Skip to content

Commit

Permalink
Optimize echo, whirlpool. Faster x16/x17/xevan
Browse files Browse the repository at this point in the history
  • Loading branch information
@KL0nLutiy
KL0nLutiy committed May 9, 2018
1 parent 3775915 commit 98195d7
Show file tree
Hide file tree
Showing 4 changed files with 102 additions and 250 deletions.
173 changes: 45 additions & 128 deletions kernel/wolf-echo.cl
View file
Original file line number Diff line number Diff line change
Expand Up @@ -19,97 +19,92 @@ uint4 Echo_AES_Round(const __local uint *AES0, const __local uint *AES1, const _
{
uint4 Y;
Y.s0 = AES0[(BYTE0(X.s0))] ^ AES1[BYTE1(X.s1)] ^ AES2[BYTE2(X.s2)] ^ AES3[BYTE3(X.s3)];
Y.s1 = AES0[(BYTE0(X.s1))] ^ AES1[BYTE1(X.s2)] ^ AES2[BYTE2(X.s3)] ^ AES3[BYTE3(X.s0)];
Y.s2 = AES0[(BYTE0(X.s2))] ^ AES1[BYTE1(X.s3)] ^ AES2[BYTE2(X.s0)] ^ AES3[BYTE3(X.s1)];
Y.s3 = AES0[(BYTE0(X.s3))] ^ AES1[BYTE1(X.s0)] ^ AES2[BYTE2(X.s1)] ^ AES3[BYTE3(X.s2)];
return(Y);
Y.s1 = AES0[(BYTE0(X.s1))] ^ AES1[BYTE1(X.s2)] ^ AES2[BYTE2(X.s3)] ^ AES3[BYTE3(X.s0)];
Y.s2 = AES0[(BYTE0(X.s2))] ^ AES1[BYTE1(X.s3)] ^ AES2[BYTE2(X.s0)] ^ AES3[BYTE3(X.s1)];
Y.s3 = AES0[(BYTE0(X.s3))] ^ AES1[BYTE1(X.s0)] ^ AES2[BYTE2(X.s1)] ^ AES3[BYTE3(X.s2)];
return(Y);
}

uint4 Echo_AES_Round_Small(const __local uint *AES0, const uint4 X)
{
uint4 Y;
Y.s0 = AES0[(BYTE0(X.s0))] ^ rotate(AES0[BYTE1(X.s1)], 8U) ^ rotate(AES0[BYTE2(X.s2)], 16U) ^ rotate(AES0[BYTE3(X.s3)], 24U);
Y.s1 = AES0[(BYTE0(X.s1))] ^ rotate(AES0[BYTE1(X.s2)], 8U) ^ rotate(AES0[BYTE2(X.s3)], 16U) ^ rotate(AES0[BYTE3(X.s0)], 24U);
Y.s2 = AES0[(BYTE0(X.s2))] ^ rotate(AES0[BYTE1(X.s3)], 8U) ^ rotate(AES0[BYTE2(X.s0)], 16U) ^ rotate(AES0[BYTE3(X.s1)], 24U);
Y.s3 = AES0[(BYTE0(X.s3))] ^ rotate(AES0[BYTE1(X.s0)], 8U) ^ rotate(AES0[BYTE2(X.s1)], 16U) ^ rotate(AES0[BYTE3(X.s2)], 24U);
return(Y);
Y.s1 = AES0[(BYTE0(X.s1))] ^ rotate(AES0[BYTE1(X.s2)], 8U) ^ rotate(AES0[BYTE2(X.s3)], 16U) ^ rotate(AES0[BYTE3(X.s0)], 24U);
Y.s2 = AES0[(BYTE0(X.s2))] ^ rotate(AES0[BYTE1(X.s3)], 8U) ^ rotate(AES0[BYTE2(X.s0)], 16U) ^ rotate(AES0[BYTE3(X.s1)], 24U);
Y.s3 = AES0[(BYTE0(X.s3))] ^ rotate(AES0[BYTE1(X.s0)], 8U) ^ rotate(AES0[BYTE2(X.s1)], 16U) ^ rotate(AES0[BYTE3(X.s2)], 24U);
return(Y);
}

