RADIANCE Digest Volume 1, Number 5



Dear Radiance Users,

For those of you still with us after last week's fiasco, here is a culling
of electronic mail exchanged between me and some of you over the last month.
Once again, I have given headings by subject to make it easier to browse
without reading everything.

A couple of reminders.  First, you may pick up previous issues of the
Radiance Digest at hobbes.lbl.gov (128.3.12.38) with anonymous ftp
from the pub/digest directory.  Second, you must write to me if you want
your name removed from this mailing list -- please take a few minutes now
to express your outrage at getting such unwelcomed junk mail so that
the tension doesn't build up and make the blood vessels on your forehead
bulge out in an unbecoming fashion.

I particularly recommend looking at the section on the Radiance picture
format to anyone interested in translating Radiance images.

-Greg

	MAC		- MacIntosh and Radiance
	PICTURE		- Radiance Picture file format
	HPUX		- Hewlett-Packard UNIX and Radiance 1.4
	DAYLIGHT	- Radiance and Daylight Simulation
	AMIGA		- Radiance on the Amiga 2000
	COLORPICT	- Using the Colorpict primitive
	DOS		- Radiance under DOS?
	RVIEW		- Rview and memory
	LIGHTS		- Non-standard Light Sources

============================
MAC	MacIntosh programs with Radiance and A/UX

Date: Fri, 23 Aug 91 13:07:34 NZT
From: pdbourke%ccu1.aukuni.ac.nz@Csa2.lbl.gov
Subject: Fractal Landscape generator
To: GJWard@Csa2.lbl.gov

I've just written the "old fractal landscape" generator using the subdivision
technique (as opposed to spectral techniques). A Mac II application!

It does the following:
- surface parameters, x,y range and typical z variation
- roughness parameter
- sea (lake) level
- seed specification, lets the same landscape be generated on request
- 3 point colour ramp mapped to height
- DXF, Super3D, RayShade, and Radiance
- view on the Mac screen of wireframe coloured model
- variable resolution up to 256x256 cells
- automatic triangulation of non-planar facets

------------------------------
Paul D. Bourke
pdbourke@ccu1.aukuni.ac.nz


Date: Sun, 25 Aug 91 15:54:59 EDT
From: David Brainard  
To: greg@lesosun1.epfl.ch
Subject: Re:  Mac version

I see.  I will definitely have a Big Mac (as it were)
and am trying to decide whether to load it with A/UX
or not.  You are the first person I've had any contact
with who seems to be using it.  Do you find it a reasonable
Unix?  Do you find that native Mac applications run on
top of it without too many problems?   My prior was that
running Unix on the MAC would provide the worst of both
worlds, rather than the best.  But I am willing to be 
convinced otherwise.  Thanks for any advice you can give.

David

Date: Mon, 26 Aug 91 09:40:37 +0200
From: greg (Greg Ward)
To: brainard@cvs.rochester.edu
Subject: Re:  Mac version

Hi David,

My advice is free, and it's worth what you pay for it!

I won't claim that A/UX is the best of both worlds.  I have certainly
had my share of frustrations trying to get around little annoying
compiler bugs and poor virtual memory performance on the UNIX side,
but I've dealt with worse UNIX implementations, certainly.  The
worst thing I can say about A/UX is that it's a System V derivative,
with all the associated problems.  Berkeley, Berkeley, yeah, yeah, yeah!

As for the Mac side, I haven't tried it out with every software package.
I have tried Microsoft Word 4.0, which seems fairly stable, MacDraw II,
Adobe Photoshop (excellent software in my opinion), Aldus Freehand,
Studio/8, Versaterm, Mathematica, HyperCard, and a few others I can't
think of right now.  I haven't tried Excel or any database systems, but
WingZ seems to work (though I haven't used it really at all).  Programs
that I've tried which failed are MacWrite, SuperPaint and Architrion II.
Sometimes you'll get a warning if the 32-bit clean bit isn't set on the
application and it will run anyway, but only if the code actually is
32-bit clean and they just forgot to set the bit, as is the case with
HyperCard 2.0.  In general, 24-bit applications don't work, even in the
special 24-bit mode A/UX Finder.  I don't see this as being much of a
drawback, since all applications have to be 32-bit clean to run under
System 7, anyway.

The other loss (that I don't really consider a loss) is all the
OS utilities and INIT's and so forth that work under the MacOS.
You will be very sorry if you install any of these under A/UX.
Also, specialty programs for formatting hard drives and other
hardware-specific tasks must be run under the regular OS.

Keep in mind that installing A/UX doesn't mean giving up your
regular MacIntosh.  All you really give up is about 100 Mbytes of
disk space somewhere.  You can still run without A/UX, only
booting it when you have some UNIX program you need to run.
Also, since A/UX is cognizant of the Mac volumes, it's not
necessary to install your Mac applications or data in two places.

In conclusion, I think the A/UX development team have done a
decent job getting these two very different systems to cooperate
with each other, and I've been generally satisfied with the
improvements that come with each new release -- something I
can't say for the Sun operating system!

