Home My Page Projects Code Snippets Project Openings diderot

# SCM Repository

[diderot] Diff of /branches/lamont/TODO
 [diderot] / branches / lamont / TODO

# Diff of /branches/lamont/TODO

revision 1120, Thu May 5 08:18:26 2011 UTC revision 1388, Fri Jun 24 14:45:33 2011 UTC
# Line 1  Line 1
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 [DONE]
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:2] 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 higher-order tensor values (matrices, etc).  [GLK:3] 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 re-sized
20      syntax [DONE]  runtime buffer
typechecking
IL and codegen
21
22  test/uninit.diderot:  tensor fields: convolution on general tensor images (order > 1)
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    SHORT-ISH TERM ========= (to make using Diderot less annoying to
26    ========================  program in, and slow to execute)
27
28  expand trace in mid to low translation  value-numbering optimization [DONE]
29
30  value-numbering 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] Proper handling of stabilize method
37  MEDIUM TERM ================== (including needed for streamlines & tractography)
38  ==============================  allow "*" to represent "modulate": per-component multiplication of
39    vectors, and vectors only (not tensors of order 2 or higher).  Once
40    sequences are implemented this should be removed: the operation is not
41    invariant WRT basis so it is not a legit vector computation.
42
43  [GLK:1] evals & evecs for symmetric tensor[3,3] (requires sequences)  implicit type promotion of integers to reals where reals are
44    required (e.g. not exponentiation "^")
45
46  [GLK:2] Save Diderot output to nrrd, instead of "mip.txt"  [GLK:4] Save Diderot output to nrrd, instead of "mip.txt"
47    For grid of strands, save to similarly-shaped array    For grid of strands, save to similarly-shaped array
48    For list of strands, save to long 1-D (or 2-D for non-scalar output) list    For list of strands, save to long 1-D (or 2-D for non-scalar output) list
49    For ragged things (like tractography output), will need to save both    For ragged things (like tractography output), will need to save both
50      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
51      to index into complete list      to index into complete list
52
53  [GLK:3] Use of Teem's "hest" command-line parser for getting  [GLK:5] Use of Teem's "hest" command-line parser for getting
54  any input variables that are not defined in the source file  any "input" variables that are not defined in the source file. [DONE]
55
56  [GLK:4] ability to declare a field so that probe positions are  [GLK:6] ability to declare a field so that probe positions are
57  *always* "inside"; with various ways of mapping the known image values  *always* "inside"; with various ways of mapping the known image values
58  to non-existant index locations.  One possible syntax emphasizes that  to non-existant index locations.  One possible syntax emphasizes that
59  there is a index mapping function that logically precedes convolution:  there is a index mapping function that logically precedes convolution:
# Line 76  Line 62
62    F = bspln3 ⊛ (img ◦ mirror)    F = bspln3 ⊛ (img ◦ mirror)
63  where "◦" or "∘" is used to indicate function composition  where "◦" or "∘" is used to indicate function composition
64
65  extend norm (|exp|) to all tensor types [DONE for vectors and matrices]  Level of differentiability in field type should be statement about how
66    much differentiation the program *needs*, rather than what the kernel
67  ability to emit/track/record variables into dynamically re-sized  *provides*.  The needed differentiability can be less than or equal to
68  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, X-Y are already supported.
Nearly every Diderot program would be simplified by this.
69
70  Want: non-trivial field expressions & functions:  Use ∇⊗ etc. syntax
71    image(2)[2] Vimg = load(...);      syntax [DONE]
72    field#0(2)[] Vlen = |Vimg ⊛ bspln3|;      typechecking
73  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.
74
75  tensor fields: convolution on general tensor images  Add type aliases for color types
76        rgb = real{3}
77        rgba = real{4}
78
79  ==============================  ==============================
80  other MEDIUM TERM ============ (needed for particles)  MEDIUM TERM ================== (*needed* for particles)
81  ==============================  ==============================
82
[GLK:5] run-time death of strands; test/iso2d.diderot provides
great initial test for "die" command.

