1 
NOTE: GLK's approximate ranking of 5 most important tagged with 
*************************************************** 
2 
[GLK:1], [GLK:2], ... 
*************************************************** 
3 

THIS TODO HAS BEEN MOVED TO THE DIDEROT WIKI: 
4 


5 
============================== 
http://diderotwiki.cs.uchicago.edu/index.php/Todo 

other SHORT TERM ============= (including needed for LIC) 


============================== 

6 


7 
Add a clamp function, which takes three arguments; either three scalars: 
PLEASE USE THAT PAGE TO UPDATE PROBLEMS AND PROGESS 
8 
clamp(x, minval, maxval) = max(minval, min(maxval, x)) 
*************************************************** 
9 
or three vectors of the same size: 
*************************************************** 

clamp([x,y], minvec, maxvec) = [max(minvec[0], min(maxvec[0], x)), 


max(minvec[1], min(maxvec[1], y))] 


This would be useful in many current Diderot programs. 


One question: clamp(x, minval, maxval) is the argument order 


used in OpenCL and other places, but clamp(minval, maxval, x) 


would be more consistent with lerp(minout, maxout, x). 

10 


11 
Level of differentiability in field type should be statement about how 
NOTE: GLK's approximate ranking of 8 most important tagged with 
12 
much differentiation the program *needs*, rather than what the kernel 
[GLK:1], [GLK:2], ... 
13 
*provides*. The needed differentiability can be less than or equal to 

14 
the provided differentiability. 
======================== 
15 

SHORT TERM ============= (*needed* for streamlines & tractography) 
16 

======================== 
17 


18 
[GLK:1] Add sequence types (needed for evals & evecs) 
[GLK:2] Add sequence types (needed for evals & evecs) 
19 
syntax 
syntax 
20 
types: ty '{' INT '}' 
types: ty '{' INT '}' 
21 
value construction: '{' e1 ',' … ',' en '}' 
value construction: '{' e1 ',' … ',' en '}' 
22 
indexing: e '{' e '}' 
indexing: e '{' e '}' 
23 


24 
IL support for higherorder tensor values (matrices, etc). 
[GLK:3] evals & evecs for symmetric tensor[2,2] and 
25 
tensor construction [DONE] 
tensor[3,3] (requires sequences) 

tensor indexing [DONE] 


tensor slicing 


verify that hessians work correctly [DONE] 

26 


27 
Use ∇⊗ etc. syntax 
ability to emit/track/record variables into dynamically resized 
28 
syntax [DONE] 
runtime output buffer 

typechecking 


IL and codegen 

29 


30 
test/uninit.diderot: 
[GLK:4] tensor fields from tensor images: Initially need at least 
31 
documents need for better compiler error messages when output variables 
convolution on tensor[2,2] and tensor[3,3] (the same componentwise 
32 
are not initialized; the current messages are very cryptic 
convolution as for vectors). 
33 


34 
determinant ("det") for tensor[3,3] 
======================== 
35 

SHORTISH TERM ========= (to make using Diderot less annoying to 
36 

======================== program in, and slow to execute) 
37 


38 
expand trace in mid to low translation 
Allow ".ddro" file extensions in addition to ".diderot" 
39 


40 
valuenumbering optimization 
Be able to output values of type tensor[2,2] and tensor[3,3]; 
41 

(currently only scalars & vectors). Want to add some regression tests 
42 

based on this and currently can't 
43 


44 
Add type aliases for color types 
[GLK:1] Proper handling of stabilize method 

rgb = real{3} 


rgba = real{4} 

45 


46 
============================== 
Convolution on general tensor images (order > 2) 
47 
MEDIUM TERM ================== (including needed for streamlines & tractography) 

48 
============================== 
allow "*" to represent "modulate": percomponent multiplication of 
49 

vectors, and vectors only (not tensors of order 2 or higher). Once 
50 

sequences are implemented this should be removed: the operation is not 
51 

invariant WRT basis so it is not a legit vector computation. 
52 


53 
[GLK:1] evals & evecs for symmetric tensor[3,3] (requires sequences) 
implicit type promotion of integers to reals where reals are 
54 

required (e.g. not exponentiation "^") 
55 


56 
[GLK:2] Save Diderot output to nrrd, instead of "mip.txt" 
[Nick working on this] Save Diderot output to nrrd, instead of "mip.txt" 
57 
For grid of strands, save to similarlyshaped array 
For grid of strands, save to similarlyshaped array 
58 
For list of strands, save to long 1D (or 2D for nonscalar output) list 
For list of strands, save to long 1D (or 2D for nonscalar output) list 
59 
For ragged things (like tractography output), will need to save both 
For ragged things (like tractography output), will need to save both 
60 
complete list of values, as well as list of start indices and lengths 
complete list of values, as well as list of start indices and lengths 
61 
to index into complete list 
to index into complete list 
62 