-Greg

[A sad postscript to this message -- it seems that the new Adobe
	Photoshop version 2.0 doesn't like A/UX.]

================================
PICTURE		Radiance Picture Format

Date: Fri, 13 Sep 91 16:56:02 NZT
From: russells@ccu1.aukuni.ac.nz
Subject: Radiance picture format
To: greg@lesosun1.epfl.ch

Hi,

Can you supply the format of the Radiance picture files, as produced
by rpict etal. I am planning to write an Macintosh application
to display them, using MacTCP to bring them from the Unix system
without intermediate FTPing and conversions.

Also, is it possible for Radiance to do parametric models. For instance
you set a variable and use instances of that variable throughout
your model files. This would be good for my Super 3D (a Mac modeller)
to Radiance translator for handling "one-pixel wide" lines.

Thanks in advance,

Russell Street
russells@ccu1.aukuni.ac.nz
(arch2 in a former life)

Date: Fri, 13 Sep 91 09:19:31 +0200
From: greg (Greg Ward)
To: russells@ccu1.aukuni.ac.nz
Subject: Re:  Radiance picture format

Hi Russell,

Thanks for writing the Super 3D translator.  Paul just wrote to me
about it, but I haven't had a chance yet to try it out.  I need to
find a copy of Super 3D first!

At the end of this mail I will put a shar file of the routines you
need to read and write Radiance pictures.  The format has been
enhanced slightly for the next release (in an upward compatible
way), so you should definitely use these newer routines.

The file format, like most binary files used by Radiance, contains
an ascii information header that is terminated by an empty line.
This header typically contains the commands used to generate the
file along with variables indicating exposure, view parameters, and
so on.  Next there is a single line that indicates the resolution and
pixel scanning order of the image.  For Radiance pictures, the pixels
are order as English text, left to right and top to bottom.  This is
indicated with a line of the form:

-Y M +X N

Where M and N are the y and x resolutions, respectively.  The x and y
image coordinates are always the same, starting with (0,0) at the lower
left corner, (N,0) at the lower right, and (0,M) at the upper left.
The y resolution appears first in our specification because it is the
major sort, and is preceded by a minus sign because it is decreasing in
the file.

Finally, the floating point scanlines follow.  Each pixel is represented by
at most 4 bytes.  The first three bytes are the red, green and blue mantissas
(in that order), and the fourth byte is a common exponent.  The floating point
color (R,G,B)=(1.,.5,.25) would be represented by the bytes (128,64,32,129).
The conversion back to floating point is possible using the ldexp() library
routine, or it's better to use the colr_color() routine included in color.c.

The scanlines are usually run-length encoded.  My previous scheme (release 1.4
and earlier) used a simple count for repeated pixels.  My new scheme is
more complicated and encodes the four components separately.  I don't
recommend writing your own routine to decode it -- use what's in color.c.

A skeletal program to read a Radiance picture file and convert to 24-bit
gamma-corrected color looks like this:

#include 
#include "color.h"
main(argc, argv)
int argc;
char *argv[];
{
	char *fname;		/* Radiance input file name */
	FILE *fp;		/* input stream pointer */
	int xres,yres;		/* x and y image resolution */
	int y;
	register COLR *scanin;
	register int x;

	if (argc < 2) {
		fp = stdin;
		fname = "";
	} else if ((fp = fopen(fname=argv[1], "r")) == NULL) {
		perror(fname);
		exit(1);
	}
	if (checkheader(fp, COLRFMT, NULL) < 0 ||
			fgetresolu(&xres, &yres, fp) != (YMAJOR|YDECR))
		fprintf(stderr, "%s: not a Radiance picture\n", fname);
		exit(1);
	}
	if ((scanin = (COLR *)malloc(xres*sizeof(COLR))) == NULL) {
		perror(argv[0]);
		exit(1);
	}
	setcolrgam(2.2);		/* set appropriate gamma correction */
	for (y = yres-1; y >= 0; y--) {
		if (freadcolrs(scanin, xres, fp) < 0) {
			fprintf(stderr, "%s: read error\n", fname);
			exit(1);
		}
		colrs_gambs(scanin, xres);
		for (x = 0; x < xres; x++) {
			/*
			 * Do something with the bytes:
			 *	scanin[x][RED]
			 *	scanin[x][GRN]
			 *	scanin[x][BLU]
			 */
		}
	}
	free((char *)scanin);
	exit(0);
}

You will find all the routines you need in ray/src/common.  The
checkheader() routine is in the module header.c, fgetresolu is in resolu.c,
freadcolrs() is in color.c, and setcolrgam() and colrs_gambs() are
in the module colrops.c.

If you want to convert the file to 8-bit color, the process is quite a
bit more complicated.  I suggest you take a look at the ra_pr program
in the ray/src/px directory to get a better idea of what is involved.

For communicating with another program across MacTCP, I suggest reading
and unpacking the scanlines on the remote (UNIX) host and transferring
fixed-length packets (perhaps a scanline at a time) to your MacIntosh
display program.

