tllw / sray Goto Github PK

The rendering preview window shows the first frame of an animation always
(weird, I swear this used to work correctly previously).

Forgotten to implement glossy reflection and refraction. It should be
trivial to add.

The renderer should consult a user-supplied configuration file to override
the default options. Command-line options should in turn override
config-file settings. A command-line argument should be provided to specify
an arbitrary config file instead of searching at the default locations
(like /etc/sray.conf, $HOME/.sray.conf, etc.

Implement a plugin system, and separate object types into loadable plugins.

At this point the renderer just clamps the floating point framebuffer
values to the interval [0, 1] before outputting the low dynamic-range
image. This is unacceptable for a photorealistic renderer.

User-controllable or automatic non-linear tone-mapping should be performed
on the floating point framebuffer, and additionally HDR output should be
implemented.

Irradiance caching can be used to sparsely sample large surfaces where
illumination varies slowly during final gather, thus greatly increasing GI
performance.

The implementation is probably non-trivial.

At the moment photon shooting is single-threaded, which impacts performance
heavily on SMP systems.

It should be nearly trivial to make photon shooting use the thread pool.

3ds2sray should be more than capable of outputting keyframes present in the
3ds input file.

What steps will reproduce the problem?

1. Install libvmath-svn rev 63
2. Try to build sray

You'll get this errors


g++ -pedantic -Wall -Wno-strict-aliasing -Wno-deprecated -g -O2    `pkg-config 
--cflags vmath imago`   -c -o src/main.o src/main.cc
g++ -pedantic -Wall -Wno-strict-aliasing -Wno-deprecated -g -O2    `pkg-config 
--cflags vmath imago`   -c -o src/light.o src/light.cc
g++ -pedantic -Wall -Wno-strict-aliasing -Wno-deprecated -g -O2    `pkg-config 
--cflags vmath imago`   -c -o src/object.o src/object.cc
g++ -pedantic -Wall -Wno-strict-aliasing -Wno-deprecated -g -O2    `pkg-config 
--cflags vmath imago`   -c -o src/material.o src/material.cc
g++ -pedantic -Wall -Wno-strict-aliasing -Wno-deprecated -g -O2    `pkg-config 
--cflags vmath imago`   -c -o src/aabb.o src/aabb.cc
src/material.cc: In constructor 'Material::Material()':
src/material.cc:85:34: error: no matching function for call to 
'Material::set_attribute(const char [9], double, int)'
src/material.h:86:7: note: candidate is: void Material::set_attribute(const 
char*, const Color&, Texture*)
src/material.cc:86:33: error: no matching function for call to 
'Material::set_attribute(const char [8], double, int)'
src/material.h:86:7: note: candidate is: void Material::set_attribute(const 
char*, const Color&, Texture*)
src/material.cc:87:33: error: no matching function for call to 
'Material::set_attribute(const char [8], double, int)'
src/material.h:86:7: note: candidate is: void Material::set_attribute(const 
char*, const Color&, Texture*)
src/material.cc:88:29: error: no matching function for call to 
'Material::set_attribute(const char [4], double, int)'
src/material.h:86:7: note: candidate is: void Material::set_attribute(const 
char*, const Color&, Texture*)
src/material.cc: In member function 'MatAttrib* Material::find_attribute(const 
char*)':
src/material.cc:177:27: error: no matching function for call to 
'MatAttrib::MatAttrib(const char*&, int)'
src/material.cc:29:1: note: candidates are: MatAttrib::MatAttrib(const char*, 
const Color&, Texture*)
src/material.cc:24:1: note:                 MatAttrib::MatAttrib()
src/material.h:29:17: note:                 MatAttrib::MatAttrib(const 
MatAttrib&)
src/material.cc: In member function 'const MatAttrib* 
Material::find_attribute(const char*) const':
src/material.cc:191:27: error: no matching function for call to 
'MatAttrib::MatAttrib(const char*&, int)'
src/material.cc:29:1: note: candidates are: MatAttrib::MatAttrib(const char*, 
const Color&, Texture*)
src/material.cc:24:1: note:                 MatAttrib::MatAttrib()
src/material.h:29:17: note:                 MatAttrib::MatAttrib(const 
MatAttrib&)
make: *** [src/material.o] Error 1
make: *** Waiting for unfinished jobs....
src/main.cc: In function 'void render_block(block*)':
src/main.cc:316:36: error: no match for 'operator=' in '*(hist + ((long 
unsigned int)(((long unsigned int)x) * 32ul))) = ((float)i / 
(float)opt.options::max_samples)'
/usr/include/vmath/vector.h:208:15: note: candidate is: Vector4& 
Vector4::operator=(const Vector4&)
make: *** [src/main.o] Error 1


With libvmath-svn rev 55, sray builds with no problems.

There's no particular reason why obj2sray should depend on libhenge.
Cannibalize the OBJ-loading code from henge and drop the dependency.

This should also fix with the missing-texture problem. The converter
shouldn't try to open textures and verify their presence. This happens now
as a side-effect of using the henge 3D engine to load the OBJ before
re-saving it from the internal henge data structures.

When rendering animations, it seems like photon mapping effects such as
caustics are only calculated correctly for the first frame.

The DataCache class provides per-thread key-data caching for lazy
evaluation algorithms throughout the renderer (currently used for the
calculation transformation matrices from the xform hierarchy, and
axis-aligned bounding box calculation).

The problem is that the current implementation is a sordid hack that relies
on pthread_t (returned by pthread_self) being an integer.

Currently multithreaded rendering is disabled for MacOSX to side-step the
issue.

Must implement properly with TSD, or find another way to avoid threading
issues with the data caches.

Right now final gather does not use the photon map to importance-sample the
half-space above the surface. Rectify that omission.

Depth of field is not currently implemented.

Some form of bump mapping would be really useful.