could Quakefourge be a
> single-player engine as well?(maybe with some code cut n paste from
> GLQuake?)

I don't know about all the past releases of QuakeForge but the QuakeForge
code in CVS includes both single player "nq" (i.e. "Normal Quake") and
multiplayer-only ("QuakeWorld" or "qw") engines.

What kind of DOOM game do you have in mind?  You could run into some
copyright snags there.
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_

use of engine

[adam p. <apc...@op...>]

Hi; I would like to use a GL quake engine (win32) for a DOOM game; but =
there are many source ports to chose from.
Could Quakefourge be a single-player engine as well?(maybe with some =
code cut n paste from GLQuake?)

Re: loading tga textures

[Ulrich G. <uga...@li...>]

On Tue, 15 Jan 2002, Seth Galbraith wrote:

> 
> (Did I mention that it would also be nice to be able to draw 3D objects in
> menus and HUDs?)

Did you!??!!?  What I wanna know is why didn't I!?!?!  Its been nagging me
for years!  Why can't i load a model as a menu graphic?  Like using models
for a pointing device,  or just nice graphics.  Imagine choosing your
character in a menu and being able to see the actual model...  Not hard to
imagine?  That's because quake is the only engine (exaggeration) that
doesn't have this.  

*<><><><><><><><><><><><><><><><><><><>*
//     ZANCRUS: Ulrich Galbraith      \\
\\     E-Mail:  uga...@kr...      //
//                                    \\
\\     ICQ:              3093981      //
*<><><><><><><><><><><><><><><><><><><>*

>   __    __    _                                    _    __    __
> _/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
>  \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
> _/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_
> 
> 
> 
> 
> -- 
> QuakeForge Development	[ qua...@li... ]
> https://lists.sourceforge.net/lists/listinfo/quake-devel
> 

loading tga textures

[Seth G. <sga...@li...>]

I have implemented loading tga textures for sprites, gfx.wad images, .lmp
files, and textures from bsp files.  The last type were trickiest because
you have to load the tga files after the whole bsp file is loaded, since
both LoatTGA and the bsp file loader use Hunk_TempAlloc.

Actually the implementation of tga texture loading for bsp files is quite
simple.  Just look at the Mod_LoadExternalTextures function.

Alias model (.mdl) skins will probably need something similar
(Mod_LoadExternalSkins?) and I think sprites need it too - it's probably
just dumb luck that it didn't crash my single frame sprite test case.

I'm a little confused about how to get the sprite and alias model (.spr
and .mdl) file names, skin/frame names and gl_texnums (i.e. the info I
need to load the tga textures) from the model_t *mod in Mod_RealLoadModel.

If I knew how to do that, I could fix up sprites and implement tga skins
pretty easily.


Of course some folks will see tga textures as an obvious cheat (though
your textures are in a pak file it is less obvious.)  It has been
suggested that checksums might discourage the casual cheater.  I'm not
able to implement something like that, but I don't think I've done
anything that would make it particularly difficult to add.


It's also been suggested that generating external texture names by
slapping ".tga" on the end is a bit sloppy.  I suppose the way to avoid
that is a bit different for each case.

I think textures should be in standard image formats where possible (i.e:
.tga, .pcx or .png) with a 256 color fallback option if needed for 8 bit
software rendering.

I know standard image formats help casual cheaters, but they also
lubricate the art path and to me that's a lot more important (there are
lots of cheaters but not many artists.)

I also understand that external textures don't fit everybody's idea of a
nice map, sprite or model format.  But I think it is actually better than
putting the textures in the model/map/sprite file because the textures can
be re-used in different models/map/sprites.

So, what I would suggest is making the image loading part of the client
side QC code, so for example the csQC would say something like "try to use
gfx/conback.tga for the console background, then fall back to
gfx/conback.pcx if that doesn't work" - not in plain english of course.  
This way you wouldn't get any weird results if somebody does something
silly like trying to use a 32 bit tga for the non-image .lmp files.

(Did I mention that it would also be nice to be able to draw 3D objects in
menus and HUDs?)
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_

[Openquartz-general] lighting skewed faces

[<42...@in...>]

Seth you hare a wizard! :)

 Checkout our new engine particle (tomazquake based) engine here:
 http://qbism.telefragged.com/tei/fxshots1.shtml

 This work generating particles on sky and water surfaces. I think can be 
applied to light. Any iddea?


