Merge branch 'hotfix/1.2.2'

emfomy · Oct 12, 2017 · 45da290 · 45da290
2 parents b2b732e + aec102e
commit 45da290
Show file tree

Hide file tree

Showing 52 changed files with 629 additions and 587 deletions.
diff --git a/CMakeLists.txt b/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 2)
-set(ISVD_PATCH_VERSION 0)
+set(ISVD_PATCH_VERSION 2)
 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)

diff --git a/README.md b/README.md
@@ -60,7 +60,7 @@ The following table are the main options
 | `ISVD_INDEX_TYPE`      | the selection index type.          | Options: `32/64`               |
 | `ISVD_OMP`             | the selection OpenMP library.      | Options: `OFF/GOMP/IOMP`       |
 | `ISVD_USE_GPU`         | enable GPU support.                |                                |
-| `ISVD_VERBOSE_TEST`    | enable verbose unit tests.         |                                |
+| `ISVD_TEST_VERBOSE`    | enable verbose unit tests.         |                                |
 | `GTEST_ROOT`           | the root path of Google Test.      | Require `ISVD_BUILD_TEST`      |
 | `INTEL_ROOT`           | the root path of Intel libraries.  | Require `ISVD_BLAS = MKL`      |
 | `MKL_ROOT`             | the root path of Intel MKL.        | Require `ISVD_BLAS = MKL`      |
@@ -81,7 +81,7 @@ The following table are the main Makefile rules
 
 ### Test installation
 
-* Set `ISVD_BUILD_TEST` using **ccmake**. (Also recommended to unset `ISVD_VERBOSE_TEST` if GPU is enabled).
+* Set `ISVD_BUILD_TEST` using **ccmake**. (Also recommended to unset `ISVD_TEST_VERBOSE` if GPU is enabled).
 * Run **make check**
   * Known issue: **RealSingle_WenYinIntegration.Test.#** / **s_integrate_wen_yin_#** fail the test.
 

diff --git a/check/CMakeLists.txt b/check/CMakeLists.txt
@@ -101,4 +101,4 @@ endmacro()
 # Set double unit tests
 isvd_check_fn("${ISVD_S_TYPES}")
 isvd_check_fn("${ISVD_D_TYPES}")
-unset(isvd_check_fn)
+unset(isvd_check_fn)
diff --git a/check/check.cmake b/check/check.cmake
@@ -28,7 +28,7 @@ macro(_ADD_CHECK checktype)
   _add_check_predo("${checktype}" "")
 
   # Add test
-  if(ISVD_VERBOSE_TEST)
+  if(ISVD_TEST_VERBOSE)
     gtest_add_tests($<TARGET_FILE:${checktarget}> "" ${checkmain} ${files})
   else()
     add_test(NAME ${checkname} COMMAND $<TARGET_FILE:${checktarget}>)
@@ -49,7 +49,7 @@ macro(_ADD_MPI_CHECK checktype listprocs)
   _add_check_predo("${checktype}")
 
   # Add test
-  if(ISVD_VERBOSE_TEST)
+  if(ISVD_TEST_VERBOSE)
     gtest_add_mpi_tests($<TARGET_FILE:${checktarget}> "${listprocs}" "" ${checkmain} ${files})
   else()
     foreach(procs ${listprocs})

diff --git a/check/check/libisvd/core/stage/@x@_postprocess_symmetric.cxx b/check/check/libisvd/core/stage/@x@_postprocess_symmetric.cxx
@@ -155,6 +155,8 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
   const isvd_int_t mj  = param.nrow_proc;
   const isvd_int_t mb  = param.nrow_each;
   const isvd_int_t Pmb = param.nrow_total;
+  const isvd_int_t nb  = param.ncol_each;
+  const isvd_int_t Pnb = param.ncol_total;
 
   // Create matrices
   isvd_val_t *a;
@@ -182,12 +184,15 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
   isvd_val_t *ut_ = isvd_@x@malloc(l * Pmb);
   isvd_int_t ldut_ = l;
 
