diff --git a/Cargo.toml b/Cargo.toml
index 09bb3698..13e340d7 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -41,25 +41,25 @@ codegen-units = 1
 bench = false # Don't disturb criterion command line parsing
 
 [[bench]]
-name = "single_thread_single_byte"
+name = "chunks"
 harness = false
 
 [[bench]]
-name = "single_thread_two_bytes"
+name = "const_push_pop"
 harness = false
 
 [[bench]]
-name = "single_thread_with_chunks"
+name = "push_pop"
 harness = false
 
 [[bench]]
-name = "two_threads"
+name = "single_thread_single_byte"
 harness = false
 
 [[bench]]
-name = "two_threads_with_chunks"
+name = "single_thread_two_bytes"
 harness = false
 
 [[bench]]
-name = "two_threads_const"
+name = "single_thread_with_chunks"
 harness = false
diff --git a/README.md b/README.md
index b119e8d8..26be18e0 100644
--- a/README.md
+++ b/README.md
@@ -37,9 +37,12 @@ Performance
 
 Measuring the performance of a data structure for inter-thread communication
 can be quite brittle and the results depend on many factors.
-A few performance comparisons between competing crates are shown in
-[issue #39](https://github.com/mgeier/rtrb/issues/39),
-but like all benchmarks, they are deeply flawed and to be taken with a grain of salt.
+A few benchmarks for performance comparisons between competing crates
+are available at the sub-directory
+[`performance-comparison/`](https://github.com/mgeier/rtrb/tree/main/performance-comparison)
+and some results thereof are shown in
+[issue #39](https://github.com/mgeier/rtrb/issues/39).
+Like all benchmarks, they are deeply flawed and to be taken with a grain of salt.
 You should make your own measurements that are relevant to your usage patterns.
 Feel free to share your own results by commenting on that issue.
 
diff --git a/benches/two_threads_with_chunks.rs b/benches/chunks.rs
similarity index 96%
rename from benches/two_threads_with_chunks.rs
rename to benches/chunks.rs
index d8266db4..3b261378 100644
--- a/benches/two_threads_with_chunks.rs
+++ b/benches/chunks.rs
@@ -1,6 +1,6 @@
 #![allow(clippy::incompatible_msrv)]
 
-macro_rules! create_two_threads_with_chunks_benchmark {
+macro_rules! create_chunks_benchmark {
     ($($id:literal, $create:expr, $push:expr, $pop:expr, ::)+) => {
 
 use std::convert::TryFrom as _;
@@ -36,7 +36,9 @@ $(
     assert_eq!(dst, [40, 20, 30]);
 )+
 
-    let mut group_large = criterion.benchmark_group("large-with-chunks");
+    let mut group_large = criterion.benchmark_group("chunks-large");
+    group_large.plot_config(criterion::PlotConfiguration::default()
+        .summary_scale(criterion::AxisScale::Logarithmic));
 $(
     group_large.bench_function($id, |b| {
         b.iter_custom(|iters| {
@@ -139,7 +141,7 @@ $(
 
     let queue_sizes = [4096];
 
-    let mut group_write_chunk = criterion.benchmark_group("eager-write-chunk");
+    let mut group_write_chunk = criterion.benchmark_group("chunks-eager-write");
     group_write_chunk.plot_config(criterion::PlotConfiguration::default()
         .summary_scale(criterion::AxisScale::Logarithmic));
 $(
@@ -209,7 +211,7 @@ $(
 )+
     group_write_chunk.finish();
 
-    let mut group_read_chunk = criterion.benchmark_group("eager-read-chunk");
+    let mut group_read_chunk = criterion.benchmark_group("chunks-eager-read");
     group_read_chunk.plot_config(criterion::PlotConfiguration::default()
         .summary_scale(criterion::AxisScale::Logarithmic));
 $(
@@ -275,7 +277,7 @@ criterion_main!(benches);
 
 use std::io::{Read as _, Write as _};
 
-create_two_threads_with_chunks_benchmark!(
+create_chunks_benchmark!(
     "write-read",
     rtrb::RingBuffer::new,
     |p, s| match p.write(s) {
diff --git a/benches/two_threads_const.rs b/benches/const_push_pop.rs
similarity index 97%
rename from benches/two_threads_const.rs
rename to benches/const_push_pop.rs
index ed6b7bc0..9870dc30 100644
--- a/benches/two_threads_const.rs
+++ b/benches/const_push_pop.rs
@@ -1,4 +1,4 @@
-macro_rules! create_two_threads_const_benchmark {
+macro_rules! create_const_push_pop_benchmark {
     ($($id:literal, $create:tt, $push:expr, $pop:expr,::)+) => {
 
 use std::hint::black_box;
@@ -109,7 +109,7 @@ criterion_main!(benches);
     };
 }
 
