Merge branch 'hotfix/1.3.1'

CMakeLists.txt

@@ -9,7 +9,7 @@ include(cmake/old.cmake)
 set(ISVD_NAME "Integrated Singular Value Decomposition")
 set(ISVD_MAJOR_VERSION 1)
 set(ISVD_MINOR_VERSION 3)
-set(ISVD_PATCH_VERSION 0)
+set(ISVD_PATCH_VERSION 1)
 set(ISVD_VERSION "${ISVD_MAJOR_VERSION}.${ISVD_MINOR_VERSION}.${ISVD_PATCH_VERSION}")
 if(NOT CMAKE_VERSION VERSION_LESS 3.0.0)
   cmake_policy(SET CMP0048 OLD)

README.md

@@ -22,11 +22,11 @@ This library is a C implementation of the Integrated Singular Value Decompositio
 * [CMake](https://cmake.org) 2.8.11+ (CMake 3.0+ recommended).
 * C/C++ compiler with C99/C++98 standard support ([GCC](https://gcc.gnu.org) 4.4+ recommended).
 * [BLAS](http://www.netlib.org/blas) & [LAPACK](http://www.netlib.org/lapack) library (Used for multi-core parallelization).
-* MPI Library v2.2+ ([MPICH](http://www.mpich.org) or [OpenMPI](https://www.open-mpi.org)).
+* MPI Library with 2.2+ standard support ([MPICH](http://www.mpich.org) 1.2+ or [OpenMPI](https://www.open-mpi.org) 1.7+).
 
 ## Optional
 * [Intel® Math Kernel Library](https://software.intel.com/en-us/intel-mkl) (**Recommended** for better performance, used for BLAS & LAPACK).
-* [MAGMA](http://icl.cs.utk.edu/magma/) 2+ (Used for BLAS & LAPACK with GPU support).
+* [MAGMA](http://icl.cs.utk.edu/magma/) 2.0+ (Used for BLAS & LAPACK with GPU support).
 * [OpenMP](http://openmp.org) Library (**Recommended** for better performance, used for multi-thread parallelization).
 * [Google Test Library](https://github.com/google/googletest) (Used for code testing).
 * [Doxygen](http://www.stack.nl/~dimitri/doxygen/) (Used for documentation).
@@ -141,7 +141,7 @@ Please link exactly one library for all categories.
 
 ## How to use 64-bit integer?
 
-* Set `ISVD_USE_ILP64` with `ccmake` before building libraries.
+* Set `ISVD_USE_ILP64` with `ccmake` before building libraries; or
 * Add `-DISVD_USE_ILP64` to compile flag.
 * Make sure your LAPACK&BLAS / Intel MKL uses 64bit integer.
 

cmake/display.cmake

@@ -76,75 +76,77 @@ disp("Build type:                     " "${CMAKE_BUILD_TYPE}")
 message(STATUS "")
 
 # Display compilers
-disp("Use C   Compiler:               " "${CMAKE_C_COMPILER}")
-disp("Use C++ Compiler:               " "${CMAKE_CXX_COMPILER}")
+disp("C   compiler:                   " "${CMAKE_C_COMPILER}")
+disp("C++ compiler:                   " "${CMAKE_CXX_COMPILER}")
 
 message(STATUS "")
 
 string(TOUPPER "CMAKE_C_FLAGS_${CMAKE_BUILD_TYPE}" cflag_build)
 string(TOUPPER "CMAKE_CXX_FLAGS_${CMAKE_BUILD_TYPE}" cxxflag_build)
 
-disp("Use C   flags:                  " "${CMAKE_C_FLAGS} ${${cflag_build}}")
-disp("Use C++ flags:                  " "${CMAKE_CXX_FLAGS} ${${cxxflag_build}}")
-disp("Use libraries:                  " "${DEFAULT_LIBRARY}")
+disp("C   flags:                      " "${CMAKE_C_FLAGS} ${${cflag_build}}")
+disp("C++ flags:                      " "${CMAKE_CXX_FLAGS} ${${cxxflag_build}}")
+disp("System libraries:               " "${DEFAULT_LIBRARY}")
 
 message(STATUS "")
 
