Skip to main content
Add updates.
Source Link
user53032
user53032

The program is correct, but it may have some room for improvement. ItPowerPC and Power 8 is not our area of expertise, so there could be a lot of room for improvement. The code was tested on GCC112, which is a little endian machine running Linux. It was also tested on GCC119, which is a big endian machine running AIX. Both machines provide GCC and IBM's XL C/C++.


We applied J.H.'s feedback and checked in the source file. However, we left the original question in place so J.H.'s answer did not lose context. The updated file can be found at fips197-p8.c.

The program is correct, but it may have some room for improvement. It was tested on GCC112, which is a little endian machine running Linux. It was also tested on GCC119, which is a big endian machine running AIX. Both machines provide GCC and IBM's XL C/C++.

The program is correct, but it may have some room for improvement. PowerPC and Power 8 is not our area of expertise, so there could be a lot of room for improvement. The code was tested on GCC112, which is a little endian machine running Linux. It was also tested on GCC119, which is a big endian machine running AIX. Both machines provide GCC and IBM's XL C/C++.


We applied J.H.'s feedback and checked in the source file. However, we left the original question in place so J.H.'s answer did not lose context. The updated file can be found at fips197-p8.c.

Source Link
user53032
user53032

Power 8 AES implementation using built-ins

I need a review of an AltiVec/Power 8 AES implementation using built-ins to access in-core crypto. The program below is an implementation of FIPS 197, Appendix B. The hard-coded values are due to the fixed key and message from the appendix.

The program is correct, but it may have some room for improvement. It was tested on GCC112, which is a little endian machine running Linux. It was also tested on GCC119, which is a big endian machine running AIX. Both machines provide GCC and IBM's XL C/C++.

I had to work around several XL C/C++ bugs. That's why oddities like this are present:

uint8x16_p8 Load8x16(const uint8_t src[16])
{
#if defined(__xlc__) || defined(__xlC__)
    /* IBM XL C/C++ compiler */
    uint8_t* s = (uint8_t*)src;
# if (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
    return vec_xl_be(0, s);
# else
    return vec_xl(0, s);
# endif
#else
    /* GCC, Clang, etc */
    return (uint8x16_p8)vec_vsx_ld(0, src);
#endif
}

The requirements are:

  1. GCC and IBM XL C/C++ compiler support
  2. Big-endian and little-endian support
  3. The input key can be unaligned. The user supplies it.
  4. The input message can be unaligned. The user supplies it.
  5. The output buffer can be unaligned. The user supplies it.
  6. The subkeys must be 16-byte aligned to allow use of vec_ld. The library controls them, so this is easy to comply with.

The program is available at Noloader/AES-Power8 github. It is reproduced below.

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdint.h>

#if defined(__ALTIVEC__)
# include <altivec.h>
#endif

#if defined(__xlc__) || defined(__xlC__)
// #include <builtins.h>
typedef vector unsigned char uint8x16_p8;
typedef vector unsigned int uint64x2_p8;
#elif defined(__GNUC__)
typedef vector unsigned char uint8x16_p8;
typedef vector unsigned long long uint64x2_p8;
#endif

/* Avoid 0 == 0 when __BYTE_ORDER__ is not defined */
#ifndef __BYTE_ORDER__
# define __BYTE_ORDER__ -16
#endif

uint8x16_p8 Load8x16(const uint8_t src[16])
{
#if defined(__xlc__) || defined(__xlC__)
   /* IBM XL C/C++ compiler */
   uint8_t* s = (uint8_t*)src;
# if (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
   return vec_xl_be(0, s);
# else
   return vec_xl(0, s);
# endif
#else
   /* GCC, Clang, etc */
   return (uint8x16_p8)vec_vsx_ld(0, src);
#endif
}

void Save8x16(const uint8x16_p8 src, uint8_t dest[16])
{
#if defined(__xlc__) || defined(__xlC__)
   /* IBM XL C/C++ compiler */
# if (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
   vec_xst_be(src, 0, dest);
# else
   vec_xst(src, 0, dest);
# endif
#else
   /* GCC, Clang, etc */
   vec_vsx_st(src, 0, dest);
#endif
}

