27 June 2017

  minor changes to the comments in timing.out, no code changes
---
24 June 2017

  compile timing.c with no optimize for closer comparison with PLC code
---
23 June 2017

  timing.c: added tests using 17 16-bit pieces
---
22 June 2017

  timing.c: changed compile-time options into run-time options;
  improved efficiency of the 24- and 32-bit routines

  upgraded the compiler and generated new timing results
---
21 June 2017

  timing.c: implemented 11 24-bit pieces without using 64-bit data.

  for timing results see timing.out
---
20 June 2017

  timing.c: added optional Comba product-scanning multiply order.
  It makes mul() slower, but makes mul32() and mul24() faster.

  fixed bug in timing.c cpu() CPU time measurement.
---
19 June 2017

  timing.c: added USE64 option .c to use unsigned long long for intermediate
  multiply results.  

  added test using 11 24-bit pieces with 64-bit intermediate results.
---
18 June 2017

  added timing.c to compare speed of multiply for 22 12-bit pieces, with
  reduce at the end, vs. 8 32-bit pieces, with reduction during the multiply,
  assuming no 64-bit data type
---
12 June 2017

  added signature verification: verify() in impl.c, tested at end of key.c

  removed unnecessary calls to adjust() at end of Point_mul()
---

11 June 2017

  fixed bug in Point_decode(): was setting result point t = 0, should be t = x*y
---

9 May 2017

  added print_P() to print a Point, from Ken Fischer
---

28 April 2017

  improvements and bug fixes to Point_decode()

  added ltm() to check if u < m, used by adjust() and Point_decode()

  added dec() to decrement a value, used by Point_decode()

  added decode.c and test_decode.sh to test Point_decode() with some large y values

  made default n=1024 in keytest.sh to run all test cases
---

26 April 2017

  added Point_decode() and neg()

  added constants: m1 = m-1, m58 = (m-5)/8, sqrtm1 = 2^((m-1)/4), d

  modified Point_encode() to optionally test Point_decode()

  modified key.c to test Point_decode()
---

2 January 2017

  added Point_double(), used by Point_mul() to perform p+p

  added sqr2(), used in Point_double()
---

June 2016

  initial version