before next stage

doc/explain.md

@@ -0,0 +1,14 @@
+# Chart with suffix -1.png
+
+In those three chart, we know that they all have similar instruction, the distribution wasn't similar to linklist, but the average time is.
+
+It's because two reason, one cause high baseline time, another is the data structure I use.
+
+The first reason which cause high baseline time is bad IO optimization, It read from file for every prefix, even worse, I open file to write the chart before calling `usize end=rdtsc_64bits();`, which block.
+
+The second reason is the addition of HashMap, rather than placing linklist(xor linklist) in vector(slice), I placed linklist in a custom hashmap, moreover, when the key is missing in the hashmap, the `hash_get()` function iterating all elements in hashmap, and hash collusion was discovered in one of the bucket, so there are several operation took significantly more time.
+
+# Chart with suffix -2.png
+
+After optimize some IO, I get correct result for combination of hashmap and linklist.
+