Merge pull request #11 from nakatamaho/v2.0

V2.0

ChangeLog

@@ -1,3 +1,64 @@
+2022/05/16    Nakata Maho <[email protected]> 2.0 branch
+            * Disable Rgejsv, Cgejsv, Rgesvj and Cgesvj by dd/qd, since we cannot
+              gurantee the precisions.
+
+2022/05/16    Nakata Maho <[email protected]> 2.0 branch
+            * Rtest_rfp.in, Ctest_rfp.in: Test for linear equation routines RFP format
+              have passed for all precisions.
+
+2022/05/10    Nakata Maho <[email protected]> 2.0 branch
+            * csd.in: Test for CS decomposition routines
+              have passed for all precisions (except for GMP; it doesn't support csd).
+
+2022/05/09    Nakata Maho <[email protected]> 2.0 branch
+            * gsv.in: Test for Generalized SVD routines
+              have passed for all precisions.
+            * Csg.in: MPFR tests have also passed. Thus, tests of the Generalized
+              Nonsymmetric Eigenvalue Problem routines have passed for all precisions.
+
+2022/05/08    Nakata Maho <[email protected]> 2.0 branch
+            * Cgd.in: Tests for complex Nonsymmetric generalized Eigenvalue/Schur Form Driver/Expert Driver
+              have passed for all precisions.
+
+2022/05/06    Nakata Maho <[email protected]> 2.0 branch
+            * svd.in: Tests of the complex Singular Value Decomposition routines
+              have passed for all precisions except for dd and qd.
+
+2022/05/04    Nakata Maho <[email protected]> 2.0 branch
+            * Cgg.in: Tests of the complex Generalized Nonsymmetric Eigenvalue Problem routines
+              have passed for all precisions.
+
+2022/05/03    Nakata Maho <[email protected]> 2.0 branch
+            * Cgg.in: Tests of the Generalized Nonsymmetric Eigenvalue Problem routines
+              have passed for all precisions expect for _Float128 and _Float64x.
+            * Csg.in: Tests of the Generalized Nonsymmetric Eigenvalue Problem routines
+              have passed for all precisions expect for mpfr.
+
+2022/05/02    Nakata Maho <[email protected]> 2.0 branch
+            * Ced.in: Tests for Complex Nonsymmetric Schur Form, Eigenvalue expert/expert drivers
+              have passed for all precisions.
+
+2022/05/01    Nakata Maho <[email protected]> 2.0 branch
+            * se2.in: Tests for Hermitian Eigenvalue Problem routines for complex numbers have passed
+              for all precisions.
+
+2022/04/30    Nakata Maho <[email protected]> 2.0 branch
+            * sep.in: Tests for Hermitian Eigenvalue Problem routines for complex numbers have passed
+              for all precisions.
+
+2022/04/26    Nakata Maho <[email protected]> 2.0 branch
+            * nep.in: Test for Nonsymmetric Eigenvalue Problem routines for complex have passed
+              for all precisions.
+            * lse.in: Test for Constrained Linear Least Squares routines for complex have passed
+              for all precisions.
+            * Csb.in: Test for Hermitian Eigenvalue Problem routines have passed
+            * glm.in: Tests of the Generalized Linear Regression Model routines for complex have passed
+            * gqr.in: testing Generalized QR and RQ routines for complex have passed
+            * Cbb.in: reduction of a general band matrix to real bidiagonal form tests have passed
+
+2022/04/24    Nakata Maho <[email protected]> 2.0 branch
+            * Cec.in test (Testing complex Eigen Condition Routines) have passed for all precisions.
+
 2022/03/21    Nakata Maho <[email protected]> 2.0 branch
             * Ctest.in test have passed for all precisions, except for linear fitting.
               Some tests using SVDs still fails, however, they will be fixed when eig tests

README.md

@@ -206,6 +206,7 @@ This is the release process for MPLAPACK 3.0.0
 * version 1.0.0 https://github.com/nakatamaho/mplapack/blob/master/doc/Release1.0.0.md
 
 # History
+* 2022/05/24 2.0 All tests for complex and RFP have passed! alpha version is merged into HEAD.
 * 2022/03/21 2.0 (develoment ongoing). Now complex lin tests have passed for all precisions.
 * 2021/11/1 1.0.1 release. Fixing dd and qd arithmetic by Inte One API.
 * 2021/10/1 1.0.0 release. Huge improvement; all real LAPACK routines are available; SVD, eigenproblem solver for non symmetric matrices are added. manual is available:  https://raw.githubusercontent.com/nakatamaho/mplapack/master/doc/manual/manual.pdf 

configure.ac

@@ -2,9 +2,9 @@ dnl
 dnl Configure for MPLAPACK
 dnl
 
-m4_define([MPLAPACK_VER_MAJOR], [1])
+m4_define([MPLAPACK_VER_MAJOR], [2])
 m4_define([MPLAPACK_VER_MINOR], [0])
-m4_define([MPLAPACK_VER_PATCH], [1])
+m4_define([MPLAPACK_VER_PATCH], [0])
 m4_define([MPLAPACK_VERSION], [MPLAPACK_VER_MAJOR.MPLAPACK_VER_MINOR.MPLAPACK_VER_PATCH])
 AC_INIT(mplapack, [MPLAPACK_VERSION])
 
