lp:adol-c
- Get this branch:
- bzr branch lp:adol-c
Branch merges
Branch information
Import details
This branch is an import of the Subversion branch from https://projects.coin-or.org/svn/ADOL-C/trunk.
Last successful import was .
Recent revisions
- 219. By utke
-
int and size_t are incompatible on some platforms - so be consistent and
list it in the manual as declared in the code with size_t - 215. By utke
-
various typos (advance -> advanced) , and rephrased the section on differetiating external functions trying to clarify things
- 213. By kulshres
-
Merge branch 'master' of 'gitclone' into svn
It may require some work for hessian patterns and advector to work
properly, since they were changed simultaneously by Andrea and ISigned-off-by: Kshitij Kulshreshtha <email address hidden>
The following commits were merged:
commit aeb288e939c7657
aeaf75a8cbb833f 9096ee2ce0
Author: Kshitij Kulshreshtha <email address hidden>
Date: Mon Aug 6 16:46:20 2012 +0200correct compilation error in message printing
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 0d6e681b19d116b
f0aca1d918fb567 aed49e1766
Author: Benjamin Letschert <email address hidden>
Date: Tue May 29 15:53:45 2012 +0200for floor_op and ceil_op in INDO+TIGHT mode ind_dom should be set zero
commit 35689c69c49e02f
b1b211e4f873851 ce734c839e
Author: Kshitij Kulshreshtha <email address hidden>
Date: Fri May 4 16:34:49 2012 +0200add a comment for clarification
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 710b93301d98610
239f17dc40ba8cb a6651f2ea5
Author: Kshitij Kulshreshtha <email address hidden>
Date: Fri May 4 14:21:49 2012 +0200add copyright notices
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 830f2cc75ba823c
2343f4a524eb3e9 0ab02bc278
Author: Kshitij Kulshreshtha <email address hidden>
Date: Thu May 3 13:59:32 2012 +0200Use the block allocation and don't write all locations to tape
Now that we ensure block structure while allocating an advector
we don't need to write all locations to tape in order to do
subscripting operations. We can use arithmetic.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 8599c889e5cd684
a7066600835df5f 0c033eb339
Author: Kshitij Kulshreshtha <email address hidden>
Date: Thu May 3 13:57:39 2012 +0200Allocate advector ensuring contiguous locations
use a dummy blocker to call the function before calling the vector
constructor. Also use the fill constructor since the initialise
constructor is experimental (for some reason) and a fill constructor
is called internally by gcc anyway. This way we won't have to ensure a
n+1 block and get holes.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit d62777e6070cf8b
80b7b32616a8741 86e6463919
Author: Kshitij Kulshreshtha <email address hidden>
Date: Thu May 3 13:50:48 2012 +0200Always put the last freed block to the front
This will make it efficienter if a block is deallocated right
one after the other. There will only be a search the first time
and all others will just work with the front of the list.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 33ae71a35fc04da
062c4ba85659cba f474caf1b8
Author: Kshitij Kulshreshtha <email address hidden>
Date: Fri Apr 27 17:36:41 2012 +0200changes in StoreManagemerL
ocintBlock to make it work right add StoreManagerLoc
intBlock to be the default StoreManager in
tape_handling. cpp All existing ADOL-C programs need to be tested again.
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit af30b922ee0a267
555fdcdfe66d082 3aa98688be
Author: Benjamin Letschert <email address hidden>
Date: Fri Apr 27 13:09:56 2012 +0200Adding new storem anagement StoreManagerLoc
intBlock to storemanager.h
and to tape_handling.cppcommit 7875317a03b08a2
729879edfaec7ae 5cd1363ef2
Author: Kshitij Kulshreshtha <email address hidden>
Date: Wed Feb 29 08:49:25 2012 +0100convert tape stats to size_t from uint
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 11e535ae503f596
1d0c62336d6942c 96e4b7c3ba
Author: Kshitij Kulshreshtha <email address hidden>
Date: Thu Jan 12 13:22:27 2012 +0100Revert "add missing arguments to ext_fct calls"
This reverts commit 00622f92b5444e8
1f0c3760eeee0e1 4d4e8e4787. This was temporary and wasn't supposed to be uploaded, some parts will
go into trunk.commit 37ca7b37f353815
3054dee40d3a75d 560b3aa81f
Author: Kshitij Kulshreshtha <email address hidden>
Date: Thu Jan 12 13:21:13 2012 +0100Revert "add safeguards and indo mode to ext_diff"
This reverts commit 507df83d988e3d6
4848eb035036909 4b0b2ce742. This was temporary for debugging and wasn't supposed to be uploaded
commit 507df83d988e3d6
4848eb035036909 4b0b2ce742
Author: Kshitij Kulshreshtha <email address hidden>
Date: Sat Dec 31 23:46:23 2011 +0100add safeguards and indo mode to ext_diff
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 00622f92b5444e8
1f0c3760eeee0e1 4d4e8e4787
Author: Kshitij Kulshreshtha <email address hidden>
Date: Sat Dec 31 00:47:17 2011 +0100add missing arguments to ext_fct calls
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 1c1c7347867e583
509716f7e2a8395 875d0e3d53
Author: Benjamin Letschert <email address hidden>
Date: Fri Dec 16 14:18:24 2011 +0100Bugfix in copy_index_domain
commit 69df20b8f0240e8
0515c7f0df456f1 24cef70cf8
Author: Kshitij Kulshreshtha <email address hidden>
Date: Mon Dec 5 11:32:58 2011 +0100shift around some code and add another constructor
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit b2c13089d4f5b45
e7626e04c17b2f7 bee08adc6a
Author: Kshitij Kulshreshtha <email address hidden>
Date: Tue Nov 29 17:29:47 2011 +0100const no more needed after last bugfix
add check for gradient w.r.t. rhs vector.
