|
| 1 | +## SeDuMi 1.3.5 (2021-06-15) |
| 2 | + |
| 3 | +> https://github.com/sqlp/sedumi/compare/v1.3.4...v1.3.5 |
| 4 | +
|
| 5 | +- Merge of "MEX-free implementation of pretransfo and posttransfo" branch by |
| 6 | + Michael C. Grant. |
| 7 | +- Improved `install_sedumi.m` function. |
| 8 | +- Use GitHub Actions for CI tests. |
| 9 | +- Bugs fixed: |
| 10 | + - [#64: function handle problem in minpsdeig.m](https://github.com/sqlp/sedumi/issues/64) |
| 11 | + - [#18: Typo in eigK.m](https://github.com/sqlp/sedumi/issues/18) |
| 12 | + - [#15: results are not apparently deterministic?](https://github.com/sqlp/sedumi/issues/15) |
| 13 | + |
| 14 | +## SeDuMi 1.3.4 (2020-01-09) |
| 15 | + |
| 16 | +> https://github.com/sqlp/sedumi/compare/v1.3.2...v1.3.4 |
| 17 | +> |
| 18 | +> Note: version 1.3.3 was not officially released. |
| 19 | +
|
| 20 | +- Many many bug fixes in seven years. |
| 21 | + |
| 22 | + |
| 23 | +## SeDuMi 1.3.2 (2013-07-24, Michael Grant) |
| 24 | + |
| 25 | +> https://github.com/sqlp/sedumi/compare/v1.3.1...v1.3.2 |
| 26 | +
|
| 27 | +- Merged changes from CVX and SeDuMi/Octave into the distribution. |
| 28 | +- Further changes to suppress Matlab Code Analyzer warnings. |
| 29 | +- Removed `matrix.h` dependency from `givens.h` for Octave compatibility; |
| 30 | + fixed assertion problem with `givensrot.c`. |
| 31 | + |
| 32 | + |
| 33 | +## SeDuMi 1.3.1 (2013-04-04) |
| 34 | + |
| 35 | +> https://github.com/sqlp/sedumi/compare/v1.3.0...v1.3.1 |
| 36 | +
|
| 37 | +- Corrected bugs in `sedumi.m` and `getada.m`. |
| 38 | +- Updated .m files as per mlint suggestions. |
| 39 | +- Commented `length(b)` vs `length(c)` check in pretransfo. |
| 40 | +- Recompiled windows mex files, updated `install_sedumi.m` for MATLAB v8 |
| 41 | + |
| 42 | + |
| 43 | +## New features in SeDuMi 1.1 since 1.05R5 |
| 44 | + |
| 45 | +- 2010-02-15 Fix: global variables are used for some large data structures |
| 46 | + (ADA, At). |
| 47 | + |
| 48 | +- 2010-02-15 Fix: `getada1` and `getada2` have been backported to Matlab, |
| 49 | + which actually increases speed, due to parallel BLAS. |
| 50 | + |
| 51 | +- 2009-11-25 Fix: computation in `getsymbada.m` has been reorganized for speed. |
| 52 | + |
| 53 | +- 2009-11-15 Fix: the `psdeig`, `psdinvscale`, `psdjmulv`, `psdfactor`, |
| 54 | + `psdscale` and `triumtriu` mex files were replaced with a pure Matlab |
| 55 | + equivalent. |
| 56 | + |
| 57 | +- 2009-09-05 Fix: SeDuMi now uses wall clock time instead of cputime, as the |
| 58 | + CPU time is misleading in parallel processing environments. |
| 59 | + It is reported in `info.walltime`. `info.cpusec` contains the total cpu time, |
| 60 | + which may be more than the wall time. |
| 61 | + |
| 62 | +- 2009-08-05 Fix: Based on feedback from Didier Henrion and others, the rank |
| 63 | + and infeasibility detection code has been modified. |
| 64 | + |
| 65 | +- 2009-05-10 Fix: Michael C. Grant modified some files to compile under 64-bit |
| 66 | + versions. |
| 67 | + |
| 68 | +- 2009-05-10 Fix: A new installer script is provided by Michael C. Grant. |
| 69 | + |
| 70 | +- 2009-05-10 Fix: Some BLAS calls are fixed in `blkaux.c`. |
| 71 | + |
| 72 | +- 2008-04-09 Fix: BLAS is used to improve performance. |
| 73 | + |
| 74 | +- 2008-04-09 Fix: SeDuMi now works on 64-bit operating systems. |
| 75 | + |
| 76 | +- 2006-10-14 Fix: A bugfix was contributed by Michael C. Grant regarding the |
| 77 | + handling of 2-dimensional Lorentz cones. |
| 78 | + |
| 79 | +- 2006-10-10 Fix: A new, completely rewritten SDPA format reader is provided in |
| 80 | + the conversion folder. It is fast and it follows the SDPA standard. |
| 81 | + |
| 82 | +- 2006-10-10 Fix: Michael Grant corrected a bug in the postprocessing phase. |
| 83 | + |
| 84 | +- 2006-10-10 Fix: The semidefinite preprocessing assumed that there are no |
| 85 | + free variables. |
| 86 | + |
| 87 | +- 2006-10-10 Fix: Matlab R2006b does not support `fprintf` with `fid=0`, |
| 88 | + (i.e., no output). A workaround is provided. Thanks to Johan Löfberg. |
| 89 | + |
| 90 | +- 2006-10-10 Fix: The solutions were incorrect if the error measures were turned on. |
| 91 | + |
| 92 | +- 2006-07-12 Fix: A bug was discovered by Paul Tseng when using free variables |
