diff --git a/readme.md b/readme.md
--- a/readme.md
+++ b/readme.md
@@ -30,9 +30,9 @@ Please report your issues to https://tra
 
 ## Performance ##
 
-As it is, the code is not very fast. Let's assume we have a secret of length _m_. For each byte, the splitting takes _n_ evaluations of a polynomial of order _k_ over Galois field 256, leading to _O(n\*k\*m)_ finite field multiplications. Reconstruction of the constant parameters during joining takes _O(k\*k + k\*m)_ multiplications.
+Being written in pure Python, the code is not especially fast. It is therefore recommended to split rather keys to encrypted files than the files themselves.
 
-Benchmark results, all values mean _seconds per byte_ of the secret length:
+Benchmark results, as obtained by running `shamira benchmark`. All values mean _seconds per byte_ of the secret length:
 <table>
     <tr>
         <th>k / n parameters</th>
@@ -41,24 +41,22 @@ Benchmark results, all values mean _seco
     </tr>
     <tr>
         <td>2 / 3 (a Raspberry Pi 3)</td>
-        <td>6.08e-05</td>
-        <td>0.000435</td>
+        <td>0.000142</td>
+        <td>0.000448</td>
     </tr>
     <tr>
         <td>2 / 3 (a laptop)</td>
-        <td>5.02e-06</td>
-        <td>4.12e-05</td>
+        <td>7.88e-06</td>
+        <td>4.28e-05</td>
     </tr>
     <tr>
         <td>254 / 254 (a Raspberry Pi 3)</td>
-        <td>0.226</td>
-        <td>0.0314</td>
+        <td>0.0268</td>
+        <td>0.0287</td>
     </tr>
     <tr>
         <td>254 / 254 (a laptop)</td>
-        <td>0.0125</td>
-        <td>0.00175</td>
+        <td>0.00183</td>
+        <td>0.00156</td>
     </tr>
 </table>
-
-While the speeds are not awful, for longer secrets I recommend encrypting them with a random key of your choice and splitting only the key. Anyway, you can run your own benchmark with `shamira benchmark`