New testing/timing structure

CMakeLists.txt

@@ -1030,6 +1030,7 @@ endif(CHAMELEON_ENABLE_EXAMPLE AND NOT CHAMELEON_SIMULATION)
 # Testing executables
 if(CHAMELEON_ENABLE_TESTING AND NOT CHAMELEON_SIMULATION)
   add_subdirectory(testing)
+  add_subdirectory(new-testing)
 endif(CHAMELEON_ENABLE_TESTING AND NOT CHAMELEON_SIMULATION)
 
 # Timing executables

cmake_modules/local_subs.py

@@ -42,6 +42,7 @@ subs = {
         ('int',                  'float',                'double',               'CHAMELEON_Complex32_t', r'\bCHAMELEON_Complex64_t'),
         ('ChamPattern',          'ChamRealFloat',        'ChamRealDouble',       'ChamComplexFloat',    r'\bChamComplexDouble' ),
         ('ChamPattern',          'ChamRealFloat',        'ChamRealDouble',       'ChamRealFloat',       r'\bChamRealDouble'    ),
+        ('int',                  'float',                'double',               'complex32',           r'\bcomplex64'         ),
 
         # ----- Additional BLAS
         ('',                     'sTile',                'dTile',                'cTile',                'zTile'               ),

include/chameleon/types.h

@@ -122,6 +122,8 @@ static inline int chameleon_min( int a, int b ) {
     if ( a < b ) return a; else return b;
 }
 
