Add changelog

diff --git a/doc/news/changes/minor/20170324MartinKronbichler b/doc/news/changes/minor/20170324MartinKronbichler
new file mode 100644 (file)
index 0000000..1ae9f0e
--- /dev/null
+++ b/doc/news/changes/minor/20170324MartinKronbichler
@@ -0,0 +1,6 @@
+New: The tensor product kernels used by FEEvaluation have gained a faster
+variant for evaluating gradients derived from spectral identities. This
+improves performance of the Laplace operator evaluation by 10-20% in the
+computation-bound case.
+<br>
+(Martin Kronbichler, 2017/03/24)

diff --git a/doc/news/changes/minor/20170324MartinKronbichler-b b/doc/news/changes/minor/20170324MartinKronbichler-b
new file mode 100644 (file)
index 0000000..60a5e4e
--- /dev/null
+++ b/doc/news/changes/minor/20170324MartinKronbichler-b
@@ -0,0 +1,6 @@
+New: The FEEvaluation::read_dof_values and
+FEEvaluation::distribute_local_to_global now use dedicate gather (AVX2,
+AVX-512) and scatter (AVX-512) instructions for faster vector access if those
+are available.
+<br>
+(Martin Kronbichler, 2017/03/24)