 # Display libraries
 
 if(ISVD_USE_PACKAGE AND ISVD_OMP)
-  disp("Use OpenMP:                     " "${OpenMP_LIBRARIES}")
+  disp("OpenMP:                         " "${OpenMP_LIBRARIES}")
 else()
-  disp("Use OpenMP:                     " "OFF")
+  disp("OpenMP:                         " "OFF")
 endif()
 
 if(ISVD_USE_PACKAGE)
-  disp("Use MPI:                        " "${MPI_C_LIBRARIES}")
+  disp("MPI:                            " "${MPI_C_LIBRARIES}")
 else()
-  disp("Use MPI:                        " "OFF")
+  disp("MPI:                            " "OFF")
 endif()
 
 if(ISVD_USE_PACKAGE)
   if(NOT ISVD_USE_MKL)
     set(LAPACK_LIBS ${LAPACK_LIBRARIES})
     list(REMOVE_DUPLICATES LAPACK_LIBS)
-    disp("Use BLAS:                       " "${LAPACK_LIBS}")
+    disp("BLAS:                           " "${LAPACK_LIBS}")
   else()
-    disp("Use BLAS: (Intel MKL)           " "${MKL_LIBS}")
+    disp("BLAS: (Intel MKL)               " "${MKL_LIBS}")
   endif()
 endif()
 
 if(ISVD_USE_PACKAGE AND ISVD_USE_GPU)
-  disp("Use GPU:                        " "${CUDA_TOOLKIT_ROOT_DIR}")
-  disp("Use MAGMA:                      " "${MAGMA_LIBRARIES}")
+  disp("GPU:                            " "${CUDA_TOOLKIT_ROOT_DIR}")
+  disp("MAGMA:                          " "${MAGMA_LIBRARIES}")
 else()
-  disp("Use GPU:                        " "OFF")
+  disp("GPU:                            " "OFF")
 endif()
 
 if(ISVD_BUILD_TEST)
-  disp("Use Google Test:                " "${GTEST_BOTH_LIBRARIES}")
+  disp("Google Test:                    " "${GTEST_BOTH_LIBRARIES}")
 else()
-  disp("Use Google Test:                " "OFF")
+  disp("Google Test:                    " "OFF")
 endif()
 
 if(ISVD_BUILD_DOC)
-  disp("Use Doxygen:                    " "${DOXYGEN_EXECUTABLE}")
+  disp("Doxygen:                        " "${DOXYGEN_EXECUTABLE}")
 else()
-  disp("Use Doxygen:                    " "OFF")
+  disp("Doxygen:                        " "OFF")
 endif()
 
 message(STATUS "")
 
 if(ISVD_USE_PACKAGE)
   disp("MPI processes (demo only):      " "${MPI_PROCS}")
-  disp("OpenMP threads (demo only):     " "${OMP_THRDS}")
+  if(ISVD_OMP)
+    disp("OpenMP threads (demo only):     " "${OMP_THRDS}")
+  endif()
 endif()
 
 message(STATUS "")
 message(STATUS "================================================================================")
 message(STATUS "")
 
-if(ISVD_USE_PACKAGE)
+if(ISVD_USE_PACKAGE AND CMAKE_WARNING)
   if(ISVD_BLAS STREQUAL "BLAS" AND ISVD_OMP)
     message(${DEPRECATION} "${Esc}[1;33mPlain BLAS library might not support OpenMP multithreading.${Esc}[0m")
   endif()

cmake/options.cmake

@@ -13,6 +13,7 @@ option(ISVD_BUILD_DOC     "Build documentation."        "ON")
 option(ISVD_USE_GPU       "Enable GPU support."         "OFF")
 option(ISVD_TEST_DEEP     "Enable deep unit tests."     "ON")
 option(ISVD_TEST_VERBOSE  "Enable verbose unit tests."  "ON")
+option(CMAKE_WARNING      "Enable CMake warning."       "ON")
 
 set(ISVD_INDEX_TYPE "32" CACHE STRING "Index type. [32/64]")
 set_property(CACHE ISVD_INDEX_TYPE PROPERTY STRINGS "32;64")

cmake/packages.cmake

@@ -124,7 +124,7 @@ else()
 endif()
 
 # MPI
-find_package(MPI ${findtype})
+find_package(MPI 2.2 ${findtype})
 function(ISVD_SET_TARGET_MPI target lang)
   target_include_directories(${target} ${SYSTEM} PRIVATE "${MPI_${lang}_INCLUDE_PATH}")
   target_link_libraries(${target} ${MPI_${lang}_LIBRARIES})

demo/cdemo.c

@@ -74,11 +74,12 @@ int main( int argc, char **argv ) {
 
   /// [allocate-matrix]
   // Allocate matrix
-  const isvd_int_t k = 20, p = 12, l = k+p, N = 16;
+  const isvd_int_t k = 20, p = 12, N = 16;
   const isvd_Param param = isvd_createParam(m, n, k, p, N, mpi_root, MPI_COMM_WORLD);
   const isvd_int_t mb  = param.nrow_each;
   const isvd_int_t Pmb = param.nrow_total;
   const isvd_int_t Pnb = param.ncol_total;
+  const isvd_int_t l   = param.dim_sketch;
 
   double *s  = isvd_dMalloc(l);
 

demo/ctest.c

@@ -28,16 +28,8 @@ int main( int argc, char **argv ) {
     isvd_PrintEnvironment(MPI_COMM_WORLD);
   }
 