63 
[GLK:3] Use of Teem's "hest" commandline parser for getting 
[GLK:6] ability to declare a field so that probe positions are 

any input variables that are not defined in the source file 





[GLK:4] ability to declare a field so that probe positions are 

64 
*always* "inside"; with various ways of mapping the known image values 
*always* "inside"; with various ways of mapping the known image values 
65 
to nonexistant index locations. One possible syntax emphasizes that 
to nonexistant index locations. One possible syntax emphasizes that 
66 
there is a index mapping function that logically precedes convolution: 
there is a index mapping function that logically precedes convolution: 
69 
F = bspln3 ⊛ (img ◦ mirror) 
F = bspln3 ⊛ (img ◦ mirror) 
70 
where "◦" or "∘" is used to indicate function composition 
where "◦" or "∘" is used to indicate function composition 
71 


72 
extend norm (exp) to all tensor types [DONE for vectors and matrices] 
Level of differentiability in field type should be statement about how 
73 

much differentiation the program *needs*, rather than what the kernel 
74 
ability to emit/track/record variables into dynamically resized 
*provides*. The needed differentiability can be less than or equal to 
75 
runtime buffer 
the provided differentiability. 
76 


77 
Want: allow X *= Y, X /= Y, X += Y, X = Y to mean what they do in C, 
Use ∇⊗ etc. syntax 
78 
provided that X*Y, X/Y, X+Y, XY are already supported. 
syntax [DONE] 
79 
Nearly every Diderot program would be simplified by this. 
typechecking 
80 

IL and codegen 
81 


82 
Want: nontrivial field expressions & functions: 
Add type aliases for color types 
83 
image(2)[2] Vimg = load(...); 
rgb = real{3} 
84 
field#0(2)[] Vlen = Vimg ⊛ bspln3; 
rgba = real{4} 

to get a scalar field of vector length, or 


field#2(2)[] F = Fimg ⊛ bspln3; 


field#0(2)[] Gmag = ∇F; 


to get a scalar field of gradient magnitude, or 


field#2(2)[] F = Fimg ⊛ bspln3; 


field#0(2)[] Gmsq = ∇F•∇F; 


to get a scalar field of squared gradient magnitude, which is simpler 


to differentiate. However, there is value in having these, even if 


the differentiation of them is not supported (hence the indication 


of "field#0" for these above) 





Want: ability to apply "normalize" to a field itself, e.g. 


field#0(2)[2] V = normalize(Vimg ⊛ ctmr); 


so that V(x) = normalize((Vimg ⊛ ctmr)(x)). 


Having this would simplify expression of standard LIC method, and 


would also help express other vector field expressions that arise 


in vector field feature exraction. 

85 


86 
tensor fields: convolution on general tensor images 
Revisit how images are created within the language. 
87 

The "load" operator should probably go away, and its strangs 
88 

that strings are there only as a way to refer to nrrd filenames 
89 


90 
============================== 
============================== 
91 
other MEDIUM TERM ============ (needed for particles) 
MEDIUM TERM ================== (*needed* for particles) 
92 
============================== 
============================== 
93 


94 
Put small 1D and 2D fields, when reconstructed specifically by tent 
[Lamont working on this] runtime birth of strands 

and when differentiation is not needed, into faster texture buffers. 


test/illustvr.diderot is good example of program that uses multiple 


such 1D fields basically as lookuptablebased function evaluation 





runtime birth of strands 

95 


96 
"initially" supports lists 
"initially" supports lists 
97 


98 
"initially" supports lists of positions output from 
"initially" supports lists of positions output from different 
99 
different initalization Diderot program 
initalization Diderot program (or output from the same program; 
100 

e.g. using output of iso2d.diderot for one isovalue to seed the input 
101 

to another invocation of the same program) 
102 


103 

[Lamont working on this] Communication between strands: they have to 
104 

be able to learn each other's state (at the previous iteration). 
105 

Early version of this can have the network of neighbors be completely 
106 

static (for running one strand/pixel image computations). Later 
107 

version with strands moving through the domain will require some 
108 

spatial data structure to optimize discovery of neighbors. 
109 


110 

============================ 
111 

MEDIUMISH TERM ============ (to make Diderot more useful/effective) 
112 

============================ 
113 


114 

[GLK:5] Want codegeneration working for tensors of order three. 
115 

Order three matters for edge detection in scalar fields (to get 
116 

second derivatives of gradient magnitude), second derivatives 
117 

of vector fields (for some feature extraction), and first 
118 

derivatives of diffusion tensor fields. 
119 


120 
spatial data structure that permits strands' queries of neighbors 
Python/ctypes interface to runtime 
121 


