11 files changed, 223 insertions, 0 deletions
diff --git a/nist/make-package.sh b/nist/make-package.sh
index 124da4b..234532a 100755
--- a/nist/make-package.sh
+++ b/nist/make-package.sh
@@ -65,6 +65,7 @@ add-variant ()
 
     implementations=(
         ref
+        add_felicsref
         add_threshold
         add_tweakeyloop
     )
diff --git a/src/add_felicsref/cipher.c b/src/add_felicsref/cipher.c
new file mode 120000
index 0000000..a2ac6a3
--- /dev/null
+++ b/src/add_felicsref/cipher.c
@@ -0,0 +1 @@
+../ref/cipher.c
+\ No newline at end of file
diff --git a/src/add_felicsref/cipher.h b/src/add_felicsref/cipher.h
new file mode 120000
index 0000000..eab258b
--- /dev/null
+++ b/src/add_felicsref/cipher.h
@@ -0,0 +1 @@
+../ref/cipher.h
+\ No newline at end of file
diff --git a/src/add_felicsref/constants.h b/src/add_felicsref/constants.h
new file mode 120000
index 0000000..67df0f3
--- /dev/null
+++ b/src/add_felicsref/constants.h
@@ -0,0 +1 @@
+../ref/constants.h
+\ No newline at end of file
diff --git a/src/add_felicsref/implem.mk b/src/add_felicsref/implem.mk
new file mode 120000
index 0000000..eb789fb
--- /dev/null
+++ b/src/add_felicsref/implem.mk
@@ -0,0 +1 @@
+../ref/implem.mk
+\ No newline at end of file
diff --git a/src/add_felicsref/lilliput-ae-utils.h b/src/add_felicsref/lilliput-ae-utils.h
new file mode 120000
index 0000000..b46625b
--- /dev/null
+++ b/src/add_felicsref/lilliput-ae-utils.h
@@ -0,0 +1 @@
+../ref/lilliput-ae-utils.h
+\ No newline at end of file
diff --git a/src/add_felicsref/lilliput-ae.h b/src/add_felicsref/lilliput-ae.h
new file mode 120000
index 0000000..66c8314
--- /dev/null
+++ b/src/add_felicsref/lilliput-ae.h
@@ -0,0 +1 @@
+../ref/lilliput-ae.h
+\ No newline at end of file
diff --git a/src/add_felicsref/lilliput-i.c b/src/add_felicsref/lilliput-i.c
new file mode 120000
index 0000000..46688d4
--- /dev/null
+++ b/src/add_felicsref/lilliput-i.c
@@ -0,0 +1 @@
+../ref/lilliput-i.c
+\ No newline at end of file
diff --git a/src/add_felicsref/lilliput-ii.c b/src/add_felicsref/lilliput-ii.c
new file mode 120000
index 0000000..09abf10
--- /dev/null
+++ b/src/add_felicsref/lilliput-ii.c
@@ -0,0 +1 @@
+../ref/lilliput-ii.c
+\ No newline at end of file
diff --git a/src/add_felicsref/tweakey.c b/src/add_felicsref/tweakey.c
new file mode 100644
index 0000000..635c179
--- /dev/null
+++ b/src/add_felicsref/tweakey.c
@@ -0,0 +1,213 @@
+/*
+Implementation of the Lilliput-AE tweakable block cipher.
+
+Authors, hereby denoted as "the implementer":
+    Kévin Le Gouguec,
+    2019.
+
+For more information, feedback or questions, refer to our website:
+https://paclido.fr/lilliput-ae
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+
+---
+
+This file provides an implementation of Lilliput-TBC's tweakey schedule,
+similar to the reference implementation save for a few manual optimizations:
+
+- unused multiplication functions were removed using preprocessor
+  conditionals based on the number of lanes;
+
+- the loop over an array of function pointers was unrolled.
+
+These handmade optimizations have been found to significantly decrease code
+size and execution time on GCC versions used in the FELICS framework.
+
+This suggests that the compiler does not detect dead code nor does it
+recognize unrolling opportunities, despite the multiplication functions
+being static and thus limited in scope to the compilation unit.
