1 
NOTE: GLK's approximate ranking of 5 most important tagged with 
NOTE: GLK's approximate ranking of 8 most important tagged with 
2 
[GLK:1], [GLK:2], ... 
[GLK:1], [GLK:2], ... 
3 


4 
============================== 
======================== 
5 
other SHORT TERM ============= (including needed for LIC) 
SHORT TERM ============= (*needed* for streamlines & tractography) 
6 
============================== 
======================== 
7 


8 
Add a clamp function, which takes three arguments; either three scalars: 
Remove CL from compiler 

clamp(x, minval, maxval) = max(minval, min(maxval, x)) 


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). 





Level of differentiability in field type should be statement about how 


much differentiation the program *needs*, rather than what the kernel 


*provides*. The needed differentiability can be less than or equal to 


the provided differentiability. 

9 


10 
[GLK:1] Add sequence types (needed for evals & evecs) 
[GLK:3] Add sequence types (needed for evals & evecs) 
11 
syntax 
syntax 
12 
types: ty '{' INT '}' 
types: ty '{' INT '}' 
13 
value construction: '{' e1 ',' … ',' en '}' 
value construction: '{' e1 ',' … ',' en '}' 
14 
indexing: e '{' e '}' 
indexing: e '{' e '}' 
15 


16 
IL support for higherorder tensor values (matrices, etc). 
[GLK:4] evals & evecs for symmetric tensor[2,2] and 
17 
tensor construction [DONE] 
tensor[3,3] (requires sequences) 

tensor indexing [DONE] 


tensor slicing 


verify that hessians work correctly [DONE] 

18 


19 
Use ∇⊗ etc. syntax 
ability to emit/track/record variables into dynamically resized 
20 
syntax [DONE] 
runtime buffer 

typechecking 


IL and codegen 

21 


22 
test/uninit.diderot: 
tensor fields: convolution on general tensor images 

documents need for better compiler error messages when output variables 


are not initialized; the current messages are very cryptic 

23 


24 
determinant ("det") for tensor[3,3] 
======================== 
25 

SHORTISH TERM ========= (to make using Diderot less annoying to 
26 

======================== program in, and slow to execute) 
27 


28 
expand trace in mid to low translation 
valuenumbering optimization [DONE] 
29 


30 
valuenumbering optimization 
Allow ".ddro" file extensions in addition to ".diderot" 
31 


32 
Add type aliases for color types 
Be able to output values of type tensor[2,2] and tensor[3,3]; 
33 
rgb = real{3} 
(currently only scalars & vectors). Want to add some regression tests 
34 
rgba = real{4} 
based on this and currently can't 
35 


36 
============================== 
[GLK:1] Add a clamp function, which takes three arguments; either 
37 
MEDIUM TERM ================== (including needed for streamlines & tractography) 
three scalars: 
38 
============================== 
clamp(lo, hi, x) = max(lo, min(hi, x)) 
39 

or three vectors of the same size: 
40 

clamp(lo, hi, [x,y]) = [max(lo[0], min(hi[0], x)), 
41 

max(lo[1], min(hi[1], y))] 
42 

This would be useful in many current Diderot programs. 
43 

One question: clamp(x, lo, hi) is the argument order used in OpenCL 
44 

and other places, but clamp(lo, hi, x) is much more consistent with 
45 

lerp(lo, hi, x), hence GLK's preference 
46 

[DONE] 
47 


48 

[GLK:2] Proper handling of stabilize method 
49 


50 

allow "*" to represent "modulate": percomponent multiplication of 
51 

vectors, and vectors only (not tensors of order 2 or higher). Once 
52 

sequences are implemented this should be removed: the operation is not 
53 

invariant WRT basis so it is not a legit vector computation. 
54 


55 
[GLK:1] evals & evecs for symmetric tensor[3,3] (requires sequences) 
implicit type promotion of integers to reals where reals are 
56 

required (e.g. not exponentiation "^") 
57 


58 
[GLK:2] Save Diderot output to nrrd, instead of "mip.txt" 
[GLK:5] Save Diderot output to nrrd, instead of "mip.txt" 
59 
For grid of strands, save to similarlyshaped array 
For grid of strands, save to similarlyshaped array 
60 
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 
61 
For ragged things (like tractography output), will need to save both 
For ragged things (like tractography output), will need to save both 
62 
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 
63 
to index into complete list 
to index into complete list 
64 


