1 | $Id$ |
---|
2 | /* -------------------------------------------------------------------------- |
---|
3 | CppAD: C++ Algorithmic Differentiation: Copyright (C) 2003-15 Bradley M. Bell |
---|
4 | |
---|
5 | CppAD is distributed under multiple licenses. This distribution is under |
---|
6 | the terms of the |
---|
7 | Eclipse Public License Version 1.0. |
---|
8 | |
---|
9 | A copy of this license is included in the COPYING file of this distribution. |
---|
10 | Please visit http://www.coin-or.org/CppAD/ for information on other licenses. |
---|
11 | -------------------------------------------------------------------------- */ |
---|
12 | |
---|
13 | $begin whats_new_15$$ |
---|
14 | $dollar @$$ |
---|
15 | $spell |
---|
16 | ind |
---|
17 | const |
---|
18 | enum |
---|
19 | ctor |
---|
20 | dimensioned |
---|
21 | cskip_op |
---|
22 | adolc |
---|
23 | zdouble |
---|
24 | expm1 |
---|
25 | eigen |
---|
26 | expm1 |
---|
27 | atanh |
---|
28 | acosh |
---|
29 | asinh |
---|
30 | src |
---|
31 | op |
---|
32 | onetape |
---|
33 | CppAD |
---|
34 | colpack |
---|
35 | hes |
---|
36 | cppad |
---|
37 | Jacobian |
---|
38 | gettimeofday |
---|
39 | cmake |
---|
40 | hpp |
---|
41 | dir |
---|
42 | includedirs |
---|
43 | libdirs |
---|
44 | datadir |
---|
45 | docdir |
---|
46 | std |
---|
47 | cxx |
---|
48 | erf |
---|
49 | var |
---|
50 | NumRes |
---|
51 | chrono |
---|
52 | aeps |
---|
53 | CondExpGt |
---|
54 | hasnan |
---|
55 | Adolc |
---|
56 | resize |
---|
57 | bool |
---|
58 | alloc |
---|
59 | adouble |
---|
60 | gmtime |
---|
61 | asctime |
---|
62 | ipopt |
---|
63 | fadbad |
---|
64 | sacado |
---|
65 | unreferenced |
---|
66 | azmul |
---|
67 | retape |
---|
68 | $$ |
---|
69 | |
---|
70 | $section CppAD Changes and Additions During 2015$$ |
---|
71 | |
---|
72 | $head Introduction$$ |
---|
73 | This section contains a list of the changes to CppAD during 2015 |
---|
74 | (in reverse order by date). |
---|
75 | The purpose of this section is to |
---|
76 | assist you in learning about changes between various versions of CppAD. |
---|
77 | |
---|
78 | $head 11-08$$ |
---|
79 | $list number$$ |
---|
80 | Fixed a bug in the $cref RevSparseJac$$ routine. |
---|
81 | To be specific, the argument $cref/r/RevSparseJac/r/$$ |
---|
82 | was transposed from what the documentation said. |
---|
83 | (This has no effect in the usual case where $icode r$$ is the identity.) |
---|
84 | $lnext |
---|
85 | Added the $cref bool_sparsity.cpp$$ examples which show how |
---|
86 | to conserve memory when computing sparsity patterns. |
---|
87 | $lnext |
---|
88 | Modified the $cref ipopt_solve$$ procedure to take |
---|
89 | advantage of the memory conserving sparsity pattern calculations |
---|
90 | when $cref/retape/ipopt_solve/options/Retape/$$ is false. |
---|
91 | $lnext |
---|
92 | Added the $cref/bit_per_unit/CppAD_vector/vectorBool/bit_per_unit/$$ |
---|
93 | function to the $code vectorBool$$ class. |
---|
94 | (This aids the memory conservation mentioned above.) |
---|
95 | $lend |
---|
96 | |
---|
97 | |
---|
98 | $head 11-06$$ |
---|
99 | It is often difficult to determine what cause a $code nan$$ result |
---|
100 | during an operation with an $cref ADFun$$ object. |
---|
101 | The new feature |
---|
102 | $cref/get_check_for_nan/check_for_nan/get_check_for_nan/$$ was |
---|
103 | added to make this easier. |
---|
104 | |
---|
105 | $head 10-21$$ |
---|
106 | There was a mistake in the documentation for $cref index_sort$$, |
---|
107 | the argument $cref/ind/index_sort/ind/$$ is not $code const$$. |
---|
108 | |
---|
109 | $head 10-16$$ |
---|
110 | Add a $cref/PrintFor optimization/WishList/PrintFor Optimization/$$ |
---|
111 | item to the wish list. |
---|
112 | |
---|
113 | $head 10-06$$ |
---|
114 | $list number$$ |
---|
115 | Add |
---|
116 | $cref/CPPAD_USE_CPLUSPLUS_2011 |
---|
117 | /preprocessor |
---|
118 | /Documented Here |
---|
119 | /CPPAD_USE_CPLUSPLUS_2011 |
---|
120 | /$$, |
---|
121 | $code CPPAD_NUMERIC_LIMITS$$, and |
---|
122 | $code CPPAD_STANDARD_MATH_UNARY$$, to |
---|
123 | the $cref preprocessor$$ section. |
---|
124 | In addition, remove checking that all user API preprocessor symbols |
---|
125 | are in this section form the $cref WishList$$. |
---|
126 | $lnext |
---|
127 | Alphabetize and make some corrections to |
---|
128 | $cref/list of examples/ListAllExamples/$$. |
---|
129 | $lnext |
---|
130 | The documentation for some of the $cref deprecated$$ features |
---|
131 | was missing the date when they were deprecated. |
---|
132 | This has been fixed; e.g., see |
---|
133 | $cref/Deprecated 2012-12-26/auto_tools/Deprecated 2012-12-26/$$. |
