Funniest website of the week/month/year?

http://damnyouautocorrect.com/category/best-of-dyac

Diagram of the Day #5

Its wang time again. This wang is importance sampled!
Woooo. No wonder that eye is looking at it.
Bucking the page 3 trend, this comes from p.13 of the siggraph 2010 "IS in production rendering course notes". Good work!

Diagram of the Day #4

This one comes from p.3 of :

http://langley.atmos.colostate.edu/courses/at622/lectures/supp_lecture_4A.pdf

iomanip

C++ std iomanip is super handy!
e.g.

"setw(4) << setfill('0')"

http://www.cplusplus.com/reference/iostream/manipulators/

Diagram of the Day 3

thanks to @fxmonkeydan for this little beauty :

these just make me laugh...

 


this week on Jamie Oliver inside a badgers arse :
"Inside a Badgers Arse Risotto".

RSL shadeop plugin memory

So there are 5 ways to manage memory allocation in an RSL shadeop.

1. per plugin init - lifetime from first call to end of frame, not threadsafe, no access to args

RSLEXPORT RslFunctionTable RslPublicFunctions(myFunctions,
                                             initStringBuffer,
                                             deleteStringBuffer);

2. per function init - lifetime from first call to end of frame, not threadsafe, access to args

static RslFunction myFunctions[] = {
   { "string appendTx(string)", appendTx, initStringBuffer, deleteStringBuffer },
   { NULL }

3. Set / GetGlobalStorate - lifetime from first call to end of frame, not threadsafe

4. Set / GetThreadData - lifetime of thread, threadsafe

5. Set / GetLocalData - lifetime of current shaded grid, threadsafe

Not bloody makefiles again ?

Replacing symlinks at compile time, so that republished shadeop DSOs don't fuck with existing shaders:

Makefile :
# here's the symlink path
MY_SO = $(PRMAN_PLUGDIR)MY.so

# resolve that to what the symlink points at
_MY_SOA := $(shell readlink $(MY_SO))

# double slashes in the path will mess up my shader macro's, so remove them
_MY_SO := $(subst //,/,$(_MY_SOA))

#now we can pass that into SL file scope using the pre-processor define flag, -D
shader .... -D_MYSO=$(_MY_SO)

Eat it, Makey - in your face, ha!

SIGGRAPH 2010

I am flying off to LA to attend my first ever SIGGRAPH conference tomorrow. I couldn't be more excited - its something i've always wanted to do since discovering SIGGRAPH papers existed when I was studying computer graphics at UMIST. ( 15+ yrs ago ! ).
I'm not sure why exactly but i've always been a fan of Jim Blinn's articles and work, and always thought it would be quite cool to attend a talk of his or meet him at Siggraph perhaps.
Any road up I am also due to give a short presentation at Pixar's exhibition booth about rendering the Olympus Map room clouds for Clash of the Titans. I am experiencing terror and excitement in equal measure. This is a new experience, usually its just the terror on its own. Fingers crossed it all goes to plan.

Diagram of the Day 2

Todays winner comes from the cutting edge of towel research.

( again, page 3 seems a popular choice ) :

http://reference.kfupm.edu.sa/content/a/b/abstract_a_layered_method_for_realistic__76571.pdf

Diagram of the Day 1

How they ever managed to sneak a picture of a cock and pair of balls past the moderator, i'll never know...

http://www.cemyuksel.com/research/dualscattering/dualscattering.pdf

( from page 3 ):

Makefile and GLSL

So it turns out, as cool as GLSL is, there's no support for #include directives.
Arse.
Includes are quite handy so they are.

Right, so what we need now is some existing pre-processor to expand the includes in-place.
What about gcc -E ( or g++ -E ) ? Weeelll, they are ok, but they leave place markers everywhere, it doesn't read very nicely. Also you'd have to name all your glsl shader files .c or .cpp or something icky like that.

