bug #698: rewrite LinSpaced for integer scalar types to avoid overflow and guarantee an even spacing when possible.

Otherwise, the "high" bound is implicitly lowered to the largest value allowing for an even distribution.
This changeset also disable vectorization for this integer path.

1 files changed, 8 insertions, 0 deletions

diff --git a/doc/snippets/DenseBase_LinSpacedInt.cpp b/doc/snippets/DenseBase_LinSpacedInt.cpp
new file mode 100644
index 000000000..0d7ae068e
--- /dev/null
+++ b/doc/snippets/DenseBase_LinSpacedInt.cpp
@@ -0,0 +1,8 @@
+cout << "Even spacing inputs:" << endl;
+cout << VectorXi::LinSpaced(8,1,4).transpose() << endl;
+cout << VectorXi::LinSpaced(8,1,8).transpose() << endl;
+cout << VectorXi::LinSpaced(8,1,15).transpose() << endl;
+cout << "Uneven spacing inputs:" << endl;
+cout << VectorXi::LinSpaced(8,1,7).transpose() << endl;
+cout << VectorXi::LinSpaced(8,1,9).transpose() << endl;
+cout << VectorXi::LinSpaced(8,1,16).transpose() << endl;