+typedef double cham_fixdbl_t;
+
 END_C_DECLS
 
 #endif /* _chameleon_types_h_ */

new-testing/CMakeLists.txt

+###
+#
+# @file CMakeLists.txt
+#
+# @copyright 2009-2014 The University of Tennessee and The University of
+#                      Tennessee Research Foundation. All rights reserved.
+# @copyright 2012-2019 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
+#                      Univ. Bordeaux. All rights reserved.
+#
+###
+#
+#  @project CHAMELEON
+#  CHAMELEON is a software package provided by:
+#     Inria Bordeaux - Sud-Ouest,
+#     Univ. of Tennessee,
+#     King Abdullah Univesity of Science and Technology
+#     Univ. of California Berkeley,
+#     Univ. of Colorado Denver.
+#
+# @version 0.9.2
+#  @author Cedric Castagnede
+#  @author Emmanuel Agullo
+#  @author Mathieu Faverge
+#  @author Lucas Barros de Assis
+#  @date 2014-11-16
+#
+###
+if (CHAMELEON_SIMULATION)
+  message(ERROR "new-testing directory should not be included when simulation is enabled")
+endif()
+
+# Generate chameleon auxiliary testing sources for all possible precisions
+# --------------------------------------------------------------------
+set(NEWTESTING_HDRS_GENERATED "")
+set(ZHDR
+  testing_zauxiliary.h
+  testing_zcheck.h
+ )
+
+precisions_rules_py(NEWTESTING_HDRS_GENERATED "${ZHDR}"
+    PRECISIONS "s;d;c;z;ds;zc" )
+
+add_custom_target(new-testing_include ALL SOURCES ${NEWTESTING_HDRS_GENERATED})
+set(CHAMELEON_SOURCES_TARGETS "${CHAMELEON_SOURCES_TARGETS};new-testing_include" CACHE INTERNAL "List of targets of sources")
+
+# Generate chameleon testing sources for all possible precisions
+# ----------------------------------------------------------
+set(ZSRC
+    testing_zauxiliary.c
+    testing_zcheck.c
+    ##################
+    # LAPACK
+    ##################
+    #testing_zlaset.c
+    testing_zlacpy.c
+    testing_zlange.c
+    testing_zlanhe.c
+    testing_zlansy.c
+    testing_zlantr.c
+    testing_zgeadd.c
+    testing_ztradd.c
+    testing_zlascal.c
+    testing_zgemm.c
+    testing_zhemm.c
+    testing_zherk.c
+    testing_zher2k.c
+    testing_zsymm.c
+    testing_zsyrk.c
+    testing_zsyr2k.c
+    testing_ztrmm.c
+    testing_ztrsm.c
+    testing_zpotrf.c
+    testing_zpotrs.c
+    testing_zposv.c
+    testing_ztrtri.c
+    testing_zlauum.c
+    testing_zpotri.c
+    testing_zsytrf.c
+    testing_zsytrs.c
+    testing_zsysv.c
+    testing_zgetrf.c
+    testing_zgetrs.c
+    testing_zgesv.c
+    testing_zgeqrf.c
+    testing_zungqr.c
+    testing_zunmqr.c
+    testing_zgelqf.c
+    testing_zunglq.c
+    testing_zunmlq.c
+    # testing_zgeqrs.c
+    # testing_zgelqs.c
+    testing_zgels.c
+    testing_zgeqrf_hqr.c
+    testing_zungqr_hqr.c
+    testing_zunmqr_hqr.c
+    testing_zgelqf_hqr.c
+    testing_zunglq_hqr.c
+    testing_zunmlq_hqr.c
+    # testing_zgeqrs_hqr.c
+    # testing_zgelqs_hqr.c
+    testing_zgels_hqr.c
+    )
+
+# Add include and link directories
+# --------------------------------
+include_directories(${CMAKE_CURRENT_SOURCE_DIR})
+include_directories(${CMAKE_CURRENT_BINARY_DIR})
+
+# Define what libraries we have to link with
+# ------------------------------------------
+unset(libs_for_tests)
+# testing executables depend on chameleon and cblas, lapacke, mpi (already chameleon's dependencies)
+list(APPEND libs_for_tests chameleon)
+# message(STATUS "libs testings: ${libs_for_tests}")
+
+# Create one testing per precision with all files
+# -----------------------------------------------
+foreach(_precision ${CHAMELEON_PRECISION} )
+
+  precisions_rules_py(${_precision}SRC_GENERATED "${ZSRC}"
+    PRECISIONS "${_precision}" )
+
+  add_executable(${_precision}new-testing
+    ${${_precision}SRC_GENERATED}
+    values.c
+    run_list.c
+    parameters.c
+    )
+  add_dependencies(${_precision}new-testing
+    chameleon_include
+    coreblas_include
+    control_include
+    new-testing_include
+    )
+  set_property(TARGET ${_precision}new-testing PROPERTY LINKER_LANGUAGE Fortran)
+  target_link_libraries(${_precision}new-testing ${libs_for_tests})
+
+  install(TARGETS ${_precision}new-testing
+    DESTINATION bin/new-testing)
+
+endforeach()
+
+# Force generation of sources
+# ---------------------------
+set(NEWTESTING_SRCS)
+foreach(_precision ${CHAMELEON_PRECISION})
+  list(APPEND NEWTESTING_SRCS ${${_precision}SRC_GENERATED})
+endforeach()
+add_custom_target(new-testing_sources ALL SOURCES ${NEWTESTING_SRCS})
+set(CHAMELEON_SOURCES_TARGETS "${CHAMELEON_SOURCES_TARGETS};new-testing_sources" CACHE INTERNAL "List of targets of sources")
+
+#-------- Tests ---------
+include(CTestLists.cmake)
+
+# copy input files
+file(COPY ${CMAKE_CURRENT_SOURCE_DIR}/input DESTINATION ${CMAKE_CURRENT_BINARY_DIR})
+
+###
+### END CMakeLists.txt
+###