-  const isvd_int_t m = 100;
-  const isvd_int_t k = 10;
-
-  isvd_d_val_t *q = isvd_dMalloc(k * m);
-  const isvd_int_t ldq = k;
-
-  isvd_d_val_t *tau = isvd_dMalloc(k);
-
-  isvd_dGelqf(k, m, q, ldq, tau);
-  isvd_dOrglq(k, m, k, q, ldq, tau);
+  const isvd_int_t m = 100, n = 1000, k = 20, p = 12, N = 16;
+  const isvd_Param param = isvd_createParam(m, n, k, p, N, mpi_root, MPI_COMM_WORLD);
 
   isvd_Finalize();
 

doxygen/tutorial/core/example.dox

@@ -46,9 +46,11 @@
   Above example code can be compiled by
 
   ```
-  gcc demo/cdemo.c -I<include-path> -L<library-path> -Wl,-rpath <library-path> -lisvd_core -lisvd_la_mkl_gomp -lisvd_gpu_none
+  gcc demo/cdemo.c <path-to-source>/ext/mmio/mmio.c -include <path-to-source>/ext/mmio/mmio.h -I<include-path> -L<library-path> -Wl,-rpath <library-path> -lisvd_core -lisvd_la_mkl_gomp -lisvd_gpu_none
   ```
 
+  where <include-path> is the folder containing the header files (located at <path-to-source>/build/include), and <library-path> is the folder containing the libraries (located at <path-to-source>/build/lib).
+
   One may also use **make run_cdemo** to compile and run it.
 
   The outputs of the example are shown as below. (Note that the detailed number might be different due to the executing environment.)

doxygen/tutorial/core/stage.dox

@@ -48,7 +48,7 @@
 
   The postprocessing stage forms a rank-𝑘 approximate SVD of \f$\boldsymbol{A}\f$ in the range of \f$\overline{\boldsymbol{Q}}\f$.
 
-  - The \ref isvd_dPostprocessTallSkinnyQr "Tall-Skinny QR Orthogonalization" finds the orthonormal basis using the Tall-Skinny QR (TSQR) decomposition.
+  - The \ref isvd_dPostprocessTallSkinnyQr "Tall-Skinny QR Postprocessing" finds the orthonormal basis using the Tall-Skinny QR (TSQR) decomposition.
   - The \ref isvd_dPostprocessGramian "Gramian Postprocessing" finds approximate SVD using the eigen-decomposition of \f$\overline{\boldsymbol{Q}}^\top \boldsymbol{A} \boldsymbol{A}^\top \overline{\boldsymbol{Q}}\f$.
   - The \ref isvd_dPostprocessSymmetric "Symmetric Postprocessing" finds approximate SVD using the eigen-decomposition of \f$\overline{\boldsymbol{Q}}^\top \boldsymbol{A} \overline{\boldsymbol{Q}}\f$ for symmetric input matrix \f$\boldsymbol{A}\f$.
 */

src/include/c/isvd/def.h

@@ -93,7 +93,7 @@ typedef int omp_int_t;
 /// \brief  The MPI communicator type.
 #if !defined(DOXYGEN_SHOULD_SKIP_THIS)
 #if !defined(__cplusplus)
-typedef void* isvd_MpiComm;
+typedef int isvd_MpiComm;
 #else  // __cplusplus
 #include <mpi.h>
 typedef MPI_Comm isvd_MpiComm;

src/lib/libisvd/core/stage/@x@_sketch.h

@@ -128,7 +128,7 @@ static void sketchBlockRow(
   // Random generate
 
   isvd_int_t seed_ = seed;
-  isvd_assert_pass(MPI_Bcast(&seed_, sizeof(isvd_VSLStreamStatePtr), MPI_BYTE, mpi_root, param.mpi_comm));
+  isvd_assert_pass(MPI_Bcast(&seed_, sizeof(seed_), MPI_BYTE, mpi_root, param.mpi_comm));
   isvd_v@x@RngGaussianDriver(seed_, 0, n * Nl, omegat, 0.0, 1.0);
 
   // ====================================================================================================================== //

src/lib/libisvd/gpu/stage/@x@_postprocess_gpu.h

@@ -284,9 +284,9 @@ static void projectBlockRow(
     isvd_Free(zt_);
   }
 
-  // magma_free(a_gpu);
-  // magma_free(qt_gpu);
-  // magma_free(zt_gpu);
+  magma_free(a_gpu);
+  magma_free(qt_gpu);
+  magma_free(zt_gpu);
 
 }
 #endif  // DOXYGEN_SHOULD_SKIP_THIS