---|
134 | $lend |
---|
135 | |
---|
136 | |
---|
137 | $head 10-04$$ |
---|
138 | $list number$$ |
---|
139 | $cref base_require$$: |
---|
140 | Add the macro |
---|
141 | $cref/CPPAD_NUMERIC_LIMITS/base_limits/CPPAD_NUMERIC_LIMITS/$$ |
---|
142 | to aid in setting the numeric limits for a user defined $icode Base$$ class. |
---|
143 | $lnext |
---|
144 | $cref base_require$$: |
---|
145 | The $cref/quiet_NaN/numeric_limits/quiet_NaN/$$ function has been added |
---|
146 | to the CppAD $code numeric_limits$$. |
---|
147 | Note the reason for not using |
---|
148 | $cref/std::numeric_limits/numeric_limits/std::numeric_limits/$$. |
---|
149 | $lnext |
---|
150 | The $cref/nan(zero)/nan/nan(zero)/$$ function computes a $code nan$$ by |
---|
151 | dividing zero by zero which results in a warning when using some compilers. |
---|
152 | This function has been deprecated and the corresponding |
---|
153 | $cref WishList$$ item has been removed. |
---|
154 | $lnext |
---|
155 | Move documentation for $cref zdouble$$ to $cref deprecated$$ section |
---|
156 | and documentation for $cref numeric_limits$$ to $cref ADValued$$. |
---|
157 | $lnext |
---|
158 | Remove all uses of, and references to, |
---|
159 | $cref zdouble$$ from the $cref/examples/Example/$$. |
---|
160 | $lend |
---|
161 | |
---|
162 | $head 10-03$$ |
---|
163 | $cref base_require$$: |
---|
164 | It is no longer necessary to define the specialization for |
---|
165 | $codei%CppAD::epsilon<%Base>()%$$ for each $icode Base$$ type. |
---|
166 | |
---|
167 | $head 10-02$$ |
---|
168 | There was a bug in $code test_more/azmul.cpp$$ whereby the vector |
---|
169 | $code z$$ had the wrong dimension (in two places). This has been fixed. |
---|
170 | |
---|
171 | $head 09-28$$ |
---|
172 | $list number$$ |
---|
173 | Use the current $cref atomic_option$$ setting to determine |
---|
174 | which type of sparsity patterns to use for |
---|
175 | $cref/dependency/dependency.cpp/$$ calculations during |
---|
176 | $cref optimize$$ procedure. |
---|
177 | It used to be that the |
---|
178 | $cref/bool_sparsity_enum/atomic_option/atomic_sparsity/bool_sparsity_enum/$$ |
---|
179 | was used when |
---|
180 | $cref/pack_sparsity_enum/atomic_option/atomic_sparsity/pack_sparsity_enum/$$ |
---|
181 | was specified. |
---|
182 | $lnext |
---|
183 | It is not longer an error to take the derivative of the square root function, |
---|
184 | because the result may be the part of a |
---|
185 | $cref/conditional expression/CondExp/$$ that is not used. |
---|
186 | $lnext |
---|
187 | Update the $cref WishList$$ section. |
---|
188 | $lend |
---|
189 | |
---|
190 | |
---|
191 | $head 09-27$$ |
---|
192 | $list number$$ |
---|
193 | It is no longer necessary to use the $cref zdouble$$ class when |
---|
194 | computing with $cref/multiple levels of AD/mul_level/$$ |
---|
195 | $cref/conditional expressions/CondExp/$$ and |
---|
196 | $cref/reverse mode/reverse/$$. |
---|
197 | $lnext |
---|
198 | The $code zdouble$$ class has been deprecated. |
---|
199 | Use the $cref azmul$$ function for absolute zero (when it is needed). |
---|
200 | $lend |
---|
201 | |
---|
202 | $head 09-25$$ |
---|
203 | $cref base_require$$: |
---|
204 | $cref/absolute zero multiplication/base_require/Absolute Zero, azmul/$$ |
---|
205 | is now required for user defined base types. |
---|
206 | This makes it possible to combine |
---|
207 | $cref/conditional expression/CondExp/$$, |
---|
208 | $cref/multiple levels/mul_level/$$, |
---|
209 | $cref/reverse/Reverse/$$, and |
---|
210 | a base type that has standard ieee multiplication; e.g., $code double$$. |
---|
211 | In other words, not all multiplications will need to have an |
---|
212 | absolute zero (as is the case with the $cref zdouble$$ base class. |
---|
213 | |
---|
214 | $head 09-24$$ |
---|
215 | Fix some Visual Studio 2013 C++ level four $code /W4$$ warnings |
---|
216 | (previous warnings were are level 3). |
---|
217 | In addition, disable warning |
---|
218 | $code 4100$$ unreferenced formal parameter, |
---|
219 | and warning |
---|
220 | $code 4127$$ conditional expression is constant. |
---|
221 | |
---|
222 | $head 09-23$$ |
---|
223 | CppAD can optionally test its use with the external packages |
---|
224 | $cref/eigen/get_eigen.sh/$$, |
---|
225 | $cref/ipopt/get_ipopt.sh/$$, and |
---|
226 | $cref/colpack/get_colpack.sh/$$. |
---|
227 | In addition, it can compare its $cref speed$$ with the external AD packages |
---|
228 | $cref/adolc/get_adolc.sh/$$, |
---|
229 | $cref/fadbad/get_fadbad.sh/$$, and |
---|
230 | $cref/sacado/get_sacado.sh/$$. |
---|
