sse2.c

Go to the documentation of this file.

 
#if SIMD_SUPPORT_SSE2
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdint.h>
 
#include <emmintrin.h>
 
#include <ascii-chat/video/simd/sse2.h>
#include <ascii-chat/video/simd/ascii_simd.h>
#include <ascii-chat/common.h>
#include <ascii-chat/output_buffer.h>
#include <ascii-chat/util/overflow.h>
 
//=============================================================================
// Image-based API (matches NEON architecture)
//=============================================================================
 
// Simple monochrome ASCII function (matches scalar image_print performance)
char *render_ascii_image_monochrome_sse2(const image_t *image, const char *ascii_chars) {
  if (!image || !image->pixels || !ascii_chars) {
    return NULL;
  }
 
  const int h = image->h;
  const int w = image->w;
 
  if (h <= 0 || w <= 0) {
    return NULL;
  }
 
  // Get cached UTF-8 character mappings
  utf8_palette_cache_t *utf8_cache = get_utf8_palette_cache(ascii_chars);
  if (!utf8_cache) {
    log_error("Failed to get UTF-8 palette cache");
    return NULL;
  }
 
  // Buffer size for UTF-8 characters
  const size_t max_char_bytes = 4;
  const size_t len = (size_t)h * ((size_t)w * max_char_bytes + 1);
 
  char *output;
  output = SAFE_MALLOC(len, char *);
 
  char *pos = output;
  const rgb_pixel_t *pixels = (const rgb_pixel_t *)image->pixels;
 
  // Pure SSE2 processing - matches NEON approach
  for (int y = 0; y < h; y++) {
    const rgb_pixel_t *row = &pixels[y * w];
    int x = 0;
 
    // Process 16 pixels at a time with SSE2 (full 128-bit register capacity)
    for (; x + 15 < w; x += 16) {
      // Manual deinterleave RGB components (SSE2 limitation vs NEON's vld3q_u8)
      uint8_t r_array[16], g_array[16], b_array[16];
      for (int j = 0; j < 16; j++) {
        r_array[j] = row[x + j].r;
        g_array[j] = row[x + j].g;
        b_array[j] = row[x + j].b;
      }
 
      // Load full 16 bytes into SSE2 registers (process in two 8-pixel batches)
      __m128i r_vec_lo = _mm_loadl_epi64((__m128i *)(r_array + 0)); // First 8 pixels
      __m128i r_vec_hi = _mm_loadl_epi64((__m128i *)(r_array + 8)); // Second 8 pixels
      __m128i g_vec_lo = _mm_loadl_epi64((__m128i *)(g_array + 0));
      __m128i g_vec_hi = _mm_loadl_epi64((__m128i *)(g_array + 8));
      __m128i b_vec_lo = _mm_loadl_epi64((__m128i *)(b_array + 0));
      __m128i b_vec_hi = _mm_loadl_epi64((__m128i *)(b_array + 8));
 
      // Process first 8 pixels
      __m128i r_16_lo = _mm_unpacklo_epi8(r_vec_lo, _mm_setzero_si128());
      __m128i g_16_lo = _mm_unpacklo_epi8(g_vec_lo, _mm_setzero_si128());
      __m128i b_16_lo = _mm_unpacklo_epi8(b_vec_lo, _mm_setzero_si128());
 
      __m128i luma_r_lo = _mm_mullo_epi16(r_16_lo, _mm_set1_epi16(77));
      __m128i luma_g_lo = _mm_mullo_epi16(g_16_lo, _mm_set1_epi16(150));
      __m128i luma_b_lo = _mm_mullo_epi16(b_16_lo, _mm_set1_epi16(29));
 
      __m128i luma_sum_lo = _mm_add_epi16(luma_r_lo, luma_g_lo);
      luma_sum_lo = _mm_add_epi16(luma_sum_lo, luma_b_lo);
      luma_sum_lo = _mm_add_epi16(luma_sum_lo, _mm_set1_epi16(128));
      luma_sum_lo = _mm_srli_epi16(luma_sum_lo, 8);
 
      // Process second 8 pixels
      __m128i r_16_hi = _mm_unpacklo_epi8(r_vec_hi, _mm_setzero_si128());
      __m128i g_16_hi = _mm_unpacklo_epi8(g_vec_hi, _mm_setzero_si128());
      __m128i b_16_hi = _mm_unpacklo_epi8(b_vec_hi, _mm_setzero_si128());
 
      __m128i luma_r_hi = _mm_mullo_epi16(r_16_hi, _mm_set1_epi16(77));
      __m128i luma_g_hi = _mm_mullo_epi16(g_16_hi, _mm_set1_epi16(150));
      __m128i luma_b_hi = _mm_mullo_epi16(b_16_hi, _mm_set1_epi16(29));
 
      __m128i luma_sum_hi = _mm_add_epi16(luma_r_hi, luma_g_hi);
      luma_sum_hi = _mm_add_epi16(luma_sum_hi, luma_b_hi);
      luma_sum_hi = _mm_add_epi16(luma_sum_hi, _mm_set1_epi16(128));
      luma_sum_hi = _mm_srli_epi16(luma_sum_hi, 8);
 
