crypto: camellia - simplify key setup and CAMELLIA_ROUNDSM macro
authorJussi Kivilinna <jussi.kivilinna@mbnet.fi>
Sat, 3 Mar 2012 11:59:00 +0000 (13:59 +0200)
committerHerbert Xu <herbert@gondor.apana.org.au>
Wed, 14 Mar 2012 09:25:54 +0000 (17:25 +0800)
camellia_setup_tail() applies 'inverse of the last half of P-function' to
subkeys, which is unneeded if keys are applied directly to yl/yr in
CAMELLIA_ROUNDSM.

Patch speeds up key setup and should speed up CAMELLIA_ROUNDSM as applying
key to yl/yr early has less register dependencies.

Quick tcrypt camellia results:
 x86_64, AMD Phenom II, ~5% faster
 x86_64, Intel Core 2, ~0.5% faster
 i386, Intel Atom N270, ~1% faster

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
crypto/camellia.c

index 64cff46ea5e4fcd60bc74c5f7516666d4b4cba66..7ae4bcddd1de336dfbabf8c1ad18c7e113d2e7c9 100644 (file)
@@ -382,7 +382,6 @@ static void camellia_setup_tail(u32 *subkey, u32 *subL, u32 *subR, int max)
 {
        u32 dw, tl, tr;
        u32 kw4l, kw4r;
-       int i;
 
        /* absorb kw2 to other subkeys */
        /* round 2 */
@@ -557,24 +556,6 @@ static void camellia_setup_tail(u32 *subkey, u32 *subL, u32 *subR, int max)
                SUBKEY_L(32) = subL[32] ^ subL[31]; /* kw3 */
                SUBKEY_R(32) = subR[32] ^ subR[31];
        }
-
-       /* apply the inverse of the last half of P-function */
-       i = 2;
-       do {
-               dw = SUBKEY_L(i + 0) ^ SUBKEY_R(i + 0); dw = rol32(dw, 8);/* round 1 */
-               SUBKEY_R(i + 0) = SUBKEY_L(i + 0) ^ dw; SUBKEY_L(i + 0) = dw;
-               dw = SUBKEY_L(i + 1) ^ SUBKEY_R(i + 1); dw = rol32(dw, 8);/* round 2 */
-               SUBKEY_R(i + 1) = SUBKEY_L(i + 1) ^ dw; SUBKEY_L(i + 1) = dw;
-               dw = SUBKEY_L(i + 2) ^ SUBKEY_R(i + 2); dw = rol32(dw, 8);/* round 3 */
-               SUBKEY_R(i + 2) = SUBKEY_L(i + 2) ^ dw; SUBKEY_L(i + 2) = dw;
-               dw = SUBKEY_L(i + 3) ^ SUBKEY_R(i + 3); dw = rol32(dw, 8);/* round 4 */
-               SUBKEY_R(i + 3) = SUBKEY_L(i + 3) ^ dw; SUBKEY_L(i + 3) = dw;
-               dw = SUBKEY_L(i + 4) ^ SUBKEY_R(i + 4); dw = rol32(dw, 8);/* round 5 */
-               SUBKEY_R(i + 4) = SUBKEY_L(i + 4) ^ dw; SUBKEY_L(i + 4) = dw;
-               dw = SUBKEY_L(i + 5) ^ SUBKEY_R(i + 5); dw = rol32(dw, 8);/* round 6 */
-               SUBKEY_R(i + 5) = SUBKEY_L(i + 5) ^ dw; SUBKEY_L(i + 5) = dw;
-               i += 8;
-       } while (i < max);
 }
 
 static void camellia_setup128(const unsigned char *key, u32 *subkey)
@@ -869,6 +850,8 @@ static void camellia_setup192(const unsigned char *key, u32 *subkey)
 
 #define CAMELLIA_ROUNDSM(xl, xr, kl, kr, yl, yr, il, ir)               \
     do {                                                               \
+       yl ^= kl;                                                       \
+       yr ^= kr;                                                       \
        ir =  camellia_sp1110[(u8)xr];                                  \
        il =  camellia_sp1110[    (xl >> 24)];                          \
        ir ^= camellia_sp0222[    (xr >> 24)];                          \
@@ -877,8 +860,7 @@ static void camellia_setup192(const unsigned char *key, u32 *subkey)
        il ^= camellia_sp3033[(u8)(xl >> 8)];                           \
        ir ^= camellia_sp4404[(u8)(xr >> 8)];                           \
        il ^= camellia_sp4404[(u8)xl];                                  \
-       il ^= kl;                                                       \
-       ir ^= il ^ kr;                                                  \
+       ir ^= il;                                                       \
        yl ^= ir;                                                       \
        yr ^= ror32(il, 8) ^ ir;                                                \
     } while (0)