1 saludo
Tei

 - - -

I am working on an enhanced light.exe for Quake 1 map compiling, and I
want it to properly light faces even when the two texture axes are not
perpendicular.  (Which won't happen in most editors, but can be done in
QuArK.)

Unfortunately light.exe seems to only pay attention to one texture axis
when casting light onto a skewed texture.  Perhaps a screenshot will
demonstrate this better than my description:

http://www.planetquake.com/gg/quarktx.gif

 * the back wall texture is skewed like this: /_/
 * the light on the back wall should look nice and round
 * the light on the near cube is okay because it isn't so skewed
 * the light on the yellow thing is slightly messed up in parts

I suspect that the problem is located in the ltface.c file of the
light.exe source code, most likely in the LightFace or CalcFaceVectors
function.

Any idea how I can fix this?

The modified light program I'm working on and the test map can be found
at http://sourceforge.net/projects/openquartz
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_



---------------------------------------------
This message was sent using Endymion MailMan.
http://www.endymion.com/products/mailman/

Transparency on map brushes

[Ulrich G. <uga...@li...>]

I'm working on some quake engine projects that require texture
transparency on map brushes.

When is QuakeForge going to support this?

*<><><><><><><><><><><><><><><><><><><>*
//     ZANCRUS: Ulrich Galbraith      \\
\\     E-Mail:  uga...@kr...      //
//                                    \\
\\     ICQ:              3093981      //
*<><><><><><><><><><><><><><><><><><><>*

lighting skewed faces

[Seth G. <sga...@li...>]

I am working on an enhanced light.exe for Quake 1 map compiling, and I
want it to properly light faces even when the two texture axes are not
perpendicular.  (Which won't happen in most editors, but can be done in
QuArK.)

Unfortunately light.exe seems to only pay attention to one texture axis
when casting light onto a skewed texture.  Perhaps a screenshot will
demonstrate this better than my description:

http://www.planetquake.com/gg/quarktx.gif

 * the back wall texture is skewed like this: /_/
 * the light on the back wall should look nice and round
 * the light on the near cube is okay because it isn't so skewed
 * the light on the yellow thing is slightly messed up in parts

I suspect that the problem is located in the ltface.c file of the
light.exe source code, most likely in the LightFace or CalcFaceVectors
function.

Any idea how I can fix this?

The modified light program I'm working on and the test map can be found
at http://sourceforge.net/projects/openquartz
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_

[PATCH] more buttons for your mouse

[John F. <jo...@fr...>]

I have a Micro$oft Intellimouse Optical which has a total of 7 buttons
if you count up down on the wheel as two. QuakeForge X11 did not
support buttons 6 and 7. 

Here is a patch against quakeforge-current (downloaded on 6 January)
to let Quakeforge support up to mouse 32 buttons. Only the X11 backend
was updated. The patch compiled and worked first time so there is
probably something wrong with it ;-)

I also more or less got rid of M_WHEEL_{UP,DOWN}, you might want to do
things differently.

Re: Warnings when compiling Quake2 source

[Bill C. <bi...@ta...>]

On Sun, Jan 06, 2002 at 12:24:01AM -0800, Todd Kirby wrote:
> I got a couple uninitalized variable warnings when I
> compile the quake2 source on my Linux 7.1 athlon box
> (gcc 2.96). 

Thanks, I've committed the fixes.

I hope quake2 works better with 2.96 than quake1 does.

Bill
-- 
Leave others their otherness. -- Aratak

Warnings when compiling Quake2 source

[Todd K. <ki...@ya...>]

I got a couple uninitalized variable warnings when I
compile the quake2 source on my Linux 7.1 athlon box
(gcc 2.96). 

Warnings are:

game/g_combat.c: In function `CheckPowerArmor':
game/g_combat.c:179: warning: `power' might be used
uninitialized in this function
make[1]: *** [releasei386/game/g_combat.o] Error 1
make[1]: Leaving directory `/home/tkirby/quake2'
make: *** [build_release] Error 2


ctf/g_combat.c: In function `CheckPowerArmor':
ctf/g_combat.c:179: warning: `power' might be used
uninitialized in this function
make[1]: *** [releasei386/ctf/g_combat.o] Error 1
make[1]: Leaving directory `/home/tkirby/quake2'
make: *** [build_release] Error 2


The following patch applied to quake2/game/g_combat.c
and quake2/ctf/g_combat.c fixed the problems:


179c179
< 	int			power;
---
> 	int			power = 0;




-Todd Kirby



=====


__________________________________________________
Do You Yahoo!?
Send FREE video emails in Yahoo! Mail!
http://promo.yahoo.com/videomail/

model rendering

[Seth G. <sga...@li...>]

I fixed a little performance problem my high precision code caused in the
software renderer by moving the if out of the for loop (just like GL)

Here are some suggestions to improve model rendering, but I have to
get to work on my Hack and Slash mod, so I don't have time to do them.

