Introduce the DTrace scripts used to measure the performance in

the project.

.gitignore

@@ -3,7 +3,6 @@
 ebin
 *.beam
 *.o
-*.d
 *.eqc
 *.so
 eqc_test/.eqc-info

bench/enacl_nif.d

@@ -0,0 +1,37 @@
+/* Dirty NIF schedule overhead */
+pid$target:beam.smp:schedule_dirty_cpu_nif:return
+{
+	s = timestamp;
+}
+
+pid$target:libsodium.so.*:randombytes:entry {
+	e = timestamp;
+}
+
+pid$target:beam.smp:execute_dirty_nif:entry
+/s != 0/
+{
+	@SchedTime = lquantize(timestamp - s, 0, 10000, 250);
+	s = 0;
+}
+
+pid$target:beam.smp:execute_dirty_nif:return
+{
+	@ExecTime = lquantize(timestamp - e, 0, 10000, 250);
+	e = 0;
+	r = timestamp;
+}
+
+pid$target:beam.smp:dirty_nif_finalizer:entry
+/r != 0/
+{
+	@ReturnTime = lquantize(timestamp - r, 0, 10000, 250);
+	r = 0;
+}
+
+END
+{
+	printa("Scheduling overhead (nanos):%@d\n", @SchedTime);
+	printa("Return overhead (nanos):%@d\n", @ReturnTime);
+	printa("Exec time (nanos):%@d\n", @ExecTime);
+}

bench/funcall_enacl.d

@@ -0,0 +1,9 @@
+erlang*:::nif-entry
+{
+	funcall_entry_ts[cpu, copyinstr(arg1)] = vtimestamp;
+}
+
+erlang*:::nif-return
+{
+	@time[cpu, copyinstr(arg1)] = lquantize((vtimestamp - funcall_entry_ts[cpu, copyinstr(arg1)] ), 0, 60000, 1000);
+}