+  isvd_val_t *vt_ = isvd_@x@malloc(l * Pnb);
+  isvd_int_t ldvt_ = l;
+
   switch ( jobuv ) {
     case GatherUV: {
 
       // Run stage
       isvd_@x@PostprocessSymmetric(param, NULL, 0, NULL, 0, dista_, ordera_,
-                                   a, lda, qt, ldqt, s, ut_, ldut_, NULL, 0, mpi_root, -2);
+                                   a, lda, qt, ldqt, s, ut_, ldut_, vt_, ldvt_, mpi_root, mpi_root);
 
       break;
     }
@@ -198,15 +203,20 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
       isvd_val_t *ut = isvd_@x@malloc(l * mb);
       isvd_int_t ldut = l;
 
+      isvd_val_t *vt = isvd_@x@malloc(l * nb);
+      isvd_int_t ldvt = l;
+
       // Run stage
       isvd_@x@PostprocessSymmetric(param, NULL, 0, NULL, 0, dista_, ordera_,
-                                   a, lda, qt, ldqt, s, ut, ldut, NULL, 0, -1, -2);
+                                   a, lda, qt, ldqt, s, ut, ldut, vt, ldvt, -1, -1);
 
       // Gather results
       MPI_Gather(ut, mb*ldut, MPI_@XTYPE@, ut_, mb*ldut, MPI_@XTYPE@, mpi_root, MPI_COMM_WORLD);
+      MPI_Gather(vt, nb*ldvt, MPI_@XTYPE@, vt_, nb*ldvt, MPI_@XTYPE@, mpi_root, MPI_COMM_WORLD);
 
       // Deallocate memory
       isvd_free(ut);
+      isvd_free(vt);
 
       break;
     }
@@ -236,19 +246,30 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
       isvd_int_t lduut_ = m;
       isvd_val_t *uut0 = isvd_@x@malloc(m * m);
       isvd_int_t lduut0 = m;
+      isvd_val_t *vvt_ = isvd_@x@malloc(n * n);
+      isvd_int_t ldvvt_ = n;
+      isvd_val_t *vvt0 = uut0;
+      isvd_int_t ldvvt0 = lduut0;
       isvd_@x@Syrk('U', 'T', m, k, 1.0, ut_, ldut_, 0.0, uut_, lduut_);
       isvd_@x@Syrk('U', 'T', m, k, 1.0, ut0, ldut0, 0.0, uut0, lduut0);
+      isvd_@x@Syrk('U', 'T', n, k, 1.0, vt_, ldvt_, 0.0, vvt_, ldvvt_);
 
       // Check results
       for ( isvd_int_t ir = 0; ir < m; ++ir ) {
         for ( isvd_int_t ic = ir; ic < m; ++ic ) {
           ASSERT_NEAR(uut_[ir+ic*lduut_], uut0[ir+ic*lduut0], @x@err) << "(ir, ic) =  (" << ir << ", " << ic << ")";
         }
       }
+      for ( isvd_int_t ir = 0; ir < n; ++ir ) {
+        for ( isvd_int_t ic = ir; ic < n; ++ic ) {
+          ASSERT_NEAR(vvt_[ir+ic*ldvvt_], vvt0[ir+ic*ldvvt0], @x@err) << "(ir, ic) =  (" << ir << ", " << ic << ")";
+        }
+      }
 
       // Deallocate memory
       isvd_free(uut_);
       isvd_free(uut0);
+      isvd_free(vvt_);
     }
   }
 
@@ -260,6 +281,7 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
   isvd_free(qt);
   isvd_free(s);
   isvd_free(ut_);
+  isvd_free(vt_);
 }
 
 TEST(@XStr@_SymmetricPostprocessing, BlockCol_ColMajor_GatherUV) {

diff --git a/check/check/libisvd/gpu/stage/@x@_postprocess_symmetric_gpu.cxx b/check/check/libisvd/gpu/stage/@x@_postprocess_symmetric_gpu.cxx
@@ -155,6 +155,8 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
   const isvd_int_t mj  = param.nrow_proc;
   const isvd_int_t mb  = param.nrow_each;
   const isvd_int_t Pmb = param.nrow_total;
+  const isvd_int_t nb  = param.ncol_each;
+  const isvd_int_t Pnb = param.ncol_total;
 
