Random: add yasmarang and cheap_rand generators.

README.md

@@ -32,7 +32,7 @@ and modules which are documented and supported.
   4.5 [Timed function](./README.md#45-timed-function) Time execution with a decorator.  
   4.6 [ESP8266 MQTT benchmark](./README.md#46-esp8266-mqtt-benchmark) Test performance of MQTT with official library.  
   4.7 [Rotary incremental encoder](./README.md#47-rotary-incremental-encoder) Fast, simple, proven algorithm.  
-  4.8 [A pseudo random number generator](./README.md#48-a-pseudo-random-number-generator)  
+  4.8 [Pseudo random number generators](./README.md#48-pseudo-random-number-generators)  
   4.9 [Verifying incrementing sequences](./README.md#49-verifying-incrementing-sequences) Test communications drivers.  
   4.10 [Bitmaps](./README.md#410-bitmaps) Non-allocating ways to access bitmaps.  
   4.11 [Functors and singletons](./README.md#411-functors-and-singletons) Useful decorators.  
@@ -213,7 +213,7 @@ The [encoder_portable.py](./encoders/encoder_portable.py) version should work on
 all MicroPython platforms. Tested on ESP8266. Note that interrupt latency on
 the ESP8266 limits performance. ESP32 has similar limitations.
 
-## 4.8 A pseudo random number generator
+## 4.8 Pseudo random number generators
 
 On the Pyboard V1.1, true random numbers may be generated rapidly with
 `pyb.rng()` which uses a hardware random number generator on the
@@ -226,7 +226,12 @@ number generator is seeded with an arbitrary initial value. On each call to the
 function it will return a random number, but (given the same seed) the sequence
 of numbers following initialisation will always be the same.
 
-See [random.py](./random/random.py) for usage and timing documentation.
+See [random.py](./random/random.py) for usage and timing documentation. The
+[yasmarang generator](./random/yasmarang.py) is also included, along with my
+own [cheap random](./random/cheap_rand.py). The latter constrains calculations
+to 30 bits, allowing its use in an ISR. It comes with no guarantees of random
+quality and the only statistical test is that the mean converges on the right
+value. 
 
 ## 4.9 Verifying incrementing sequences
 

random/cheap_rand.py

@@ -0,0 +1,37 @@
+# pseudorandom numbers for MicroPython. ISR friendly version.
+# Probably poor quality numbers but useful in test scripts
+
+# Author: Peter Hinch
+# Copyright Peter Hinch 2020 Released under the MIT license
+
+# Example usage to produce numbers between 0 and 99
+# rand = cheap_rand(100)
+# successive calls to rand() will produce the required result.
+
+def cheap_rand(modulo, seed=0x3fba2):
+    x = seed 
+    def func():
+        nonlocal x
+        x ^= (x & 0x1ffff) << 13;
+        x ^= x >> 17;
+        x ^= (x & 0x1ffffff) << 5;
+        return x % modulo
+    return func
+
+# The sum total of my statistical testing
+#import pyb, micropython, time
+#rand = cheap_rand(1000)
+#sum = 0
+#cnt = 0
+#def avg(n):
+    #global sum, cnt
+    #sum += n
+    #cnt += 1
+#def cb(t):
+    #n = rand()
+    #micropython.schedule(avg, n)
+
+#t = pyb.Timer(1, freq=20, callback=cb)
+#while True:
+    #time.sleep(1)
+    #print(sum/cnt)

random/yasmarang.py

@@ -0,0 +1,25 @@
+# yasmarang pseudorandom number generator.
+# Source http://www.literatecode.com/yasmarang
+
+# Author: Peter Hinch
+# Copyright Peter Hinch 2020 Released under the MIT license
+
+def yasmarang():
+    pad = 0xeda4baba
+    n = 69
+    d = 233
+    dat = 0
+    def func():
+        nonlocal pad, n, d, dat
+        pad = (pad + dat + d * n) & 0xffffffff
+        pad = ((pad<<3) + (pad>>29)) & 0xffffffff
+        n = pad | 2
+        d = (d ^ ((pad<<31) + (pad>>1))) & 0xffffffff
+        dat ^= ((pad & 0xff) ^ (d>>8) ^ 1) & 0xff
+        return (pad^(d<<5)^(pad>>18)^(dat<<1)) & 0xffffffff
+    return func
+
+# Test: produces same outcome as website.
+#ym = yasmarang()
+#for _ in range(20):
+    #print(hex(ym()))