Merge pull request #511 from negativeExponent/updates

[pcsx_rearmed.git] / deps / lzma-16.04 / C / AesOpt.c

/* AesOpt.c -- Intel's AES\r
2013-11-12 : Igor Pavlov : Public domain */\r
\r
#include "Precomp.h"\r
\r
#include "CpuArch.h"\r
\r
#ifdef MY_CPU_X86_OR_AMD64\r
#if _MSC_VER >= 1500\r
#define USE_INTEL_AES\r
#endif\r
#endif\r
\r
#ifdef USE_INTEL_AES\r
\r
#include <wmmintrin.h>\r
\r
void MY_FAST_CALL AesCbc_Encode_Intel(__m128i *p, __m128i *data, size_t numBlocks)\r
{\r
  __m128i m = *p;\r
  for (; numBlocks != 0; numBlocks--, data++)\r
  {\r
    UInt32 numRounds2 = *(const UInt32 *)(p + 1) - 1;\r
    const __m128i *w = p + 3;\r
    m = _mm_xor_si128(m, *data);\r
    m = _mm_xor_si128(m, p[2]);\r
    do\r
    {\r
      m = _mm_aesenc_si128(m, w[0]);\r
      m = _mm_aesenc_si128(m, w[1]);\r
      w += 2;\r
    }\r
    while (--numRounds2 != 0);\r
    m = _mm_aesenc_si128(m, w[0]);\r
    m = _mm_aesenclast_si128(m, w[1]);\r
    *data = m;\r
  }\r
  *p = m;\r
}\r
\r
#define NUM_WAYS 3\r
\r
#define AES_OP_W(op, n) { \\r
    const __m128i t = w[n]; \\r
    m0 = op(m0, t); \\r
    m1 = op(m1, t); \\r
    m2 = op(m2, t); \\r
    }\r
\r
#define AES_DEC(n) AES_OP_W(_mm_aesdec_si128, n)\r
#define AES_DEC_LAST(n) AES_OP_W(_mm_aesdeclast_si128, n)\r
#define AES_ENC(n) AES_OP_W(_mm_aesenc_si128, n)\r
#define AES_ENC_LAST(n) AES_OP_W(_mm_aesenclast_si128, n)\r
\r
void MY_FAST_CALL AesCbc_Decode_Intel(__m128i *p, __m128i *data, size_t numBlocks)\r
{\r
  __m128i iv = *p;\r
  for (; numBlocks >= NUM_WAYS; numBlocks -= NUM_WAYS, data += NUM_WAYS)\r
  {\r
    UInt32 numRounds2 = *(const UInt32 *)(p + 1);\r
    const __m128i *w = p + numRounds2 * 2;\r
    __m128i m0, m1, m2;\r
    {\r
      const __m128i t = w[2];\r
      m0 = _mm_xor_si128(t, data[0]);\r
      m1 = _mm_xor_si128(t, data[1]);\r
      m2 = _mm_xor_si128(t, data[2]);\r
    }\r
    numRounds2--;\r
    do\r
    {\r
      AES_DEC(1)\r
      AES_DEC(0)\r
      w -= 2;\r
    }\r
    while (--numRounds2 != 0);\r
    AES_DEC(1)\r
    AES_DEC_LAST(0)\r
\r
    {\r
      __m128i t;\r
      t = _mm_xor_si128(m0, iv); iv = data[0]; data[0] = t;\r
      t = _mm_xor_si128(m1, iv); iv = data[1]; data[1] = t;\r
      t = _mm_xor_si128(m2, iv); iv = data[2]; data[2] = t;\r
    }\r
  }\r
  for (; numBlocks != 0; numBlocks--, data++)\r
  {\r
    UInt32 numRounds2 = *(const UInt32 *)(p + 1);\r
    const __m128i *w = p + numRounds2 * 2;\r
    __m128i m = _mm_xor_si128(w[2], *data);\r