   // Create matrices
   isvd_val_t *a;
@@ -182,12 +184,15 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
   isvd_val_t *ut_ = isvd_@x@malloc(l * Pmb);
   isvd_int_t ldut_ = l;
 
+  isvd_val_t *vt_ = isvd_@x@malloc(l * Pnb);
+  isvd_int_t ldvt_ = l;
+
   switch ( jobuv ) {
     case GatherUV: {
 
       // Run stage
       isvd_@x@PostprocessSymmetric_gpu(param, NULL, 0, NULL, 0, dista_, ordera_,
-                                       a, lda, qt, ldqt, s, ut_, ldut_, NULL, 0, mpi_root, -2);
+                                   a, lda, qt, ldqt, s, ut_, ldut_, vt_, ldvt_, mpi_root, mpi_root);
 
       break;
     }
@@ -198,15 +203,20 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
       isvd_val_t *ut = isvd_@x@malloc(l * mb);
       isvd_int_t ldut = l;
 
+      isvd_val_t *vt = isvd_@x@malloc(l * nb);
+      isvd_int_t ldvt = l;
+
       // Run stage
       isvd_@x@PostprocessSymmetric_gpu(param, NULL, 0, NULL, 0, dista_, ordera_,
-                                       a, lda, qt, ldqt, s, ut, ldut, NULL, 0, -1, -2);
+                                   a, lda, qt, ldqt, s, ut, ldut, vt, ldvt, -1, -1);
 
       // Gather results
       MPI_Gather(ut, mb*ldut, MPI_@XTYPE@, ut_, mb*ldut, MPI_@XTYPE@, mpi_root, MPI_COMM_WORLD);
+      MPI_Gather(vt, nb*ldvt, MPI_@XTYPE@, vt_, nb*ldvt, MPI_@XTYPE@, mpi_root, MPI_COMM_WORLD);
 
       // Deallocate memory
       isvd_free(ut);
+      isvd_free(vt);
 
       break;
     }
@@ -215,7 +225,7 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
 
       // Run stage
       isvd_@x@PostprocessSymmetric_gpu(param, NULL, 0, NULL, 0, dista_, ordera_,
-                                       a, lda, qt, ldqt, s, NULL, 0, NULL, 0, -2, -2);
+                                   a, lda, qt, ldqt, s, NULL, 0, NULL, 0, -2, -2);
 
       break;
     }
@@ -236,19 +246,30 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
       isvd_int_t lduut_ = m;
       isvd_val_t *uut0 = isvd_@x@malloc(m * m);
       isvd_int_t lduut0 = m;
+      isvd_val_t *vvt_ = isvd_@x@malloc(n * n);
+      isvd_int_t ldvvt_ = n;
+      isvd_val_t *vvt0 = uut0;
+      isvd_int_t ldvvt0 = lduut0;
       isvd_@x@Syrk('U', 'T', m, k, 1.0, ut_, ldut_, 0.0, uut_, lduut_);
       isvd_@x@Syrk('U', 'T', m, k, 1.0, ut0, ldut0, 0.0, uut0, lduut0);
+      isvd_@x@Syrk('U', 'T', n, k, 1.0, vt_, ldvt_, 0.0, vvt_, ldvvt_);
 
       // Check results
       for ( isvd_int_t ir = 0; ir < m; ++ir ) {
         for ( isvd_int_t ic = ir; ic < m; ++ic ) {
           ASSERT_NEAR(uut_[ir+ic*lduut_], uut0[ir+ic*lduut0], @x@err) << "(ir, ic) =  (" << ir << ", " << ic << ")";
         }
       }
+      for ( isvd_int_t ir = 0; ir < n; ++ir ) {
+        for ( isvd_int_t ic = ir; ic < n; ++ic ) {
+          ASSERT_NEAR(vvt_[ir+ic*ldvvt_], vvt0[ir+ic*ldvvt0], @x@err) << "(ir, ic) =  (" << ir << ", " << ic << ")";
+        }
+      }
 