If anybody wants to implement any of these, I can try to help out:

 * lerp (tween/smooth) mdl animation in software as in gl
   Just implement this as another special case in sw_ralias.c
   like the high precision case, with (gl_lerp_anim->int_val)
   instead of (pmdl->ident == POLYHEADER16) and of course you
   will want to rename gl_lerp_anim or create a seperate cvar
   when the cvar is off, lerping won't affect performance

 * make software and gl alias model rendering share more code
   (also reducing redundancy between sw and sw32) this is a
   good idea because they actually both do the same stuff,
   just organized differently (e.g. in sw high precision data
   is treated like extra verts, in gl it's like extra poses.)

 * reorganize high precision data at load time so that it is
   formatted as 16 bit ints instead of 2 non-sequential bytes.
   this will affect the layout of light normals too.

 * avoid calling realloc a lot for finalverts and auxverts by
   only calling it after all the models are loaded and you know
   how many vertices the most complex model has.  In the worst
   case scenario you'll allocate a several extra poses worth of
   extra data for a model that never appears.
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_

high precision models

[Seth G. <sga...@li...>]

I fixed a bit of slowdown my high precison gl rendering code was adding by
moving the if statements out of the per-vertex loops.

I removed the hardcoded vertex limit by changing R_AliasDrawModel in the
software renderers to "realloc" the finalverts and auxverts array instead
of creating these arrays each time the function is called.

Seems like that should result in a speedup except in exceptional cases
(like drawing a bunch of different models for the first time in order from
fewest to most vertices.)
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_

high precision model GL rendering

[Seth G. <sga...@li...>]

High precision model rendering is implemented in OpenGL now.
http://www.planetquake.com/gg/v_rock2a.zip (no hassle direct download)
is a replacement for the rocket launcher model.  To try it out:

extract the model to <your quakeforge directory>/hp/progs
rename it "v_rock2.mdl"
run "<any single player qf executable> -game hp"
enter the "map start" and "impulse 9" console commands
now you should have the high precision rocket launcher equipped

Now I want to remove the last vestiges of hard coded vertex limits from
sw_ralias.c and sw32_ralias.c.  In both cases it's these two temporary
arrays in R_AliasDrawModel.  In fact here's the whole function:

#define MAXALIASVERTS 1024
void
R_AliasDrawModel (alight_t *plighting)
{
	finalvert_t finalverts[MAXALIASVERTS +
		((CACHE_SIZE - 1) / sizeof (finalvert_t)) + 1];
	auxvert_t auxverts[MAXALIASVERTS];
	r_amodels_drawn++;
	// cache align
	pfinalverts = (finalvert_t *)
		(((long) &finalverts[0] + CACHE_SIZE - 1) & ~(CACHE_SIZE - 1));
	pauxverts = &auxverts[0];
	paliashdr = Cache_Get (&currententity->model->cache);
	pmdl = (mdl_t *) ((byte *) paliashdr + paliashdr->model);
	R_AliasSetupSkin ();
	R_AliasSetUpTransform (currententity->trivial_accept);
	R_AliasSetupLighting (plighting);
	R_AliasSetupFrame ();
	if (!currententity->colormap)
		Sys_Error ("R_AliasDrawModel: !currententity->colormap");
	acolormap = currententity->colormap;
	if (acolormap == &vid.colormap8 && r_pixbytes != 1)
	{
		if (r_pixbytes == 2)
			acolormap = vid.colormap16;
		else if (r_pixbytes == 4)
			acolormap = vid.colormap32;
		else
			Sys_Error("R_AliasDrawmodel: unsupported r_pixbytes %i\n",
				r_pixbytes);
	}
	if (currententity != r_view_model)
		ziscale = (float) 0x8000 *(float) 0x10000;
	else
		ziscale = (float) 0x8000 *(float) 0x10000 *3.0;
	if (currententity->trivial_accept && pmdl->ident != POLYHEADER16)
		R_AliasPrepareUnclippedPoints ();
	else
		R_AliasPreparePoints ();
	Cache_Release (&currententity->model->cache);
}