As for a macro facility to replace variables with values in a Radiance
scene file, Paul asked me also about this earlier in the context of
pixel-thick lines.  You can certainly use a macro program such as
cpp or m4 to replace variables with values in a file, but the real
problem is that pixel-thick lines do not exist.  They should be
converted to cylinders with a non-zero radius corresponding to the
physical objects the lines were meant to represent.  This may end
up being a pixel wide in the final image, but usually it is not.
Radiance actually has no notion of pixel size in its rendering
routines, so variable replacement with the pixel size is impossible
anyway.

-Greg

------------------------------
#! /bin/sh
# This is a shell archive, meaning:
# 1. Remove everything above the #! /bin/sh line.
# 2. Save the resulting text in a file.
# 3. Execute the file with /bin/sh (not csh) to create:
#	color.h
#	header.c
#	resolu.c
#	color.c
#	colrops.c
# This archive created: Fri Sep 13 09:19:19 1991
[Deleted for the sake of brevity.]

============================================
HPUX		Hewlett-Packard UNIX and Radiance 1.4

From: jaf@beauty.graphics.cornell.edu (James A. Ferwerda)
Subject: Radiance installation on HP9000/835
To: GJWard@Csa2.lbl.gov
Date: Thu, 29 Aug 91 14:53:05 EDT

Hi,

I'm a potentially new user of Radiance, but I'm having a little trouble
getting it to compile on my HP9000/835. Included below is a copy of the
output of the makeall script. I'm writing to you for assistance before
I make any major changes to the distributed version because I'd like to
remain compatible with any updates or bug fixes, and because I figure
you know your software better than I ever will and might know a quick
fix which would save me hours of hacking. I compiled the package using
the "HP workstation" option in the makeall script, and so far the only
code change I've made was to change the line

#include  

in malloc.c to

#include 

because there is no var.h in usr/include/sys on the HP system.

I'm not a hardware or a systems person (or even a very good programmer)
but to the best of my knowledge the HP9000/835 is a RISC based machine
running a modified version of System V. (But the makeall script gave
similar messages when I used the "Other" option and indicated a RISC based
non-BSD machine).

Any insights you can give me on how to get the package up and running
on my HP machine would be greatly appreciated. Radiance looks like a
great tool for my work on surface lightness/ illumination perception
and I'd really like to be able to use it. By the way, I first heard about
Radiance from Holly Rushmeier from Georgia Tech who gave it rave reviews.
Thanks in advance for any help, and keep up the good work.

-Jim Ferwerda
jaf@graphics.cornell.edu

Date: Fri, 30 Aug 91 10:38:00 +0200
From: greg (Greg Ward)
To: jaf@beauty.graphics.cornell.edu
Subject: Re:  Radiance installation on HP9000/835

Hello Jim,

OK, so I admit that I haven't actually compiled the distribution on all
these different machines.  I'm still running on a Sun 3/60 myself.

Thanks for sending me the compilation errors.  It really is the best
way for me to correct portability problems in the code.  I recommend
the following changes:

	1) Change the #ifdef lines to #ifndef BSD at:
		line 78 of src/cv/dxfcvt/config.h and
		line 22 of src/cal/calc/calc.c

	2) Change the COMPAT=malloc.o to COMPAT=bmalloc.o in:
		src/ot/Makefile and
		src/rt/Makefile

These fixes will appear in the next release, thanks to you.
-Greg

From greg Thu Sep  5 11:03:53 1991
Date: Thu, 5 Sep 91 11:03:52 +0200
From: greg (Greg Ward)
To: jaf@beauty.graphics.cornell.edu
Subject: Re: Radiance installation on HP9000/835
Status: R

Hi Jim,

I took a look at the core and oconv is hitting a bus error in the scanf()
procedure.  I suspect the problem is memory alignment, and that the HP
is a RISC machine and we need to define ALIGN=double for bmalloc to
be compiled properly.  I recommend doing this manually and I will
correct the entry for HP workstations in the makeall script for
the next distribution.  Do:

	% cd ray/src/ot
	% rm bmalloc.o
	% cc -O -DALIGN=double -c bmalloc.c
	% cd ../rt
	% rm bmalloc.o
	% cc -O -DALIGN=double -c bmalloc.c

Then rerun makeall from the top and the fixed compilations of bmalloc.o
will be included.  Hopefully, this will fix your problems.

-Greg

P.S.  I didn't do it myself because of file permissions obviously.

From daemon Fri Sep  6 00:24:34 1991
From: James A. Ferwerda 
Subject: Success!
To: greg@lesosun1.epfl.ch
Date: Thu, 5 Sep 91 18:26:33 EDT
Mailer: Elm [revision: 64.9]
Status: R


Greg, 

Thanks for the fixes. I'm off and running, working my way through the
tutorial. Thus far I'm really impressed with how comprehensive your 
package looks and how easy it's been to get things to come up. You've
really done a great job. I'll keep you posted on how things are coming
along. Thanks.