new-testing/CTestLists.cmake

+#
+# Check testing/
+#
+set(NP 2) # Amount of MPI processes
+set(THREADS 2) # Amount of threads
+set(N_GPUS 0) # Amount of graphic cards
+set(TEST_CATEGORIES shm)
+if (CHAMELEON_USE_MPI AND MPI_C_FOUND)
+  set( TEST_CATEGORIES ${TEST_CATEGORIES} mpi )
+endif()
+if (CHAMELEON_USE_CUDA AND CUDA_FOUND)
+  set(N_GPUS 0 1)
+endif()
+
+foreach(prec ${RP_CHAMELEON_PRECISIONS})
+  set (CMD ./${prec}new-testing)
+
+  #
+  # Create the list of test based on precision and runtime
+  #
+  set( TESTS lacpy lange lantr lansy )
+  if ( ${prec} STREQUAL c OR ${prec} STREQUAL z )
+    set( TESTS ${TESTS} lanhe )
+  endif()
+  set( TESTS ${TESTS}
+    geadd tradd lascal
+    gemm symm syrk syr2k trmm trsm )
+  if ( ${prec} STREQUAL c OR ${prec} STREQUAL z )
+    set( TESTS ${TESTS}
+      hemm herk her2k )
+  endif()
+  set( TESTS ${TESTS}
+    potrf potrs posv trtri lauum )
+  if ( NOT CHAMELEON_SCHED_PARSEC )
+    set( TESTS ${TESTS} potri )
+  endif()
+  if ( ${prec} STREQUAL c OR ${prec} STREQUAL z )
+    set( TESTS ${TESTS}
+      sytrf sytrs sysv )
+  endif()
+  set( TESTS ${TESTS}
+    getrf     getrs gesv
+    geqrf     gelqf
+    geqrf_hqr gelqf_hqr)
+  if ( ${prec} STREQUAL c OR ${prec} STREQUAL z )
+    set( TESTS ${TESTS}
+      ungqr     unglq     unmqr     unmlq
+      ungqr_hqr unglq_hqr unmqr_hqr unmlq_hqr)
+  else()
+    set( TESTS ${TESTS}
+      orgqr     orglq     ormqr     ormlq
+      orgqr_hqr orglq_hqr ormqr_hqr ormlq_hqr)
+  endif()
+  set( TESTS ${TESTS}
+    #geqrs     gelqs
+    #geqrs_hqr gelqs_hqr
+    gels
+    gels_hqr )
+
+  foreach(cat ${TEST_CATEGORIES})
+    foreach(gpus ${N_GPUS})
+
+      if (${gpus} EQUAL 1)
+        set(cat ${cat}_gpu)
+      endif()
+
+      if (${cat} STREQUAL "mpi")
+        set (PREFIX mpiexec --bind-to none -n ${NP})
+      else()
+        set (PREFIX "")
+      endif()
+
+      foreach(_test ${TESTS})
+        add_test(test_${cat}_${prec}${_test} ${PREFIX} ${CMD} -c -t ${THREADS} -g ${gpus} -P 1 -o ${_test} -f input/${_test}.in )
+      endforeach()
+    endforeach()
+  endforeach()
+endforeach()

new-testing/input/geadd.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GEADD
+# nb: Tile size
+# ib: Inner tile size
+# M: Number of rows of matrix A
+# N: Number of columns of matrix B
+# LDA: Leading dimension of matrix A
+# LDB: Leading dimension of matrix B
+# trans: Wether the matrix A is not tranposed, transposed or conjugate transposed (0: not transposed, 1: transposed, 2: conjugate transposed)
+# alpha: Scalar alpha
+# beta: Scalar beta
+
+nb = 16, 17
+ib = 8
+m = 15, 17, 33
+n = 13, 19, 35
+lda = 37
+ldb = 39
+trans = 0:2

new-testing/input/gelqf.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GELQF
+
+# nb: Tile size
+# ib: Inner tile size
+# m: Number of rows of the A matrix
+# n: Number of columns of the A matrix
+# lda: Leading dimension of the A matrix
+# RH: Size of each subdomain when using RH
+
+nb = 16, 17
+ib = 8
+m = 13, 17, 35
+n = 15, 19, 33
+lda = 41
+qra = 0, 3
\ No newline at end of file

