/**

 * FreeRDP: A Remote Desktop Protocol Implementation

 * RemoteFX Codec Library - SSE2 Optimizations

 *

 * Copyright 2011 Stephen Erisman

 * Copyright 2011 Norbert Federa <norbert.federa@thincast.com>

 *

 * Licensed under the Apache License, Version 2.0 (the "License");

 * you may not use this file except in compliance with the License.

 * You may obtain a copy of the License at

 *

 *     http://www.apache.org/licenses/LICENSE-2.0

 *

 * Unless required by applicable law or agreed to in writing, software

 * distributed under the License is distributed on an "AS IS" BASIS,

 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

 * See the License for the specific language governing permissions and

 * limitations under the License.

 */

#include <winpr/assert.h>

#include <winpr/cast.h>

#include <winpr/platform.h>

#include <freerdp/config.h>

#include "../rfx_types.h"

#include "rfx_sse2.h"

#include "../rfx_quantization.h"

#include "../../core/simd.h"

#include "../../primitives/sse/prim_avxsse.h"

#if defined(SSE_AVX_INTRINSICS_ENABLED)

#include <stdio.h>

#include <stdlib.h>

#include <string.h>

#include <winpr/sysinfo.h>

#include <xmmintrin.h>

#include <emmintrin.h>

#ifdef _MSC_VER

#define __attribute__(...)

#endif

#define CACHE_LINE_BYTES 64

#ifndef __clang__

#define ATTRIBUTES __gnu_inline__, __always_inline__, __artificial__

#else

#define ATTRIBUTES __gnu_inline__, __always_inline__

#endif

static inline void __attribute__((ATTRIBUTES)) mm_prefetch_buffer(char* WINPR_RESTRICT buffer,

                                                                  size_t num_bytes)

{

  __m128i* buf = (__m128i*)buffer;

  for (size_t i = 0; i < (num_bytes / sizeof(__m128i)); i += (CACHE_LINE_BYTES / sizeof(__m128i)))

  {

    _mm_prefetch((char*)(&buf[i]), _MM_HINT_NTA);

  }

}

/* rfx_decode_ycbcr_to_rgb_sse2 code now resides in the primitives library. */

/* rfx_encode_rgb_to_ycbcr_sse2 code now resides in the primitives library. */

static inline void __attribute__((ATTRIBUTES))

rfx_quantization_decode_block_sse2(INT16* WINPR_RESTRICT buffer, const size_t buffer_size,

                                   const UINT32 factor)

{

  __m128i* ptr = (__m128i*)buffer;

  const __m128i* buf_end = (__m128i*)(buffer + buffer_size);

  if (factor == 0)

    return;

  do

  {

    const __m128i la = LOAD_SI128(ptr);

    const __m128i a = _mm_slli_epi16(la, WINPR_ASSERTING_INT_CAST(int, factor));

    STORE_SI128(ptr, a);

    ptr++;

  } while (ptr < buf_end);

}

WINPR_ATTR_NODISCARD

static BOOL rfx_quantization_decode_sse2(INT16* WINPR_RESTRICT buffer,

                                         const UINT32* WINPR_RESTRICT quantVals,

                                         size_t nrQuantValues)

{

  WINPR_ASSERT(buffer);

  WINPR_ASSERT(quantVals);

  WINPR_ASSERT(nrQuantValues == NR_QUANT_VALUES);

  for (size_t x = 0; x < nrQuantValues; x++)

  {

    const UINT32 val = quantVals[x];

    if (val < 1)

      return FALSE;

  }

  mm_prefetch_buffer((char*)buffer, 4096 * sizeof(INT16));

  rfx_quantization_decode_block_sse2(&buffer[0], 1024, quantVals[8] - 1);    /* HL1 */

  rfx_quantization_decode_block_sse2(&buffer[1024], 1024, quantVals[7] - 1); /* LH1 */

  rfx_quantization_decode_block_sse2(&buffer[2048], 1024, quantVals[9] - 1); /* HH1 */

  rfx_quantization_decode_block_sse2(&buffer[3072], 256, quantVals[5] - 1);  /* HL2 */

  rfx_quantization_decode_block_sse2(&buffer[3328], 256, quantVals[4] - 1);  /* LH2 */

  rfx_quantization_decode_block_sse2(&buffer[3584], 256, quantVals[6] - 1);  /* HH2 */

  rfx_quantization_decode_block_sse2(&buffer[3840], 64, quantVals[2] - 1);   /* HL3 */

  rfx_quantization_decode_block_sse2(&buffer[3904], 64, quantVals[1] - 1);   /* LH3 */

  rfx_quantization_decode_block_sse2(&buffer[3968], 64, quantVals[3] - 1);   /* HH3 */

  rfx_quantization_decode_block_sse2(&buffer[4032], 64, quantVals[0] - 1);   /* LL3 */

  return TRUE;