-Jim

From: nfotis%theseas.ntua.gr@Csa2.lbl.gov (Nikolaos)
Subject: Troubles with Radiance1R4
To: gjward@Csa2.lbl.gov (Greg Ward)
Date: Sat, 7 Sep 91 8:33:14 EET DST

Dear Mr. Ward,

I just set to play with Radiance, but with bad results
(on a HP-9000/720. Here's its `uname -a` output:
HP-UX kentayro A.B8.01 A 9000/720 29904182  )

The first trouble: malloc.c doesn't compile with the 5th option
of makeall:

cc: "malloc.c", line 270: error 1588: "v_pageshift" undefined.
cc: "malloc.c", line 270: error 1531: Invalid member of struct or union.
*** Error code 1

I did the following changes:

--
#ifndef NOVMEM
#ifndef BSD
/* For HP-UX 8.01, I changed the line:

#include 
to: 
#include 

ARRGH!! This damned machine has not .v_pageshift structure!?! .

I have checked the /usr/include directory, and much to my horror, I have found
various page sizes:

/usr/include > fgrep "page size" *
a.out.h:#define EXEC_PAGESIZE   4096 not always the same as the MMU page size

/usr/include/sys > fgrep "page size" *
sys/framebuf.h: locked page size = 2 ** 11 
sys/unistd.h:#  define _SC_PAGE_SIZE   3001  PAGE_SIZE: Software page size

/usr/include/machine > fgrep "page size" machine
machine/param.h:#define NBPG_PA83 2048      2kb page size for PA-RISC 1.0 
*/

/* So I try with the following test: */
int
getpagesize()

{			/* I think that machine/param.h has the right number */
#include 

return(NBPG_PA83);	/* This is supposed to be ok for PA-RISC 1.0, but
				I don't know about 1.1 (i.e. Snakes) */

}

/* I played with various combinations, but the examples on "Radiance tutorisl"
keep crashing (the examples with rview give bus error/core dumps. Fighting
with HP's debugger showed that the program was inside readobj.c, immediately
after line 165 :

   165                          if (fscanf(fp, "%lf", &fa->farg[i]) != 1)

#ifdef REALSYSV
int
getpagesize()			/* use SYSV var structure to get page size */

etc...


That's all I can do for the moment (I should go to the bed, you see... )

Have a nice weekend,
Nick.
-- 
Nikolaos Fotis			National Technical Univ. of Athens, Greece
16 Esperidon St.,		UUCP:	mcsun!ariadne!theseas!nfotis
Halandri, GR - 152 32		or InterNet : nfotis@theseas.ntua.gr
Athens, GREECE			FAX: (+30 1) 77 84 578

Date: Thu, 12 Sep 91 09:11:54 +0200
From: greg (Greg Ward)
To: nfotis%theseas.ntua.gr@Csa2.lbl.gov
Subject: Re:  Troubles with Radiance1R4

Yes, someone else has had trouble with HP workstations and my version
of malloc.  The truth is, you don't need my version of malloc and it
would probably make life simpler if you replaced the appearance of
malloc.o with bmalloc.o in the Makefiles of the ot and rt directories.
Also, you should compile bmalloc.c with the option -DALIGN=double for
RISC architectures, something I should have included for HP workstations
in the makeall script but didn't out of ignorance.  I recommend compiling
bmalloc.c by hand in the ot and rt subdirectories, replacing malloc.o with
bmalloc.o in the associated Makefile's, and rerunning makeall afterwards.

I hope that this fixes your problems.
-Greg

========================================
DAYLIGHT	Radiance and Daylight Simulation

Date: Wed,  4 Sep 91 11:57:27 Z
From: Environmental Design Unit 
To: greg@lesosun1.epfl.ch
Subject: Radiance

Dear Greg,

I have some more questions about Radiance and daylight factor calculation.

a) What is the latest version of Radiance currently avialable, and how
   does is it differ from v1.3.1 with respect to df calc?

b) Any joy yet in dealing with adjacent spaces?

c) Is there a way of modelling external structures so that both the direct
   AND diffuse daylight entering a space is modified? (A "fudge factor"
   in the sky model?).

d) Specular reflections from (pseudo?) glazing elements - are they a
   function of incidence angle?  (Reflections at grazing incidence
   dominate for deep-well atria with glass 'walls').

Thanks in advance.

-John Mardaljevic

ps.  I realise I might be asking some difficult questions.

pps. It might not be a bad thing to warn any Radiance users moving in
     the direction of df calcs about the dangers of using insufficiently
     small source polygons for window elements (df>100% !!).