231 | The scripts that download and install a local copy of these external packages |
---|
232 | have been modified to automatically skip installation |
---|
233 | when it has already been done. |
---|
234 | |
---|
235 | $head 09-21$$ |
---|
236 | Improve discussion of |
---|
237 | $cref/windows download and testing |
---|
238 | /download |
---|
239 | /Compressed Archives |
---|
240 | /Windows File Extraction and Testing |
---|
241 | /$$. |
---|
242 | |
---|
243 | $head 09-20$$ |
---|
244 | $list number$$ |
---|
245 | Add the $cref/cppad_profile_flag/cmake/cppad_profile_flag/$$ |
---|
246 | to the list of possible $code cmake$$ command arguments. |
---|
247 | $lnext |
---|
248 | More of the warnings generated by Visual Studio 2013 have been fixed. |
---|
249 | One remaining warning is about $code asctime$$ and $code gmtime$$ not being |
---|
250 | thread safe. |
---|
251 | $lend |
---|
252 | |
---|
253 | $head 09-19$$ |
---|
254 | $list number$$ |
---|
255 | There was a bug in the $cref/numeric_limits/base_alloc.hpp/numeric_limits/$$ |
---|
256 | section of the example user defined base type. |
---|
257 | This has been fixed. |
---|
258 | $lnext |
---|
259 | There were some compile and link errors when |
---|
260 | running the tests using Visual Studio 2013. |
---|
261 | These have been fixed. |
---|
262 | $lnext |
---|
263 | Many of the warnings generated by Visual Studio 2013 have been fixed. |
---|
264 | $lend |
---|
265 | |
---|
266 | $head 09-16$$ |
---|
267 | The conditional expressions, $cref CondExp$$, were not working |
---|
268 | for the type $code< CppAD::AD<adouble> >$$ where $code adouble$$ |
---|
269 | is the ADOL-C AD type. |
---|
270 | This has been fixed by adding a call to |
---|
271 | $cref/CPPAD_COND_EXP_REL/base_adolc.hpp/CondExpRel/$$ in |
---|
272 | $code base_adolc.hpp$$. |
---|
273 | |
---|
274 | $head 09-03$$ |
---|
275 | $list number$$ |
---|
276 | There was a bug in the $cref/vectorBool/CppAD_vector/vectorBool/$$ |
---|
277 | $cref/assignment/CppAD_vector/Assignment/$$. |
---|
278 | To be specific, |
---|
279 | it not allow a size zero vector to be assigned using a vector any other size. |
---|
280 | This has been fixed. |
---|
281 | $lnext |
---|
282 | The addition of the |
---|
283 | $cref/pack/atomic_option/atomic_sparsity/pack_sparsity_enum/$$ option |
---|
284 | on 08-31 introduced a bug in the calculation of $cref RevSparseHes$$. |
---|
285 | The $cref checkpoint.cpp$$ example was changed to demonstrate this |
---|
286 | problem and the bug was fixed. |
---|
287 | $lend |
---|
288 | |
---|
289 | $head 09-02$$ |
---|
290 | The $cref/dependency pattern/dependency.cpp/Dependency Pattern/$$ |
---|
291 | was not being computed correctly for the |
---|
292 | $cref sign$$, $cref Discrete$$, and $cref VecAD$$ operations. |
---|
293 | This has been fixed. |
---|
294 | This could have caused problems using |
---|
295 | $cref checkpoint$$ functions that used any of these operations. |
---|
296 | |
---|
297 | $head 08-31$$ |
---|
298 | $list number$$ |
---|
299 | Mention the fact that using checkpoint functions can make |
---|
300 | $cref/recordings faster/checkpoint/Purpose/Faster Recording/$$. |
---|
301 | $lnext |
---|
302 | Add the |
---|
303 | $cref/pack/atomic_option/atomic_sparsity/pack_sparsity_enum/$$ |
---|
304 | sparsity option for $cref atomic_base$$ operations. |
---|
305 | $lnext |
---|
306 | Add the pack sparsity option to |
---|
307 | $cref/checkpoint/checkpoint/sparsity/$$ functions. |
---|
308 | $lnext |
---|
309 | Added the $cref atomic_sparsity.cpp$$ example. |
---|
310 | $lnext |
---|
311 | Remove mention of OpenMP from $cref/thread_alloc::thread_num/ta_thread_num/$$ |
---|
312 | ($cref thread_alloc$$ never was OpenMP specific). |
---|
313 | $lend |
---|
314 | |
---|
315 | $head 08-30$$ |
---|
316 | $list number$$ |
---|
317 | The $cref/sparsity/atomic_ctor/atomic_base/sparsity/$$ |
---|
318 | argument was added to the $code atomic_base$$ constructor and the |
---|
319 | $cref/checkpoint/checkpoint/sparsity/$$ constructor. |
---|
320 | $lnext |
---|
321 | Make $cref atomic_norm_sq.cpp$$ an example with no set sparsity |
---|
322 | and $cref atomic_reciprocal.cpp$$ an example with no bool sparsity. |
---|
323 | $lnext |
---|
324 | Improve discussion of $code Independent$$ and |
---|
325 | $cref/parallel mode/Independent/Parallel Mode/$$. |
---|
326 | $lend |
---|
327 | |
---|
328 | $head 08-29$$ |
---|
329 | Some asserts in the $cref checkpoint$$ implementation were not using |
---|
330 | the CppAD $cref ErrorHandler$$. This has been fixed. |
---|
331 | |
---|
332 | $head 08-28$$ |
---|