Turns out you can run "cpp -P" and that will pre-process the includes nicely. It works on any old file name too - so no problem calling them .glsl
How do I automate it though ? A script ? A makefile ?
Well, since you're preprocessing code, its near enough to a compile to use a makefile.
This is where the (un)fun begins... actually lets cut a long story short, i'm already sick of makefiles without recounting blow by blow details.

basically you'd do this:

GLSL_SHADERS := $(wildcard hwshaders/*.glsl)
GLSL_SHADERSOUT:= $(GLSL_SHADERS:.glsl=.glslo)

%.glslo : %.glsl
  @echo $< $@
  @rm -rf $@ ; cpp -P $< $@

glsl : $(GLSL_SHADERSOUT) 

and then run

make glsl

Which would expect the glsl files to be one directory below the makefile in a dir called hwshaders.
Also I found out that if you are using a makefile conditional if, then the @ thingy to silence command echo doesn't really work. What did work for me was using the .SILENT rule.

.SILENT: safepub

safepub:
make clean
make
echo
if test -e /some/path/or/other/my.file; \
then \
echo File Exists \
else \
echo Doesn't Exist \
fi

so let that be a lesson to ye.

dPds works !

This was posted on the renderman forums, so i'm pretty sure its common knowledge.
Basically if you are in need of a consistent local basis transform for the usual stuff like normal maps and mapped anisotropic direction, you'll need a reliable tangent vector, sometimes referred to (by me, mainly) as dPds.

The RSL code is as follows:


void dPdst( float ss,tt; point PP;
output varying vector dPds;
output varying vector dPdt; )
{
float Dus = Du(ss);
float Dut = Du(tt);
float Dvs = Dv(ss);
float Dvt = Dv(tt);
vector DuP = Du(PP);
vector DvP = Dv(PP);
float det = (Dus*Dvt-Dvs*Dut);
dPds = (DuP*Dvt - DvP*Dut) / det;
dPdt = (DvP*Dus - DuP*Dvs) / det;
}


9 things i hate about my iPhone

These are the 10 9 most irritating things about my iPhone that stop at 'quite good' where it could have been truly great.

1 - Battery Life : shocking.
2 - Its expensive. Very very expensive.
3 - It can't bluetooth photos to non-iPhones. Basic stuff, and it can't even do that.
4 - You can't use your own sound files for ring tones. You have to buy them
5 - You can't copy files, such as pdfs, to it for offline reading. That's rubbish!!
5 - The calendar can't repeat on useful intervals such as 'last working day of month' or 'first tuesday in month'. Basic, basic, *basic* functionality that it can't even do.
6 - The pin number screen-lock is so irritating. If I want my data secure, I have to keep unlocking it all the time, even with the least annoying unlock option. Its so annoying when you are in a hurry.
7 - Touch sensitive keypad and predictive text frequently cause typos. The more of a hurry you are in, the worse it gets since it is so over-sensitive to mis-keying (or not adding the letter even when you hit a button with your finger).
8 - You can't write your own software for it unless you have a mac. Why should that be? Its not even running OSX !
9 - The sim card slot is like a chinese maths puzzle.


And that ladies and gentlemen, is why I think Apple are such smug bastards.

handy shell alias

pf is a wildcard running task search :
alias pf "ps -ef | grep \!* | egrep -v 'grep |bin/pf'"


rmane set prman env :
alias rmane 'setenv RMANTREE /opt/pixar/RenderManProServer-\!:1 ; setenv PRMANVER \!:1_64 ; setenv PRMAN_SHADER_PATH ${BUILDROOT}prman-${PRMANVER}/shaders ; setenv PRMAN_DSO_PATH ${BUILDROOT}prman-${PRMANVER} ; echo RMANTREE set to $RMANTREE ; echo PRMANVER set to $PRMANVER ; echo PRMAN_SHADER_PATH set to $PRMAN_SHADER_PATH ; echo PRMAN_DSO_PATH set to $PRMAN_DSO_PATH '


rmanv queries prman env :
alias rmanv 'env | grep RMAN'


ribgrep :
alias ribgrep '~/ribgrepext \!*'


where ribgrepext is a bash script like so:

ribgrepext :
#!/bin/bash

if [ ! $# == 2 ]; then
$RMANTREE/bin/catrib $1 | grep -o -e '[^\"]*\.exr\|[^\"]*\.dsh\|[^\"]*\.shd\|[^\"]*\.ptc\|[^\"]*\.optc\|[^\"]*\.bkm\|[^\"]*\.tex\|[^\"]*\.unk' | sort | uniq
else
$RMANTREE/bin/catrib $1 | grep -o -e [^\"]*\.${2} | sort | uniq
fi