Date: Wed, 4 Sep 91 15:53:25 +0200
From: greg (Greg Ward)
To: edu@leicester-poly.ac.uk
Subject: Re:  Radiance

Hi John,

a)
I think it's really the next release of Radiance that you want.  Version 1.4
has relatively few advantages in the daylight area over 1.3.1.  The release
on which I am currently laboring (probably 2.0), has many more of the
calculation capabilities that are needed for difficult daylight modeling
and analysis.  In particular, there is a daylight factor calculation and
visualization script that produces contour plots of the workplane (hallelujah)
and additional capabilities built into Radiance itself for the simulation
of daylight reflected from mirrored surfaces and so on.

b)
I have not tried it out yet, but I think Radiance is now ready to tackle
adjacent spaces to atria.  The key is a new program called mkillum that
calculates in a separate pass the distribution of light from windows or
other such "secondary" light sources.  In the process, mkillum accounts
for all interactions including external obstructions and interreflections.

c)
In older releases of Radiance, the only way to account properly for
external obstructions (other than computing the distribution yourself)
is to use the interreflection calculation to compute the contribution
from the window, ie. do not make the windows into type illum.  This
is even still the best method for offices with very large windows.
(Avoiding the problem you mentioned in your P.S.)

d)
Radiance surfaces obey Fresnel's laws where reflection goes
to one and transmission goes to zero at grazing for specular surfaces.
However, to properly account for reflected sunlight from atria walls,
you must use the next release of Radiance with code for finding virtual
light sources.  Previous releases cannot find secondary rays from such
tiny sources as the sun.

If you want to be a beta test site for version 2.0, you need only ask.
I would be happy to put together a current distribution for you to
test.  I plan to make an official release near the end of this year.

-Greg

=======================================
AMIGA		Radiance on the Amiga 2000

Date: Thu, 29 Aug 91 13:51:24 MED
From: bojsen@dc.dth.dk (Per Bojsen)
To: greg@hobbes.lbl.gov
Subject: Distribution of a port of the RADIANCE package

Greg,

I'm currently porting your RADIANCE package to the Amiga.  I would
like to distribute at least the binaries along with the data files
necessary, but preferably the whole package including the patched
sources.  Is it permitted to redistribute the package with pacthed
sources?  If not, what parts of the package may be redistributed, if
any?

--------------------------------------------------------------------------------
Per Bojsen                The VLSI Research Group        EMail: bojsen@dc.dth.dk
MoDAG                 Technical University of Denmark
--------------------------------------------------------------------------------

Date: Thu, 29 Aug 91 14:02:57 +0200
From: greg (Greg Ward)
To: bojsen@dc.dth.dk
Subject: Re:  Distribution of a port of the RADIANCE package

Hello Per Bojsen,

First off, let me thank you for porting the software.  I haven't used
an Amiga myself, but I like what I've seen on it and it sounds like
a great value.

Could you tell me a little about your experience bringing the software
over?  Did you have to throw much out?  Do you have a display driver?
Did you get rview to work?

Since no one has asked to redistribute a modified version of Radiance
before, I think I will have to ask around and find out if it would
be acceptable.  Our main concern I suppose is protecting the reputation
of Lawrence Berkeley Lab (if it has one) against irresponsible changes.
I'll try and get back to you by the end of next week.

-Greg

Date: Wed, 4 Sep 91 15:42:26 +0200
From: her%compel.dk%dkuug.dk@Csa2.lbl.gov (Helge Egelund Rasmussen)

New RADIANCE 1R4 user...
Mail address:
	Helge E. Rasmussen
	Compel A/S
	Hvidovrevej 80
	DK-2610 Roedovre
	Denmark
Phone: +45 36 72 33 00
E-mail: her@compel.dk
Machine: 386, Amiga
System: Interactive Unix, AmigaDos
Application:
	Hobby, I've been working with lots of different renderes on the
	Amiga.
	I'm in the process of porting Radiance to the Amiga.
	At the moment, nearly everything works except programs which use
	the pipe system call (f.ex. pinterp). I've created a rview device 
	driver for the Amiga HAM-E 'framebuffer', and created a picture 
	converter to the Amiga IFF24 bit graphics format.
	At the moment, I'm working on a converter which can convert 
	Imagine objects and scenes to Radiance scenes (Imagine is a commercial
	Amiga based render/animation package which has a rather good 
	'triangle' based 3d editor).

	I can upload the patches for the Amiga version to hobbes if you are
	interested.

Date: Wed, 4 Sep 91 17:17:48 +0200
From: greg (Greg Ward)
To: bojsen@dc.dth.dk, her@compel.dk
Subject: Re:  Distribution of a port of the RADIANCE package

Dear Per Bojsen and Helge Rasmussen,

Thank you both for your work porting Radiance to the Amiga.  It is a shock
to me that anyone has attempted this, let alone two people from the same
country at the same time!

I have asked those in the know at LBL about the official policies on
redistribution of software, and there don't appear to be any.  Therefore,
I am going to make up my own policies, which I hope will be acceptable to
everyone.

First off, I don't think we really need TWO ports of Radiance to the
Amiga, so I would like the two of you to have a little discussion and
fight it out among you to decide which and what to include in a distribution.

Second, I would like hobbes.lbl.gov to be used as the distribution site,
at least for the time being.  Before the end of the year, I hope to set
up a site here in Switzerland for distribution on the European continent
that is identical to the one in California.  I have created a new directory
in the anonymous ftp pub/ports subdirectory called "amiga".  I would like
it very much if you (collectively) would deposit your patches for
the Amiga plus any driver programs or routines you have written.  Please
do not duplicate the main distribution as I would like people to continue
to draw from the original for the sake of future compatibility.  Please also
include a README file describing the contents of your directory so that other
folks who are not so gifted (including myself) can figure out what to
do with it.

