LICM pass (#87)

* LICM pass

* Uses loop analysis to detect loop tree
* For each loop (starting with the inner ones), create a pre-header and move there loop-invariant instructions
* An instruction is loop invariant if it does not use as argument a value defined earlier in the loop
* File tests to check LICM's correctness
* Optimized pre-header creation
If the loop already has a natural pre-header, we use it instead of creating a new one.
The natural pre-header of a loop is the only predecessor of the header it doesn't dominate.

filetests/licm/nested_loops.cton

@@ -0,0 +1,52 @@
+test licm
+
+function nested_loops(i32) -> i32 {
+
+ebb0(v0: i32):
+    v1 = iconst.i32 1
+    v2 = iconst.i32 2
+    v3 = iadd v1, v2
+    v4 = isub v0, v1
+    jump ebb1(v4,v4)
+
+ebb1(v10: i32,v11: i32):
+    brz v11, ebb2(v10)
+    v12 = iconst.i32 1
+    v15 = iadd v12, v4
+    v13 = isub v11, v12
+    jump ebb1(v10,v13)
+
+ebb2(v20: i32):
+    brz v20, ebb3(v20)
+    jump ebb0(v20)
+
+ebb3(v30: i32):
+    return v30
+
+}
+
+; sameln:function nested_loops(i32) -> i32 {
+; nextln: ebb4(v12: i32):
+; nextln:     v1 = iconst.i32 1
+; nextln:     v2 = iconst.i32 2
+; nextln:     v3 = iadd v1, v2
+; nextln:     v7 = iconst.i32 1
+; nextln:     jump ebb0(v12)
+; nextln: 
+; nextln: ebb0(v0: i32):
+; nextln:     v4 = isub v0, v1
+; nextln:     v8 = iadd.i32 v7, v4
+; nextln:     jump ebb1(v4, v4)
+; nextln: 
+; nextln: ebb1(v5: i32, v6: i32):
+; nextln:     brz v6, ebb2(v5)
+; nextln:     v9 = isub v6, v7
+; nextln:     jump ebb1(v5, v9)
+; nextln: 
+; nextln: ebb2(v10: i32):
+; nextln:     brz v10, ebb3(v10)
+; nextln:     jump ebb0(v10)
+; nextln: 
+; nextln: ebb3(v11: i32):
+; nextln:     return v11
+; nextln: }