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 78ea1e83eb17163
fa0703e7b2b2896 59c4c6d42b
Author: Kshitij Kulshreshtha <email address hidden>
Date: Mon Nov 28 17:13:23 2011 +0100bugfix in reverse mode during ref_copyout
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 3a7c70061118c89
ad66afeec45d957 dbdaf002b3
Author: Kshitij Kulshreshtha <email address hidden>
Date: Mon Nov 28 17:12:39 2011 +0100Add one more assignment operator
this one is used when an adubref is copied into another adubref
as in from one advector to another advector, as this needs taping
however we don't need to override copy-constructor, because
construction can only be done when returning adubref.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 9c150d9aa825cf1
e5964cd15c7d38b 9a34984fce
Author: Kshitij Kulshreshtha <email address hidden>
Date: Fri Nov 25 12:49:14 2011 +0100changed aTmp*TRES to (aTmp==
0)?0:(aTmp* TRES) and similar occurences to avoid nans by 0*inf Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 0efabbcb41d8b17
f0fe55a35ca63ba 93913bccbe
Author: Kshitij Kulshreshtha <email address hidden>
Date: Fri Nov 25 12:32:36 2011 +0100compile with -std=gnu99 to get both _GNU_SOURCE and _ISOC99_SOURCE
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 45fa79bd4d8ae0f
4ef1e39b32f9627 1134e886f5
Author: Kshitij Kulshreshtha <email address hidden>
Date: Fri Nov 25 12:20:02 2011 +0100cast result of trunc to avoid warnings
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit b97ae47a80f7c00
34b8f29d14656f7 5abcd51b4d
Author: Kshitij Kulshreshtha <email address hidden>
Date: Thu Nov 24 17:53:11 2011 +0100counting down with unsigned variables is dangerous
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 5dd6a95ce0403f9
ac4b262a9546f86 e95a5489ac
Author: Kshitij Kulshreshtha <email address hidden>
Date: Thu Nov 24 17:52:40 2011 +0100reverse the order of indep vars and add reverse call
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 2d794980215e1ae
41af0e3cc6ed8a5 e97507a2f7
Author: Kshitij Kulshreshtha <email address hidden>
Date: Thu Nov 24 17:49:25 2011 +0100Add functions to print L and R and compute scalar product
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 889d35f671f2d2e
ea4dda182b0d903 746994fbcd
Author: Kshitij Kulshreshtha <email address hidden>
Date: Thu Nov 24 17:47:31 2011 +0100bugfixes in luexam
use less adoubles, loop variables for constant sized loops should
remain inactive size_t. counting down with size_t is bad.const-ness of advectors matters during active subscripting,
espacially when the lhs is a subscripted advector too, unless
there is a expression involving badouble on the rhs.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 0dadd11bf743e72
83a26613c150377 8c6997654a
Author: Kshitij Kulshreshtha <email address hidden>
Date: Wed Nov 23 17:18:05 2011 +0100tweak the example a bit
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 898afff0d43d869
41125dd2e71857c 619e3f28a4
Author: Kshitij Kulshreshtha <email address hidden>
Date: Wed Nov 23 17:17:43 2011 +0100If location buffer must contain n+2 free spaces
in case location buffer goes onto disk then check and write that
block before doing a subscript operation
if location buffer is smaller than n+2 then we cannot continue.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 1c2e4771972f0ce
4db38bd959c23d1 8c061be568
Author: Kshitij Kulshreshtha <email address hidden>
Date: Wed Nov 23 15:38:50 2011 +0100one more cast
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 71b506ae55f8d67
54f5c2a1f800220 cb5a5bb974
Author: Kshitij Kulshreshtha <email address hidden>
Date: Wed Nov 23 15:27:53 2011 +0100Compilation fixes for sparse configuration
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 73334f20d51e707
ee709a3602060aa 3351dc6c29
Author: Kshitij Kulshreshtha <email address hidden>
Date: Tue Nov 22 18:23:37 2011 +0100add an example that uses advector with permutations
Lu factorization with pivoting is implemented here. The same traced
procedure can be used for different matrices that require different
pivots.Zos_forward has been verified, WIP to verify derivatives.
