1 
NOTE: GLK's approximate ranking of 5 most important tagged with 
NOTE: GLK's approximate ranking of 5 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 
============================== 
======================== 




Add a clamp function, which takes three arguments; either three scalars: 


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. 

7 


8 
[GLK:1] Add sequence types (needed for evals & evecs) 
[GLK:1] Add sequence types (needed for evals & evecs) 
9 
syntax 
syntax 
10 
types: ty '{' INT '}' 
types: ty '{' INT '}' 
11 
value construction: '{' e1 ',' … ',' en '}' 
value construction: '{' e1 ',' … ',' en '}' 
12 
indexing: e '{' e '}' 
indexing: e '{' e '}' 
13 

[GLK:1] evals & evecs for symmetric tensor[3,3] (requires sequences) 
14 


15 
IL support for higherorder tensor values (matrices, etc). 
ability to emit/track/record variables into dynamically resized 
16 
tensor construction [DONE] 
runtime buffer 

tensor indexing [DONE] 


tensor slicing 


verify that hessians work correctly [DONE] 





Use ∇⊗ etc. syntax 


syntax [DONE] 


typechecking 


IL and codegen 





test/uninit.diderot: 


documents need for better compiler error messages when output variables 


are not initialized; the current messages are very cryptic 

17 


18 
determinant ("det") for tensor[3,3] 
tensor fields: convolution on general tensor images 
19 


20 
expand trace in mid to low translation 
======================== 
21 

SHORTISH TERM ========= (to make using Diderot less annoying/slow) 
22 

======================== 
23 


24 
valuenumbering optimization 
valuenumbering optimization 
25 


26 
Add type aliases for color types 
proper handling of stabilize method 

rgb = real{3} 


rgba = real{4} 





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


MEDIUM TERM ================== (including needed for streamlines & tractography) 


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





[GLK:1] evals & evecs for symmetric tensor[3,3] (requires sequences) 

27 


28 
[GLK:2] Save Diderot output to nrrd, instead of "mip.txt" 
[GLK:2] Save Diderot output to nrrd, instead of "mip.txt" 
29 
For grid of strands, save to similarlyshaped array 
For grid of strands, save to similarlyshaped array 
39 
*always* "inside"; with various ways of mapping the known image values 
*always* "inside"; with various ways of mapping the known image values 
40 
to nonexistant index locations. One possible syntax emphasizes that 
to nonexistant index locations. One possible syntax emphasizes that 
41 
there is a index mapping function that logically precedes convolution: 
there is a index mapping function that logically precedes convolution: 
42 
F = bspln3 ⊛ (img ◦ clamp) 
F = bspln3 ⊛ (img clamp) 
43 
F = bspln3 ⊛ (img ◦ repeat) 
F = bspln3 ⊛ (img ◦ repeat) 
44 
F = bspln3 ⊛ (img ◦ mirror) 
F = bspln3 ⊛ (img ◦ mirror) 
45 
where "◦" or "∘" is used to indicate function composition 
where "◦" or "∘" is used to indicate function composition 
46 


47 
extend norm (exp) to all tensor types [DONE for vectors and matrices] 
Use ∇⊗ etc. syntax 
48 

syntax [DONE] 
49 
ability to emit/track/record variables into dynamically resized 
typechecking 
50 
runtime buffer 
IL and codegen 




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. 

51 


52 
Want: nontrivial field expressions & functions: 
Add a clamp function, which takes three arguments; either three scalars: 
53 
image(2)[2] Vimg = load(...); 
clamp(x, minval, maxval) = max(minval, min(maxval, x)) 
54 
field#0(2)[] Vlen = Vimg ⊛ bspln3; 
or three vectors of the same size: 
55 
to get a scalar field of vector length, or 
clamp([x,y], minvec, maxvec) = [max(minvec[0], min(maxvec[0], x)), 
56 
field#2(2)[] F = Fimg ⊛ bspln3; 
max(minvec[1], min(maxvec[1], y))] 
57 
field#0(2)[] Gmag = ∇F; 
This would be useful in many current Diderot programs. 
58 
to get a scalar field of gradient magnitude, or 
One question: clamp(x, minval, maxval) is the argument order 
59 
field#2(2)[] F = Fimg ⊛ bspln3; 
used in OpenCL and other places, but clamp(minval, maxval, x) 
60 
field#0(2)[] Gmsq = ∇F•∇F; 
would be more consistent with lerp(minout, maxout, x). 