| 93 | + together with rotated Lorentz cones. |
| 94 | + |
| 95 | +- 2005-06-22 Fix: A bug about free variable handling has been reported by |
| 96 | + Johan Löfberg, the fixed version is now posted. |
| 97 | + |
| 98 | +- 2005-06-21 Fix: A bug about complex variables has been reported by |
| 99 | + Michael C. Grant, the correct version is now available. |
| 100 | + |
| 101 | +- 2005-06-10 Fix: `&&` and `||` were replaced by `&` and `|` to make the |
| 102 | + package compatible with Matlab 6.0 (R12). |
| 103 | + |
| 104 | +- 2005-05-10 Feature: The binaries are now built from Matlab, see `Install.txt` |
| 105 | + for details. |
| 106 | + |
| 107 | +- 2005-03-02 Fix: The `'See also'` references at the end of the help portions |
| 108 | + are now clickable. |
| 109 | + |
| 110 | +- 2005-03-02 Fix: If ADA is actually dense then the symbolic Cholesky |
| 111 | + factorization and the minimum degree ordering are not performed. |
| 112 | + Also in this case `SYMBADA` is created directly as a fully dense matrix of |
| 113 | + ones stored as sparse. |
| 114 | + |
| 115 | +- 2005-03-02 Feature: Included a simple heuristic to turn step-differentiation on/off. |
| 116 | + |
| 117 | +- 2005-03-02 Feature: New options to control preprocessing: `pars.free`, `pars.sdp`. |
| 118 | + |
| 119 | +- 2005-03-02 Feature: Preprocessing routines are now included. |
| 120 | + Diagonal SDP blocks are converted into nonnegative variables. |
| 121 | + Free variables are handled in a quadratic cone, |
| 122 | + split free variables are detected in the linear part. |
| 123 | + The sparsity structure of A is corrected at the end of preprocessing. |
| 124 | + |
| 125 | +- 2005-03-02 Fix: New default values. |
| 126 | + `pars.eps = 1e-8`, `pars.stepdif = 2`, `pars.cg.qprec = 1`. |
| 127 | + |
| 128 | +- 2005-02-02 Feature: If `pars.errors=1` then SeDuMi outputs the six errors |
| 129 | + measures defined in the 7th DIMACS challenge paper that is available at |
| 130 | + <http://plato.asu.edu/dimacs/node3.html>. |
| 131 | + These are in `info.err1`, ..., `info.err6`. |
| 132 | + |
| 133 | +- 2005-01-20 Fix: Minor speed improvement in the `sdmaux*` files (loop unrolling). |
| 134 | + |
| 135 | +- 2005-01-10 Fix: Minor speed improvement in `popK.m`. |
| 136 | + |
| 137 | +- 2005-01-10 Fix: `blkchol` is now invoked directly from `SeDuMi.m` without |
| 138 | + `sparchol`. |
| 139 | + |
| 140 | +- 2005-01-10 Fix: A small change in `Amul.m` resulted in drastic speed |
| 141 | + improvement for middle-sized problems. |
| 142 | + Additionally, `sum(c.*x)` is much faster than `c'*x` if `c` and `x` are sparse. |
| 143 | + The moral is that taking the transpose of a sparse matrix is slow. |
| 144 | + |
| 145 | +- 2004-01-03 Fix: Fixed a small bug in `wregion.m` concerning division by zero. |
| 146 | + (Johan Löfberg) |
| 147 | + |
| 148 | +- 2004-11-29 Fix: Many if-then statements checking for data validity in the C |
| 149 | + code were replaced by `mxAssert` statements. |
| 150 | + The only exceptions are memory failure checks. |
| 151 | + |
| 152 | +- 2004-11-29 Feature: Timing is changed. |
| 153 | + So far SeDuMi reported the CPU time spent in the main IPM loop. |
| 154 | + Now `info.timing` is a vector containing the time spent in the preprocessing, |
| 155 | + IPM and postprocessing, respectively. |
| 156 | + The old `info.cpusec` contains the total time of the algorithm. |
| 157 | + This does not make much difference now, since 99% of the time is spent in |
| 158 | + the IPM loop, but if more stuff is added to the pre- and postprocessing, |
| 159 | + this can be an issue. |
| 160 | + |
| 161 | +- 2004-11-11 Fix: Cleaned up the code with M-Lint. |
| 162 | + All unused variables and never used assignments are removed. |
| 163 | + All `|` and `&` are replaced with `||` and `&&` whenever it was possible. |
| 164 | + This speeds up conditional statements in the preprocessing phase. |
| 165 | + Unused .m and .c files are removed from the distribution. |
| 166 | + |
| 167 | +- 2004-11-03 Feature: pars.stopat can now be a vector. |
| 168 | + The algorithm enters debug mode at each iteration specified in `pars.stopat`. |
| 169 | + The order does not matter, invalid values are ignored. |
0 commit comments