Third, I don't really want the Amiga binaries all stored on hobbes because
it would put quite a burden on my disk space and potentially on the network
if everybody and his brother wants a copy.  Therefore, I would be delighted
if you would distribute the executables yourself to interested parties via
floppy disk or whatever transfer medium you prefer.  (I have been using
floppies myself to distribute the MacIntosh A/UX version.)  I have no
objection if you want to redistribute the source code as well, so long as
you make it clear what version you are distributing and that the main site
has the most recent version.

Thanks again guys.  Bravo!  Good work!  (And all that.)

-Greg

Date: Mon, 9 Sep 91 17:36:40 MED
From: bojsen@dc.dth.dk (Per Bojsen)
To: greg@lesosun1.epfl.ch
Cc: her@compel.dk
Subject: Distribution of a port of the RADIANCE package

Hello Greg,

> Thank you both for your work porting Radiance to the Amiga.  It is a shock
> to me that anyone has attempted this, let alone two people from the same
> country at the same time!
>
In fact it is a surprise for me too!  I don't know Helge personally,
but I have seen him appear on USENET.

> First off, I don't think we really need TWO ports of Radiance to the
> Amiga, so I would like the two of you to have a little discussion and
> fight it out among you to decide which and what to include in a distribution.
>
I agree with that.  The discussion has started.  One thing we have to
agree upon is whether we will support old versions of the Amiga
operating systems, or only the newest.  As soon as we have compared
our ports and agreed upon the patches we'll let you know!

						Per.

========================================
COLORPICT	Using the Colorpict primitive

Date: Wed, 4 Sep 91 15:01:52 -0700
From: chet@cs.uoregon.edu (Chet Haase)
To: GJWard@Csa2.lbl.gov
Subject: Colorpict function

Hi.  I'm trying to get the Colorpict pattern to work right now and can't
seem to manage it.  I see it being used in example pictures (such as
model.new), but the source for those pictures is not in our distribution
(1.2?).  Is there more documentation or examples elsewhere that
I could get ahold of?   The main error that's occurring is:
	rview: rfuncname: undefined function,
where rfuncname is the rfunc I've defined in the .cal file (which it is
finding).   But then, I'm not really sure what I should be using for these
functions without a good example to work from (all I want to do is a straight 
mapping of the image onto a polygonal surface in another image, so I'm not
sure what parameters I should be using), so the source of the problem
may be elsewhere.

Thanks for your help.

Incidentally, I've been using your Architrion translator that you sent me help
on last Fall and it works great.  Thanks.

Chet Haase
CIS Department
University of Oregon

Date: Thu, 5 Sep 91 10:28:04 +0200
From: greg (Greg Ward)
To: chet@cs.uoregon.edu
Subject: Re:  Colorpict function

Hi Chet,

You've discovered the worst documented part of Radiance -- function files!
One day, I hope to make all this clear to people, but as you can see it
is very muddy water.

First off, rfuncname is just an example to get you to pick your own name
as appropriate for your material.  Each of the primary color functions is
a function of the three input primaries, thus allowing any mapping.  A
straightfoward mapping (as defined in rayinit.cal) is:

	red(r,g,b) = r;
	green(r,g,b) = g;
	blue(r,g,b) = b;

You may want to use the clip_r, clip_g and clip_b functions instead to
prevent reflectances greater than one (a no-no in any physical simulation).

An example of this should be contained in the file "picture" in the directory
model.new.  The actual files used, picture.cal and pine.pic, should be found
in the Radiance library location (ray/lib in the distribution).  The additional
arguments at the end of the colorpict define a transformation to get from
the world coordinates to the coordinates desired for the picture, pic_u and 
pic_v.

If you can't find the files or have other specific questions, I'll be happy to
help.  Otherwise, you could tell me exactly what you want and I could do it
for you as the most appropriate example.

-Greg

Date: Thu, 5 Sep 91 13:27:40 -0700
From: chet@cs.uoregon.edu (Chet Haase)
To: GJWard@Csa2.lbl.gov
Subject: colorpict revisited

I couldn't find any of the examples you listed except picture.cal and the 
picture only for model.new; I couldn't find the model.new directory or
the pine file in either the radiance directory on the system or the latest 
distribution tape we have (1.3.1).  So while I kind of understand what
I'm supposed to do, I'm apparently not getting the format or usage right
because I keep getting the same errors.  




