remove valgrind, update docs

scenario3-syscall-storm/README.md

@@ -134,6 +134,24 @@ Look at:
 - `cpu-migrations`
 - `instructions per cycle`
 
+## Exercise 6: Benchmarking with hyperfine
+
+Use `hyperfine` for proper statistical benchmarking with warmup runs:
+
+```bash
+hyperfine --warmup 1 './read_slow testfile' './read_fast testfile'
+```
+
+This gives you:
+- Mean execution time with standard deviation
+- Min/max times
+- Direct comparison between implementations
+
+For all three implementations:
+```bash
+hyperfine --warmup 1 './read_slow testfile' './read_stdio testfile' './read_fast testfile'
+```
+
 ## Further Exploration
 
 1. What happens with `read(fd, buf, 4096)` vs `read(fd, buf, 65536)`?