Revision: 55184
          http://brlcad.svn.sourceforge.net/brlcad/?rev=55184&view=rev
Author:   starseeker
Date:     2013-04-18 20:57:28 +0000 (Thu, 18 Apr 2013)
Log Message:
-----------
Move dvec test into tests subdirectory

Modified Paths:
--------------
    brlcad/trunk/src/librt/CMakeLists.txt
    brlcad/trunk/src/librt/tests/CMakeLists.txt

Added Paths:
-----------
    brlcad/trunk/src/librt/tests/test_dvec.cpp

Removed Paths:
-------------
    brlcad/trunk/src/librt/test_dvec.cpp

Modified: brlcad/trunk/src/librt/CMakeLists.txt
===================================================================
--- brlcad/trunk/src/librt/CMakeLists.txt       2013-04-18 20:51:30 UTC (rev 
55183)
+++ brlcad/trunk/src/librt/CMakeLists.txt       2013-04-18 20:57:28 UTC (rev 
55184)
@@ -247,7 +247,6 @@
   )
 
 set(librt_ignored_files
-  test_dvec.cpp
   columnparse.c
   librt_private.h
   namegen.c
@@ -315,9 +314,6 @@
 #BRLCAD_ADDEXEC(test_root3-subd test_root3-subd.cpp 
"librt;libwdb;libbrep;libbu" NOSTRICT NO_INSTALL)
 #BRLCAD_ADDEXEC(test_surfacetree test_surfacetree.cpp 
"librt;libwdb;libbrep;libbu" NOSTRICT NO_INSTALL)
 
-BRLCAD_ADDEXEC(test_dvec test_dvec.cpp "librt" NOSTRICT NO_INSTALL LOCAL)
-ADD_TEST(test_dvec test_dvec)
-
 add_subdirectory(tests)
 