Put small 1-D and 2-D fields, reconstructed by tent, into
faster texture buffers

83  run-time birth of strands  run-time birth of strands
84
85  "initially" supports lists  "initially" supports lists
86
87  "initially" supports lists of positions output from  "initially" supports lists of positions output from different
88  different initalization Diderot program  initalization Diderot program (or output from the same program;
89    e.g. using output of iso2d.diderot for one isovalue to seed the input
90    to another invocation of the same program)
91
92    Communication between strands: they have to be able to learn each
93    other's state (at the previous iteration).  Early version of this can
94    have the network of neighbors be completely static (for running one
95    strand/pixel image computations).  Later version with strands moving
96    through the domain will require some spatial data structure to
97    optimize discovery of neighbors.
98
99    ============================
100    MEDIUM-ISH TERM ============ (to make Diderot more useful/effective)
101    ============================
102
103  spatial data structure that permits strands' queries of neighbors  Python/ctypes interface to run-time
104
105  proper handling of stabilize method  support for Python interop and GUI
106
107  test/vr-kcomp2.diderot: Add support for code like  Allow integer exponentiation ("^2") to apply to square matrices,
108    to represent repeated matrix multiplication
109
110          (F1 if x else F2)@pos  Put small 1-D and 2-D fields, when reconstructed specifically by tent
111    and when differentiation is not needed, into faster texture buffers.
112    test/illust-vr.diderot is good example of program that uses multiple
113    such 1-D fields basically as lookup-table-based function evaluation
114
115  This will require duplication of the continuation of the conditional  expand trace in mid to low translation [DONE]
116  (but we should only duplicate over the live-range of the result of the
117  conditional.  extend norm (|exp|) to all tensor types [DONE for vectors and matrices]
118
119    determinant ("det") for tensor[3,3]
120
121  add ":" for tensor dot product (contracts out two indices  add ":" for tensor dot product (contracts out two indices
122  instead of one like •), valid for all pairs of tensors with  instead of one like •), valid for all pairs of tensors with
123  at least two indices  at least two indices
124
125  ==============================  test/uninit.diderot:
126  other MEDIUM TERM ============  documents need for better compiler error messages when output variables
127  ==============================  are not initialized; the current messages are very cryptic
128
129  want: warnings when "D" (reserved for differentiation) is declared as  want: warnings when "D" (reserved for differentiation) is declared as
130  a variable name (get confusing error messages now)  a variable name (get confusing error messages now)
131
support for Python interop and GUI