65 
[GLK:3] Use of Teem's "hest" commandline parser for getting 
[GLK:6] Use of Teem's "hest" commandline parser for getting 
66 
any input variables that are not defined in the source file 
any "input" variables that are not defined in the source file. 
67 


68 
[GLK:4] ability to declare a field so that probe positions are 
[GLK:7] ability to declare a field so that probe positions are 
69 
*always* "inside"; with various ways of mapping the known image values 
*always* "inside"; with various ways of mapping the known image values 
70 
to nonexistant index locations. One possible syntax emphasizes that 
to nonexistant index locations. One possible syntax emphasizes that 
71 
there is a index mapping function that logically precedes convolution: 
there is a index mapping function that logically precedes convolution: 
74 
F = bspln3 ⊛ (img ◦ mirror) 
F = bspln3 ⊛ (img ◦ mirror) 
75 
where "◦" or "∘" is used to indicate function composition 
where "◦" or "∘" is used to indicate function composition 
76 


77 
extend norm (exp) to all tensor types [DONE for vectors and matrices] 
Level of differentiability in field type should be statement about how 
78 

much differentiation the program *needs*, rather than what the kernel 
79 
ability to emit/track/record variables into dynamically resized 
*provides*. The needed differentiability can be less than or equal to 
80 
runtime buffer 
the provided differentiability. 




Want: allow X *= Y, X /= Y, X += Y, X = Y to mean what they do in C, 


provided that X*Y, X/Y, X+Y, XY are already supported. 


Nearly every Diderot program would be simplified by this. 

81 


82 
Want: nontrivial field expressions & functions: 
Introduce "∇•" and "∇×" operators 
83 
image(2)[2] Vimg = load(...); 
syntax [DONE] 
84 
field#0(2)[] Vlen = Vimg ⊛ bspln3; 
typechecking 
85 
to get a scalar field of vector length, or 
IL and codegen 

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. 

86 


87 
tensor fields: convolution on general tensor images 
Add type aliases for color types 
88 

rgb = real{3} 
89 

rgba = real{4} 
90 


91 
============================== 
============================== 
92 
other MEDIUM TERM ============ (needed for particles) 
MEDIUM TERM ================== (*needed* for particles) 
93 
============================== 
============================== 
94 



Put small 1D and 2D fields, when reconstructed specifically by tent 


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 




95 
runtime birth of strands 
runtime birth of strands 
96 


97 
"initially" supports lists 
"initially" supports lists 
98 


99 
"initially" supports lists of positions output from 
"initially" supports lists of positions output from different 
100 
different initalization Diderot program 
initalization Diderot program (or output from the same program; 
101 

e.g. using output of iso2d.diderot for one isovalue to seed the input 
102 

to another invocation of the same program) 
103 


104 

Communication between strands: they have to be able to learn each 
105 

other's state (at the previous iteration). Early version of this can 
106 

have the network of neighbors be completely static (for running one 
107 

strand/pixel image computations). Later version with strands moving 
108 

through the domain will require some spatial data structure to 
109 

optimize discovery of neighbors. 
110 


111 

============================ 
112 

MEDIUMISH TERM ============ (to make Diderot more useful/effective) 
113 

============================ 
114 


115 
spatial data structure that permits strands' queries of neighbors 
Python/ctypes interface to runtime 
116 


117 
proper handling of stabilize method 
support for Python interop and GUI 
118 