}

static inline void __attribute__((ATTRIBUTES))

rfx_quantization_encode_block_sse2(INT16* WINPR_RESTRICT buffer, const unsigned buffer_size,

                                   const INT16 factor)

{

  __m128i* ptr = (__m128i*)buffer;

  const __m128i* buf_end = (const __m128i*)(buffer + buffer_size);

  if (factor == 0)

    return;

  const __m128i half = _mm_set1_epi16(WINPR_ASSERTING_INT_CAST(INT16, 1 << (factor - 1)));

  do

  {

    const __m128i la = LOAD_SI128(ptr);

    __m128i a = _mm_add_epi16(la, half);

    a = _mm_srai_epi16(a, factor);

    STORE_SI128(ptr, a);

    ptr++;

  } while (ptr < buf_end);

}

WINPR_ATTR_NODISCARD

static BOOL rfx_quantization_encode_sse2(INT16* WINPR_RESTRICT buffer,

                                         const UINT32* WINPR_RESTRICT quantization_values,

                                         size_t quantVals)

{

  WINPR_ASSERT(buffer);

  WINPR_ASSERT(quantization_values);

  WINPR_ASSERT(quantVals == NR_QUANT_VALUES);

  for (size_t x = 0; x < quantVals; x++)

  {

    const UINT32 val = quantization_values[x];

    if (val < 6)

      return FALSE;

    if (val > INT16_MAX)

      return FALSE;

  }

  mm_prefetch_buffer((char*)buffer, 4096 * sizeof(INT16));

  rfx_quantization_encode_block_sse2(

      buffer, 1024, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[8] - 6)); /* HL1 */

  rfx_quantization_encode_block_sse2(

      buffer + 1024, 1024, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[7] - 6)); /* LH1 */

  rfx_quantization_encode_block_sse2(

      buffer + 2048, 1024, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[9] - 6)); /* HH1 */

  rfx_quantization_encode_block_sse2(

      buffer + 3072, 256, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[5] - 6)); /* HL2 */

  rfx_quantization_encode_block_sse2(

      buffer + 3328, 256, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[4] - 6)); /* LH2 */

  rfx_quantization_encode_block_sse2(

      buffer + 3584, 256, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[6] - 6)); /* HH2 */

  rfx_quantization_encode_block_sse2(

      buffer + 3840, 64, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[2] - 6)); /* HL3 */

  rfx_quantization_encode_block_sse2(

      buffer + 3904, 64, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[1] - 6)); /* LH3 */

  rfx_quantization_encode_block_sse2(

      buffer + 3968, 64, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[3] - 6)); /* HH3 */

  rfx_quantization_encode_block_sse2(

      buffer + 4032, 64, WINPR_ASSERTING_INT_CAST(INT16, quantization_values[0] - 6)); /* LL3 */

  rfx_quantization_encode_block_sse2(buffer, 4096, 5);

  return TRUE;

}

static inline void __attribute__((ATTRIBUTES))

rfx_dwt_2d_decode_block_horiz_sse2(INT16* WINPR_RESTRICT l, INT16* WINPR_RESTRICT h,

                                   INT16* WINPR_RESTRICT dst, size_t subband_width)

{

  INT16* l_ptr = l;

  INT16* h_ptr = h;

  INT16* dst_ptr = dst;

  int first = 0;

  int last = 0;

  __m128i dst1;

  __m128i dst2;

  for (size_t y = 0; y < subband_width; y++)

  {

    /* Even coefficients */

    for (size_t n = 0; n < subband_width; n += 8)

    {

      /* dst[2n] = l[n] - ((h[n-1] + h[n] + 1) >> 1); */

      __m128i l_n = LOAD_SI128(l_ptr);

      __m128i h_n = LOAD_SI128(h_ptr);

      __m128i h_n_m = LOAD_SI128(h_ptr - 1);

      if (n == 0)

      {

        first = _mm_extract_epi16(h_n_m, 1);

        h_n_m = _mm_insert_epi16(h_n_m, first, 0);

      }

      __m128i tmp_n = _mm_add_epi16(h_n, h_n_m);

      tmp_n = _mm_add_epi16(tmp_n, _mm_set1_epi16(1));

      tmp_n = _mm_srai_epi16(tmp_n, 1);

      const __m128i dst_n = _mm_sub_epi16(l_n, tmp_n);