I'll describe my situation a bit more clearly (hopefully):

I've got a picture called rgb.pic of a monitor screen and I'd like to map
it into a scene in which I've defined a monitor.  Using the functions in
rayinit.cal and picture.cal as models, I've defined my own functions in
rgbpic.cal:
{
	rgbpic.cal
}

clip_r(r,g,b) = min(r,1);
clip_g(r,g,b) = min(g,1);
clip_b(r,g,b) = min(b,1);
rgb_u = 1;
rgb_v = 1;
 (Note - these are perhaps silly functions for doing what I want, but at this
stage I just want to get the thing to compile)

I then try to map the picture with colorpict using this .cal file like so
void colorpict rgbpic
7
	clip_r(0,0,0) clip_g(0,0,0) clip_b(0,0,0) rgb.pic
	rgbpic.cal rgb_u rgb_v 
0
0

rgbpic plastic rgbimage
0
0
5 1 1 1 .05 0

rgbimage polygon rgbpicture
0
0
12
	.13 .481 .13
	1.21 .481 .13
	1.21 .481 .96
	.13 .481 .96


When I attempt to run rview on the .oct file, however, I get the error:
rview: clip_r(0,0,0): undefined function

Since it does find the .cal file (I've got RAYPATH defined correctly),
I don't understand why it's not using the function I've defined.



SO, the main problems I'm having are:
1) the above function error and how to avoid it
2) what functions/values I should be using for this purpose - I don't really
want to do anything funky with textures or colors, I simply want to map
the picture directly over what's in the scene.

If that model.new file would show me more about how to do this, I'd love to
see it.  The closest example I've found here is the source for the tennis
ball picture, but it wasn't quite enough for me to get the colorpict usage
down...


Thanks for your help.

Chet.

Date: Fri, 6 Sep 91 10:05:26 +0200
From: greg (Greg Ward)
To: chet@cs.uoregon.edu
Subject: Re:  colorpict revisited

Hi Chet,

I'll offer the following changes to the file, and hopefully you can
get this to work.

----------------------
rgbpic.cal:
{
	rgbpic.cal
}

clip_r(r,g,b) = min(r,1);
clip_g(r,g,b) = min(g,1);
clip_b(r,g,b) = min(b,1);
rgb_u = (Px-.13)/(.96-.13); 	{ was rgb_u = 1; }
rgb_v = (Pz-.13)/(.96-.13);	{ was rgb_v = 1; }

----------------------
Radiance file:

#
# Took out the arguments for the following:
#
void colorpict rgbpic
7
	clip_r clip_g clip_b rgb.pic
	rgbpic.cal rgb_u rgb_v 
0
0

rgbpic plastic rgbimage
0
0
5 1 1 1 .05 0

rgbimage polygon rgbpicture
0
0
12
	.13 .481 .13
	1.21 .481 .13
	1.21 .481 .96
	.13 .481 .96

-------------------------

The coordinate mapping I have made is based on the location, orientation,
and size of the polygon in your file.  If any of these things change, you
will have to change the coordinate mapping.  The simpler way to do this
is to use picture.cal and add a transformation to the colorpict primitive,
but since you're just trying to get it to work for now, I wanted to stay
as close to your original as possible.

The scalefactors for pictures is determined based on the aspect ratio.  I'm
assuming that your picture has square pixels and will fill the polygon you
have supplied, thus the horizontal (x) dimension is larger.  Quoting from
the Radiance reference manual (ray.1):

	The dimensions of the image data are determined by the picture
	such that the smaller dimension is always 1, and the other
	is the ratio between the larger and the smaller.
	For example, a 500x338 picture would have coordinates (u,v)
	in the rectangle between (0,0) and (1.48,1).

Hope this works!
-Greg

Date: Fri, 6 Sep 91 16:31:05 -0700
From: Chet Haase 
To: greg@lesosun1.epfl.ch
Subject: Re: colorpict revisited

That was the kind of help I needed - problem solved.

Thanks,
Chet.

===============================================
DOS		Radiance under DOS?

Date: Thu, 5 Sep 91 17:26:24 PDT
From: Donald Yett 
To: greg@hobbes.lbl.gov
Subject: Radiance questions

Hi, I just grabbed the Radiance package and related files from the ftp site.. 
I do have a few questions in hand.

1). Has this ever been ported to (please don't flame me!) DOS?

2). Is there a translator to convert the output to (yea I know it's limited
format) GIF?

In this day and age of 40-MIPS / 160 MFLOPS PC's I think it is a valid
question, even though I don't condone people wasting their money just to run
DOS!

(Yea I said 160 MFLOPS!  Although that board would cost the end-user about
$15k...)

Date: Fri, 6 Sep 91 09:22:26 +0200
From: greg (Greg Ward)
To: dyett@phad.hsc.usc.edu
Subject: Re:  Radiance questions

No, no one I know of has ported it to DOS yet, but there is now an Amiga
version they tell me and I use it on the Mac under A/UX and plenty of
folks have gotten it on their IBM RS/2 running AIX.

You are not the first to ask me this question, but since DOS is limited
in so many ways with virtually no standard graphics interface, I
haven't felt it was worth my time to attempt a port.  Even if I did
port the software to a DOS platform that could handle it, I'then get
hundreds of people coming to me with questions like, "I tried to get
Radiance to run on my IBM AT and it said something about a memory
error.  Do I need a hard drive?"  So, I don't think I'll be porting
Radiance to MS-DOS in my lifetime.  Any takers?

I have done some work on translators lately, but have not written anything
for GIF.  I have now a Poskanzer Pixmap translator and one for the TIFF
format, but gave up on Utah RLE format because it was too complicated and
GIF because it's only 8 bits and rather nasty itself.