122 
proper handling of stabilize method 
support for Python interop and GUI 
123 


124 
test/vrkcomp2.diderot: Add support for code like 
Allow integer exponentiation ("^2") to apply to square matrices, 
125 

to represent repeated matrix multiplication 
126 


127 
(F1 if x else F2)@pos 
Put small 1D and 2D fields, when reconstructed specifically by tent 
128 

and when differentiation is not needed, into faster texture buffers. 
129 

test/illustvr.diderot is good example of program that uses multiple 
130 

such 1D fields basically as lookuptablebased function evaluation 
131 


132 
This will require duplication of the continuation of the conditional 
extend norm (exp) to all tensor types [DONE for vectors and matrices] 
133 
(but we should only duplicate over the liverange of the result of the 

134 
conditional. 
determinant ("det") for tensor[3,3] 
135 


136 
add ":" for tensor dot product (contracts out two indices 
add ":" for tensor dot product (contracts out two indices 
137 
instead of one like •), valid for all pairs of tensors with 
instead of one like •), valid for all pairs of tensors with 
138 
at least two indices 
at least two indices 
139 


140 
============================== 
test/uninit.diderot: 
141 
other MEDIUM TERM ============ 
documents need for better compiler error messages when output variables 
142 
============================== 
are not initialized; the current messages are very cryptic 
143 


144 
want: warnings when "D" (reserved for differentiation) is declared as 
want: warnings when "D" (reserved for differentiation) is declared as 
145 
a variable name (get confusing error messages now) 
a variable name (get confusing error messages now) 
146 



support for Python interop and GUI 





Python/ctypes interface to runtime 





============================== 


LONG TERM ==================== 

147 
============================== 
============================== 
148 

LONG TERM ==================== (make Diderot more interesting/attractive from 
149 

============================== a research standpoint) 
150 


151 

IL support for higherorder tensor values (matrices, etc). 
152 

tensor construction [DONE] 
153 

tensor indexing [DONE] 
154 

tensor slicing 
155 


156 
Better handling of variables that determines the scope of a variable 
Better handling of variables that determines the scope of a variable 
157 
based on its actual use, instead of where the user defined it. So, 
based on its actual use, instead of where the user defined it. So, 
159 
scope. Also prune out useless variables, which should include field 
scope. Also prune out useless variables, which should include field 
160 
variables after the translation to midil. 
variables after the translation to midil. 
161 