       // Deallocate memory
       isvd_free(uut_);
       isvd_free(uut0);
+      isvd_free(vvt_);
     }
   }
 
@@ -260,6 +281,7 @@ static void test( char dista, char ordera, const JobUV jobuv ) {
   isvd_free(qt);
   isvd_free(s);
   isvd_free(ut_);
+  isvd_free(vt_);
 }
 
 TEST(@XStr@_SymmetricPostprocessing_Gpu, BlockCol_ColMajor_GatherUV) {

diff --git a/check/lib/CMakeLists.txt b/check/lib/CMakeLists.txt
@@ -22,14 +22,22 @@ isvd_set_target(checkisvd_core_la)
 isvd_set_target_omp(checkisvd_core_la CXX)
 isvd_set_target_mpi(checkisvd_core_la CXX)
 isvd_set_target_blas(checkisvd_core_la)
-isvd_set_target_gtest(checkisvd_core_la)
+if(ISVD_TEST_DEEP)
+  isvd_set_target_gtest(checkisvd_core_la)
+else()
+  target_compile_definitions(checkisvd_core_la PRIVATE "NDEBUG")
+endif()
 
 # checkisvd_gpu_none
 file(GLOB_RECURSE files "${CMAKE_CURRENT_CONFIG_DIR}/libisvd/nogpu/*")
 add_library(checkisvd_gpu_none EXCLUDE_FROM_ALL ${files})
 isvd_set_target(checkisvd_gpu_none)
 isvd_set_target_mpi(checkisvd_gpu_none CXX)
-isvd_set_target_gtest(checkisvd_gpu_none)
+if(ISVD_TEST_DEEP)
+  isvd_set_target_gtest(checkisvd_gpu_none)
+else()
+  target_compile_definitions(checkisvd_gpu_none PRIVATE "NDEBUG")
+endif()
 
 # checkisvd_gpu_magma
 if(ISVD_USE_GPU)
@@ -38,5 +46,9 @@ if(ISVD_USE_GPU)
   isvd_set_target(checkisvd_gpu_magma)
   isvd_set_target_mpi(checkisvd_gpu_magma CXX)
   isvd_set_target_gpu(checkisvd_gpu_magma)
-  isvd_set_target_gtest(checkisvd_gpu_magma)
+  if(ISVD_TEST_DEEP)
+    isvd_set_target_gtest(checkisvd_gpu_magma)
+  else()
+    target_compile_definitions(checkisvd_gpu_magma PRIVATE "NDEBUG")
+  endif()
 endif()
diff --git a/cmake/display.cmake b/cmake/display.cmake
@@ -24,8 +24,9 @@ message(STATUS "Set 'ISVD_BUILD_LIB'            to build libraries.")
 message(STATUS "Set 'ISVD_BUILD_TEST'           to build unit tests.")
 message(STATUS "Set 'ISVD_INDEX_TYPE'           to select index type. [32/64]")
 message(STATUS "Set 'ISVD_OMP'                  to select OpenMP library. [OFF/GOMP/IOMP]")
+message(STATUS "Set 'ISVD_TEST_DEEP'            to enable deep unit tests.")
+message(STATUS "Set 'ISVD_TEST_VERBOSE'         to enable verbose unit tests.")
 message(STATUS "Set 'ISVD_USE_GPU'              to enable GPU support.")
-message(STATUS "Set 'ISVD_VERBOSE_TEST'         to enable verbose unit tests.")
 message(STATUS "")
 message(STATUS "Set 'GTEST_ROOT'                for the root path of Google Test.     (Require 'ISVD_BUILD_TEST')")
 message(STATUS "Set 'INTEL_ROOT'                for the root path of Intel libraries. (Require 'ISVD_BLAS = MKL')")
@@ -52,7 +53,11 @@ if(NOT ISVD_USE_ILP64)
 else()
   disp("Integer type:                   " "64bit integer (ILP64)")
 endif()
-disp("Verbose unit tests:             " "${ISVD_VERBOSE_TEST}")
+
+message(STATUS "")
+
+disp("Deep unit tests:                " "${ISVD_TEST_DEEP}")
+disp("Verbose unit tests:             " "${ISVD_TEST_VERBOSE}")
 
 message(STATUS "")
 message(STATUS "================================================================================")

diff --git a/cmake/options.cmake b/cmake/options.cmake
@@ -6,12 +6,13 @@ else()
 endif()
 
