Home My Page Projects Code Snippets Project Openings diderot

# SCM Repository

[diderot] Diff of /branches/ein16/TODO
 [diderot] / branches / ein16 / TODO

# Diff of /branches/ein16/TODO

revision 1155, Sun May 8 14:43:30 2011 UTC revision 1445, Mon Jul 11 16:21:47 2011 UTC
# Line 1  Line 1
1  NOTE: GLK's approximate ranking of 5 most important tagged with  THIS TODO HAS BEEN MOVED TO THE DIDEROT WIKI.  PLEASE USE THAT ONE.
[GLK:1], [GLK:2], ...
2
3  ==============================  NOTE: GLK's approximate ranking of 8 most important tagged with
4  other SHORT TERM =============  (including needed for LIC)  [GLK:1], [GLK:2], ...
==============================
5
6  Add a clamp function, which takes three arguments; either three scalars:  ========================
7    clamp(x, minval, maxval)  = max(minval, min(maxval, x))  SHORT TERM ============= (*needed* for streamlines & tractography)
8  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).
9
10  Level of differentiability in field type should be statement about how  [GLK:2] Add sequence types (needed for evals & evecs)
much differentiation the program *needs*, rather than what the kernel
*provides*.  The needed differentiability can be less than or equal to
the provided differentiability.

[GLK:1] 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 output buffer
typechecking
IL and codegen
21
22  test/uninit.diderot:  [GLK:4] tensor fields from tensor images: Initially need at least
23  documents need for better compiler error messages when output variables  convolution on tensor[2,2] and tensor[3,3] (the same component-wise
24  are not initialized; the current messages are very cryptic  convolution as for vectors).
25
26  determinant ("det") for tensor[3,3]  ========================
27    SHORT-ISH TERM ========= (to make using Diderot less annoying to
28    ========================  program in, and slow to execute)
29
30  expand trace in mid to low translation  Allow ".ddro" file extensions in addition to ".diderot"
31
32  value-numbering optimization  Be able to output values of type tensor[2,2] and tensor[3,3];
33    (currently only scalars & vectors).  Want to add some regression tests
34    based on this and currently can't
35
36  Add type aliases for color types  [GLK:1] Proper handling of stabilize method
rgb = real{3}
rgba = real{4}
37
38  ==============================  Convolution on general tensor images (order > 2)
39  MEDIUM TERM ================== (including needed for streamlines & tractography)
40  ==============================  allow "*" to represent "modulate": per-component multiplication of
41    vectors, and vectors only (not tensors of order 2 or higher).  Once
42    sequences are implemented this should be removed: the operation is not
43    invariant WRT basis so it is not a legit vector computation.
44
45  [GLK:1] evals & evecs for symmetric tensor[3,3] (requires sequences)  implicit type promotion of integers to reals where reals are
46    required (e.g. not exponentiation "^")
47
48  [GLK:2] Save Diderot output to nrrd, instead of "mip.txt"  [Nick working on this] Save Diderot output to nrrd, instead of "mip.txt"
49    For grid of strands, save to similarly-shaped array    For grid of strands, save to similarly-shaped array
50    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
51    For ragged things (like tractography output), will need to save both    For ragged things (like tractography output), will need to save both
52      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
53      to index into complete list      to index into complete list
54
55  [GLK:3] Use of Teem's "hest" command-line 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
56  *always* "inside"; with various ways of mapping the known image values  *always* "inside"; with various ways of mapping the known image values
57  to non-existant index locations.  One possible syntax emphasizes that  to non-existant index locations.  One possible syntax emphasizes that
58  there is a index mapping function that logically precedes convolution:  there is a index mapping function that logically precedes convolution:
# Line 76  Line 61
61    F = bspln3 ⊛ (img ◦ mirror)    F = bspln3 ⊛ (img ◦ mirror)
62  where "◦" or "∘" is used to indicate function composition  where "◦" or "∘" is used to indicate function composition
63
64  extend norm (|exp|) to all tensor types [DONE for vectors and matrices]  Level of differentiability in field type should be statement about how
65    much differentiation the program *needs*, rather than what the kernel
66  ability to emit/track/record variables into dynamically re-sized  *provides*.  The needed differentiability can be less than or equal to
67  runtime buffer  the provided differentiability.
68
69  Want: allow X *= Y, X /= Y, X += Y, X -= Y to mean what they do in C,  Use ∇⊗ etc. syntax
70  provided that X*Y, X/Y, X+Y, X-Y are already supported.      syntax [DONE]
71  Nearly every Diderot program would be simplified by this.      typechecking
72        IL and codegen
73
74  [GLK:5] Want: non-trivial field expressions & functions:  Add type aliases for color types
75    image(2)[2] Vimg = load(...);      rgb = real{3}
76    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.
77
78  tensor fields: convolution on general tensor images  Revisit how images are created within the language.
79    The "load" operator should probably go away, and its strangs
80    that strings are there only as a way to refer to nrrd filenames
81
82  ==============================  ==============================
83  other MEDIUM TERM ============ (needed for particles)  MEDIUM TERM ================== (*needed* for particles)
84  ==============================  ==============================
85
86  Put small 1-D and 2-D fields, when reconstructed specifically by tent  [Lamont working on this] run-time birth of strands
and when differentiation is not needed, into faster texture buffers.
test/illust-vr.diderot is good example of program that uses multiple
such 1-D fields basically as lookup-table-based function evaluation