162 

test/vrkcomp2.diderot: Add support for code like 
163 

(F1 if x else F2)@pos 
164 

This will require duplication of the continuation of the conditional 
165 

(but we should only duplicate over the liverange of the result of the 
166 

conditional. 
167 


168 

[GLK:7] Want: nontrivial field expressions & functions. 
169 

scalar fields from scalar fields F and G: 
170 

field#0(2)[] X = (sin(F) + 1.0)/2; 
171 

field#0(2)[] X = F*G; 
172 

scalar field of vector field magnitude: 
173 

image(2)[2] Vimg = load(...); 
174 

field#0(2)[] Vlen = Vimg ⊛ bspln3; 
175 

field of normalized vectors (for LIC and vector field feature extraction) 
176 

field#2(2)[2] F = ... 
177 

field#0(2)[2] V = normalize(F); 
178 

scalar field of gradient magnitude (for edge detection)) 
179 

field#2(2)[] F = Fimg ⊛ bspln3; 
180 

field#0(2)[] Gmag = ∇F; 
181 

scalar field of squared gradient magnitude (simpler to differentiate): 
182 

field#2(2)[] F = Fimg ⊛ bspln3; 
183 

field#0(2)[] Gmsq = ∇F•∇F; 
184 

There is value in having these, even if the differentiation of them is 
185 

not supported (hence the indication of "field#0" for these above) 
186 


187 

Introduce region types (syntax region(d), where d is the dimension of the 
188 

region. One useful operator would be 
189 

dom : field#k(d)[s] > region(d) 
190 

Then the inside test could be written as 
191 

pos ∈ dom(F) 
192 

We could further extend this approach to allow geometric definitions of 
193 

regions. It might also be useful to do inside tests in world space, 
194 

instead of image space. 
195 


196 
co vs contra index distinction 
co vs contra index distinction 
197 


198 
some indication of tensor symmetry 
Permit field composition: 
199 

field#2(3)[3] warp = bspln3 ⊛ warpData; 
200 

field#2(3)[] F = bspln3 ⊛ img; 
201 

field#2(3)[] Fwarp = F ◦ warp; 
202 

So Fwarp(x) = F(warp(X)). Chain rule can be used for differentation. 
203 

This will be instrumental for expressing nonrigid registration 
204 

methods (but those will require covscontra index distinction) 
205 


206 

Allow the convolution to be specified either as a single 1D kernel 
207 

(as we have it now): 
208 

field#2(3)[] F = bspln3 ⊛ img; 
209 

or, as a tensor product of kernels, one for each axis, e.g. 
210 

field#0(3)[] F = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img; 
211 

This is especially important for things like timevarying fields 
212 

and the use of scalespace in field visualization: one axis of the 
213 

must be convolved with a different kernel during probing. 
214 

What is very unclear is how, in such cases, we should notate the 
215 

gradient, when we only want to differentiate with respect to some 
216 

subset of the axes. One ambitious idea would be: 
217 

field#0(3)[] Ft = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img; // 2D timevarying field 
218 

field#0(2)[] F = lambda([x,y], Ft([x,y,42.0])) // restriction to time=42.0 
219 

vec2 grad = ∇F([x,y]); // 2D gradient 
220 


221 

representation of tensor symmetry 
222 
(have to identify the group of index permutations that are symmetries) 
(have to identify the group of index permutations that are symmetries) 
223 


224 
dot works on all tensors 
dot works on all tensors 
225 


226 
outer works on all tensors 
outer works on all tensors 
227 


228 

Help for debugging Diderot programs: need to be able to uniquely 
229 

identify strands, and for particular strands that are known to behave 
230 

badly, do something like printf or other logging of their computations 
231 

and updates. 
232 


233 

Permit writing dimensionally general code: Have some statement of the 
234 

dimension of the world "W" (or have it be learned from one particular 
235 

field of interest), and then able to write "vec" instead of 
236 

"vec2/vec3", and perhaps "tensor[W,W]" instead of 
237 

"tensor[2,2]/tensor[3,3]" 
238 


239 

Traits: all things things that have boilerplate code (especially 
240 

volume rendering) should be expressed in terms of the unique 
241 

computational core. Different kinds of streamline/tractography 
242 

computation will be another example, as well as particle systems. 
243 


244 
Einstein summation notation 
Einstein summation notation 
245 


246 
"tensor comprehension" (like list comprehension) 
"tensor comprehension" (like list comprehension) 
247 


248 

Fields coming from different sources of data: 
249 

* triangular or tetrahedral meshes over 2D or 3D domains (of the 
250 

source produced by finiteelement codes; these will come with their 
251 

own specialized kinds of reconstruction kernels, called "basis 
252 

functions" in this context) 
253 

* Large point clouds, with some radial basis function around each point, 
254 

which will be tuned by parameters of the point (at least one parameter 
255 

giving some notion of radius) 
256 


257 
====================== 
====================== 
258 
BUGS ================= 
BUGS ================= 
259 
====================== 
====================== 
260 



test/read2vecs.diderot: 


// HEY (BUG?) shouldn't it be a type error to load this 2D array of 


// 2vectors into a 2D *scalar* field? Instead, get: 


// uncaught exception Fail [Fail: Error in compiling lic.diderot] 


// raised at driver/main.sml:31.3931.76 


image(2)[] Vimg = load("../data/vorttest.nrrd"); 




261 
test/zslice2.diderot: 
test/zslice2.diderot: 
262 
// HEY (bug) bspln5 leads to problems ... 
// HEY (bug) bspln5 leads to problems ... 
263 
// uncaught exception Size [size] 
// uncaught exception Size [size] 
264 
// raised at ctarget/ctarget.sml:47.1547.19 
// raised at ctarget/ctarget.sml:47.1547.19 
265 
//field#4(3)[] F = img ⊛ bspln5; 
//field#4(3)[] F = img ⊛ bspln5; 
266 