void BigSubBytesSmall(const __local uint *restrict AES0, uint4 *restrict W, uchar rnd)
void BigSubBytesSmall(const __local uint *restrict AES0, uint4 *restrict W, uint k0)
{
#pragma unroll
for(int x = 0; x < 16; ++x)
{
uint4 tmp;
tmp = Echo_AES_Round_Small(AES0, W[x]);
tmp.s0 ^= (rnd << 4) + x + 512;
tmp.s0 ^= k0 | x | 0x200;
W[x] = Echo_AES_Round_Small(AES0, tmp);
}
}

void BigSubBytesSmall80(const __local uint *restrict AES0, uint4 *restrict W, uchar rnd)
void BigSubBytesSmall80(const __local uint *restrict AES0, uint4 *restrict W, uint k0)
{
#pragma unroll
for(int x = 0; x < 16; ++x)
{
uint4 tmp;
tmp = Echo_AES_Round_Small(AES0, W[x]);
tmp.s0 ^= (rnd << 4) + x + 640;
tmp.s0 ^= (k0 | x) + 0x280;
W[x] = Echo_AES_Round_Small(AES0, tmp);
}
}

//#if defined(WOLF_X11_CL)
#if 0
void BigSubBytes(const __local uint *restrict AES0, const __local uint *restrict AES1, const __local uint *restrict AES2, const __local uint *restrict AES3, uint *restrict W, uchar rnd)
#else
void BigSubBytes(const __local uint *restrict AES0, const __local uint *restrict AES1, const __local uint *restrict AES2, const __local uint *restrict AES3, uint4 *restrict W, uchar rnd)
#endif
void BigSubBytes0(const __local uint *restrict AES0, const __local uint *restrict AES1, const __local uint *restrict AES2, const __local uint *restrict AES3, uint4 *restrict W, uint k0)
{
//#if defined(WOLF_X11_CL)
#if 0

#if defined(WOLF_X11_CL)
#pragma unroll 8
#else
#pragma unroll 4
#endif
#pragma unroll
for(int x = 0; x < 16; ++x)
{
const uint idx = x << 2;
uint tmp[4];

#if defined(WOLF_X11_CL)
#pragma unroll 2
#else
#pragma unroll
#endif
for(int i = 0; i < 4; ++i)
tmp[i] = AES0[BYTE0(W[idx + i])] ^ AES1[BYTE1(W[idx + ((i + 1) & 3)])] ^ AES2[BYTE2(W[idx + ((i + 2) & 3)])] ^ AES3[BYTE3(W[idx + ((i + 3) & 3)])];

tmp[0] ^= (rnd << 4) + x + 512;

#if defined(WOLF_X11_CL)
#pragma unroll 2
#else
#pragma unroll
#endif
for(int i = 0; i < 4; ++i)
W[idx + i] = AES0[BYTE0(tmp[i])] ^ AES1[BYTE1(tmp[(i + 1) & 3])] ^ AES2[BYTE2(tmp[(i + 2) & 3])] ^ AES3[BYTE3(tmp[(i + 3) & 3])];
uint4 tmp;
tmp = Echo_AES_Round(AES0, AES1, AES2, AES3, W[x]);
tmp.s0 ^= (k0 | x);
W[x] = Echo_AES_Round(AES0, AES1, AES2, AES3, tmp);
}
}

#else
void BigSubBytes(const __local uint *restrict AES0, const __local uint *restrict AES1, const __local uint *restrict AES2, const __local uint *restrict AES3, uint4 *restrict W, uint k0)
{
#pragma unroll
for(int x = 0; x < 16; ++x)
{
uint4 tmp;
tmp = Echo_AES_Round(AES0, AES1, AES2, AES3, W[x]);
tmp.s0 ^= k0 | x | 0x200;
W[x] = Echo_AES_Round(AES0, AES1, AES2, AES3, tmp);
}
}

void BigSubBytes80(const __local uint *restrict AES0, const __local uint *restrict AES1, const __local uint *restrict AES2, const __local uint *restrict AES3, uint4 *restrict W, uint k0)
{
#pragma unroll
for(int x = 0; x < 16; ++x)
{
uint4 tmp;
tmp = Echo_AES_Round(AES0, AES1, AES2, AES3, W[x]);
tmp.s0 ^= (rnd << 4) + x + 512;
tmp.s0 ^= (k0 | x) + 0x280;
W[x] = Echo_AES_Round(AES0, AES1, AES2, AES3, tmp);
}
}

