Changement de l'implémentation de référence

Et suppression de l'implémentation  add_tweakeysequences, qui n'a plus
aucun intérêt (plus lente et plus grosse que les deux autres).

1 files changed, 46 insertions, 17 deletions

diff --git a/src/ref/tweakey.c b/src/ref/tweakey.c
index b1f349e..b2822b0 100644
--- a/src/ref/tweakey.c
+++ b/src/ref/tweakey.c
@@ -14,7 +14,7 @@ http://creativecommons.org/publicdomain/zero/1.0/
 
 This file provides an implementation of Lilliput-TBC's tweakey schedule,
 where multiplications by matrices M and M_R to the power n are performed
-by applying functions for M and M_R n times.
+by functions expressing the exponentiated matrices with shifts and XORs.
 */
 
 #include <stdint.h>
@@ -76,18 +76,34 @@ static void _multiply_M(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
 
 static void _multiply_M2(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
 {
-    uint8_t M_x[LANE_BYTES];
-    _multiply_M(x, M_x);
-    _multiply_M(M_x, y);
+    uint8_t x_M_5 = x[5]<<3 ^ x[4];
+    uint8_t x_M_4 = x[4]>>3 ^ x[3];
+
+    y[7] = x[5];
+    y[6] = x_M_5;
+    y[5] = x_M_5<<3 ^ x_M_4;
+    y[4] = x_M_4>>3 ^ x[2];
+    y[3] = x[6]<<2  ^ x[1];
+    y[2] = x[5]<<2  ^ x[0];
+    y[1] = x[7];
+    y[0] = x[6];
 }
 
 static void _multiply_M3(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
 {
-    uint8_t M_x[LANE_BYTES];
-    uint8_t M2_x[LANE_BYTES];
-    _multiply_M(x, M_x);
-    _multiply_M(M_x, M2_x);
-    _multiply_M(M2_x, y);
+    uint8_t x_M_5  = x[5]<<3  ^ x[4];
+    uint8_t x_M_4  = x[4]>>3  ^ x[3];
+    uint8_t x_M2_5 = x_M_5<<3 ^ x_M_4;
+    uint8_t x_M2_4 = x_M_4>>3 ^ x[2];
+
+    y[7] = x_M_5;
+    y[6] = x_M2_5;
+    y[5] = x_M2_5<<3 ^ x_M2_4;
+    y[4] = x_M2_4>>3 ^ x[6]<<2 ^ x[1];
+    y[3] = x[5]<<2   ^ x[0];
+    y[2] = x_M_5<<2  ^ x[7];
+    y[1] = x[6];
+    y[0] = x[5];
 }
 
 static void _multiply_MR(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
@@ -104,18 +120,31 @@ static void _multiply_MR(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
 
 static void _multiply_MR2(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
 {
-    uint8_t MR_x[LANE_BYTES];
-    _multiply_MR(x, MR_x);
-    _multiply_MR(MR_x, y);
+    uint8_t x_MR_4 = x[5] ^ x[6]<<3;
+
+    y[0] = x[2];
+    y[1] = x[3]    ^ x[4]>>3;
+    y[2] = x[4]    ^ x_MR_4>>3;
+    y[3] = x_MR_4;
+    y[4] = x[3]<<2 ^ x[6]      ^ x[7]<<3;
+    y[5] = x[4]<<2 ^ x[7];
+    y[6] = x[0];
+    y[7] = x[1];
 }
 
 static void _multiply_MR3(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
 {
-    uint8_t MR_x[LANE_BYTES];
-    uint8_t MR2_x[LANE_BYTES];
-    _multiply_MR(x, MR_x);
-    _multiply_MR(MR_x, MR2_x);
-    _multiply_MR(MR2_x, y);
+    uint8_t x_MR_4 = x[5]     ^ x[6]<<3;
+    uint8_t x_MR2_4 = x[3]<<2 ^ x[6]    ^ x[7]<<3;
+
+    y[0] = x[3]      ^ x[4]>>3;
+    y[1] = x[4]      ^ x_MR_4>>3;
+    y[2] = x_MR_4    ^ x_MR2_4>>3;
+    y[3] = x_MR2_4;
+    y[4] = x[0]<<3   ^ x[4]<<2 ^ x[7];
+    y[5] = x_MR_4<<2 ^ x[0];
+    y[6] = x[1];
+    y[7] = x[2];
 }
 
 typedef void (*matrix_multiplication)(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES]);