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) 

61 


62 
Want: ability to apply "normalize" to a field itself, e.g. 
Level of differentiability in field type should be statement about how 
63 
field#0(2)[2] V = normalize(Vimg ⊛ ctmr); 
much differentiation the program *needs*, rather than what the kernel 
64 
so that V(x) = normalize((Vimg ⊛ ctmr)(x)). 
*provides*. The needed differentiability can be less than or equal to 
65 
Having this would simplify expression of standard LIC method, and 
the provided differentiability. 

would also help express other vector field expressions that arise 


in vector field feature exraction. 

66 


67 
tensor fields: convolution on general tensor images 
Add type aliases for color types 
68 

rgb = real{3} 
69 

rgba = real{4} 
70 


71 
============================== 
============================== 
72 
other MEDIUM TERM ============ (needed for particles) 
MEDIUM TERM ================== (*needed* for particles) 
73 
============================== 
============================== 
74 



[GLK:5] runtime death of strands; test/iso2d.diderot provides 


great initial test for "die" command. 





Put small 1D and 2D fields, reconstructed by tent, into 


faster texture buffers 




75 
runtime birth of strands 
runtime birth of strands 
76 


77 
"initially" supports lists 
"initially" supports lists 
79 
"initially" supports lists of positions output from 
"initially" supports lists of positions output from 
80 
different initalization Diderot program 
different initalization Diderot program 
81 


82 
spatial data structure that permits strands' queries of neighbors 
Communication between strands: they have to be able to learn each 
83 

other's state (at the previous iteration). Early version of this can 
84 

have the network of neighbors be completely static (for running one 
85 

strand/pixel image computations). Later version with strands moving 
86 

through the domain will require some spatial data structure to 
87 

optimize discovery of neighbors. 
88 


89 

============================ 
90 

MEDIUMISH TERM ============ (to make Diderot more useful/effective) 
91 

============================ 
92 


93 
proper handling of stabilize method 
Python/ctypes interface to runtime 
94 


95 
test/vrkcomp2.diderot: Add support for code like 
support for Python interop and GUI 
96 


97 
(F1 if x else F2)@pos 
Alow X *= Y, X /= Y, X += Y, X = Y to mean what they do in C, 
98 

provided that X*Y, X/Y, X+Y, XY are already supported. 
99 

Nearly every Diderot program would be simplified by this. 
100 


101 
This will require duplication of the continuation of the conditional 
Put small 1D and 2D fields, when reconstructed specifically by tent 
102 
(but we should only duplicate over the liverange of the result of the 
and when differentiation is not needed, into faster texture buffers. 
103 
conditional. 
test/illustvr.diderot is good example of program that uses multiple 
104 

such 1D fields basically as lookuptablebased function evaluation 
105 


106 

expand trace in mid to low translation 
107 


108 

extend norm (exp) to all tensor types [DONE for vectors and matrices] 
109 


110 

determinant ("det") for tensor[3,3] 
111 


112 
add ":" for tensor dot product (contracts out two indices 
add ":" for tensor dot product (contracts out two indices 
113 
instead of one like •), valid for all pairs of tensors with 
instead of one like •), valid for all pairs of tensors with 
114 
at least two indices 
at least two indices 
115 


116 
============================== 
test/uninit.diderot: 
117 
other MEDIUM TERM ============ 
documents need for better compiler error messages when output variables 
118 
============================== 
are not initialized; the current messages are very cryptic 
119 