My best advice is to pick up the pbmplus package which offers translation
between many different image format "standards", including Targa and GIF,
so you can get from Radiance to the format you want.  The pbmplus distribution
is available via anonymous ftp from export.lcs.mit.edu (18.30.0.238)
in the file "contrib/pbmplus.tar.Z".  Not everything works perfectly,
but it's the best package I know of in the public domain.  Personally,
I prefer PhotoShop from Adobe, which does image manipulation as well as
import/export from many formats.

-Greg

==========================================
RVIEW		Rview and memory

Date: Thu, 5 Sep 91 09:04:54 EST
From: vanwyk@arc.cmu.edu (Skip Van Wyk)
To: greg@lesosun1.epfl.ch
Status: RO

Greg, I got the conf model up and running.  I have 16MB on this
Sparc2.  Though it is also configured as a server for 8 accounts, right
now they are inactive.

Anyway, the conf scene cooked away for about 2.5 hours before I went
home last night; it looked good, even in 8bit.  Sometime during the
night, I got the message

rview: system - out of memory in refine:  Not enough memory
*** Error code 2

What kind of memory do you have?  And would everything have worked if I
logged out?

--Skip

Date: Fri, 6 Sep 91 10:11:57 +0200
From: greg (Greg Ward)
To: vanwyk@arc.cmu.edu
Subject: rview and memory

Hi Skip,

The problem is that you shouldn't be using rview to do your rendering.
You should use it to figure out what view you want, write it out with
the view command, then use rpict with the -vf option to read the view
and render the file in the background.  Rview is meant to be a previewer,
and uses up tons of memory as it goes to higher resolutions.  I've made
an enhancement for the next release that keeps rview from bombing when
it runs out of memory, but it will still run out of memory if you haven't
enough swap space on your disk.  A simple example of an rpict command is:

	% rpict -vf myview.vp -av .04 .04 .04 electric.oct > myview.pic &

Afterwards, you can logout and come back in the morning to see if the
process is still running (using ps).  The values for -av I selected for
the conference room model in particular, and they would be different for
a different scene.  (I'm also assuming that you did a "view myview.vp"
command in rview or you can use one of the provided view files in the
vf subdirectory instead.)

-Greg

===================================================
LIGHTS		Non-standard Light Sources

Date: Sat, 7 Sep 1991 15:18 EDT
From: elci@pluto.gs.com (Reha Elci)
Subject: radiance 4.0 questions + problems
To: GJWARD@Csa2.lbl.gov

First, I'd like to say that this is really a great package for learning and
production! Thanks for making this available on the net. I have a DECStation
with a PXG card; and the only problem I had so far is that ximage does not
work (probably does not recognize the visual); it produces no picture and
a lot of XServer errors (bad value). Currently I do a pvalue the pic file into
an rle file to display it.

But I have a question as well; cylinders as light sources are not supported
neither is "glow" applied to plastic or dielectric. So how does one go 
about doing neon lights or glowing rubys? Those examples would truly
demonstrate the power of radiosity. Testing for a maximum radius for shadow
testing (like you support on glow) would even make it better! Is there a
work around? Thanks for all your help.

Reha Elci

PS: Please add me to the newuser list; the automatic mailing failed since
my mail does not go thru to internet directly. Thanks.

Date: Thu, 12 Sep 91 11:19:22 +0200
From: greg (Greg Ward)
To: elci@pluto.gs.com
Subject: Re:  radiance 4.0 questions + problems

Dear Reha,

I am sorry but I don't think I can help you with your ximage problems.
I have found it very difficult to debug such things remotely, and X11
servers seem to be particularly flakey when it comes to color.  For
example, I know very well that 24-bit color does not work properly
in ximage, but I have no hardware to test it on and the hardware I
have tried seems unreliable.

I do plan to make cylinders usable as light sources in the near future,
but until then the only way to model neon tubes is to break the tubes
into many small polygons.  You can use gensurf to help you with that.
An example to create a tube of length 1 and radius .05:

	gensurf neon lamp '.05*sin(2*PI*t)' '.05*cos(2*PI*t)' 's' 20 6

You would then define the neon material using type glow with a maximum
radius of .5 (if you only wanted to illuminate nearby objects):

	void glow neon
	0
	0
	4 20 1 2 .5

As for glowing jewels, I'm not sure what you suggest is really physical.
Gems usually "glow" because of light reflected many times within the
stone and scintillation processes (in rubies for example).  You can
simulate this in Radiance by giving the red transmission coefficient
for a dielectric a value greater than one.  Note that giving values
greater than one for all three coefficients would mean that the
material was generating radiation -- which would be wrong in the
case of non-radioactive materials.

Hope this helps.
-Greg


Back to Top of Digest Volume 1, Number 5
Return to RADIANCE Home Page
Return to RADIANCE Digests Overview


All trademarks and product names mentioned herein are the property of their registered owners.
All written material is the property of its respective contributor.
Neither the contributors nor their employers are responsible for consequences arising from any use or misuse of this information.
We make no guarantee that any of this information is correct.