333 | Free $cref checkpoint$$ function sparsity patters during |
---|
334 | $cref forward$$ operations that use its atomic operation. |
---|
335 | (They kept between sparsity calculations because they do not change.) |
---|
336 | |
---|
337 | $head 08-26$$ |
---|
338 | Fix a bug in $cref RevSparseJac$$ when used to compute sparsity pattern |
---|
339 | for a subset of the rows in a $cref checkpoint$$ function. |
---|
340 | |
---|
341 | $head 08-25$$ |
---|
342 | Reduce the amount of memory required for $cref checkpoint$$ functions |
---|
343 | (since sparsity patterns are now being held so they do not need to be |
---|
344 | recalculated). |
---|
345 | |
---|
346 | $head 08-20$$ |
---|
347 | Added an example that computes the sparsity pattern for a subset |
---|
348 | of the $cref/Jacobian/sparsity_sub.cpp/ForSparseJac/$$ and a subset |
---|
349 | of the $cref/Hessian/sparsity_sub.cpp/RevSparseHes/$$. |
---|
350 | |
---|
351 | $head 08-17$$ |
---|
352 | $list number$$ |
---|
353 | Do some optimization of the |
---|
354 | $cref checkpoint$$ feature so that sparsity patterns are |
---|
355 | stored and not recalculated. |
---|
356 | $lnext |
---|
357 | Fix a warning (introduced on 08-11) where the $code CppAD::vector$$ |
---|
358 | $cref/data/CppAD_vector/data/$$ function was being shadowed by |
---|
359 | a local variable. |
---|
360 | $lnext |
---|
361 | The source code control for CppAD has a link to $code compile$$, |
---|
362 | instead of real file. |
---|
363 | This sometimes caused problems with the |
---|
364 | deprecated $cref auto_tools$$ install procedure and has been fixed. |
---|
365 | $lend |
---|
366 | |
---|
367 | $head 08-16$$ |
---|
368 | $list number$$ |
---|
369 | Improve the documentation for checkpoint functions. |
---|
370 | To be specific, change the $cref/syntax/checkpoint/Syntax/$$ to use |
---|
371 | the name $icode atom_fun$$. |
---|
372 | In addition, include the fact that $icode atom_fun$$ must not |
---|
373 | be destructed for as along as the corresponding atomic operations are used. |
---|
374 | $lnext |
---|
375 | Add the $cref/size_var/checkpoint/size_var/$$ function |
---|
376 | to the checkpoint objects. |
---|
377 | $lend |
---|
378 | |
---|
379 | |
---|
380 | $head 08-09$$ |
---|
381 | Add the preservation of data to the specifications of a $code CppAD::vector$$ |
---|
382 | during a $cref/resize/CppAD_vector/resize/$$ when the |
---|
383 | capacity of the vector does not change. |
---|
384 | In addition, added and example of this to $cref cppad_vector.cpp$$. |
---|
385 | |
---|
386 | $head 08-06$$ |
---|
387 | The $cref zdouble$$ |
---|
388 | $cref/numeric_limits/base_limits/$$ |
---|
389 | were not being computed properly. |
---|
390 | This has been fixed. |
---|
391 | |
---|
392 | $head 07-31$$ |
---|
393 | Added the $cref sparse_sub_hes.cpp$$ example, |
---|
394 | a way to compute the sparsity for a subset of variables without |
---|
395 | using $cref/multiple levels of AD/mul_level/$$. |
---|
396 | |
---|
397 | $head 06-16$$ |
---|
398 | $list number$$ |
---|
399 | There were some |
---|
400 | $cref/unknown/cppad_assert/Unknown/$$ asserts |
---|
401 | when the sparsity pattern $icode p$$ in |
---|
402 | $cref/sparse_jacobian/sparse_jacobian/p/$$ and |
---|
403 | $cref/sparse_hessian/sparse_hessian/p/$$ was not properly dimensioned. |
---|
404 | These have been changed to |
---|
405 | $cref/known/cppad_assert/Known/$$ asserts to give better |
---|
406 | error reporting. |
---|
407 | $lnext |
---|
408 | In the special case where sparse Hessian $cref/work/sparse_hessian/work/$$ or |
---|
409 | sparse Jacobian $cref/work/sparse_hessian/work/$$ was specified and the |
---|
410 | set of elements to be computed was empty, the work vector is empty after |
---|
411 | the call (and it appears to need to be calculated on subsequent calls). |
---|
412 | This resulted in a bug when the sparsity pattern was not provided |
---|
413 | on subsequent calls (and has been fixed). |
---|
414 | $lend |
---|
415 | |
---|
416 | $head 06-11$$ |
---|
417 | $list number$$ |
---|
418 | Some C++11 features were not being taken advantage of after the change on |
---|
419 | $cref/05-10/whats_new_15/05-10/$$. To be specific, move semantics, |
---|
420 | the high resolution clock, and null pointers. |
---|
421 | This has been fixed. |
---|
422 | $lnext |
---|
423 | In the example $cref zdouble.cpp$$, the vector $code a1z$$ was not |
---|
424 | properly dimensioned. |
---|
425 | This has been fixed and the dimensions of all the variables have been |
---|
426 | clarified. |
---|
427 | $lend |
---|
428 | |
---|
429 | |
---|
430 | $head 06-09$$ |
---|
