[PATCH] D40565: [LibFuzzer] Improve comments on `CounterToFeature()` function.

[Dan Liew via Phabricator via llvm-commits]

delcypher created this revision.

This is based on discussion in https://reviews.llvm.org/D40376 .

The comments try to explain the reason for the current implementation
and note that it might change in the future, so clients should not
rely on this particular implementation.


https://reviews.llvm.org/D40565

Files:
  lib/fuzzer/FuzzerTracePC.h


Index: lib/fuzzer/FuzzerTracePC.h
===================================================================
--- lib/fuzzer/FuzzerTracePC.h
+++ lib/fuzzer/FuzzerTracePC.h
@@ -196,9 +196,20 @@
       Handle8bitCounter(FirstFeature, P - Begin, V);
 }
 
-// Given a non-zero Counters returns a number in [0,7].
+// Given a non-zero Counter returns a number in the range [0,7].
 template<class T>
 unsigned CounterToFeature(T Counter) {
+    // Returns a feature number by placing Counters into buckets as illustrated
+    // below.
+    //
+    // Counter bucket: [1] [2] [3] [4-7] [8-15] [16-31] [32-127] [128+]
+    // Feature number:  0   1   2    3     4       5       6       7
+    //
+    // This is a heuristic taken from AFL (see
+    // http://lcamtuf.coredump.cx/afl/technical_details.txt).
+    //
+    // This implementation may change in the future so clients should
+    // not rely on it.
     assert(Counter);
     unsigned Bit = 0;
     /**/ if (Counter >= 128) Bit = 7;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D40565.124593.patch
Type: text/x-patch
Size: 979 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20171128/9a2ab182/attachment.bin>