I'd like these arrays to be either dynamically sized or resized,
with no hardcoded limit.  What is the best way to do that?
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_

high precision model software rendering

[Seth G. <sga...@li...>]

I have commited the code that renders high precison models in
software.  GL rendering should be easy, and fortunately my sister's car
broke down and I can't drive so I don't have to fix her computer tommorow
and can work on QF instead :-)

Also I want to remove the last vestiges of hard coded vertex limits from
sw_ralias.c and sw32_ralias.c.  In both cases it's these two temporary
arrays in R_AliasDrawModel.  In fact here's the whole function:

#define MAXALIASVERTS 1024
void
R_AliasDrawModel (alight_t *plighting)
{
	finalvert_t finalverts[MAXALIASVERTS +
		((CACHE_SIZE - 1) / sizeof (finalvert_t)) + 1];
	auxvert_t auxverts[MAXALIASVERTS];
	r_amodels_drawn++;
	// cache align
	pfinalverts = (finalvert_t *)
		(((long) &finalverts[0] + CACHE_SIZE - 1) & ~(CACHE_SIZE - 1));
	pauxverts = &auxverts[0];
	paliashdr = Cache_Get (&currententity->model->cache);
	pmdl = (mdl_t *) ((byte *) paliashdr + paliashdr->model);
	R_AliasSetupSkin ();
	R_AliasSetUpTransform (currententity->trivial_accept);
	R_AliasSetupLighting (plighting);
	R_AliasSetupFrame ();
	if (!currententity->colormap)
		Sys_Error ("R_AliasDrawModel: !currententity->colormap");
	acolormap = currententity->colormap;
	if (acolormap == &vid.colormap8 && r_pixbytes != 1)
	{
		if (r_pixbytes == 2)
			acolormap = vid.colormap16;
		else if (r_pixbytes == 4)
			acolormap = vid.colormap32;
		else
			Sys_Error("R_AliasDrawmodel: unsupported r_pixbytes %i\n",
				r_pixbytes);
	}
	if (currententity != r_view_model)
		ziscale = (float) 0x8000 *(float) 0x10000;
	else
		ziscale = (float) 0x8000 *(float) 0x10000 *3.0;
	if (currententity->trivial_accept && pmdl->ident != POLYHEADER16)
		R_AliasPrepareUnclippedPoints ();
	else
		R_AliasPreparePoints ();
	Cache_Release (&currententity->model->cache);
}

I'd like these arrays to be either dynamically sized or resized,
with no hardcoded limit.  What is the best way to do that?
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_

Re: high precision models

[Seth G. <sga...@li...>]

On Fri, 4 Jan 2002, Bill Currie wrote:

> I was having a look at that last night. By using a memory format that combines
> the highprec verts into 16 bit words rather tyan bytes, that asm function
> would only need a trivial change for highprec. It would, however, need
> duplicating.
> 
> Lerping, on the other hand, would be a bit trickier. I haven't come up with
> anything for that yet. (I hadn't realised sw alias model rendering was /that/
> hairy).

I think it's just a case of the right conditional statements in the right
places, i.e. if interpolation is off, and high precision is not needed,
then use the old function, otherwise use a new function.

high precision models would not HAVE to have their own asm optimizations -
at least not right away (and getting this implemented quickly is important
to me as I have other things to work on) but someone could optimize that
side of things later (perhaps once the have more high precision models to
test it with!)
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_

Re: high precision models

[Bill C. <bi...@ta...>]

On Fri, Jan 04, 2002 at 08:38:44AM -0800, Seth Galbraith wrote:
> I think we'll have to drop that assembler function (or at least make it
> conditional) to get either interpolation or high precision models.  How
> big of a difference does this assembler function make?

I was having a look at that last night. By using a memory format that combines
the highprec verts into 16 bit words rather tyan bytes, that asm function
would only need a trivial change for highprec. It would, however, need
duplicating.

Lerping, on the other hand, would be a bit trickier. I haven't come up with
anything for that yet. (I hadn't realised sw alias model rendering was /that/
hairy).

Bill
-- 
Leave others their otherness. -- Aratak