#endif
void BigSubBytes128(const __local uint *restrict AES0, const __local uint *restrict AES1, const __local uint *restrict AES2, const __local uint *restrict AES3, uint4 *restrict W, uint k0)
{
#pragma unroll
for(int x = 0; x < 16; ++x)
{
uint4 tmp;
tmp = Echo_AES_Round(AES0, AES1, AES2, AES3, W[x]);
tmp.s0 ^= (k0 | x) + 0x400;
W[x] = Echo_AES_Round(AES0, AES1, AES2, AES3, tmp);
}
}

void BigShiftRows(uint4 *WV)
Expand All @@ -134,56 +129,12 @@ void BigShiftRows(uint4 *WV)
WV[7] = tmp;
}

uint NewIdx(uint absidx, uint shiftamt)
{
return((absidx + (shiftamt << 2)) & 15);
}

//#ifdef WOLF_X11_CL
#if 0
void BigMixColumns(uint *W)
#else
void BigMixColumns(uint4 *WV)
#endif
{
//#ifdef WOLF_X11_CL
#if 0
for(int y = 0; y < 64; y += 16)
{
#pragma unroll
for(int x = 0; x < 4; ++x)
{
const uint a = W[y + x];
const uint b = W[y + x + 4];
const uint c = W[y + x + 8];
const uint d = W[y + x + 12];

const uint ab = a ^ b;
const uint bc = b ^ c;
const uint cd = c ^ d;

const uint t1 = ab & 0x80808080U;
const uint t2 = bc & 0x80808080U;
const uint t3 = cd & 0x80808080U;

const uint abx = ((t1 >> 7) * 27) ^ ((ab ^ t1) << 1);
const uint bcx = ((t2 >> 7) * 27) ^ ((bc ^ t2) << 1);
const uint cdx = ((t3 >> 7) * 27) ^ ((cd ^ t3) << 1);

W[y + x] = abx ^ bc ^ d;
W[y + x + 4] = bcx ^ a ^ cd;
W[y + x + 8] = cdx ^ ab ^ d;
W[y + x + 12] = abx ^ bcx ^ cdx ^ ab ^ c;
}
}

#else

#pragma unroll
for(int x = 0; x < 16; x += 4)
{
const uint4 a = WV[x], b = WV[x + 1], c = WV[x + 2], d = WV[x + 3];
//const uint4 a = WV[NewIdx(x + 0, 0)], b = WV[NewIdx(x + 1, 1)], c = WV[NewIdx(x + 2, 2)], d = WV[NewIdx(x + 3, 3)];

const uint4 ab = a ^ b;
const uint4 bc = b ^ c;
Expand All @@ -202,40 +153,6 @@ void BigMixColumns(uint4 *WV)
WV[x + 2] = cdx ^ ab ^ d;
WV[x + 3] = abx ^ bcx ^ cdx ^ ab ^ c;
}

/*uint4 a[4], b[4], c[4], d[4];
#pragma unroll
for(int x = 0; x < 16; x += 4)
{
a[x >> 2] = WV[NewIdx(x + 0, 0)];
b[x >> 2] = WV[NewIdx(x + 1, 1)];
c[x >> 2] = WV[NewIdx(x + 2, 2)];
d[x >> 2] = WV[NewIdx(x + 3, 3)];
}
#pragma unroll
for(int x = 0; x < 16; x += 4)
{
const uint4 ab = a[x >> 2] ^ b[x >> 2];
const uint4 bc = b[x >> 2] ^ c[x >> 2];
const uint4 cd = c[x >> 2] ^ d[x >> 2];
const uint4 t1 = ab & 0x80808080U;
const uint4 t2 = bc & 0x80808080U;
const uint4 t3 = cd & 0x80808080U;
const uint4 abx = ((t1 >> 7) * 27) ^ ((ab ^ t1) << 1);
const uint4 bcx = ((t2 >> 7) * 27) ^ ((bc ^ t2) << 1);
const uint4 cdx = ((t3 >> 7) * 27) ^ ((cd ^ t3) << 1);
WV[x] = abx ^ bc ^ d[x >> 2];
WV[x + 1] = bcx ^ a[x >> 2] ^ cd;
WV[x + 2] = cdx ^ ab ^ d[x >> 2];
WV[x + 3] = abx ^ bcx ^ cdx ^ ab ^ c[x >> 2];
}*/

#endif
}

#endif
#endif
Loading

0 comments on commit 98195d7

Please sign in to comment.