      STORE_SI128(l_ptr, dst_n);

      l_ptr += 8;

      h_ptr += 8;

    }

    l_ptr -= subband_width;

    h_ptr -= subband_width;

    /* Odd coefficients */

    for (size_t n = 0; n < subband_width; n += 8)

    {

      /* dst[2n + 1] = (h[n] << 1) + ((dst[2n] + dst[2n + 2]) >> 1); */

      __m128i h_n = LOAD_SI128(h_ptr);

      h_n = _mm_slli_epi16(h_n, 1);

      __m128i dst_n = LOAD_SI128(l_ptr);

      __m128i dst_n_p = LOAD_SI128(l_ptr + 1);

      if (n == subband_width - 8)

      {

        last = _mm_extract_epi16(dst_n_p, 6);

        dst_n_p = _mm_insert_epi16(dst_n_p, last, 7);

      }

      __m128i tmp_n = _mm_add_epi16(dst_n_p, dst_n);

      tmp_n = _mm_srai_epi16(tmp_n, 1);

      tmp_n = _mm_add_epi16(tmp_n, h_n);

      dst1 = _mm_unpacklo_epi16(dst_n, tmp_n);

      dst2 = _mm_unpackhi_epi16(dst_n, tmp_n);

      STORE_SI128(dst_ptr, dst1);

      STORE_SI128(dst_ptr + 8, dst2);

      l_ptr += 8;

      h_ptr += 8;

      dst_ptr += 16;

    }

  }

}

static inline void __attribute__((ATTRIBUTES))

rfx_dwt_2d_decode_block_vert_sse2(INT16* WINPR_RESTRICT l, INT16* WINPR_RESTRICT h,

                                  INT16* WINPR_RESTRICT dst, size_t subband_width)

{

  INT16* l_ptr = l;

  INT16* h_ptr = h;

  INT16* dst_ptr = dst;

  const size_t total_width = subband_width + subband_width;

  /* Even coefficients */

  for (size_t n = 0; n < subband_width; n++)

  {

    for (size_t x = 0; x < total_width; x += 8)

    {

      /* dst[2n] = l[n] - ((h[n-1] + h[n] + 1) >> 1); */

      const __m128i l_n = LOAD_SI128(l_ptr);

      const __m128i h_n = LOAD_SI128(h_ptr);

      __m128i tmp_n = _mm_add_epi16(h_n, _mm_set1_epi16(1));

      if (n == 0)

        tmp_n = _mm_add_epi16(tmp_n, h_n);

      else

      {

        const __m128i h_n_m = LOAD_SI128(h_ptr - total_width);

        tmp_n = _mm_add_epi16(tmp_n, h_n_m);

      }

      tmp_n = _mm_srai_epi16(tmp_n, 1);

      const __m128i dst_n = _mm_sub_epi16(l_n, tmp_n);

      STORE_SI128(dst_ptr, dst_n);

      l_ptr += 8;

      h_ptr += 8;

      dst_ptr += 8;

    }

    dst_ptr += total_width;

  }

  h_ptr = h;

  dst_ptr = dst + total_width;

  /* Odd coefficients */

  for (size_t n = 0; n < subband_width; n++)

  {

    for (size_t x = 0; x < total_width; x += 8)

    {

      /* dst[2n + 1] = (h[n] << 1) + ((dst[2n] + dst[2n + 2]) >> 1); */

      __m128i h_n = LOAD_SI128(h_ptr);

      __m128i dst_n_m = LOAD_SI128(dst_ptr - total_width);

      h_n = _mm_slli_epi16(h_n, 1);

      __m128i tmp_n = dst_n_m;

      if (n == subband_width - 1)

        tmp_n = _mm_add_epi16(tmp_n, dst_n_m);

      else

      {

        const __m128i dst_n_p = LOAD_SI128(dst_ptr + total_width);

        tmp_n = _mm_add_epi16(tmp_n, dst_n_p);

      }

      tmp_n = _mm_srai_epi16(tmp_n, 1);

      const __m128i dst_n = _mm_add_epi16(tmp_n, h_n);

      STORE_SI128(dst_ptr, dst_n);

      h_ptr += 8;

      dst_ptr += 8;

    }

    dst_ptr += total_width;

  }

}

static inline void __attribute__((ATTRIBUTES))

rfx_dwt_2d_decode_block_sse2(INT16* WINPR_RESTRICT buffer, INT16* WINPR_RESTRICT idwt,

                             size_t subband_width)

{

  mm_prefetch_buffer((char*)idwt, 4ULL * subband_width * sizeof(INT16));

  /* Inverse DWT in horizontal direction, results in 2 sub-bands in L, H order in tmp buffer idwt.

