Modify ECC points to do all math in Montgomery form, rather than

converting back and forth. This gives a 10 to 20% speedup on a Core
i7. In addition, the CurveGFp no longer contains a Barrett reducer,
saving 3 BigInts worth of memory.

Add a #if'ed out alternative to point multiplication using the
Montgomery ladder technique. It runs in (more or less) constant time,
but rather significantly slower than the 4 bit window technique
currently used.

Tweak the window sizes to match the theoretical optimums.

1 files changed, 6 insertions, 1 deletions

diff --git a/src/math/numbertheory/pow_mod.cpp b/src/math/numbertheory/pow_mod.cpp
index a66a1f7df..bf6b29275 100644
--- a/src/math/numbertheory/pow_mod.cpp
+++ b/src/math/numbertheory/pow_mod.cpp
@@ -118,7 +118,12 @@ size_t Power_Mod::window_bits(size_t exp_bits, size_t,
                               Power_Mod::Usage_Hints hints)
    {
    static const size_t wsize[][2] = {
-      { 2048, 7 }, { 1024, 6 }, { 256, 5 }, { 128, 4 }, { 64, 3 }, { 0, 0 }
+      { 1434, 7 },
+      {  539, 6 },
+      {  197, 4 },
+      {   70, 3 },
+      {   25, 2 },
+      {    0, 0 }
    };
 
    size_t window_bits = 1;