431 | Add an $cref/abort_op_index/Independent/abort_op_index/$$ |
---|
432 | item to the wish list. It has since been removed |
---|
433 | (domain errors may not affect the results due to |
---|
434 | $cref/conditional expressions/CondExp/$$). |
---|
435 | |
---|
436 | |
---|
437 | $head 06-07$$ |
---|
438 | Add a $cref/absolute zero/base_require/Absolute Zero, azmul/$$ item |
---|
439 | and a $cref numeric_limits$$ item to the wish list. |
---|
440 | The absolute zero item has been completed and the |
---|
441 | numeric limit item was modified on implementation. |
---|
442 | Remove the multiple directions with list item. |
---|
443 | |
---|
444 | $head 05-26$$ |
---|
445 | |
---|
446 | $subhead cond_exp_1$$ |
---|
447 | There was a problem using |
---|
448 | $cref/conditional expressions/CondExp/$$ |
---|
449 | with $cref/multiple levels of AD/mul_level/$$ where |
---|
450 | the result of the conditional expression might not be determined |
---|
451 | during forward mode. |
---|
452 | This would generate an assert of the form: |
---|
453 | $codei% |
---|
454 | Error detected by false result for |
---|
455 | IdenticalPar(%side%) |
---|
456 | at line %number% in the file |
---|
457 | %...%/cppad/local/cskip_op.hpp |
---|
458 | %$$ |
---|
459 | where $icode side$$ was $code left$$ or $code right$$ |
---|
460 | and $icode number$$ was the line number of an assert in $code cskip_op.hpp$$. |
---|
461 | This has been fixed. |
---|
462 | |
---|
463 | $subhead cond_exp_2$$ |
---|
464 | There was a problem with using |
---|
465 | $cref/conditional expressions/CondExp/$$ and $cref/reverse mode/reverse/$$ |
---|
466 | with $cref/multiple levels of AD/mul_level/$$. |
---|
467 | This was problem was represented by the file $code bug/cond_exp_2.sh$$. |
---|
468 | |
---|
469 | $list number$$ |
---|
470 | The problem above has been fixed by adding the base type $code zdouble$$, see |
---|
471 | $cref/CppAD motivation/zdouble/Motivation/CppAD/$$ for this new type. |
---|
472 | (It is no longer necessary to use $code zdouble$$ to get an absolute zero |
---|
473 | because CppAD now uses $cref azmul$$ where an absolute zero is required.) |
---|
474 | $lnext |
---|
475 | The sections |
---|
476 | $cref mul_level$$, |
---|
477 | $cref change_param.cpp$$, |
---|
478 | $cref mul_level.cpp$$, |
---|
479 | and $cref mul_level_ode.cpp$$ were changed to use $cref zdouble$$. |
---|
480 | $lnext |
---|
481 | The $cref/adolc/adolc_prefix/$$ multi-level examples |
---|
482 | $cref mul_level_adolc.cpp$$ and $cref mul_level_adolc_ode.cpp$$ |
---|
483 | were changed to mention the limitations because Adolc does not have an |
---|
484 | $cref/absolute zero/zdouble/Absolute Zero/$$. |
---|
485 | $lnext |
---|
486 | The example above were also changed so that AD variable names that |
---|
487 | indicated the level of AD for the variable. |
---|
488 | $lnext |
---|
489 | $cref base_require$$: |
---|
490 | The base type requirements were modified to include mention of |
---|
491 | $cref/absolute zero/base_require/Absolute Zero, azmul/$$. |
---|
492 | In addition, the base type requirements |
---|
493 | $cref/API warning/base_require/API Warning/$$ is now more informative. |
---|
494 | $lend |
---|
495 | |
---|
496 | $head 05-11$$ |
---|
497 | Reorganize the $cref unary_standard_math$$ documentation. |
---|
498 | |
---|
499 | $head 05-10$$ |
---|
500 | $list number$$ |
---|
501 | Add the exponential minus one function $cref log1p$$. |
---|
502 | $lnext |
---|
503 | $cref base_require$$: |
---|
504 | If you are defining your own base type, |
---|
505 | note that $cref/log1p/base_std_math/erf, asinh, acosh, atanh, expm1, log1p/$$ |
---|
506 | was added to the base type requirements. |
---|
507 | $lnext |
---|
508 | Use the single preprocessor flag |
---|
509 | $code CPPAD_USE_CPLUSPLUS_2011$$ to signal that the functions |
---|
510 | $cref/erf, asinh, acosh, atanh, expm1, log1p |
---|
511 | /base_std_math |
---|
512 | /erf, asinh, acosh, atanh, expm1, log1p |
---|
513 | /$$ |
---|
514 | are part of the base type requirements. |
---|
515 | $lend |
---|
516 | |
---|
517 | $head 05-09$$ |
---|
518 | $list number$$ |
---|
519 | Add the exponential minus one function $cref expm1$$. |
---|
520 | If you are defining your own base type, |
---|
521 | note that $cref/expm1/base_std_math/erf, asinh, acosh, atanh, expm1, log1p/$$ |
---|
522 | was added to the base type requirements. |
---|
523 | $lnext |
---|
524 | Fix some warnings about comparing signed and unsigned integers |
---|
525 | when using $cref/eigen/cppad_testvector/eigen/$$ for the CppAD test vector. |
---|
526 | (The eigen vector $code size()$$ function returns an $code int$$ instead of a |