   */

  /* The 4 sub-bands are stored in HL(0), LH(1), HH(2), LL(3) order. */

  /* The lower part L uses LL(3) and HL(0). */

  /* The higher part H uses LH(1) and HH(2). */

  INT16* ll = buffer + 3ULL * subband_width * subband_width;

  INT16* hl = buffer;

  INT16* l_dst = idwt;

  rfx_dwt_2d_decode_block_horiz_sse2(ll, hl, l_dst, subband_width);

  INT16* lh = buffer + 1ULL * subband_width * subband_width;

  INT16* hh = buffer + 2ULL * subband_width * subband_width;

  INT16* h_dst = idwt + 2ULL * subband_width * subband_width;

  rfx_dwt_2d_decode_block_horiz_sse2(lh, hh, h_dst, subband_width);

  /* Inverse DWT in vertical direction, results are stored in original buffer. */

  rfx_dwt_2d_decode_block_vert_sse2(l_dst, h_dst, buffer, subband_width);

}

static void rfx_dwt_2d_decode_sse2(INT16* WINPR_RESTRICT buffer, INT16* WINPR_RESTRICT dwt_buffer)

{

  WINPR_ASSERT(buffer);

  WINPR_ASSERT(dwt_buffer);

  mm_prefetch_buffer((char*)buffer, 4096 * sizeof(INT16));

  rfx_dwt_2d_decode_block_sse2(&buffer[3840], dwt_buffer, 8);

  rfx_dwt_2d_decode_block_sse2(&buffer[3072], dwt_buffer, 16);

  rfx_dwt_2d_decode_block_sse2(&buffer[0], dwt_buffer, 32);

}

static inline void __attribute__((ATTRIBUTES))

rfx_dwt_2d_encode_block_vert_sse2(INT16* WINPR_RESTRICT src, INT16* WINPR_RESTRICT l,

                                  INT16* WINPR_RESTRICT h, size_t subband_width)

{

  const size_t total_width = subband_width << 1;

  for (size_t n = 0; n < subband_width; n++)

  {

    for (size_t x = 0; x < total_width; x += 8)

    {

      __m128i src_2n = LOAD_SI128(src);

      __m128i src_2n_1 = LOAD_SI128(src + total_width);

      __m128i src_2n_2 = src_2n;

      if (n < subband_width - 1)

        src_2n_2 = LOAD_SI128(src + 2ULL * total_width);

      /* h[n] = (src[2n + 1] - ((src[2n] + src[2n + 2]) >> 1)) >> 1 */

      __m128i h_n = _mm_add_epi16(src_2n, src_2n_2);

      h_n = _mm_srai_epi16(h_n, 1);

      h_n = _mm_sub_epi16(src_2n_1, h_n);

      h_n = _mm_srai_epi16(h_n, 1);

      STORE_SI128(h, h_n);

      __m128i h_n_m = h_n;

      if (n != 0)

        h_n_m = LOAD_SI128(h - total_width);

      /* l[n] = src[2n] + ((h[n - 1] + h[n]) >> 1) */

      __m128i l_n = _mm_add_epi16(h_n_m, h_n);

      l_n = _mm_srai_epi16(l_n, 1);

      l_n = _mm_add_epi16(l_n, src_2n);

      STORE_SI128(l, l_n);

      src += 8;

      l += 8;

      h += 8;

    }

    src += total_width;

  }

}

static inline void __attribute__((ATTRIBUTES))

rfx_dwt_2d_encode_block_horiz_sse2(INT16* WINPR_RESTRICT src, INT16* WINPR_RESTRICT l,

                                   INT16* WINPR_RESTRICT h, size_t subband_width)

{

  for (size_t y = 0; y < subband_width; y++)

  {

    for (size_t n = 0; n < subband_width; n += 8)

    {

      /* The following 3 Set operations consumes more than half of the total DWT processing

       * time! */

      const INT16 src16 = (INT16)(((n + 8) == subband_width) ? src[14] : src[16]);

      __m128i src_2n =

          _mm_set_epi16(src[14], src[12], src[10], src[8], src[6], src[4], src[2], src[0]);

      __m128i src_2n_1 =

          _mm_set_epi16(src[15], src[13], src[11], src[9], src[7], src[5], src[3], src[1]);