new-testing/input/gelqf_hqr.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GELQF_HQR
+
+# nb: Tile size
+# ib: Inner tile size
+# m: Number of rows of the A matrix
+# n: Number of columns of the A matrix
+# lda: Leading dimension of the A matrix
+# qra: Size of TS domain
+# qrp: Size of high level tree for distributed mode
+# llvl: Tree used for low level reduction insides nodes
+# hlv: Tree used for high level reduction between nodes, only if qrp > 1
+# domino: Enable/Disable the domino between upper and lower trees
+
+nb = 8, 9
+ib = 3
+m = 13, 17, 35
+n = 15, 19, 33
+lda = 41
+qra = 2
+qrp = -1
+llvl = -1
+hlvl = -1
+domino = -1
\ No newline at end of file

new-testing/input/gelqs.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GELQS
+
+# nb: Tile size
+# ib: Inner tile size
+# m: Number of rows of the A matrix
+# n: Number of columns of the A matrix
+# lda: Leading dimension of the A matrix
+# RH: Size of each subdomain when using RH
+
+nb = 16, 17
+ib = 8
+m = 13, 17, 35
+n = 15, 19, 33
+lda = 41
+qra = 0, 3
\ No newline at end of file

new-testing/input/gels.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GELS
+# nb: Tile size
+# ib: Inner tile size
+# M: Number of rows of matrix A
+# N: Number of columns of matrix B
+# K: Number of columns of matrix A and rows of Matrix B
+# LDA: Leading dimension of matrix A
+# LDB: Leading dimension of matrix B
+# trans: Wether the system involves A (ChamNoTrans) or A^H *ChamTransConj)
+# RH: Size of each subdomain when using RH
+
+nb = 16, 17
+ib = 8
+m = 17, 31, 35
+n = 15, 29, 33
+k = 11, 27, 37
+lda = 43
+ldb = 45
+trans = 0, 2
+qra = 0, 3
\ No newline at end of file

new-testing/input/gels_hqr.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GELS
+# nb: Tile size
+# ib: Inner tile size
+# M: Number of rows of matrix A
+# N: Number of columns of matrix B
+# K: Number of columns of matrix A and rows of Matrix B
+# LDA: Leading dimension of matrix A
+# LDB: Leading dimension of matrix B
+# trans: Wether the system involves A (ChamNoTrans) or A^H *ChamTransConj). Currently only ChamNoTrans is supported.
+# qra: Size of TS domain
+# qrp: Size of high level tree for distributed mode
+# llvl: Tree used for low level reduction insides nodes
+# hlv: Tree used for high level reduction between nodes, only if qrp > 1
+# domino: Enable/Disable the domino between upper and lower trees
+
+nb = 8, 9
+ib = 3
+m = 17, 31, 35
+n = 15, 29, 33
+k = 11, 27, 37
+lda = 43
+ldb = 45
+trans = 0
+qra = 2
+qrp = -1
+llvl = -1
+hlvl = -1
+domino = -1
\ No newline at end of file

new-testing/input/gemm.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GEMM
+# nb: Tile size
+# ib: Inner tile size
+# M: Number of rows of matrix A and C
+# N: Number of columns of matrix B and C
+# K: Number of columns of matrix A and rows of matrix B
+# LDA: Leading dimension of matrix A
+# LDB: Leading dimension of matrix B
+# LDC: Leading dimension of matrix C
+# transA: Wether the matrix A is not tranposed, transposed or conjugate transposed (0: not transposed, 1: transposed, 2: conjugate transposed)
+# transB: Wether the matrix B is not tranposed, transposed or conjugate transposed (0: not transposed, 1: transposed, 2: conjugate transposed)
+# alpha: Scalar alpha
+# beta: Scalar beta
+
+nb = 16, 17
+ib = 8
+m = 15, 25, 37
+n = 13, 23, 35
+k = 14, 24, 34
+lda = 41
+ldb = 43
+ldc = 41
+transA = 0:2
+transB = 0:2