 # Local Variables:

Deleted: brlcad/trunk/src/librt/test_dvec.cpp
===================================================================
--- brlcad/trunk/src/librt/test_dvec.cpp        2013-04-18 20:51:30 UTC (rev 
55183)
+++ brlcad/trunk/src/librt/test_dvec.cpp        2013-04-18 20:57:28 UTC (rev 
55184)
@@ -1,188 +0,0 @@
-/*                   T E S T _ D V E C . C P P
- * BRL-CAD
- *
- * Copyright (c) 2008-2013 United States Government as represented by
- * the U.S. Army Research Laboratory.
- *
- * This library is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public License
- * version 2.1 as published by the Free Software Foundation.
- *
- * This library is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with this file; see the file named COPYING for more
- * information.
- */
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <iostream>
-#include <assert.h>
-#include "vmath.h"
-#include "brep.h"
-#include "dvec.h"
-
-
-const int COUNT = 1000000;
-
-int
-main(int argc, char** argv)
-{
-    if (argc > 1)
-       printf("Usage: %s\n", argv[0]);
-
-#ifdef __x86_vector__
-    printf("using x86 vectorization\n");
-#elif defined(__fpu_vector__)
-    printf("using fpu vectorization\n");
-#endif
-
-    {
-       // test correctness
-       double _a[8] VEC_ALIGN = {0, 1, 2, 3, 4, 5, 6, 7};
-       double _b[8] VEC_ALIGN = {2, 4, 6, 8, 10, 12, 14, 16};
-       dvec<8> a(_a);
-       dvec<8> b(_b);
-
-       dvec<8> c = a + b;
-       double _c[8] VEC_ALIGN = {2, 5, 8, 11, 14, 17, 20, 23};
-       assert(c == dvec<8>(_c));
-
-       dvec<8> d = a - b;
-       double _d[8] VEC_ALIGN = {-2, -3, -4, -5, -6, -7, -8, -9};
-       assert(d == dvec<8>(_d));
-
-       dvec<8> e = a * b;
-       double _e[8] VEC_ALIGN = {0, 4, 12, 24, 40, 60, 84, 112};
-       assert(e == dvec<8>(_e));
-
-       dvec<8> f = a / b;
-       double _f[8] VEC_ALIGN = {0, 0.25, 0.333333333333, 0.375, 0.40, 
0.4166666666, 0.42857142, 0.4375};
-       assert(f == dvec<8>(_f));
-
-       dvec<8> g = a.madd(c, b);
-       double _g[8] VEC_ALIGN = {2, 9, 22, 41, 66, 97, 134, 177};
-       assert(g == dvec<8>(_g));
-
-       dvec<8> h = a.madd(2.0, b);
-       double _h[8] VEC_ALIGN = {2, 6, 10, 14, 18, 22, 26, 30};
-       assert(h == dvec<8>(_h));
-
-       srand(time(NULL));
-       double total = 0.0;
-       double rval[] = {
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0,
-           rand()/10000.0 };
-
-       clock_t start = clock();
-       for (int i = 0; i < COUNT; i++) {
-           double l_a[8] VEC_ALIGN = {rval[i%8], rval[i%8+1], rval[i%8+2], 
rval[i%8+3],
-                                      rval[i%8+4], rval[i%8+5], rval[i%8+6], 
rval[i%8+7]};
-           double l_b[8] VEC_ALIGN = {rval[i%8+1], rval[i%8+2], rval[i%8+3], 
rval[i%8+4],
-                                      rval[i%8+5], rval[i%8+6], rval[i%8+7], 
rval[i%8+8]};
-
-           dvec<8> la(l_a);
-           dvec<8> lb(l_b);
-
-           dvec<8> lc = la + lb;
-           dvec<8> ld = lc - la;
-           dvec<8> le = ld * lc;
-           dvec<8> lf = le / la;
-           dvec<8> lg = la.madd(rval[i%16], lf);
-           dvec<8> lh = lg.madd(lf, la+lb*lc-ld/le*lf);
-           total += lh[7];
-       }
-       printf("dvec<8> time: %3.4g\n", 
(double)(clock()-start)/(double)CLOCKS_PER_SEC);
-    }
-
-/* test correctness */
-    vec2d a(100.0, -100.0);
-    vec2d b(200.0, -200.0);
-
-    vec2d c = a + b;
-    assert(vequals(c, vec2d(300.0, -300.0)));
-
-    vec2d d = a - b;
-    assert(vequals(d, vec2d(-100.0, 100.0)));
-
-    vec2d e = a * b;
-    assert(vequals(e, vec2d(20000.0, 20000.0)));
-
-    vec2d f = b / a;
-    assert(vequals(f, vec2d(2.0, 2.0)));
-
-    vec2d g = a.madd(20.0, b);
-    assert(vequals(g, vec2d(2200.0, -2200.0)));
-
-    vec2d h = a.madd(d, b);
-    assert(vequals(h, vec2d(-9800.0, -10200.0)));
-
-/* simple performance test */
-
-    srand(time(NULL));
-    double total = 0.0;
-    double rval[] = {
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0,
-       rand()/10000.0 };
-
-    clock_t start = clock();
-    for (int i = 0; i < COUNT; i++) {
-       vec2d la(rval[i%12], rval[i%12+2]);
-       vec2d lb(rval[i%12+1], rval[i%12+3]);
-
-       vec2d lc = la + lb;
-       vec2d ld = lc - la;
-       vec2d le = ld * lc;
-       vec2d lf = le / la;
-       vec2d lg = la.madd(rval[i%12], lf);
-       vec2d lh = lg.madd(lg, la + lb - lc + ld - le + lf);
-       total += lh.x() + lh.y();
-    }
-    printf("vec2d time: %3.4g\n", 
(double)(clock()-start)/(double)CLOCKS_PER_SEC);
-
-    /* return 'fail' if no time elapsed */
-    return (total > 0) ? 0 : 1;
-}
-
-
-// Local Variables:
-// tab-width: 8
-// mode: C++
-// c-basic-offset: 4
-// indent-tabs-mode: t
-// c-file-style: "stroustrup"
-// End:
-// ex: shiftwidth=4 tabstop=8