      __m128i src_2n_2 =

          _mm_set_epi16(src16, src[14], src[12], src[10], src[8], src[6], src[4], src[2]);

      /* h[n] = (src[2n + 1] - ((src[2n] + src[2n + 2]) >> 1)) >> 1 */

      __m128i h_n = _mm_add_epi16(src_2n, src_2n_2);

      h_n = _mm_srai_epi16(h_n, 1);

      h_n = _mm_sub_epi16(src_2n_1, h_n);

      h_n = _mm_srai_epi16(h_n, 1);

      STORE_SI128(h, h_n);

      __m128i h_n_m = LOAD_SI128(h - 1);

      if (n == 0)

      {

        int first = _mm_extract_epi16(h_n_m, 1);

        h_n_m = _mm_insert_epi16(h_n_m, first, 0);

      }

      /* l[n] = src[2n] + ((h[n - 1] + h[n]) >> 1) */

      __m128i l_n = _mm_add_epi16(h_n_m, h_n);

      l_n = _mm_srai_epi16(l_n, 1);

      l_n = _mm_add_epi16(l_n, src_2n);

      STORE_SI128(l, l_n);

      src += 16;

      l += 8;

      h += 8;

    }

  }

}

static inline void __attribute__((ATTRIBUTES))

rfx_dwt_2d_encode_block_sse2(INT16* WINPR_RESTRICT buffer, INT16* WINPR_RESTRICT dwt,

                             size_t subband_width)

{

  mm_prefetch_buffer((char*)dwt, 4ULL * subband_width * sizeof(INT16));

  /* DWT in vertical direction, results in 2 sub-bands in L, H order in tmp buffer dwt. */

  INT16* l_src = dwt;

  INT16* h_src = dwt + 2ULL * subband_width * subband_width;

  rfx_dwt_2d_encode_block_vert_sse2(buffer, l_src, h_src, subband_width);

  /* DWT in horizontal direction, results in 4 sub-bands in HL(0), LH(1), HH(2), LL(3) order,

   * stored in original buffer. */

  /* The lower part L generates LL(3) and HL(0). */

  /* The higher part H generates LH(1) and HH(2). */

  INT16* ll = buffer + 3ULL * subband_width * subband_width;

  INT16* hl = buffer;

  INT16* lh = buffer + 1ULL * subband_width * subband_width;

  INT16* hh = buffer + 2ULL * subband_width * subband_width;

  rfx_dwt_2d_encode_block_horiz_sse2(l_src, ll, hl, subband_width);

  rfx_dwt_2d_encode_block_horiz_sse2(h_src, lh, hh, subband_width);

}

static void rfx_dwt_2d_encode_sse2(INT16* WINPR_RESTRICT buffer, INT16* WINPR_RESTRICT dwt_buffer)

{

  WINPR_ASSERT(buffer);

  WINPR_ASSERT(dwt_buffer);

  mm_prefetch_buffer((char*)buffer, 4096 * sizeof(INT16));

  rfx_dwt_2d_encode_block_sse2(buffer, dwt_buffer, 32);

  rfx_dwt_2d_encode_block_sse2(buffer + 3072, dwt_buffer, 16);

  rfx_dwt_2d_encode_block_sse2(buffer + 3840, dwt_buffer, 8);

}

#endif

void rfx_init_sse2_int(RFX_CONTEXT* WINPR_RESTRICT context)

{

#if defined(SSE_AVX_INTRINSICS_ENABLED)

  WLog_VRB(PRIM_TAG, "SSE2/SSE3 optimizations");

  PROFILER_RENAME(context->priv->prof_rfx_quantization_decode, "rfx_quantization_decode_sse2")

  PROFILER_RENAME(context->priv->prof_rfx_quantization_encode, "rfx_quantization_encode_sse2")

  PROFILER_RENAME(context->priv->prof_rfx_dwt_2d_decode, "rfx_dwt_2d_decode_sse2")

  PROFILER_RENAME(context->priv->prof_rfx_dwt_2d_encode, "rfx_dwt_2d_encode_sse2")

  context->quantization_decode = rfx_quantization_decode_sse2;

  context->quantization_encode = rfx_quantization_encode_sse2;

  context->dwt_2d_decode = rfx_dwt_2d_decode_sse2;

  context->dwt_2d_encode = rfx_dwt_2d_encode_sse2;

#else

  WINPR_UNUSED(context);

  WLog_VRB(PRIM_TAG, "undefined WITH_SIMD or SSE2 intrinsics not available");

#endif

}