    numRounds2--;\r
    do\r
    {\r
      m = _mm_aesdec_si128(m, w[1]);\r
      m = _mm_aesdec_si128(m, w[0]);\r
      w -= 2;\r
    }\r
    while (--numRounds2 != 0);\r
    m = _mm_aesdec_si128(m, w[1]);\r
    m = _mm_aesdeclast_si128(m, w[0]);\r
\r
    m = _mm_xor_si128(m, iv);\r
    iv = *data;\r
    *data = m;\r
  }\r
  *p = iv;\r
}\r
\r
void MY_FAST_CALL AesCtr_Code_Intel(__m128i *p, __m128i *data, size_t numBlocks)\r
{\r
  __m128i ctr = *p;\r
  __m128i one;\r
  one.m128i_u64[0] = 1;\r
  one.m128i_u64[1] = 0;\r
  for (; numBlocks >= NUM_WAYS; numBlocks -= NUM_WAYS, data += NUM_WAYS)\r
  {\r
    UInt32 numRounds2 = *(const UInt32 *)(p + 1) - 1;\r
    const __m128i *w = p;\r
    __m128i m0, m1, m2;\r
    {\r
      const __m128i t = w[2];\r
      ctr = _mm_add_epi64(ctr, one); m0 = _mm_xor_si128(ctr, t);\r
      ctr = _mm_add_epi64(ctr, one); m1 = _mm_xor_si128(ctr, t);\r
      ctr = _mm_add_epi64(ctr, one); m2 = _mm_xor_si128(ctr, t);\r
    }\r
    w += 3;\r
    do\r
    {\r
      AES_ENC(0)\r
      AES_ENC(1)\r
      w += 2;\r
    }\r
    while (--numRounds2 != 0);\r
    AES_ENC(0)\r
    AES_ENC_LAST(1)\r
    data[0] = _mm_xor_si128(data[0], m0);\r
    data[1] = _mm_xor_si128(data[1], m1);\r
    data[2] = _mm_xor_si128(data[2], m2);\r
  }\r
  for (; numBlocks != 0; numBlocks--, data++)\r
  {\r
    UInt32 numRounds2 = *(const UInt32 *)(p + 1) - 1;\r
    const __m128i *w = p;\r
    __m128i m;\r
    ctr = _mm_add_epi64(ctr, one);\r
    m = _mm_xor_si128(ctr, p[2]);\r
    w += 3;\r
    do\r
    {\r
      m = _mm_aesenc_si128(m, w[0]);\r
      m = _mm_aesenc_si128(m, w[1]);\r
      w += 2;\r
    }\r
    while (--numRounds2 != 0);\r
    m = _mm_aesenc_si128(m, w[0]);\r
    m = _mm_aesenclast_si128(m, w[1]);\r
    *data = _mm_xor_si128(*data, m);\r
  }\r
  *p = ctr;\r
}\r
\r
#else\r
\r
void MY_FAST_CALL AesCbc_Encode(UInt32 *ivAes, Byte *data, size_t numBlocks);\r
void MY_FAST_CALL AesCbc_Decode(UInt32 *ivAes, Byte *data, size_t numBlocks);\r
void MY_FAST_CALL AesCtr_Code(UInt32 *ivAes, Byte *data, size_t numBlocks);\r
\r
void MY_FAST_CALL AesCbc_Encode_Intel(UInt32 *p, Byte *data, size_t numBlocks)\r
{\r
  AesCbc_Encode(p, data, numBlocks);\r
}\r
\r
void MY_FAST_CALL AesCbc_Decode_Intel(UInt32 *p, Byte *data, size_t numBlocks)\r
{\r
  AesCbc_Decode(p, data, numBlocks);\r
}\r
\r
void MY_FAST_CALL AesCtr_Code_Intel(UInt32 *p, Byte *data, size_t numBlocks)\r
{\r
  AesCtr_Code(p, data, numBlocks);\r
}\r
\r
#endif\r