---|
527 | $code size_t$$.) |
---|
528 | $lend |
---|
529 | |
---|
530 | $head 05-08$$ |
---|
531 | $list number$$ |
---|
532 | Add the inverse hyperbolic sine function $cref atanh$$. |
---|
533 | If you are defining your own base type, |
---|
534 | note that $cref/atanh/base_std_math/erf, asinh, acosh, atanh, expm1, log1p/$$ |
---|
535 | was added to the base type requirements. |
---|
536 | $lnext |
---|
537 | Fix a bug in the implementation of the $code acosh$$ |
---|
538 | multiple direction forward mode $cref forward_dir$$ |
---|
539 | (when compiler has $cref/acosh/acosh/CPPAD_USE_CPLUSPLUS_2011/$$). |
---|
540 | $lend |
---|
541 | |
---|
542 | $head 05-07$$ |
---|
543 | Add the inverse hyperbolic sine function $cref acosh$$. |
---|
544 | If you are defining your own base type, |
---|
545 | note that $cref/acosh/base_std_math/erf, asinh, acosh, atanh, expm1, log1p/$$ |
---|
546 | was added to the base type requirements. |
---|
547 | |
---|
548 | $head 05-05$$ |
---|
549 | Add the inverse hyperbolic sine function $cref asinh$$. |
---|
550 | If you are defining your own base type, |
---|
551 | note that $cref/asinh/base_std_math/erf, asinh, acosh, atanh, expm1, log1p/$$ |
---|
552 | was added to the base type requirements. |
---|
553 | |
---|
554 | $head 04-18$$ |
---|
555 | In the sparse jacobian and sparse hessian calculations, |
---|
556 | If $icode work$$ is present, and has already been computed, |
---|
557 | the sparsity pattern $icode p$$ is not used. |
---|
558 | This has been added to the documentation; see |
---|
559 | $cref/sparse jacobian/sparse_jacobian/work/p/$$ and |
---|
560 | $cref/sparse hessian/sparse_hessian/work/p/$$ documentation |
---|
561 | for $icode work$$ and $icode p$$. |
---|
562 | |
---|
563 | $head 03-13$$ |
---|
564 | Remove the syntax |
---|
565 | $codei% |
---|
566 | AD<%Base%> %y% = %x% |
---|
567 | %$$ |
---|
568 | for the $cref/AD constructor/ad_ctor/$$ documentation because it does not |
---|
569 | work when the constructor is $cref/explicit/ad_ctor/x/explicit/$$. |
---|
570 | Also document the restriction that the constructor in the |
---|
571 | $cref/assignment/ad_assign/$$ must be implicit. |
---|
572 | |
---|
573 | $head 03-06$$ |
---|
574 | The developers of the |
---|
575 | $href%https://github.com/kaskr/adcomp%TMB%$$ package reported that |
---|
576 | for large $cref ADFun$$ tapes, the $cref optimize$$ routine uses |
---|
577 | a large amount of memory because it allocates a standard set for |
---|
578 | each variable on the tape. These sets are only necessary for variables in |
---|
579 | $cref/conditional expressions/condexp/$$ that can be skipped once |
---|
580 | the independent variables have a set value. |
---|
581 | The problem has been reduced by using a NULL pointer for the empty set |
---|
582 | and similar changes. It still needs more work. |
---|
583 | |
---|
584 | $head 02-28$$ |
---|
585 | It used to be the case that the |
---|
586 | $cref/Reverse mode/reverse/$$ would propagate $cref nan$$ |
---|
587 | through the $cref/conditional expression/condexp/$$ |
---|
588 | case that is not used. |
---|
589 | For example, if |
---|
590 | $codep |
---|
591 | Independent(ax); |
---|
592 | AD<double> aeps = 1e-10; |
---|
593 | ay[0] = CondExpGt( ax[0], aeps, 1.0/ax[0], 1.0/aeps ); |
---|
594 | ADFun<double> f(ax, ay); |
---|
595 | $$ |
---|
596 | The corresponding reverse mode calculation, |
---|
597 | at $code x[0] = 0.0$$, would result in |
---|
598 | $codep |
---|
599 | Error detected by false result for |
---|
600 | ! ( hasnan(value) && check_for_nan_ ) |
---|
601 | $$ |
---|
602 | This has been fixed so that only the conditional expression case that is used |
---|
603 | affects the reverse mode results. |
---|
604 | The example $cref cond_exp.cpp$$ was changed to reflect this |
---|
605 | (a check for $code nan$$ was changed to a check for zero). |
---|
606 | Note that this fix only works when |
---|
607 | $cref/IdenticalPar/base_identical/Identical/IdenticalPar/$$ is true for the |
---|
608 | base type of the result in the conditional expression; e.g., |
---|
609 | one can still get a $code nan$$ effect from the case that is not selected |
---|
610 | when using $code AD< AD<double> >$$ conditional expressions. |
---|
611 | |
---|
612 | $head 02-18$$ |
---|
613 | If the compiler supports the c++11 feature |
---|
614 | $code std::chrono:high_resolution_clock$$ then use it for |
---|
615 | the $cref elapsed_seconds$$ function. |
---|
616 | |
---|
617 | $head 02-16$$ |
---|
618 | The new example $cref sub_sparse_hes.cpp$$ shows one way to |
---|