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit c54da9e8cd03e99
d9362cc1a7eb551 101b5a4cb8
Author: Kshitij Kulshreshtha <email address hidden>
Date: Tue Nov 22 18:20:35 2011 +0100some more inline helpers for advector
the destructor need not contain anything.
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit d403b2d2b7e591d
b7c353c8c3ce551 e7d6174df0
Author: Kshitij Kulshreshtha <email address hidden>
Date: Tue Nov 22 18:19:58 2011 +0100implement condassign for adubref
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 696e18b5dcf9534
f7609243f085a50 8fba178b6c
Author: Kshitij Kulshreshtha <email address hidden>
Date: Tue Nov 22 18:19:09 2011 +0100bugfix in -= operator for adubref
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit b7dd8285c6fe826
2d07d0a5393c15b cca7eeae96
Author: Kshitij Kulshreshtha <email address hidden>
Date: Mon Nov 21 13:47:48 2011 +0100implement reference operations in forward and reverse mode
the referenced location is recorded in the taylor buffer of the
reference variable, and read from the taylor buffer before any
operation is performed. The operations are then performed directly
on the referenced location.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit d546c517676a420
65a98df782898f9 e3472836b2
Author: Kshitij Kulshreshtha <email address hidden>
Date: Mon Nov 21 13:45:49 2011 +0100add new operations for the tape to work with references
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 07508c193f53c6c
5912485dcb517ca 7c67b75234
Author: Kshitij Kulshreshtha <email address hidden>
Date: Mon Nov 21 13:45:17 2011 +0100Add a class to work as a reference to adouble and adub
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit e60d733c951a8a1
d621a1401744fa0 3db78e9b92
Author: Kshitij Kulshreshtha <email address hidden>
Date: Wed Mar 2 15:32:28 2011 +0100introduce active subscript operation in forward and reverse
subscript operation as previously defined on advector was not quite
correct in case of computed index that changed while being evaluated
from tape. In this case we need the tape to record the subscript
operation so that the correct taylor and adjoint values are propagated
even if the index changed. This is especially true for index computation
via the lookupindex() function, where the index depends on the arguments
x and y which may change their values.This active subscript is a copy_value operation not a copy_ref one.
So the result should not be used as an lvalue for any further operation.
In case the result is used as an lvalue the original entry in the
vector will not be altered, only the copied out version will be altered.Also the safe mode of integer propagation for sparsity pattern
computation is not supported, only the tight mode is supported.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 5764beb91000483
e02d4c6b52e67d1 77fff04a42
Author: Kshitij Kulshreshtha <email address hidden>
Date: Wed Mar 2 13:59:12 2011 +0100zeroing out the store is not necessary at all
and saves us some time by reducing memory access.
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 29d3460e0c7a6d2
839a25d8896c37f ff98ad6a3c
Author: Kshitij Kulshreshtha <email address hidden>
Date: Wed Mar 2 13:57:33 2011 +0100oplate.h is not needed after installation
so don't install it, and don't make it public api. it is totally
internal to ADOL-C itself.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit 0e3ac08ba9c6ac0
aa7d3820def9e62 e079a83e44
Author: Kshitij Kulshreshtha <email address hidden>
Date: Tue Mar 1 16:00:17 2011 +0100add missing include
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit b7364c7c2164981
ce9e198b69f6aef 8684388b43
Author: Kshitij Kulshreshtha <email address hidden>
Date: Tue Mar 1 15:46:56 2011 +0100Add new files to buildsystem.
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
commit aaefbdef3a6e007
92ce0c57acff67e 4f4ee3224b
Author: Kshitij Kulshreshtha <email address hidden>
Date: Tue Mar 1 15:38:59 2011 +0100New class advector for index lookup and subscripting
This is a container using std::vector<
adouble> . The subscripting
operations return references to adoubles, nothing needs to be written
onto the tape for this since the following operation using the adouble
reference will do the right thing.
The index lookup operation uses a sequence of n condassign_s ops which
are taped, and returns the index which matches the requirement as another
adouble.Signed-off-by: Kshitij Kulshreshtha <email address hidden>
- 212. By kulshres
-
remove DOS EOL characters
Signed-off-by: Kshitij Kulshreshtha <email address hidden>
Branch metadata
- Branch format:
- Branch format 7
- Repository format:
- Bazaar repository format 2a (needs bzr 1.16 or later)