run-time birth of strands
87
88  "initially" supports lists  "initially" supports lists
89
90  "initially" supports lists of positions output from  "initially" supports lists of positions output from different
91  different initalization Diderot program  initalization Diderot program (or output from the same program;
92    e.g. using output of iso2d.diderot for one isovalue to seed the input
93    to another invocation of the same program)
94
95    [Lamont working on this] Communication between strands: they have to
96    be able to learn each other's state (at the previous iteration).
97    Early version of this can have the network of neighbors be completely
98    static (for running one strand/pixel image computations).  Later
99    version with strands moving through the domain will require some
100    spatial data structure to optimize discovery of neighbors.
101
102    ============================
103    MEDIUM-ISH TERM ============ (to make Diderot more useful/effective)
104    ============================
105
106    [GLK:5] Want code-generation working for tensors of order three.
107    Order three matters for edge detection in scalar fields (to get
108    second derivatives of gradient magnitude), second derivatives
109    of vector fields (for some feature extraction), and first
110    derivatives of diffusion tensor fields.
111
112  spatial data structure that permits strands' queries of neighbors  Python/ctypes interface to run-time
113
114  proper handling of stabilize method  support for Python interop and GUI
115
116  test/vr-kcomp2.diderot: Add support for code like  Allow integer exponentiation ("^2") to apply to square matrices,
117    to represent repeated matrix multiplication
118
119          (F1 if x else F2)@pos  Put small 1-D and 2-D fields, when reconstructed specifically by tent
120    and when differentiation is not needed, into faster texture buffers.
121    test/illust-vr.diderot is good example of program that uses multiple
122    such 1-D fields basically as lookup-table-based function evaluation
123
124  This will require duplication of the continuation of the conditional  extend norm (|exp|) to all tensor types [DONE for vectors and matrices]
125  (but we should only duplicate over the live-range of the result of the
126  conditional.  determinant ("det") for tensor[3,3]
127
128  add ":" for tensor dot product (contracts out two indices  add ":" for tensor dot product (contracts out two indices
129  instead of one like •), valid for all pairs of tensors with  instead of one like •), valid for all pairs of tensors with
130  at least two indices  at least two indices
131
132  ==============================  test/uninit.diderot:
133  other MEDIUM TERM ============  documents need for better compiler error messages when output variables
134  ==============================  are not initialized; the current messages are very cryptic
135
136  want: warnings when "D" (reserved for differentiation) is declared as  want: warnings when "D" (reserved for differentiation) is declared as
137  a variable name (get confusing error messages now)  a variable name (get confusing error messages now)
138
support for Python interop and GUI

Python/ctypes interface to run-time

Allow the convolution to be specified either as a single 1D kernel
(as we have it now):
field#2(3)[] F = bspln3 ⊛ img;
or, as a tensor product of kernels, one for each axis, e.g.
field#0(3)[] F = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img;
This is especially important for things like time-varying data, or
other multi-dimensional fields where one axis of the domain is very
different from the rest.  What is very unclear is how, in such cases,
we should notate the gradient, when we only want to differentiate with
respect to some of the axes.