uint64x2_p8 Load64x2(const uint8_t src[16])
{
#if defined(__xlc__) || defined(__xlC__)
   /* IBM XL C/C++ compiler */
   uint8_t* s = (uint8_t*)src;
# if (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
   return (uint64x2_p8)vec_xl_be(0, s);
# else
   return (uint64x2_p8)vec_xl(0, s);
# endif
#else
   /* GCC, Clang, etc */
# if (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
   __attribute__((aligned(16)))
   static const uint8_t m[16] =
      /* Looks a lot like an epi8 mask ... */
      { 15,14,13,12,  11,10,9,8,  7,6,5,4, 3,2,1,0 };
   __attribute__((aligned(16)))
   static const uint8_t z[16] =
      { 0,0,0,0,  0,0,0,0,  0,0,0,0,  0,0,0,0 };

   const uint8x16_p8 mask = vec_ld(0, m);
   const uint8x16_p8 zero = vec_ld(0, z);

   return (uint64x2_p8)vec_perm(Load8x16(src), zero, mask);
# else
   return (uint64x2_p8)vec_vsx_ld(0, src);
# endif
#endif
}

void Save64x2(const uint64x2_p8 src, uint8_t dest[16])
{
#if defined(__xlc__) || defined(__xlC__)
   /* IBM XL C/C++ compiler */
# if (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
   vec_xst_be((uint8x16_p8)src, 0, dest);
# else
   vec_xst((uint8x16_p8)src, 0, dest);
# endif
#else
   /* GCC, Clang, etc */
# if (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
   __attribute__((aligned(16)))
   static const uint8_t m[16] =
      /* Looks a lot like an epi8 mask ... */
      { 15,14,13,12,  11,10,9,8,  7,6,5,4, 3,2,1,0 };
   __attribute__((aligned(16)))
   static const uint8_t z[16] =
      { 0,0,0,0,  0,0,0,0,  0,0,0,0,  0,0,0,0 };

   const uint8x16_p8 mask = vec_ld(0, m);
   const uint8x16_p8 zero = vec_ld(0, z);

   uint8x16_p8 data = vec_perm((uint8x16_p8)src, zero, mask);
   vec_st(data, 0, dest);
# else
   vec_vsx_st((uint8x16_p8)src, 0, dest);
# endif
#endif
}

void Print64x2(const uint64x2_p8 block, const char* label)
{
   __attribute__((aligned(16)))
   uint8_t t[16];
   Save64x2(block, t);

   if (label && label[0])
      printf("%s: ", label);

   for (unsigned int i = 0; i<16; ++i) {
      printf("%02X ", t[i]);
      if (i == 7) printf(" ");
   }
   printf("\n");
}

void Print8x16(const uint8x16_p8 block, const char* label)
{
   __attribute__((aligned(16)))
   uint8_t t[16];
   Save8x16(block, t);

   if (label && label[0])
      printf("%s: ", label);

   for (unsigned int i = 0; i<16; ++i) {
      printf("%02X ", t[i]);
      if (i == 7) printf(" ");
   }
   printf("\n");
}

int main(int argc, char* argv[])
{
   /* Initial input, key and a few rounds for testing */
   __attribute__((aligned(16)))
   const uint8_t input[16] = {
      0x32, 0x43, 0xf6, 0xa8, 0x88, 0x5a, 0x30, 0x8d, 0x31, 0x31, 0x98, 0xa2, 0xe0, 0x37, 0x07, 0x34
   };
   __attribute__((aligned(16)))
   const uint8_t key[16] = {
      0x2b, 0x7e, 0x15, 0x16, 0x28, 0xae, 0xd2, 0xa6, 0xab, 0xf7, 0x15, 0x88, 0x9 , 0xcf, 0x4f, 0x3c
   };
   __attribute__((aligned(16)))
   const uint8_t subkeys[10][16] = {
      {0xA0, 0xFA, 0xFE, 0x17, 0x88, 0x54, 0x2c, 0xb1, 0x23, 0xa3, 0x39, 0x39, 0x2a, 0x6c, 0x76, 0x05},
      {0xF2, 0xC2, 0x95, 0xF2, 0x7a, 0x96, 0xb9, 0x43, 0x59, 0x35, 0x80, 0x7a, 0x73, 0x59, 0xf6, 0x7f},
      {0x3D, 0x80, 0x47, 0x7D, 0x47, 0x16, 0xFE, 0x3E, 0x1E, 0x23, 0x7E, 0x44, 0x6D, 0x7A, 0x88, 0x3B},
      {0xEF, 0x44, 0xA5, 0x41, 0xA8, 0x52, 0x5B, 0x7F, 0xB6, 0x71, 0x25, 0x3B, 0xDB, 0x0B, 0xAD, 0x00},
      {0xD4, 0xD1, 0xC6, 0xF8, 0x7C, 0x83, 0x9D, 0x87, 0xCA, 0xF2, 0xB8, 0xBC, 0x11, 0xF9, 0x15, 0xBC},
      {0x6D, 0x88, 0xA3, 0x7A, 0x11, 0x0B, 0x3E, 0xFD, 0xDB, 0xF9, 0x86, 0x41, 0xCA, 0x00, 0x93, 0xFD},
      {0x4E, 0x54, 0xF7, 0x0E, 0x5F, 0x5F, 0xC9, 0xF3, 0x84, 0xA6, 0x4F, 0xB2, 0x4E, 0xA6, 0xDC, 0x4F},
      {0xEA, 0xD2, 0x73, 0x21, 0xB5, 0x8D, 0xBA, 0xD2, 0x31, 0x2B, 0xF5, 0x60, 0x7F, 0x8D, 0x29, 0x2F},
      {0xAC, 0x77, 0x66, 0xF3, 0x19, 0xFA, 0xDC, 0x21, 0x28, 0xD1, 0x29, 0x41, 0x57, 0x5c, 0x00, 0x6E},
      {0xD0, 0x14, 0xF9, 0xA8, 0xC9, 0xEE, 0x25, 0x89, 0xE1, 0x3F, 0x0c, 0xC8, 0xB6, 0x63, 0x0C, 0xA6}
   };