Modified: brlcad/trunk/src/librt/tests/CMakeLists.txt
===================================================================
--- brlcad/trunk/src/librt/tests/CMakeLists.txt 2013-04-18 20:51:30 UTC (rev 
55183)
+++ brlcad/trunk/src/librt/tests/CMakeLists.txt 2013-04-18 20:57:28 UTC (rev 
55184)
@@ -1,6 +1,10 @@
+BRLCAD_ADDEXEC(test_dvec test_dvec.cpp "librt" NOSTRICT NO_INSTALL LOCAL)
+add_test(test_dvec test_dvec)
+
+# NURBS testing
 BRLCAD_ADDEXEC(nurbs_tester nurbs_tests.cpp "librt;libbrep;libbu" NO_INSTALL 
LOCAL)
 
-add_test(get_closest_point-distinct_points nurbs_tester 
${CMAKE_CURRENT_SOURCE_DIR}/nurbs_surfaces.g 1)
+add_test(NURBS-get_closest_point-distinct_points nurbs_tester 
${CMAKE_CURRENT_SOURCE_DIR}/nurbs_surfaces.g 1)
 
 CMAKEFILES(nurbs_surfaces.g)
 

Copied: brlcad/trunk/src/librt/tests/test_dvec.cpp (from rev 55162, 
brlcad/trunk/src/librt/test_dvec.cpp)
===================================================================
--- brlcad/trunk/src/librt/tests/test_dvec.cpp                          (rev 0)
+++ brlcad/trunk/src/librt/tests/test_dvec.cpp  2013-04-18 20:57:28 UTC (rev 
55184)
@@ -0,0 +1,188 @@
+/*                   T E S T _ D V E C . C P P
+ * BRL-CAD
+ *
+ * Copyright (c) 2008-2013 United States Government as represented by
+ * the U.S. Army Research Laboratory.
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public License
+ * version 2.1 as published by the Free Software Foundation.
+ *
+ * This library is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this file; see the file named COPYING for more
+ * information.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <iostream>
+#include <assert.h>
+#include "vmath.h"
+#include "brep.h"
+#include "dvec.h"
+
+
+const int COUNT = 1000000;
+
+int
+main(int argc, char** argv)
+{
+    if (argc > 1)
+       printf("Usage: %s\n", argv[0]);
+
+#ifdef __x86_vector__
+    printf("using x86 vectorization\n");
+#elif defined(__fpu_vector__)
+    printf("using fpu vectorization\n");
+#endif
+
+    {
+       // test correctness
+       double _a[8] VEC_ALIGN = {0, 1, 2, 3, 4, 5, 6, 7};
+       double _b[8] VEC_ALIGN = {2, 4, 6, 8, 10, 12, 14, 16};
+       dvec<8> a(_a);
+       dvec<8> b(_b);
+
+       dvec<8> c = a + b;
+       double _c[8] VEC_ALIGN = {2, 5, 8, 11, 14, 17, 20, 23};
+       assert(c == dvec<8>(_c));
+
+       dvec<8> d = a - b;
+       double _d[8] VEC_ALIGN = {-2, -3, -4, -5, -6, -7, -8, -9};
+       assert(d == dvec<8>(_d));
+
+       dvec<8> e = a * b;
+       double _e[8] VEC_ALIGN = {0, 4, 12, 24, 40, 60, 84, 112};
+       assert(e == dvec<8>(_e));
+
+       dvec<8> f = a / b;
+       double _f[8] VEC_ALIGN = {0, 0.25, 0.333333333333, 0.375, 0.40, 
0.4166666666, 0.42857142, 0.4375};
+       assert(f == dvec<8>(_f));
+
+       dvec<8> g = a.madd(c, b);
+       double _g[8] VEC_ALIGN = {2, 9, 22, 41, 66, 97, 134, 177};
+       assert(g == dvec<8>(_g));
+
+       dvec<8> h = a.madd(2.0, b);
+       double _h[8] VEC_ALIGN = {2, 6, 10, 14, 18, 22, 26, 30};
+       assert(h == dvec<8>(_h));
+
+       srand(time(NULL));
+       double total = 0.0;
+       double rval[] = {