-create_two_threads_const_benchmark! {
+create_const_push_pop_benchmark! {
     "rtrb",
     { ($N:expr) => {
         rtrb::RingBuffer::new($N)
diff --git a/benches/two_threads.rs b/benches/push_pop.rs
similarity index 97%
rename from benches/two_threads.rs
rename to benches/push_pop.rs
index 7856ed48..10aec1f5 100644
--- a/benches/two_threads.rs
+++ b/benches/push_pop.rs
@@ -1,4 +1,4 @@
-macro_rules! create_two_threads_benchmark {
+macro_rules! create_push_pop_benchmark {
     ($($id:literal, $create:expr, $push:expr, $pop:expr,::)+) => {
 
 use std::convert::TryInto as _;
@@ -32,6 +32,8 @@ $(
 )+
 
     let mut group_large = criterion.benchmark_group("large");
+    group_large.plot_config(criterion::PlotConfiguration::default()
+        .summary_scale(criterion::AxisScale::Logarithmic));
     group_large.throughput(criterion::Throughput::Bytes(1));
 $(
     group_large.bench_function($id, |b| {
@@ -118,6 +120,8 @@ $(
     group_large.finish();
 
     let mut group_small = criterion.benchmark_group("small");
+    group_small.plot_config(criterion::PlotConfiguration::default()
+        .summary_scale(criterion::AxisScale::Logarithmic));
     group_small.throughput(criterion::Throughput::Bytes(1));
 $(
     group_small.bench_function($id, |b| {
@@ -286,7 +290,7 @@ criterion_main!(benches);
     };
 }
 
-create_two_threads_benchmark!(
+create_push_pop_benchmark!(
     "rtrb",
     rtrb::RingBuffer::new,
     |p, i| p.push(i).is_ok(),
diff --git a/performance-comparison/Cargo.toml b/performance-comparison/Cargo.toml
index 84dc02de..c55b73dc 100644
--- a/performance-comparison/Cargo.toml
+++ b/performance-comparison/Cargo.toml
@@ -33,13 +33,13 @@ codegen-units = 1
 bench = false # Don't disturb criterion command line parsing
 
 [[bench]]
-name = "two_threads"
+name = "chunks"
 harness = false
 
 [[bench]]
-name = "two_threads_with_chunks"
+name = "const_push_pop"
 harness = false
 
 [[bench]]
-name = "two_threads_const"
+name = "push_pop"
 harness = false
diff --git a/performance-comparison/README.md b/performance-comparison/README.md
new file mode 100644
index 00000000..92040dcc
--- /dev/null
+++ b/performance-comparison/README.md
@@ -0,0 +1,22 @@
+Performance Comparison between `rtrb` and Other Crates
+======================================================
+
+Run all benchmarks (takes a long time):
+
+    cargo bench
+
+Run a single benchmark file (see `benches/` subdirectory):
+
+    cargo bench --bench push_pop
+
+Run all benchmarks that match a substring:
+
+    cargo bench --bench push_pop -- eager-push
+
+List all available benchmarks:
+
+    cargo bench -- --list
+
+Some results are available in
+[issue #39](https://github.com/mgeier/rtrb/issues/39),
+feel free to add your own results there!
diff --git a/performance-comparison/benches/two_threads_with_chunks.rs b/performance-comparison/benches/chunks.rs
similarity index 90%
rename from performance-comparison/benches/two_threads_with_chunks.rs
rename to performance-comparison/benches/chunks.rs
index a709e221..597fe103 100644
--- a/performance-comparison/benches/two_threads_with_chunks.rs
+++ b/performance-comparison/benches/chunks.rs
@@ -1,6 +1,6 @@
-#[path = "../../benches/two_threads_with_chunks.rs"]
+#[path = "../../benches/chunks.rs"]
 #[macro_use]
-mod two_threads_with_chunks;
+mod chunks;
 
 use core::num::NonZeroUsize;
 
@@ -10,7 +10,7 @@ use ringbuf::traits::Consumer as _;
 use ringbuf::traits::Producer as _;
 use ringbuf::traits::Split as _;
 
-create_two_threads_with_chunks_benchmark!(
+create_chunks_benchmark!(
     "1-gil",
     |capacity| gil::spsc::channel(NonZeroUsize::new(capacity).unwrap()),
     |p, s| {
diff --git a/performance-comparison/benches/two_threads_const.rs b/performance-comparison/benches/const_push_pop.rs
similarity index 83%
rename from performance-comparison/benches/two_threads_const.rs
rename to performance-comparison/benches/const_push_pop.rs
index 8eb212c2..cbc053f8 100644
--- a/performance-comparison/benches/two_threads_const.rs
+++ b/performance-comparison/benches/const_push_pop.rs
@@ -1,10 +1,10 @@
-#[path = "../../benches/two_threads_const.rs"]
+#[path = "../../benches/const_push_pop.rs"]
 #[macro_use]
-mod two_threads_const;
+mod const_push_pop;
 
 use ringbuf::traits::*;
 
-create_two_threads_const_benchmark!(
+create_const_push_pop_benchmark!(
     "rtrb",
     { ($N:expr) => {
         rtrb::RingBuffer::new($N)
diff --git a/performance-comparison/benches/two_threads.rs b/performance-comparison/benches/push_pop.rs
similarity index 93%
rename from performance-comparison/benches/two_threads.rs
rename to performance-comparison/benches/push_pop.rs
index 0349935f..6c8fa64b 100644
--- a/performance-comparison/benches/two_threads.rs
+++ b/performance-comparison/benches/push_pop.rs
@@ -1,6 +1,6 @@
-#[path = "../../benches/two_threads.rs"]
+#[path = "../../benches/push_pop.rs"]
 #[macro_use]
-mod two_threads;
+mod push_pop;
 
 use core::num::NonZeroUsize;
 
@@ -8,7 +8,7 @@ use ringbuf::traits::Consumer as _;
 use ringbuf::traits::Producer as _;
 use ringbuf::traits::Split as _;
 
-create_two_threads_benchmark!(
+create_push_pop_benchmark!(
     "0-rtrb",
     rtrb::RingBuffer::new,
     |p, i| p.push(i).is_ok(),