Mention normalize in the manual

README.ipynb

@@ -1039,7 +1039,7 @@
     "\n",
     "`measure(:@measuring_macro, trace)` applies the given _quoted_ macro (put a `:` in front of the `@`) to every function call in `trace`. Useful macros are `@allocated`, `@elapsed` and [BenchmarkTools](https://github.com/JuliaCI/BenchmarkTools.jl)' `@belapsed`. Because measuring involves rerunning the function calls in the trace, it can be slow. A good rule-of-thumb is to use the more accurate `@belapsed` when the code to profile takes less than half a second and the trace is relatively short, and `@elapsed` when it takes less than a minute. When the trace is very large, try the `explore_worst=true` option, which will only expand the worst callee of each call (so the runtime is $O(wd)$ instead of $O(w^d)$). You can also `prune` the trace first.\n",
     "\n",
-    "Remember that due to timing fluctuations, even if `f(x)` calls `g(x)`, it does not imply that `@elapsed(f(x)) >= @elapsed(g(x))`."
+    "Remember that due to timing fluctuations, even if `f(x)` calls `g(x)`, it does not imply that `@elapsed(f(x)) >= @elapsed(g(x))`. `normalize(tr::Trace)` is useful when exploring the output of `measure`."
    ]
   },
   {
@@ -1081,7 +1081,7 @@
     "trace_pycall = @trace PyCall math[:pi];\n",
     "\n",
     "using BenchmarkTools: @belapsed\n",
-    "measure(:@belapsed, trace_pycall; normalize=true, threshold=0.005)  # only show calls that take >0.5% of total runtime"
+    "measure(:@belapsed, trace_pycall; normalize=true, threshold=0.005)  # only show/explore calls that take >0.5% of total runtime"
    ]
   },
   {