119 
test/vrkcomp2.diderot: Add support for code like 
Allow integer exponentiation ("^2") to apply to square matrices, 
120 

to represent repeated matrix multiplication 
121 


122 
(F1 if x else F2)@pos 
Alow X *= Y, X /= Y, X += Y, X = Y to mean what they do in C, 
123 

provided that X*Y, X/Y, X+Y, XY are already supported. 
124 

Nearly every Diderot program would be simplified by this. 
125 

[DONE] 
126 


127 
This will require duplication of the continuation of the conditional 
Put small 1D and 2D fields, when reconstructed specifically by tent 
128 
(but we should only duplicate over the liverange of the result of the 
and when differentiation is not needed, into faster texture buffers. 
129 
conditional. 
test/illustvr.diderot is good example of program that uses multiple 
130 

such 1D fields basically as lookuptablebased function evaluation 
131 


132 

expand trace in mid to low translation [DONE] 
133 


134 

extend norm (exp) to all tensor types [DONE for vectors and matrices] 
135 


136 

determinant ("det") for tensor[3,3] 
137 


138 
add ":" for tensor dot product (contracts out two indices 
add ":" for tensor dot product (contracts out two indices 
139 
instead of one like •), valid for all pairs of tensors with 
instead of one like •), valid for all pairs of tensors with 
140 
at least two indices 
at least two indices 
141 


142 
============================== 
test/uninit.diderot: 
143 
other MEDIUM TERM ============ 
documents need for better compiler error messages when output variables 
144 
============================== 
are not initialized; the current messages are very cryptic 
145 


146 
want: warnings when "D" (reserved for differentiation) is declared as 
want: warnings when "D" (reserved for differentiation) is declared as 
147 
a variable name (get confusing error messages now) 
a variable name (get confusing error messages now) 
148 



support for Python interop and GUI 





Python/ctypes interface to runtime 





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


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

149 
============================== 
============================== 
150 

LONG TERM ==================== (make Diderot more interesting/attractive from 
151 

============================== a research standpoint) 
152 


153 

IL support for higherorder tensor values (matrices, etc). 
154 

tensor construction [DONE] 
155 

tensor indexing [DONE] 
156 

tensor slicing 
157 

verify that hessians work correctly [DONE] 
158 


159 
Better handling of variables that determines the scope of a variable 
Better handling of variables that determines the scope of a variable 
160 
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, 
162 
scope. Also prune out useless variables, which should include field 
scope. Also prune out useless variables, which should include field 
163 
variables after the translation to midil. 
variables after the translation to midil. 
164 