120 
want: warnings when "D" (reserved for differentiation) is declared as 
want: warnings when "D" (reserved for differentiation) is declared as 
121 
a variable name (get confusing error messages now) 
a variable name (get confusing error messages now) 
122 



support for Python interop and GUI 





Python/ctypes interface to runtime 





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


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

123 
============================== 
============================== 
124 

LONG TERM ==================== (make Diderot more interesting/attractive from 
125 

============================== a research standpoint) 
126 


127 

IL support for higherorder tensor values (matrices, etc). 
128 

tensor construction [DONE] 
129 

tensor indexing [DONE] 
130 

tensor slicing 
131 

verify that hessians work correctly [DONE] 
132 


133 
Better handling of variables that determines the scope of a variable 
Better handling of variables that determines the scope of a variable 
134 
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, 
136 
scope. Also prune out useless variables, which should include field 
scope. Also prune out useless variables, which should include field 
137 
variables after the translation to midil. 
variables after the translation to midil. 
138 


139 

test/vrkcomp2.diderot: Add support for code like 
140 

(F1 if x else F2)@pos 
141 

This will require duplication of the continuation of the conditional 
142 

(but we should only duplicate over the liverange of the result of the 
143 

conditional. 
144 


145 

[GLK:5] Want: nontrivial field expressions & functions: 
146 

image(2)[2] Vimg = load(...); 
147 

field#0(2)[] Vlen = Vimg ⊛ bspln3; 
148 

to get a scalar field of vector length, or 
149 

field#2(2)[] F = Fimg ⊛ bspln3; 
150 

field#0(2)[] Gmag = ∇F; 
151 

to get a scalar field of gradient magnitude, or 
152 

field#2(2)[] F = Fimg ⊛ bspln3; 
153 

field#0(2)[] Gmsq = ∇F•∇F; 
154 

to get a scalar field of squared gradient magnitude, which is simpler 
155 

to differentiate. However, there is value in having these, even if 
156 

the differentiation of them is not supported (hence the indication 
157 

of "field#0" for these above) 
158 


159 

Want: ability to apply "normalize" to a field itself, e.g. 
160 

field#0(2)[2] V = normalize(Vimg ⊛ ctmr); 
161 

so that V(x) = normalize((Vimg ⊛ ctmr)(x)). 
162 

Having this would simplify expression of standard LIC method, and 
163 

would also help express other vector field expressions that arise 
164 

in vector field feature exraction. 
165 


166 

Permit fields composition, especially for warping images by a 
167 

smooth field of deformation vectors 
168 

field#2(3)[3] warp = bspln3 ⊛ warpData; 
169 

field#2(3)[] F = bspln3 ⊛ img; 
170 

field#2(3)[] Fwarp = F ◦ warp; 
171 

So Fwarp(x) = F(warp(X)). Chain rule can be used for differentation 
172 


173 

Allow the convolution to be specified either as a single 1D kernel 
174 

(as we have it now): 
175 

field#2(3)[] F = bspln3 ⊛ img; 
176 

or, as a tensor product of kernels, one for each axis, e.g. 
177 

field#0(3)[] F = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img; 
178 

This is especially important for things like timevarying data, or 
179 

other multidimensional fields where one axis of the domain is very 
180 

different from the rest. What is very unclear is how, in such cases, 
181 

we should notate the gradient, when we only want to differentiate with 
182 

respect to some of the axes. 
183 


184 
co vs contra index distinction 
co vs contra index distinction 
185 


186 
some indication of tensor symmetry 
some indication of tensor symmetry 
198 
BUGS ================= 
BUGS ================= 
199 
====================== 
====================== 
200 



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"); 




201 
test/zslice2.diderot: 
test/zslice2.diderot: 
202 
// HEY (bug) bspln5 leads to problems ... 
// HEY (bug) bspln5 leads to problems ... 
203 
// uncaught exception Size [size] 
// uncaught exception Size [size] 