==============================
LONG TERM ====================
139  ==============================  ==============================
140    LONG TERM ==================== (make Diderot more interesting/attractive from
141    ==============================  a research standpoint)
142
143    IL support for higher-order tensor values (matrices, etc).
144        tensor construction [DONE]
145        tensor indexing [DONE]
146        tensor slicing
147
148  Better handling of variables that determines the scope of a variable  Better handling of variables that determines the scope of a variable
149  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 172  Line 151
151  scope.  Also prune out useless variables, which should include field  scope.  Also prune out useless variables, which should include field
152  variables after the translation to mid-il.  variables after the translation to mid-il.
153
154    test/vr-kcomp2.diderot: Add support for code like
155            (F1 if x else F2)@pos
156    This will require duplication of the continuation of the conditional
157    (but we should only duplicate over the live-range of the result of the
158    conditional.
159
160    [GLK:7] Want: non-trivial field expressions & functions.
161    scalar fields from scalar fields F and G:
162      field#0(2)[] X = (sin(F) + 1.0)/2;
163      field#0(2)[] X = F*G;
164    scalar field of vector field magnitude:
166      field#0(2)[] Vlen = |Vimg ⊛ bspln3|;
167    field of normalized vectors (for LIC and vector field feature extraction)
168      field#2(2)[2] F = ...
169      field#0(2)[2] V = normalize(F);
170    scalar field of gradient magnitude (for edge detection))
171      field#2(2)[] F = Fimg ⊛ bspln3;
172      field#0(2)[] Gmag = |∇F|;
173    scalar field of squared gradient magnitude (simpler to differentiate):
174      field#2(2)[] F = Fimg ⊛ bspln3;
175      field#0(2)[] Gmsq = ∇F•∇F;
176    There is value in having these, even if the differentiation of them is
177    not supported (hence the indication of "field#0" for these above)
178
179    Introduce region types (syntax region(d), where d is the dimension of the
180    region.  One useful operator would be
181            dom : field#k(d)[s] -> region(d)
182    Then the inside test could be written as
183            pos ∈ dom(F)
184    We could further extend this approach to allow geometric definitions of
185    regions.  It might also be useful to do inside tests in world space,
187
188  co- vs contra- index distinction  co- vs contra- index distinction
189
190  some indication of tensor symmetry  Permit field composition:
191      field#2(3)[3] warp = bspln3 ⊛ warpData;
192      field#2(3)[] F = bspln3 ⊛ img;
193      field#2(3)[] Fwarp = F ◦ warp;
194    So Fwarp(x) = F(warp(X)).  Chain rule can be used for differentation.
195    This will be instrumental for expressing non-rigid registration
196    methods (but those will require co-vs-contra index distinction)
197
198    Allow the convolution to be specified either as a single 1D kernel
199    (as we have it now):
200      field#2(3)[] F = bspln3 ⊛ img;
201    or, as a tensor product of kernels, one for each axis, e.g.
202      field#0(3)[] F = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img;
203    This is especially important for things like time-varying fields
204    and the use of scale-space in field visualization: one axis of the
205    must be convolved with a different kernel during probing.
206    What is very unclear is how, in such cases, we should notate the
207    gradient, when we only want to differentiate with respect to some
208    subset of the axes.  One ambitious idea would be:
209      field#0(3)[] Ft = (bspln3 ⊗ bspln3 ⊗ tent) ⊛ img; // 2D time-varying field
210      field#0(2)[] F = lambda([x,y], Ft([x,y,42.0]))     // restriction to time=42.0
212
213    representation of tensor symmetry
214  (have to identify the group of index permutations that are symmetries)  (have to identify the group of index permutations that are symmetries)
215
216  dot works on all tensors  dot works on all tensors
217
218  outer works on all tensors  outer works on all tensors
219
220    Help for debugging Diderot programs: need to be able to uniquely
221    identify strands, and for particular strands that are known to behave
222    badly, do something like printf or other logging of their computations
224
225    Permit writing dimensionally general code: Have some statement of the
226    dimension of the world "W" (or have it be learned from one particular
227    field of interest), and then able to write "vec" instead of
228    "vec2/vec3", and perhaps "tensor[W,W]" instead of
229    "tensor[2,2]/tensor[3,3]"
230
231    Traits: all things things that have boilerplate code (especially
232    volume rendering) should be expressed in terms of the unique
233    computational core.  Different kinds of streamline/tractography
234    computation will be another example, as well as particle systems.
235
236  Einstein summation notation  Einstein summation notation
237
238  "tensor comprehension" (like list comprehension)  "tensor comprehension" (like list comprehension)
239
240    Fields coming from different sources of data:
241    * triangular or tetrahedral meshes over 2D or 3D domains (of the
242      source produced by finite-element codes; these will come with their
243      own specialized kinds of reconstruction kernels, called "basis
244      functions" in this context)
245    * Large point clouds, with some radial basis function around each point,
246      which will be tuned by parameters of the point (at least one parameter
247      giving some notion of radius)
248
249  ======================  ======================
250  BUGS =================  BUGS =================
251  ======================  ======================
# Line 194  Line 255
255  //  uncaught exception Size [size]  //  uncaught exception Size [size]
256  //    raised at c-target/c-target.sml:47.15-47.19  //    raised at c-target/c-target.sml:47.15-47.19
257  //field#4(3)[] F = img ⊛ bspln5;  //field#4(3)[] F = img ⊛ bspln5;
258

Legend:
 Removed from v.1155 changed lines Added in v.1445