+*/
+
+#include <stdint.h>
+#include <string.h>
+
+#include "constants.h"
+#include "tweakey.h"
+
+
+#define LANE_BITS  64
+#define LANE_BYTES (LANE_BITS/8)
+#define LANES_NB   (TWEAKEY_BYTES/LANE_BYTES)
+
+
+void tweakey_state_init(
+    uint8_t TK[TWEAKEY_BYTES],
+    const uint8_t key[KEY_BYTES],
+    const uint8_t tweak[TWEAK_BYTES]
+)
+{
+    memcpy(TK,             tweak, TWEAK_BYTES);
+    memcpy(TK+TWEAK_BYTES, key,   KEY_BYTES);
+}
+
+
+void tweakey_state_extract(
+    const uint8_t TK[TWEAKEY_BYTES],
+    uint8_t round_constant,
+    uint8_t round_tweakey[ROUND_TWEAKEY_BYTES]
+)
+{
+    memset(round_tweakey, 0, ROUND_TWEAKEY_BYTES);
+
+    for (size_t j=0; j<LANES_NB; j++)
+    {
+        const uint8_t *TKj = TK + j*LANE_BYTES;
+
+        for (size_t k=0; k<LANE_BYTES; k++)
+        {
+            round_tweakey[k] ^= TKj[k];
+        }
+    }
+
+    round_tweakey[0] ^= round_constant;
+}
+
+
+static void _multiply_M(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
+{
+    y[7] = x[6];
+    y[6] = x[5];
+    y[5] = x[5]<<3 ^ x[4];
+    y[4] = x[4]>>3 ^ x[3];
+    y[3] = x[2];
+    y[2] = x[6]<<2 ^ x[1];
+    y[1] = x[0];
+    y[0] = x[7];
+}
+
+static void _multiply_M2(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
+{
+    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 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];
+}
+
+#if LANES_NB >= 5
+static void _multiply_MR(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
+{
+    y[0] = x[1];
+    y[1] = x[2];
+    y[2] = x[3]    ^ x[4]>>3;
+    y[3] = x[4];
+    y[4] = x[5]    ^ x[6]<<3;
+    y[5] = x[3]<<2 ^ x[6];
+    y[6] = x[7];
+    y[7] = x[0];
+}
+
+#if LANES_NB >= 6
+static void _multiply_MR2(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
+{
+    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];
+}
+
+#if LANES_NB >= 7
+static void _multiply_MR3(const uint8_t x[LANE_BYTES], uint8_t y[LANE_BYTES])
+{
+    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];
+}
+#endif
+#endif
+#endif
+
+
+void tweakey_state_update(uint8_t TK[TWEAKEY_BYTES])
+{
+    /* Skip lane 0, as it is multiplied by the identity matrix. */
+
+    size_t j;
+    uint8_t *TKj;
+    uint8_t TKj_old[LANE_BYTES];
+
+    j = 1;
+    TKj = TK + j*LANE_BYTES;
+    memcpy(TKj_old, TKj, LANE_BYTES);
+    _multiply_M(TKj_old, TKj);
+
+    j = 2;
+    TKj = TK + j*LANE_BYTES;
+    memcpy(TKj_old, TKj, LANE_BYTES);
+    _multiply_M2(TKj_old, TKj);
+
+    j = 3;
+    TKj = TK + j*LANE_BYTES;
+    memcpy(TKj_old, TKj, LANE_BYTES);
+    _multiply_M3(TKj_old, TKj);
+
+#if LANES_NB >= 5
+    j = 4;
+    TKj = TK + j*LANE_BYTES;
+    memcpy(TKj_old, TKj, LANE_BYTES);
+    _multiply_MR(TKj_old, TKj);
+
+#if LANES_NB >= 6
+    j = 5;
+    TKj = TK + j*LANE_BYTES;
+    memcpy(TKj_old, TKj, LANE_BYTES);
+    _multiply_MR2(TKj_old, TKj);
+
+#if LANES_NB >= 7
+    j = 6;
+    TKj = TK + j*LANE_BYTES;
+    memcpy(TKj_old, TKj, LANE_BYTES);
+    _multiply_MR3(TKj_old, TKj);
+#endif
+#endif
+#endif
+}
diff --git a/src/add_felicsref/tweakey.h b/src/add_felicsref/tweakey.h
new file mode 120000
index 0000000..7f2415f
--- /dev/null
+++ b/src/add_felicsref/tweakey.h
@@ -0,0 +1 @@
+../ref/tweakey.h
+\ No newline at end of file