@@ -30,6 +30,10 @@ AC_CONFIG_FILES([packaging/PKGBUILD packaging/mplapack.spec])
 
 AC_CANONICAL_TARGET
 
+: ${CFLAGS=""}
+: ${CXXFLAGS=""}
+: ${FCFLAGS=""}
+
 AC_PROG_CC
 IS_INTELCC=0
 AC_COMPILE_IFELSE([
@@ -130,6 +134,7 @@ AS_HELP_STRING([--enable-optimization], [Enable compiler level optimization]),
     no)  enable_optimization=no  ;;
     *) AC_MSG_ERROR(bad value ${enableval} for --enable-optimization) ;;
         esac],[enable_optimization=yes])
+AM_CONDITIONAL(ENABLE_OPTIMIZATION, test x$enable_optimization = xyes)
 
 if test x"$enable_optimization" = x"yes" && test x"$enable_debug" = x"yes" ; then
         AC_MSG_ERROR([you cannot set both enable-optimization=yes and enable-debug=yes])
@@ -824,17 +829,17 @@ if test x"$enable_optimization" = x"no" && test x"$enable_debug" = x"yes" ; then
     FCFLAGS="-O0 -g"
 fi
 
-if test x"$enable_optimization" = x"yes" && test x"$enable_debug" = x"no" ; then    
+if test x"$enable_optimization" = x"yes" && test x"$enable_debug" != x"yes" ; then
     case $target in
        powerpc*-*-*)
            CFLAGS="$CFLAGS -mcpu=native -O2 -finline-functions -funroll-all-loops"
            CXXFLAGS="$CXXFLAGS -mcpu=native -O2 -finline-functions -funroll-all-loops"
            FCFLAGS="$FCFLAGS -mcpu=native -O2 -finline-functions -funroll-all-loops"
        ;;
        * ) 
-           CFLAGS="$CFLAGS -march=native -g -O2 -finline-functions -funroll-all-loops"
-           CXXFLAGS="$CXXFLAGS -march=native -g -O2 -finline-functions -funroll-all-loops"
-           FCFLAGS="$FCFLAGS -march=native -g -O2 -finline-functions -funroll-all-loops"
+           CFLAGS="$CFLAGS -march=native -O2 -finline-functions -funroll-all-loops"
+           CXXFLAGS="$CXXFLAGS -march=native -O2 -finline-functions -funroll-all-loops"
+           FCFLAGS="$FCFLAGS -march=native -O2 -finline-functions -funroll-all-loops"
        ;;
     esac
     if test x"$is_intelcxx" = x"1"; then

external/lapack/patches/print.c

@@ -1,9 +1,13 @@
 #include <stdio.h>
 #include <string.h>
 
-//  please use as following: Do not forget adding //CHAR(0) at the end.
+//  please use as follows: Do not forget adding //CHAR(0) at the end.
 //         CALL PRINTSTR('#dgelsd.f l.486'//CHAR(0))
 //         CALL PRINTVEC('s='//CHAR(0),s,n)
+//         CALL PRINTMAT('e='//CHAR(0),m,m,e,lde)
+//         for complex matrix and vectors
+//         CALL PRINTMATC('e='//CHAR(0),m,m,e,lde)
+//         CALL PRINTVECC('x='//CHAR(0),x,n)
 
 void printdummy_() { printf("\n"); }
 
@@ -19,6 +23,30 @@ void printnum_(char *s, double *A) {
     printf("%s%+21.16e\n", s, *A);
 }
 
+void printnumc(double _Complex rtmp) {
+    printf("%+21.16e%+21.16ei", __real__ rtmp, __imag__ rtmp);
+}
+
+void printmatc_(char *s, int *N, int *M, double _Complex *A, int *LDA) {
+    double _Complex tmp;
+    printf("%s[ ", s);
+    for (int i = 0; i < *N; i++) {
+        printf("[ ");
+        for (int j = 0; j < *M; j++) {
+            tmp = A[i + j * (*LDA)];
+            printnumc(tmp);
+            if (j < *M - 1)
+                printf(", ");
+        }
+        if (i < *N - 1)
+            printf("]; ");
+        else
+            printf("] ");
+    }
+    printf("]");
+    printf("\n");
+}
+
 void printmat_(char *s, int *N, int *M, double *A, int *LDA) {
     double tmp;
     printf("%s[ ", s);
@@ -72,6 +100,18 @@ void printvec_(char *s, double *A, int *lenvec) {
     printf("\n");
 }
 
+void printvecc_(char *s, double _Complex*A, int *lenvec) {
+    double _Complex tmp;
+    printf("%s[ ", s);
+    for (int i = 0; i < *lenvec; i++) {
+        printnumc(tmp);
+        if (i < *lenvec - 1)
+            printf(", ");
+    }
+    printf("]");
+    printf("\n");
+}
+
 void printveci_(char *s, int *A, int *lenvec) {
     int tmp;
     printf("%s[ ", s);