Python/ctypes interface to run-time

==============================
LONG TERM ====================
132  ==============================  ==============================
133    LONG TERM ==================== (make Diderot more interesting/attractive from
134    ==============================  a research standpoint)
135
136    IL support for higher-order tensor values (matrices, etc).
137        tensor construction [DONE]
138        tensor indexing [DONE]
139        tensor slicing
140        verify that hessians work correctly [DONE]
141
142  Better handling of variables that determines the scope of a variable  Better handling of variables that determines the scope of a variable
143  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,
# Line 162  Line 145
145  scope.  Also prune out useless variables, which should include field  scope.  Also prune out useless variables, which should include field
146  variables after the translation to mid-il.  variables after the translation to mid-il.
147
148    test/vr-kcomp2.diderot: Add support for code like
149            (F1 if x else F2)@pos
150    This will require duplication of the continuation of the conditional
151    (but we should only duplicate over the live-range of the result of the
152    conditional.
153
154    [GLK:7] Want: non-trivial field expressions & functions.
155    scalar fields from scalar fields F and G:
156      field#0(2)[] X = (sin(F) + 1.0)/2;
157      field#0(2)[] X = F*G;
158    scalar field of vector field magnitude:
160      field#0(2)[] Vlen = |Vimg ⊛ bspln3|;
161    field of normalized vectors (for LIC and vector field feature extraction)
162      field#2(2)[2] F = ...
163      field#0(2)[2] V = normalize(F);
164    scalar field of gradient magnitude (for edge detection))
165      field#2(2)[] F = Fimg ⊛ bspln3;
166      field#0(2)[] Gmag = |∇F|;
167    scalar field of squared gradient magnitude (simpler to differentiate):
168      field#2(2)[] F = Fimg ⊛ bspln3;
169      field#0(2)[] Gmsq = ∇F•∇F;
170    There is value in having these, even if the differentiation of them is
171    not supported (hence the indication of "field#0" for these above)
172
173    Introduce region types (syntax region(d), where d is the dimension of the
174    region.  One useful operator would be
175            dom : field#k(d)[s] -> region(d)
176    Then the inside test could be written as
177            pos ∈ dom(F)
178    We could further extend this approach to allow geometric definitions of
179    regions.  It might also be useful to do inside tests in world space,
181
182  co- vs contra- index distinction  co- vs contra- index distinction
183
184  some indication of tensor symmetry  Permit field composition:
185      field#2(3)[3] warp = bspln3 ⊛ warpData;
186      field#2(3)[] F = bspln3 ⊛ img;
187      field#2(3)[] Fwarp = F ◦ warp;
188    So Fwarp(x) = F(warp(X)).  Chain rule can be used for differentation.
189    This will be instrumental for expressing non-rigid registration
190    methods (but those will require co-vs-contra index distinction)
191
192    Allow the convolution to be specified either as a single 1D kernel
193    (as we have it now):
194      field#2(3)[] F = bspln3 ⊛ img;
195    or, as a tensor product of kernels, one for each axis, e.g.
196      field#0(3)[] F = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img;
197    This is especially important for things like time-varying fields
198    and the use of scale-space in field visualization: one axis of the
199    must be convolved with a different kernel during probing.
200    What is very unclear is how, in such cases, we should notate the
201    gradient, when we only want to differentiate with respect to some
202    subset of the axes.  One ambitious idea would be:
203      field#0(3)[] Ft = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img; // 2D time-varying field
204      field#0(2)[] F = lambda([x,y], Ft([x,y,42.0]))     // restriction to time=42.0
206
207    Tensors of order 3 (e.g. gradients of diffusion tensor fields, or
208    hessians of vector fields) and order 4 (e.g. Hessians of diffusion
209    tensor fields).
210
211    representation of tensor symmetry
212  (have to identify the group of index permutations that are symmetries)  (have to identify the group of index permutations that are symmetries)
213
214  dot works on all tensors  dot works on all tensors
215
216  outer works on all tensors  outer works on all tensors
217
218    Help for debugging Diderot programs: need to be able to uniquely
219    identify strands, and for particular strands that are known to behave
220    badly, do something like printf or other logging of their computations
222
223    Permit writing dimensionally general code: Have some statement of the
224    dimension of the world "W" (or have it be learned from one particular
225    field of interest), and then able to write "vec" instead of
226    "vec2/vec3", and perhaps "tensor[W,W]" instead of
227    "tensor[2,2]/tensor[3,3]"
228
229    Traits: all things things that have boilerplate code (especially
230    volume rendering) should be expressed in terms of the unique
231    computational core.  Different kinds of streamline/tractography
232    computation will be another example, as well as particle systems.
233
234  Einstein summation notation  Einstein summation notation
235
236  "tensor comprehension" (like list comprehension)  "tensor comprehension" (like list comprehension)
237
238    Fields coming from different sources of data:
239    * triangular or tetrahedral meshes over 2D or 3D domains (of the
240      source produced by finite-element codes; these will come with their
241      own specialized kinds of reconstruction kernels, called "basis
242      functions" in this context)
243    * Large point clouds, with some radial basis function around each point,
244      which will be tuned by parameters of the point (at least one parameter
245      giving some notion of radius)
246
247  ======================  ======================
248  BUGS =================  BUGS =================
249  ======================  ======================
250
// HEY (BUG?) shouldn't it be a type error to load this 2-D array of
// 2-vectors into a 2-D *scalar* field?  Instead, get:
//   uncaught exception Fail [Fail: Error in compiling lic.diderot]
//     raised at driver/main.sml:31.39-31.76