165 

test/vrkcomp2.diderot: Add support for code like 
166 

(F1 if x else F2)@pos 
167 

This will require duplication of the continuation of the conditional 
168 

(but we should only duplicate over the liverange of the result of the 
169 

conditional. 
170 


171 

[GLK:8] Want: nontrivial field expressions & functions. 
172 

scalar fields from scalar fields F and G: 
173 

field#0(2)[] X = (sin(F) + 1.0)/2; 
174 

field#0(2)[] X = F*G; 
175 

scalar field of vector field magnitude: 
176 

image(2)[2] Vimg = load(...); 
177 

field#0(2)[] Vlen = Vimg ⊛ bspln3; 
178 

field of normalized vectors (for LIC and vector field feature extraction) 
179 

field#2(2)[2] F = ... 
180 

field#0(2)[2] V = normalize(F); 
181 

scalar field of gradient magnitude (for edge detection)) 
182 

field#2(2)[] F = Fimg ⊛ bspln3; 
183 

field#0(2)[] Gmag = ∇F; 
184 

scalar field of squared gradient magnitude (simpler to differentiate): 
185 

field#2(2)[] F = Fimg ⊛ bspln3; 
186 

field#0(2)[] Gmsq = ∇F•∇F; 
187 

There is value in having these, even if the differentiation of them is 
188 

not supported (hence the indication of "field#0" for these above) 
189 


190 

Introduce region types (syntax region(d), where d is the dimension of the 
191 

region. One useful operator would be 
192 

dom : field#k(d)[s] > region(d) 
193 

Then the inside test could be written as 
194 

pos ∈ dom(F) 
195 

We could further extend this approach to allow geometric definitions of 
196 

regions. It might also be useful to do inside tests in world space, 
197 

instead of image space. 
198 


199 
co vs contra index distinction 
co vs contra index distinction 
200 


201 
some indication of tensor symmetry 
Permit field composition: 
202 

field#2(3)[3] warp = bspln3 ⊛ warpData; 
203 

field#2(3)[] F = bspln3 ⊛ img; 
204 

field#2(3)[] Fwarp = F ◦ warp; 
205 

So Fwarp(x) = F(warp(X)). Chain rule can be used for differentation. 
206 

This will be instrumental for expressing nonrigid registration 
207 

methods (but those will require covscontra index distinction) 
208 


209 

Allow the convolution to be specified either as a single 1D kernel 
210 

(as we have it now): 
211 

field#2(3)[] F = bspln3 ⊛ img; 
212 

or, as a tensor product of kernels, one for each axis, e.g. 
213 

field#0(3)[] F = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img; 
214 

This is especially important for things like timevarying fields 
215 

and the use of scalespace in field visualization: one axis of the 
216 

must be convolved with a different kernel during probing. 
217 

What is very unclear is how, in such cases, we should notate the 
218 

gradient, when we only want to differentiate with respect to some 
219 

subset of the axes. One ambitious idea would be: 
220 

field#0(3)[] Ft = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img; // 2D timevarying field 
221 

field#0(2)[] F = lambda([x,y], Ft([x,y,42.0])) // restriction to time=42.0 
222 

vec2 grad = ∇F([x,y]); // 2D gradient 
223 


224 

Tensors of order 3 (e.g. gradients of diffusion tensor fields, or 
225 

hessians of vector fields) and order 4 (e.g. Hessians of diffusion 
226 

tensor fields). 
227 


228 

representation of tensor symmetry 
229 
(have to identify the group of index permutations that are symmetries) 
(have to identify the group of index permutations that are symmetries) 
230 


231 
dot works on all tensors 
dot works on all tensors 
232 


233 
outer works on all tensors 
outer works on all tensors 
234 


235 

Help for debugging Diderot programs: need to be able to uniquely 
236 

identify strands, and for particular strands that are known to behave 
237 

badly, do something like printf or other logging of their computations 
238 

and updates. 
239 


240 

Permit writing dimensionally general code: Have some statement of the 
241 

dimension of the world "W" (or have it be learned from one particular 
242 

field of interest), and then able to write "vec" instead of 
243 

"vec2/vec3", and perhaps "tensor[W,W]" instead of 
244 

"tensor[2,2]/tensor[3,3]" 
245 


246 

Traits: all things things that have boilerplate code (especially 
247 

volume rendering) should be expressed in terms of the unique 
248 

computational core. Different kinds of streamline/tractography 
249 

computation will be another example, as well as particle systems. 
250 


251 
Einstein summation notation 
Einstein summation notation 
252 


253 
"tensor comprehension" (like list comprehension) 
"tensor comprehension" (like list comprehension) 
254 


255 

Fields coming from different sources of data: 
256 

* triangular or tetrahedral meshes over 2D or 3D domains (of the 
257 

source produced by finiteelement codes; these will come with their 
258 

own specialized kinds of reconstruction kernels, called "basis 
259 

functions" in this context) 
260 

* Large point clouds, with some radial basis function around each point, 
261 

which will be tuned by parameters of the point (at least one parameter 
262 

giving some notion of radius) 
263 


264 
====================== 
====================== 
265 
BUGS ================= 
BUGS ================= 
266 
====================== 
====================== 