619 | compute a Hessian for a subset of variables without having to compute |
---|
620 | the sparsity pattern for the entire functions. |
---|
621 | |
---|
622 | $head 02-14$$ |
---|
623 | Fix another bug in the derivative calculations for the |
---|
624 | c++11 version of the error function; see |
---|
625 | $cref/CPPAD_USE_CPLUSPLUS_2011/erf/CPPAD_USE_CPLUSPLUS_2011/$$. |
---|
626 | |
---|
627 | $head 02-11$$ |
---|
628 | Fix a bug in the optimization of conditional expressions. To be specific, |
---|
629 | if $cref/NDEBUG/faq/Speed/NDEBUG/$$ is not defined, one could get |
---|
630 | an assert with the message: |
---|
631 | $codep |
---|
632 | Error detected by false result for |
---|
633 | var_index_ >= NumRes(op_) |
---|
634 | $$ |
---|
635 | |
---|
636 | $head 02-10$$ |
---|
637 | The change on $cref/2014-12-23/whats_new_14/12-23/$$ introduced a |
---|
638 | bug when the c++11 version of the error function was used with |
---|
639 | an $cref/optimized/optimize/$$ function. see |
---|
640 | $cref/CPPAD_USE_CPLUSPLUS_2011/erf/CPPAD_USE_CPLUSPLUS_2011/$$. |
---|
641 | There was also a bug in the sparsity calculations for when |
---|
642 | this erf function was included. |
---|
643 | These bugs have been fixed. |
---|
644 | |
---|
645 | $head 02-09$$ |
---|
646 | The test $code test_more/optimize.cpp$$ was failing on some systems |
---|
647 | because an exactly equality check should have been a near equal check. |
---|
648 | This has been fixed. |
---|
649 | |
---|
650 | $head 02-07$$ |
---|
651 | On some systems, the library |
---|
652 | corresponding to $code speed/src$$ could not be found. |
---|
653 | This library is only used for testing and so has been changed to always be |
---|
654 | static (hence does not need to be found at run time). |
---|
655 | |
---|
656 | $head 02-06$$ |
---|
657 | There was a bug in the coloring method change on |
---|
658 | $cref/2015-01-07/whats_new_15/01-07/$$. |
---|
659 | To be specific, $icode%work%.color_method%$$ was not being set |
---|
660 | to $code "cppad.symmetric"$$ after $icode%work%.color_method.clear()%$$. |
---|
661 | This has been fixed. |
---|
662 | |
---|
663 | $head 02-04$$ |
---|
664 | $list number$$ |
---|
665 | Enable the same install of CppAD to be used both with and without C++11 |
---|
666 | features; e.g., with both $code g++ --std=c++11$$ and with |
---|
667 | $code g++ --std=c++98$$. Previously if the |
---|
668 | $cref/cppad_cxx_flags/cmake/cppad_cxx_flags/$$ specified C++11, |
---|
669 | then it could only be used in that way. |
---|
670 | $lnext |
---|
671 | The $cref/cmake command/cmake/CMake Command/$$ now requires the version |
---|
672 | of $code cmake$$ to be greater than or equal 2.8 |
---|
673 | (due a bug in $code cmake$$ version 2.6). |
---|
674 | $lend |
---|
675 | |
---|
676 | $head 02-03$$ |
---|
677 | Improved the searching for the boost multi-threading library |
---|
678 | which is used for by the $cref team_bthread.cpp$$ case of the |
---|
679 | $cref thread_test.cpp$$ example and test. |
---|
680 | |
---|
681 | $head 02-02$$ |
---|
682 | Improve the documentation for the |
---|
683 | $cref/cmake command/cmake/CMake Command/$$ line options |
---|
684 | $codei% |
---|
685 | cmake_install_%dir% |
---|
686 | %$$ |
---|
687 | for $icode dir$$ equal to |
---|
688 | $code prefix$$, $code postfix$$, $code includedirs$$, $code libdirs$$, |
---|
689 | $code datadir$$, and $code docdir$$. |
---|
690 | |
---|
691 | $head 01-30$$ |
---|
692 | Fix bug in $cref link_sparse_hessian$$ speed test introduced on |
---|
693 | $cref/01-09/whats_new_15/01-09/$$ below. |
---|
694 | |
---|
695 | $head 01-29$$ |
---|
696 | Fix some warnings generated by $code g++ 4.9.2$$. |
---|
697 | |
---|
698 | $head 01-26$$ |
---|
699 | The change of global variables to local in |
---|
700 | $code cppad/local/op_code.hpp$$ on $cref/2014-50-14/whats_new_14/05-14/$$ |
---|
701 | created a bug in $cref parallel_ad$$ (some local statics needed to |
---|
702 | be initialized). This has been fixed. |
---|
703 | |
---|
704 | $head 01-23$$ |
---|
705 | There was a bug in the $cref cmake$$ install detection of compiler features. |
---|
706 | One symptom of this bug was that on systems that had the $code gettimeofday$$ |
---|
707 | function, the cmake install would sometimes report |
---|
708 | $codei% |
---|
709 | cppad_has_gettimeofday = 0 |
---|
710 | %$$ |
---|
711 | This has been fixed. |
---|
712 | |
---|
713 | $head 01-21$$ |
---|
714 | The deprecated $cref auto_tools$$ procedure had a bug in the |
---|
715 | detection of when the size of an $code unsigned int$$ |
---|
716 | was the same as the size of a $code size_t$$. This has been fixed. |
