diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index f5065e5..c114ac2 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -13,6 +13,7 @@ jobs:
         os:
           - ubuntu-latest
           - macos-latest
+          - windows-latest
         rust:
           - stable
           - beta
@@ -29,6 +30,7 @@ jobs:
           override: true
 
       - run: cargo build
+      - run: cargo test
       - run: cargo bench 
       # test filter parameter for cargo bench
       - run: "[ $(cargo bench | wc -l) -gt $(cargo bench 15 | wc -l) ]"
diff --git a/src/bench.rs b/src/bench.rs
index bd9179b..510723c 100644
--- a/src/bench.rs
+++ b/src/bench.rs
@@ -18,7 +18,7 @@ pub trait Bench<'a> {
     fn clear_results(&mut self);
 }
 
-type CallBench<'a, I> = Box<dyn FnMut(&'a I)>;
+pub(crate) type CallBench<'a, I> = Box<dyn FnMut(&'a I)>;
 
 pub(crate) struct NamedBench<'a, I> {
     pub name: String,
diff --git a/src/bench_group.rs b/src/bench_group.rs
index d95d647..be35d04 100644
--- a/src/bench_group.rs
+++ b/src/bench_group.rs
@@ -6,7 +6,7 @@ use crate::{
     NamedInput,
 };
 
-/// `BenchGroup` is a group of benchmarks run together.
+/// `BenchGroup` is a group of benchmarks wich are executed together.
 ///
 pub struct BenchGroup<'a> {
     name: Option<String>,
@@ -44,7 +44,7 @@ impl<'a> BenchGroup<'a> {
         self
     }
 
-    /// Enables throughput reporting. The throughput will be valid for all inputs that are
+    /// Enables throughput reporting. The `input_size` will be used for all benchmarks that are
     /// registered afterwards.
     pub fn set_input_size(&mut self, input_size: usize) {
         self.input_size_in_bytes = Some(input_size);
diff --git a/src/bench_input_group.rs b/src/bench_input_group.rs
index c5ea671..39b34ff 100644
--- a/src/bench_input_group.rs
+++ b/src/bench_input_group.rs
@@ -15,8 +15,9 @@ pub(crate) type Alloc = &'static dyn PeakMemAllocTrait;
 ///
 /// The ownership of the inputs is transferred
 /// to the `InputGroup`. If this is not possible, use [BenchRunner](crate::BenchRunner) instead.
-pub struct InputGroup<I = ()> {
+pub struct InputGroup<I: 'static = ()> {
     inputs: Vec<OwnedNamedInput<I>>,
+    //benches: Vec<CallBench<'static, I>>,
     bench_group: BenchGroup<'static>,
 }
 
@@ -74,6 +75,7 @@ impl<I: 'static> InputGroup<I> {
         InputGroup {
             inputs,
             bench_group: BenchGroup::new(runner),
+            //benches: Vec::new(),
         }
     }
     /// Set the peak mem allocator to be used for the benchmarks.
@@ -134,9 +136,6 @@ impl<I: 'static> InputGroup<I> {
 
     /// Run the benchmarks and report the results.
     pub fn run(&mut self) {
-        //if let Some(name) = &self.name {
-        //println!("{}", name.black().on_red().invert().bold());
-        //}
         let input_name_to_ordinal: HashMap<String, usize> = self
             .inputs
             .iter()
@@ -151,9 +150,6 @@ impl<I: 'static> InputGroup<I> {
             |b| b.get_input_name(),
             |group| {
                 let input_name = group[0].get_input_name().to_owned();
-                //if let Some(input_size) = group[0].get_input_size_in_bytes() {
-                //self.bench_group.set_input_size(input_size);
-                //}
                 self.bench_group.runner.run_group(Some(&input_name), group);
             },
         );
diff --git a/src/bench_runner.rs b/src/bench_runner.rs
index 267c4f0..0cbdc6f 100644
--- a/src/bench_runner.rs
+++ b/src/bench_runner.rs
@@ -10,7 +10,10 @@ use peakmem_alloc::*;
 use yansi::Paint;
 
 pub(crate) type Alloc = &'static dyn PeakMemAllocTrait;
-pub(crate) const NUM_RUNS: usize = 32;
+
+/// Each bench is run N times in a inner loop.
+/// The outer loop is fixed. In the outer loop the order of the benchmarks in a group is shuffled.
+pub const NUM_RUNS: usize = 32;
 
 /// The main struct to run benchmarks.
 ///
@@ -198,7 +201,6 @@ impl BenchRunner {
             //
             // This has the drawback, that one bench will affect another one.
             shuffle(&mut bench_indices, iteration as u64);
-            //std::thread::yield_now();
 
             for bench_idx in bench_indices.iter() {
                 if let Some(cache_trasher) = cache_trasher {
diff --git a/src/lib.rs b/src/lib.rs
index 126acd2..9f4ea9f 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -9,7 +9,7 @@
 )]
 
 //! Binggan (餅乾, bǐng gān, means cookie in Chinese) is a benchmarking library for Rust.
-//! It is designed to be simple to use and to provide a good overview of the performance of your code and its memory consumption.
+//! It is designed to provide fast and stable results, report peak memory consumption and integrate with perf.
 //!
 //! It allows arbitrary named inputs to be passed to the benchmarks.
 //!
@@ -45,7 +45,7 @@
 //! // Run the benchmark for the group with input `Vec<usize>`
 //! fn bench_group(mut runner: InputGroup<Vec<usize>>) {
 //!     runner.set_alloc(GLOBAL); // Set the peak mem allocator. This will enable peak memory reporting.
-//!     runner.enable_perf(); // Enable perf integration. This only works on linux.
+//!     runner.config().enable_perf(); // Enable perf integration. This only works on linux.
 //!     runner.register("vec", move |data| {
 //!         test_vec(data);
 //!     });
@@ -120,10 +120,10 @@
 //!     let mut group = runner.new_group();
 //!     for (input_name, data) in inputs.iter() {
 //!         group.set_input_size(data.len() * std::mem::size_of::<usize>());
-//!         group.register_with_input("vec", input_name, data, move |data| {
+//!         group.register_with_input("vec", data, move |data| {
 //!             black_box(test_vec(data));
 //!         });
-//!         group.register_with_input("hashmap", input_name, data, move |data| {
+//!         group.register_with_input("hashmap", data, move |data| {
 //!             black_box(test_hashmap(data));
 //!         });
 //!     }