[FFmpeg-cvslog] libavutil: optimize camellia cipher
Supraja Meedinti
git at videolan.org
Mon Feb 9 22:40:45 CET 2015
ffmpeg | branch: master | Supraja Meedinti <supraja0493 at gmail.com> | Tue Feb 10 00:34:55 2015 +0530| [9a18247ec0cfec156a1006cabe81f99c0df803dd] | committer: Michael Niedermayer
libavutil: optimize camellia cipher
Before the changes :
lavu CAMELLIA size: 1048576 runs: 1024 time: 32.541 +- 0.044
After the changes:
lavu CAMELLIA size: 1048576 runs: 1024 time: 24.589 +- 0.066
Tested with crypto_bench on a Linux x86_64 OS with Intel Core i5-3210M CPU.
Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=9a18247ec0cfec156a1006cabe81f99c0df803dd
---
libavutil/camellia.c | 37 +++++++++++++++++++++++++++----------
1 file changed, 27 insertions(+), 10 deletions(-)
diff --git a/libavutil/camellia.c b/libavutil/camellia.c
index 7382414..11e57a1 100644
--- a/libavutil/camellia.c
+++ b/libavutil/camellia.c
@@ -37,6 +37,8 @@
#define Sigma5 0x10E527FADE682D1D
#define Sigma6 0xB05688C2B3E6C1FD
+static uint64_t SP[8][256];
+
typedef struct AVCAMELLIA {
uint64_t Kw[4];
uint64_t Ke[6];
@@ -140,17 +142,11 @@ static void LR128(uint64_t d[2], const uint64_t K[2], int x)
static uint64_t F(uint64_t F_IN, uint64_t KE)
{
- uint32_t Zl, Zr;
+ uint8_t y[8];
KE ^= F_IN;
- Zl = KE >> 32;
- Zr = KE & MASK32;
- Zl = (((uint32_t)SBOX1[(Zl >> 24)] << 24) | ((uint32_t)SBOX2[(Zl >> 16) & MASK8] << 16) | ((uint32_t)SBOX3[(Zl >> 8) & MASK8] << 8) |(SBOX4[Zl & MASK8]));
- Zr = (((uint32_t)SBOX2[(Zr >> 24)] << 24) | ((uint32_t)SBOX3[(Zr >> 16) & MASK8] << 16) | ((uint32_t)SBOX4[(Zr >> 8) & MASK8] << 8) |(SBOX1[Zr & MASK8]));
- Zl ^= LR32(Zr, 8);
- Zr ^= LR32(Zl, 16);
- Zl ^= RR32(Zr, 8);
- Zr ^= RR32(Zl, 8);
- return ((uint64_t)Zr << 32) | (uint64_t)Zl;
+ AV_WB64(y,KE);
+ KE=SP[0][y[0]]^SP[1][y[1]]^SP[2][y[2]]^SP[3][y[3]]^SP[4][y[4]]^SP[5][y[5]]^SP[6][y[6]]^SP[7][y[7]];
+ return KE;
}
static uint64_t FL(uint64_t FL_IN, uint64_t KE)
@@ -330,6 +326,26 @@ static void camellia_decrypt(AVCAMELLIA *cs, uint8_t *dst, const uint8_t *src, u
AV_WB64(dst + 8, D1);
}
+static void computeSP(void)
+{
+ uint64_t z;
+ int i;
+ for (i = 0; i < 256; i++) {
+ z = (uint64_t)SBOX1[i];
+ SP[0][i] = (z << 56) ^ (z << 48) ^ (z << 40) ^ (z << 24) ^ z;
+ SP[7][i] = (z << 56) ^ (z << 48) ^ (z << 40) ^ (z << 24) ^ (z << 16) ^ (z << 8);
+ z = (uint64_t)SBOX2[i];
+ SP[1][i] = (z << 48) ^ (z << 40) ^ (z << 32) ^ (z << 24) ^ (z << 16);
+ SP[4][i] = (z << 48) ^ (z << 40) ^ (z << 32) ^ (z << 16) ^ (z << 8) ^ z;
+ z = (uint64_t)SBOX3[i];
+ SP[2][i] = (z << 56) ^ (z << 40) ^ (z << 32) ^ (z << 16) ^ (z << 8);
+ SP[5][i] = (z << 56) ^ (z << 40) ^ (z << 32) ^ (z << 24) ^ (z << 8) ^ z;
+ z = (uint64_t)SBOX4[i];
+ SP[3][i] = (z << 56) ^ (z << 48) ^ (z << 32) ^ (z << 8) ^ z;
+ SP[6][i] = (z << 56) ^ (z << 48) ^ (z << 32) ^ (z << 24) ^ (z << 16) ^ z;
+ }
+}
+
struct AVCAMELLIA *av_camellia_alloc(void)
{
return av_mallocz(sizeof(struct AVCAMELLIA));
@@ -353,6 +369,7 @@ av_cold int av_camellia_init(AVCAMELLIA *cs, const uint8_t *key, int key_bits)
Kr[0] = AV_RB64(key + 16);
Kr[1] = AV_RB64(key + 24);
}
+ computeSP();
D1 = Kl[0] ^ Kr[0];
D2 = Kl[1] ^ Kr[1];
D2 ^= F(D1, Sigma1);
More information about the ffmpeg-cvslog
mailing list