 # Set options
-option(ISVD_BUILD_LIB     "Build libraries."      "ON")
-option(ISVD_BUILD_DEMO    "Build demo codes."     "ON")
-option(ISVD_BUILD_TEST    "Build unit tests."     "OFF")
-option(ISVD_BUILD_DOC     "Build documentation."  "ON")
-option(ISVD_USE_GPU       "Enable GPU support."   "OFF")
-option(ISVD_VERBOSE_TEST  "Verbose unit tests."   "ON")
+option(ISVD_BUILD_LIB     "Build libraries."            "ON")
+option(ISVD_BUILD_DEMO    "Build demo codes."           "ON")
+option(ISVD_BUILD_TEST    "Build unit tests."           "OFF")
+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")
 
 set(ISVD_INDEX_TYPE "32" CACHE STRING "Index type. [32/64]")
 set_property(CACHE ISVD_INDEX_TYPE PROPERTY STRINGS "32;64")

diff --git a/cmake/vars.cmake b/cmake/vars.cmake
@@ -21,26 +21,3 @@ list(APPEND ISVD_S_TYPES "s" "isvd_s_val_t" "Real Single"    "RealSingle")
 list(APPEND ISVD_D_TYPES "d" "isvd_d_val_t" "Real Double"    "RealDouble")
 list(APPEND ISVD_C_TYPES "c" "isvd_c_val_t" "Complex Single" "ComplexSingle")
 list(APPEND ISVD_Z_TYPES "z" "isvd_z_val_t" "Complex Double" "ComplexDouble")
-
-# BLAS definitions
-set(
-  ISVD_TYPE_MACRO_DEFINE
-  "#define CHAR1 char"
-  "#define INT   isvd_int_t"
-  "#define REAL4 isvd_s_val_t"
-  "#define REAL8 isvd_d_val_t"
-  "#define COMP4 isvd_c_val_t"
-  "#define COMP8 isvd_z_val_t"
-)
-string(REPLACE ";" "\n" ISVD_TYPE_MACRO_DEFINE "${ISVD_TYPE_MACRO_DEFINE}")
-
-set(
-  ISVD_TYPE_MACRO_UNDEF
-  "#undef CHAR1"
-  "#undef INT"
-  "#undef REAL4"
-  "#undef REAL8"
-  "#undef COMP4"
-  "#undef COMP8"
-)
-string(REPLACE ";" "\n" ISVD_TYPE_MACRO_UNDEF "${ISVD_TYPE_MACRO_UNDEF}")
diff --git a/doxygen/CMakeLists.txt b/doxygen/CMakeLists.txt
@@ -23,7 +23,7 @@ set(ISVD_DOCS_DIR "${ISVD_DOCS_DIR}" PARENT_SCOPE)
 isvd_configure_fn("${CMAKE_CURRENT_SOURCE_DIR}" "${CMAKE_CURRENT_CONFIG_DIR}")
 
 # Set install rule
-install(DIRECTORY "${ISVD_DOCS_DIR}" DESTINATION docs)
+install(DIRECTORY "${ISVD_DOCS_DIR}" DESTINATION docs OPTIONAL)
 
 # Add rule
 add_custom_target(

diff --git a/src/include/CMakeLists.txt b/src/include/CMakeLists.txt
@@ -1,4 +1,6 @@
 # The CMake setting of 'src/include/'
 
+isvd_set_config_var()
+
 # Set install rule
 install(DIRECTORY "${CMAKE_CURRENT_CONFIG_DIR}/c/" DESTINATION include/c)