---|
717 | |
---|
718 | $head 01-20$$ |
---|
719 | $list number$$ |
---|
720 | The new $cref compare_change$$ interface has been created |
---|
721 | and the old $cref CompareChange$$ function has been deprecated; |
---|
722 | see the $cref compare_change.cpp$$ example. |
---|
723 | This enables one to determine the source code during taping |
---|
724 | that corresponds to changes in the comparisons during |
---|
725 | $cref/zero order forward/forward_zero/$$ operations; see |
---|
726 | $cref/abort_op_index/Independent/abort_op_index/$$. |
---|
727 | |
---|
728 | $lnext |
---|
729 | This new $cref compare_change$$ interface can detect comparison changes |
---|
730 | even if $cref/NDEBUG/Faq/Speed/NDEBUG/$$ is defined and |
---|
731 | even if $cref/f.optimize()/optimize/$$ has been called. |
---|
732 | The deprecated function $code CompareChange$$ used to always return zero after |
---|
733 | $codei% |
---|
734 | %f%.optimize() |
---|
735 | %$$ |
---|
736 | and was not even defined when $code NDEBUG$$ was defined. |
---|
737 | There was a resulting speed effect for this; see the wish list |
---|
738 | $cref/compare change/WishList/Comparison Changes and Optimization/$$ entry. |
---|
739 | |
---|
740 | $lnext |
---|
741 | The date when some features where deprecated has been added to the |
---|
742 | documentation. For example, see |
---|
743 | $cref/Deprecated 2006-12-17/include_deprecated/Deprecated 2006-12-17/$$. |
---|
744 | $lend |
---|
745 | |
---|
746 | $head 01-09$$ |
---|
747 | $list number$$ |
---|
748 | The change 01-07 below included (but did not mention) using |
---|
749 | a sparse, instead of full, structure for the Hessian in the test. |
---|
750 | This has also been done for the |
---|
751 | $cref/sparse Jacobian/link_sparse_jacobian/$$ test. |
---|
752 | $lnext |
---|
753 | For both the |
---|
754 | $cref/sparse_jacobian/link_sparse_jacobian/$$ and |
---|
755 | $cref/sparse_hessian/link_sparse_hessian/$$ tests, |
---|
756 | the sparse function is only chosen once |
---|
757 | (it used to be different for every repeat). |
---|
758 | This reduced the amount of computation not connected what is being tested. |
---|
759 | It also make the $cref/onetape/speed_main/option_list/onetape/$$ a valid |
---|
760 | option for these tests. |
---|
761 | $lnext |
---|
762 | There was a bug in the |
---|
763 | $cref/multiple direction forward/forward_dir/$$ routine. |
---|
764 | Results for function values that are |
---|
765 | $cref/parameter/parvar/$$ were not being computed properly |
---|
766 | (all the derivatives are zero in this case). |
---|
767 | This has been fixed. |
---|
768 | $lend |
---|
769 | |
---|
770 | |
---|
771 | $head 01-07$$ |
---|
772 | The following changes were merged in from the $code color_hes$$ branch: |
---|
773 | $list number$$ |
---|
774 | Specify the type of |
---|
775 | $cref/coloring/sparse_hessian/work/color_method/$$ |
---|
776 | for the sparse hessian calculations. |
---|
777 | To be specific, instead of $code "cppad"$$ and $code "colpack"$$, |
---|
778 | the choices are $code "cppad.symmetric"$$, $code "cppad.general"$$, |
---|
779 | and $code "colpack.star"$$. |
---|
780 | This is not compatible with the change on |
---|
781 | $cref/01-02/whats_new_15/01-02/$$, which was so recent |
---|
782 | that this should not be a problem. |
---|
783 | $lnext |
---|
784 | The $cref/n_sweep/link_sparse_hessian/n_sweep/$$ values were |
---|
785 | not being returned properly by |
---|
786 | $cref cppad_sparse_hessian.cpp$$ and |
---|
787 | $cref adolc_sparse_hessian.cpp$$. |
---|
788 | The CppAD version has been fixed and the ADOL-C version |
---|
789 | has been set to zero. |
---|
790 | $lnext |
---|
791 | The $cref link_sparse_hessian$$ example case was to sparse for good |
---|
792 | testing (by mistake). |
---|
793 | This has been fixed. |
---|
794 | $lnext |
---|
795 | Add $icode n_sweep$$ to |
---|
796 | $cref/link_sparse_hessian/link_sparse_hessian/n_sweep/$$ and |
---|
797 | $cref/speed_main/speed_main/Speed Results/n_sweep/$$. |
---|
798 | $lnext |
---|
799 | Change the $code cppad$$ sparse Hessian |
---|
800 | $cref/color_method/sparse_hessian/work/color_method/$$ |
---|
801 | to take advantage of the symmetry of the Hessian |
---|
802 | (in a similar fashion to the $code colpack$$ coloring method). |
---|
803 | $lend |
---|
804 | |
---|
805 | $head 01-02$$ |
---|
806 | Added to option to uses |
---|
807 | $cref/colpack/colpack_prefix/$$ for the sparse Hessian |
---|
808 | $cref/coloring method/sparse_hessian/work/color_method/$$; |
---|
809 | see the example $cref colpack_hes.cpp$$. |
---|
810 | |
---|
811 | $end |
---|