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0,
+           rand()/10000.0 };
+
+       clock_t start = clock();
+       for (int i = 0; i < COUNT; i++) {
+           double l_a[8] VEC_ALIGN = {rval[i%8], rval[i%8+1], rval[i%8+2], 
rval[i%8+3],
+                                      rval[i%8+4], rval[i%8+5], rval[i%8+6], 
rval[i%8+7]};
+           double l_b[8] VEC_ALIGN = {rval[i%8+1], rval[i%8+2], rval[i%8+3], 
rval[i%8+4],
+                                      rval[i%8+5], rval[i%8+6], rval[i%8+7], 
rval[i%8+8]};
+
+           dvec<8> la(l_a);
+           dvec<8> lb(l_b);
+
+           dvec<8> lc = la + lb;
+           dvec<8> ld = lc - la;
+           dvec<8> le = ld * lc;
+           dvec<8> lf = le / la;
+           dvec<8> lg = la.madd(rval[i%16], lf);
+           dvec<8> lh = lg.madd(lf, la+lb*lc-ld/le*lf);
+           total += lh[7];
+       }
+       printf("dvec<8> time: %3.4g\n", 
(double)(clock()-start)/(double)CLOCKS_PER_SEC);
+    }
+
+/* test correctness */
+    vec2d a(100.0, -100.0);
+    vec2d b(200.0, -200.0);
+
+    vec2d c = a + b;
+    assert(vequals(c, vec2d(300.0, -300.0)));
+
+    vec2d d = a - b;
+    assert(vequals(d, vec2d(-100.0, 100.0)));
+
+    vec2d e = a * b;
+    assert(vequals(e, vec2d(20000.0, 20000.0)));
+
+    vec2d f = b / a;
+    assert(vequals(f, vec2d(2.0, 2.0)));
+
+    vec2d g = a.madd(20.0, b);
+    assert(vequals(g, vec2d(2200.0, -2200.0)));
+
+    vec2d h = a.madd(d, b);
+    assert(vequals(h, vec2d(-9800.0, -10200.0)));
+
+/* simple performance test */
+
+    srand(time(NULL));
+    double total = 0.0;
+    double rval[] = {
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0,
+       rand()/10000.0 };
+
+    clock_t start = clock();
+    for (int i = 0; i < COUNT; i++) {
+       vec2d la(rval[i%12], rval[i%12+2]);
+       vec2d lb(rval[i%12+1], rval[i%12+3]);
+
+       vec2d lc = la + lb;
+       vec2d ld = lc - la;
+       vec2d le = ld * lc;
+       vec2d lf = le / la;
+       vec2d lg = la.madd(rval[i%12], lf);
+       vec2d lh = lg.madd(lg, la + lb - lc + ld - le + lf);
+       total += lh.x() + lh.y();
+    }
+    printf("vec2d time: %3.4g\n", 
(double)(clock()-start)/(double)CLOCKS_PER_SEC);
+
+    /* return 'fail' if no time elapsed */
+    return (total > 0) ? 0 : 1;
+}
+
+
+// Local Variables:
+// tab-width: 8
+// mode: C++
+// c-basic-offset: 4
+// indent-tabs-mode: t
+// c-file-style: "stroustrup"
+// End:
+// ex: shiftwidth=4 tabstop=8

This was sent by the SourceForge.net collaborative development platform, the 
world's largest Open Source development site.


------------------------------------------------------------------------------
Precog is a next-generation analytics platform capable of advanced
analytics on semi-structured data. The platform includes APIs for building
apps and a phenomenal toolset for data science. Developers can use
our toolset for easy data analysis & visualization. Get a free account!
http://www2.precog.com/precogplatform/slashdotnewsletter
_______________________________________________
BRL-CAD Source Commits mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/brlcad-commits

Reply via email to