   /* Result */
   __attribute__((aligned(16))) uint8_t result[16];

#if defined(__xlc__) || defined(__xlC__)

   uint8x16_p8 s = Load8x16(input);
   uint8x16_p8 k = Load8x16(key);

   s = vec_xor(s, k);

   Print8x16(s, "State 1");

   k = Load8x16(subkeys[0]);
   s = __vcipher(s, k);

   Print8x16(s, "State 2");

   k = Load8x16(subkeys[1]);
   s = __vcipher(s, k);

   Print8x16(s, "State 3");

   k = Load8x16(subkeys[2]);
   s = __vcipher(s, k);

   Print8x16(s, "State 4");

   k = Load8x16(subkeys[3]);
   s = __vcipher(s, k);

   Print8x16(s, "State 5");

   k = Load8x16(subkeys[4]);
   s = __vcipher(s, k);

   Print8x16(s, "State 6");

   k = Load8x16(subkeys[5]);
   s = __vcipher(s, k);

   Print8x16(s, "State 7");

   k = Load8x16(subkeys[6]);
   s = __vcipher(s, k);

   Print8x16(s, "State 8");

   k = Load8x16(subkeys[7]);
   s = __vcipher(s, k);

   Print8x16(s, "State 9");

   k = Load8x16(subkeys[8]);
   s = __vcipher(s, k);

   Print8x16(s, "State 10");

   k = Load8x16(subkeys[9]);
   s = __vcipherlast(s, k);

   Save8x16(s, result);

#elif defined(__GNUC__) 

   uint64x2_p8 s = Load64x2(input);
   uint64x2_p8 k = Load64x2(key);
   s = vec_xor(s, k);

   k = Load64x2(subkeys[0]);
   s = __builtin_crypto_vcipher(s, k);

   k = Load64x2(subkeys[1]);
   s = __builtin_crypto_vcipher(s, k);

   k = Load64x2(subkeys[2]);
   s = __builtin_crypto_vcipher(s, k);

   k = Load64x2(subkeys[3]);
   s = __builtin_crypto_vcipher(s, k);

   k = Load64x2(subkeys[4]);
   s = __builtin_crypto_vcipher(s, k);

   k = Load64x2(subkeys[5]);
   s = __builtin_crypto_vcipher(s, k);

   k = Load64x2(subkeys[6]);
   s = __builtin_crypto_vcipher(s, k);

   k = Load64x2(subkeys[7]);
   s = __builtin_crypto_vcipher(s, k);

   k = Load64x2(subkeys[8]);
   s = __builtin_crypto_vcipher(s, k);

   k = Load64x2(subkeys[9]);
   s = __builtin_crypto_vcipherlast(s, k);

   Save64x2(s, result);

#endif

   printf("Output: ");
   for (unsigned int i=0; i<16; ++i)
      printf("%02X ", result[i]);
   printf("\n");

   if (result[0] == 0x39 && result[1] == 0x25 && result[2] == 0x84 && result[3] == 0x1D)
      printf("SUCCESS!!!\n");
   else
      printf("FAILURE!!!\n");

   return 0;
}