svn: ecrypt/trunk/benchmarks/aes-ctr/aes-256/hongjun/v0/aes256.h

* imported Hongjun's optimized C-code for AES-128 and AES-256.

/* aes256.h*/
/* Hongjun Wu, January 2007*/

#include "ecrypt-sync.h"
#include "aes-table.h"

#include <stdio.h>

#define first_round(ctx,x0,y0) { \
	u32 z0,t0,tem0;              \
	z0 = (x0) ^ ctx->round_key[0][0];        \
	t0 = (u8) z0;                \
	tem0 = T0[t0];               \
	(y0) = tem0 ^ ctx->first_round_output_x0; \
}

#define second_round(ctx,x0,y0,y1,y2,y3) { \
	u32 t0,t7,t10,t13; \
	u32 tem0,tem7,tem10,tem13;    \
	t0 = (u8)(x0);      \
	tem0 = T0[t0];      \
	(y0) = tem0 ^ ctx->second_round_output[0];   \
	t7 = (u8)((x0)>>24);     \
	tem7 = T3[t7];    \
	(y1) = tem7 ^ ctx->second_round_output[1];   \
	t10 = (u8)((x0)>>16);   \
	tem10 = T2[t10]; \
	(y2) = tem10 ^ ctx->second_round_output[2];  \
	t13 = (u8)((x0)>>8);    \
	tem13 = T1[t13];\
	(y3) = tem13 ^ ctx->second_round_output[3];  \
}

#define round(ctx,x0,x1,x2,x3,y0,y1,y2,y3,r) { \
	u32 t0,t1,t2,t3;    \
	u32 t4,t5,t6,t7;    \
	u32 t8,t9,t10,t11;  \
	u32 t12,t13,t14,t15;\
	u32 tem0,tem1,tem2,tem3;    \
	u32 tem4,tem5,tem6,tem7;    \
	u32 tem8,tem9,tem10,tem11;  \
	u32 tem12,tem13,tem14,tem15;\
	\
	t0 = (u8)(x0);      \
	tem0 = T0[t0];      \
	t1 = (u8)((x1)>>8); \
	tem1 = tem0 ^ T1[t1];    \
	t2 = (u8)((x2)>>16);     \
	tem2 = tem1 ^ T2[t2];    \
	t3 = (u8)((x3)>>24);     \
	tem3 = tem2 ^ T3[t3];    \
	(y0) = tem3 ^ ctx->round_key[r][0];   \
	\
	t4 = (u8)(x1);      \
	tem4 = T0[t4];      \
	t5 = (u8)((x2)>>8); \
	tem5 = tem4 ^ T1[t5];    \
	t6 = (u8)((x3)>>16);     \
	tem6 = tem5 ^ T2[t6];    \
	t7 = (u8)((x0)>>24);     \
	tem7 = tem6 ^ T3[t7];    \
	(y1) = tem7 ^ ctx->round_key[r][1];   \
	\
	t8 = (u8)(x2);          \
	tem8 = T0[t8];          \
	t9 = (u8)((x3)>>8);     \
	tem9 = tem8 ^ T1[t9];   \
	t10 = (u8)((x0)>>16);   \
	tem10 = tem9 ^ T2[t10]; \
	t11 = (u8)((x1)>>24);   \
	tem11 = tem10 ^ T3[t11];\
	(y2) = tem11 ^ ctx->round_key[r][2]; \
	\
	t12 = (u8)(x3);         \
	tem12 = T0[t12];        \
	t13 = (u8)((x0)>>8);    \
	tem13 = tem12 ^ T1[t13];\
	t14 = (u8)((x1)>>16);   \
	tem14 = tem13 ^ T2[t14];\
	t15 = (u8)((x2)>>24);   \
	tem15 = tem14 ^ T3[t15];\
	(y3) = tem15 ^ ctx->round_key[r][3]; \
}

