/[escript]/trunk/cusplibrary/CHANGELOG
ViewVC logotype

Contents of /trunk/cusplibrary/CHANGELOG

Parent Directory Parent Directory | Revision Log Revision Log


Revision 5148 - (show annotations)
Mon Sep 15 01:25:23 2014 UTC (4 years, 10 months ago) by caltinay
File size: 5946 byte(s)
Merging ripley diagonal storage + CUDA support into trunk.
Options file version has been incremented due to new options
'cuda' and 'nvccflags'.

1 #######################################
2 # Cusp v0.4.0 #
3 #######################################
4
5 Summary
6 Cusp v0.4.0 provides support for CUDA 5.5 and Thrust v1.7. Version
7 0.4.0 includes several new graph algorithms based on backend integration
8 with back40computing, a refactored smoothed aggregation solver allowing
9 user-defined functions for core components and experimental work concerning
10 sparse matrix visualization.
11
12 New Features
13 array2d formats allow access to row and column elements regardless of orientation
14 gallery method to generate grids
15 OpenGL implementation of spy to view sparse matrices
16 Generalized multilevel class
17 Added Sparse * Dense multiplication routine
18
19 Algorithms
20 cusp::graph::bfs
21 cusp::graph::connected_components
22 cusp::graph::hilbert_curve
23 cusp::graph::maximum_flow
24 cusp::graph::max_flow_to_min_cut
25 cusp::graph::pseudo_peripheral_vertex
26 cusp::graph::symmetric_rcm
27 cusp::krylov::cr
28
29 Bug Fixes
30 Resolved issues related to Thrust 1.6 & Thrust 1.7
31 Resolved issues related to CUDA 5.5
32 Silenced warnings concerning uninitialized memory in generalized SpMV
33
34 #######################################
35 # Cusp v0.3.1 #
36 #######################################
37
38 Summary
39 Minor compatibility fixes for Thrust 1.6.x
40
41 #######################################
42 # Cusp v0.3.0 #
43 #######################################
44
45 Summary
46 Cusp v0.3.0 provides support for CUDA 4.1 and Thrust v1.5. Version
47 0.3.0 includes incremental improvements to MatrixMarket I/O support
48 and array format conversions.
49
50 New Features
51 MatrixMarket I/O routines support complex matrices
52 array1d is convertible to and from all other formats (e.g. array1d<->array2d)
53
54 Bug Fixes
55 Resize final output of CSR * CSR to account for excluded zeros
56
57 Acknowledgements
58 Justin Luitjens and Simon Layton for diagnosing CSR SpMM bug
59 Kashif Rasul for Thrust v1.6 support patch
60 Filipe Maia and Steven Dalton for general support and maintenance
61
62 #######################################
63 # Cusp v0.2.0 #
64 #######################################
65
66 Summary
67 Cusp v0.2.0 provides support for CUDA 4.0 in addition to many new
68 features and performance enhancements. Support for complex numbers
69 is now available through the cusp::complex class. Lightweight views
70 to dense arrays and sparse matrix formats are now provided, allowing
71 user-allocated data to be interfaced efficiently with Cusp algorithms.
72 The performance of sparse matrix format conversions has been improved
73 dramatically on the device. Preconditioners based on algebraic multigrid
74 (AMG) and approximate inverse (AINV) methods have been added to the library.
75
76 New Features
77 Complex numbers
78 cusp::complex
79 cusp/cmath.h
80 cusp/complex.h
81
82 Sparse Matrices
83 cusp::coo_matrix_view
84 cusp::csr_matrix_view
85 cusp::dia_matrix_view
86 cusp::ell_matrix_view
87 cusp::hyb_matrix_view
88
89 Krylov methods
90 cusp::krylov::bicg
91 cusp::krylov::gmres
92 cusp::krylov::cg_m
93 cusp::krylov::bicgstab_m
94
95 Input/Output
96 cusp::io::read_matrix_market_stream
97
98 Algorithms
99 cusp::add
100 cusp::subtract
101 cusp::graph::maximal_independent_set
102 cusp::copy
103 cusp::convert
104 cusp::blas::nrm1
105 cusp::blas::nrmmax
106 cusp::relaxation::polynomial
107
108 Matrix Gallery
109 cusp::gallery::diffusion
110
111 Preconditioners
112 cusp::precond::scaled_bridson_ainv
113 cusp::precond::bridson_ainv
114 cusp::precond::nonsym_bridson_ainv
115 cusp::precond::smoothed_aggregation
116
117 Bug Fixes
118 #12 MatrixMarket file reader fails on OSX
119 #18 ensure that IndexType=long is supported
120 #27 bicgstab can't solve Ix = b
121 #31 device::spmm_coo requires huge amount of memory
122 #37 cusp::graph::maximal_independent_set should return the size of the MIS
123 #45 silence compiler warnings
124 #66 thrust hg version no longer compatible with hg version of cusp
125 #71 Several cusp::blas calls cause unexpected results by converting the output to ScalarType
126
127 Deprecations
128 cusp::print_matrix has been replaced by cusp::print
129 cusp::blas functions only accept arrays and array views
130
131 Known Issues
132 Cusp is known to cause the nvcc compiler to issue warnings of the form:
133 warning: calling a host function ... from a __device__/__global__ function ... is not allowed
134
135 Acknowledgements
136 Filipe Maia for general maintenance, testing, and implementations of cusp::complex and BiCG
137 Greg van Anders for CG-M and BiCGstab-M implementations
138 Joe Gordon for GMRES implementation
139 Jonathan Cohen for AINV preconditioners
140 Steven Dalton for improvements to conversion routines and smoothed aggregation preconditioner
141
142 #######################################
143 # Cusp v0.1.1 #
144 #######################################
145
146 Summary
147 Small fixes for compatibility with CUDA 3.1
148
149 #######################################
150 # Cusp v0.1.0 #
151 #######################################
152
153 Summary
154 Initial release
155
156 Features
157 Sparse Matrix Support
158 cusp::coo_matrix
159 cusp::csr_matrix
160 cusp::dia_matrix
161 cusp::ell_matrix
162 cusp::hyb_matrix
163 conversion between all formats
164 host<->device transfers
165 Dense Matrix Support
166 cusp::array1d
167 cusp::array2d
168 Algorithms
169 cusp::multiply(A,x,y) : y = A*x
170 cusp::transpose(A,At) : A^T
171 cusp::blas::*
172 Gallery
173 cusp::gallery::stencil
174 cusp::gallery::poisson5pt
175 Solvers
176 cusp::krylov::cg
177 cusp::krylov::bicgstab
178 Monitors
179 cusp::default_monitor
180 cusp::verbose_monitor
181 Preconditioners
182 cusp::precond::diagonal
183 User-Defined Linear Operators
184 cusp::linear_operator
185 File I/O
186 cusp::io::read_matrix_market_file
187 cusp::io::write_matrix_market_file
188 Utilities
189 cusp::print_matrix
190 cusp::is_valid_matrix etc.
191
192 Known Issues
193 Many routines assume that several arguments have the exact same type
194

  ViewVC Help
Powered by ViewVC 1.1.26