new-testing/input/geqrf.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GEQRF
+
+# nb: Tile size
+# ib: Inner tile size
+# m: Number of rows of the A matrix
+# n: Number of columns of the A matrix
+# lda: Leading dimension of the A matrix
+# RH: Size of each subdomain when using RH
+
+
+nb = 16, 17
+ib = 8
+m = 13, 17, 35
+n = 15, 19, 33
+lda = 41
+qra = 0, 3
\ No newline at end of file

new-testing/input/geqrf_hqr.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GEQRF_HQR
+
+# nb: Tile size
+# ib: Inner tile size
+# m: Number of rows of the A matrix
+# n: Number of columns of the A matrix
+# lda: Leading dimension of the A matrix
+# qra: Size of TS domain
+# qrp: Size of high level tree for distributed mode
+# llvl: Tree used for low level reduction insides nodes
+# hlv: Tree used for high level reduction between nodes, only if qrp > 1
+# domino: Enable/Disable the domino between upper and lower trees
+
+nb = 8, 9
+ib = 3
+m = 13, 17, 57
+n = 15, 19, 55
+lda = 59
+qra = 2
+qrp = -1
+llvl = -1
+hlvl = -1
+domino = -1
\ No newline at end of file

new-testing/input/geqrs.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GEQRS
+
+# nb: Tile size
+# ib: Inner tile size
+# m: Number of rows of the A matrix
+# n: Number of columns of the A matrix
+# lda: Leading dimension of the A matrix
+# RH: Size of each subdomain when using RH
+
+nb = 16, 17
+ib = 8
+m = 13, 17, 35
+n = 15, 19, 33
+lda = 41
+qra = 0, 3
\ No newline at end of file

new-testing/input/gesv.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GESV
+
+# nb: Tile size
+# ib: Inner tile size
+# n: Order of the matrix A and number of rows of matrix B
+# nrhs: The number of columns of matrix B
+# lda: Leading dimension of matrix A
+# ldb: Leading dimension of matrix B
+
+nb = 16, 17
+ib = 8
+n = 15, 21, 35
+nrhs = 13, 22, 33
+lda = 40
+ldb = 41
+seedA = 4687
\ No newline at end of file

new-testing/input/getrf.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GETRF
+
+# nb: Tile size
+# ib: Inner tile size
+# m: Number of rows of the matrix A
+# n: Number of columns of the matrix A
+# lda: Leading dimension of matrix A
+
+nb = 16, 17
+ib = 8
+m = 13, 17, 35
+n = 15, 19, 33
+lda = 41
\ No newline at end of file

new-testing/input/getrs.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# GETRS
+
+# nb: Tile size
+# ib: Inner tile size
+# m: Order of the matrix A and number of rows of matrix B
+# n: The number of columns of matrix B
+# lda: Leading dimension of matrix A
+# ldb: Leading dimension of matrix B
+
+nb = 16, 17
+ib = 8
+m = 13, 17, 35
+n = 15, 19, 33
+lda = 41
+ldb = 42
\ No newline at end of file

new-testing/input/hemm.in

+# You can enumerate each parameter's values as an explicit list separated by commas or by a range start:end[:step]
+# Not given parameters will receive default values
+
+# HEMM
+# nb: Tile size
+# ib: Inner tile size
+# side: whether the hermitian matrix A appears on the left or right side of the multiplication operation (0: left, 1: right)
+# uplo: matrix part to be used (0: Upper, 1: Lower)
+# M: Number of rows of matrix A and C
+# N: Number of columns of matrix B and C
+# LDA: Leading dimension of matrix A
+# LDB: Leading dimension of matrix B
+# LDC: Leading dimension of matrix C
+# alpha: Scalar alpha
+# beta: Scalar beta
+# bump: bump value for Hermitian matrices
+
+nb = 16, 17
+ib = 8
+side = 0:1
+uplo = 0:1
+m = 15, 25, 37
+n = 13, 23, 35
+lda = 41
+ldb = 43
+ldc = 41
+bump = 0