/* 22.14 cycles/byte*/
#define last_round(ctx,x0,x1,x2,x3,output,r) { \
	u32 t0,t1,t2,t3;    \
	u32 t4,t5,t6,t7;    \
	u32 t8,t9,t10,t11;  \
	u32 t12,t13,t14,t15;\
	\
	t0 = (u8)(x0);        \
	output[0] = Sbox[t0]; \
      t7 = (u8)((x0)>>24);  \
	output[7] = Sbox[t7]; \
	t10 = (u8)((x0)>>16);   \
	output[10] = Sbox[t10]; \
	t13 = (u8)((x0)>>8);    \
	output[13] = Sbox[t13]; \
	\
	t1 = (u8)((x1)>>8);   \
	output[1] = Sbox[t1]; \
	t4 = (u8)(x1);        \
	output[4] = Sbox[t4]; \
	t11 = (u8)((x1)>>24);     \
	output[11] = Sbox[t11]; \
	t14 = (u8)((x1)>>16);     \
	output[14] = Sbox[t14]; \
	\
	t2 = (u8)((x2)>>16);  \
	output[2] = Sbox[t2]; \
	t5 = (u8)((x2)>>8);   \
	output[5] = Sbox[t5]; \
	t8 = (u8)(x2);            \
	output[8] = Sbox[t8]; \
	t15 = (u8)((x2)>>24);     \
	output[15] = Sbox[t15]; \
      \
	t3 = (u8)((x3)>>24);  \
	output[3] = Sbox[t3]; \
	t6 = (u8)((x3)>>16);  \
	output[6] = Sbox[t6]; \
	t9 = (u8)((x3)>>8);       \
	output[9] = Sbox[t9];     \
	t12 = (u8)(x3);           \
	output[12] = Sbox[t12];   \
}

/*
	((u32*)output)[0] ^= ctx->round_key[r][0]; \
	((u32*)output)[1] ^= ctx->round_key[r][1]; \
	((u32*)output)[2] ^= ctx->round_key[r][2]; \
	((u32*)output)[3] ^= ctx->round_key[r][3]; \
}
*/


void aes256_enc_block(u32* x,u8* output, ECRYPT_ctx* ctx) {
	u32 y0,y1,y2,y3;
	u32 z0,z1,z2,z3;
	u32 a0,a1,a2,a3;
	u32 b0,b1,b2,b3;
	first_round(ctx,x[0],y0);
	second_round(ctx,y0,z0,z1,z2,z3);
	round(ctx,z0,z1,z2,z3,a0,a1,a2,a3,3);
	round(ctx,a0,a1,a2,a3,b0,b1,b2,b3,4);
	round(ctx,b0,b1,b2,b3,y0,y1,y2,y3,5);
	round(ctx,y0,y1,y2,y3,z0,z1,z2,z3,6);
	round(ctx,z0,z1,z2,z3,a0,a1,a2,a3,7);
	round(ctx,a0,a1,a2,a3,b0,b1,b2,b3,8);
	round(ctx,b0,b1,b2,b3,y0,y1,y2,y3,9);
	round(ctx,y0,y1,y2,y3,z0,z1,z2,z3,10);
	round(ctx,z0,z1,z2,z3,a0,a1,a2,a3,11);
	round(ctx,a0,a1,a2,a3,b0,b1,b2,b3,12);
	round(ctx,b0,b1,b2,b3,y0,y1,y2,y3,13);
	last_round(ctx,y0,y1,y2,y3,output,14);
}


/*compute the intermediate values for the first two rounds*/
void partial_precompute_tworounds(ECRYPT_ctx* ctx)
{
  u32 x0,x1,x2,x3,y0,y1,y2,y3;

  x0 = ctx->counter[0] ^ ctx->round_key[0][0];
  x1 = ctx->counter[1] ^ ctx->round_key[0][1];
  x2 = ctx->counter[2] ^ ctx->round_key[0][2];
  x3 = ctx->counter[3] ^ ctx->round_key[0][3];
  x0 &= 0xffffff00;
  round(ctx,x0,x1,x2,x3,y0,y1,y2,y3,1);
  ctx->first_round_output_x0 = y0 ^ T0[0];
  y0 = 0;
  round(ctx,y0,y1,y2,y3,x0,x1,x2,x3,2);
  ctx->second_round_output[0] = x0 ^ T0[0];
  ctx->second_round_output[1] = x1 ^ T3[0];
  ctx->second_round_output[2] = x2 ^ T2[0];
  ctx->second_round_output[3] = x3 ^ T1[0];
}

eSTREAM Project Powered by ViewCVS 1.0-dev

ViewCVS and CVS Help