      // Pack both halves to 8-bit
      __m128i luminance_lo = _mm_packus_epi16(luma_sum_lo, _mm_setzero_si128());
      __m128i luminance_hi = _mm_packus_epi16(luma_sum_hi, _mm_setzero_si128());
 
      // Store and convert to ASCII characters
      uint8_t luma_array[16];
      _mm_storel_epi64((__m128i *)(luma_array + 0), luminance_lo);
      _mm_storel_epi64((__m128i *)(luma_array + 8), luminance_hi);
 
      // Convert luminance to UTF-8 characters using optimized mappings
      for (int j = 0; j < 16; j++) {
        const utf8_char_t *char_info = &utf8_cache->cache[luma_array[j]];
        // Optimized: Use direct assignment for single-byte ASCII characters
        if (char_info->byte_len == 1) {
          *pos++ = char_info->utf8_bytes[0];
        } else {
          // Fallback to full memcpy for multi-byte UTF-8
          memcpy(pos, char_info->utf8_bytes, char_info->byte_len);
          pos += char_info->byte_len;
        }
      }
    }
 
    // Handle remaining pixels with optimized scalar code
    for (; x < w; x++) {
      const rgb_pixel_t pixel = row[x];
      const int luminance = (LUMA_RED * pixel.r + LUMA_GREEN * pixel.g + LUMA_BLUE * pixel.b + LUMA_THRESHOLD) >> 8;
      const utf8_char_t *char_info = &utf8_cache->cache[luminance];
      // Optimized: Use direct assignment for single-byte ASCII characters
      if (char_info->byte_len == 1) {
        *pos++ = char_info->utf8_bytes[0];
      } else {
        // Fallback to full memcpy for multi-byte UTF-8
        memcpy(pos, char_info->utf8_bytes, char_info->byte_len);
        pos += char_info->byte_len;
      }
    }
 
    // Add clear-to-end-of-line and newline (except last row)
    *pos++ = '\033';
    *pos++ = '[';
    *pos++ = 'K';
    if (y < h - 1) {
      *pos++ = '\n';
    }
  }
 
  // Null terminate
  *pos = '\0';
 
  return output;
}
 
// 256-color palette mapping (RGB to ANSI 256 color index) - copied from NEON
static inline uint8_t rgb_to_256color_sse2(uint8_t r, uint8_t g, uint8_t b) {
  return (uint8_t)(16 + 36 * (r / 51) + 6 * (g / 51) + (b / 51));
}
 
// Unified SSE2 function for all color modes (full implementation like NEON)
char *render_ascii_sse2_unified_optimized(const image_t *image, bool use_background, bool use_256color,
                                          const char *ascii_chars) {
  if (!image || !image->pixels) {
    return NULL;
  }
 
  const int width = image->w;
  const int height = image->h;
 
  if (width <= 0 || height <= 0) {
    char *empty;
    empty = SAFE_MALLOC(1, char *);
    empty[0] = '\0';
    return empty;
  }
 
  outbuf_t ob = {0};
  // Estimate buffer size based on mode (copied from NEON)
  size_t bytes_per_pixel = use_256color ? 6u : 8u; // 256-color shorter than truecolor
 
  // Calculate buffer size with overflow checking
  size_t height_times_width;
  if (checked_size_mul((size_t)height, (size_t)width, &height_times_width) != ASCIICHAT_OK) {
    log_error("Buffer size overflow: height * width overflow");
    return NULL;
  }
 
  size_t pixel_data_size;
  if (checked_size_mul(height_times_width, bytes_per_pixel, &pixel_data_size) != ASCIICHAT_OK) {
    log_error("Buffer size overflow: (height * width) * bytes_per_pixel overflow");
    return NULL;
  }
 
  size_t height_times_16;
  if (checked_size_mul((size_t)height, 16u, &height_times_16) != ASCIICHAT_OK) {
    log_error("Buffer size overflow: height * 16 overflow");
    return NULL;
  }
 
  size_t temp;
  if (checked_size_add(pixel_data_size, height_times_16, &temp) != ASCIICHAT_OK) {
    log_error("Buffer size overflow: pixel_data + height*16 overflow");
    return NULL;
  }
 
  if (checked_size_add(temp, 64u, &ob.cap) != ASCIICHAT_OK) {
    log_error("Buffer size overflow: total capacity overflow");
    return NULL;
  }
 
  ob.buf = SAFE_MALLOC(ob.cap ? ob.cap : 1, char *);
  if (!ob.buf)
    return NULL;
 
  // Get cached UTF-8 character mappings for color rendering
  utf8_palette_cache_t *utf8_cache = get_utf8_palette_cache(ascii_chars);
  if (!utf8_cache) {
    log_error("Failed to get UTF-8 palette cache for SSE2 color");
    SAFE_FREE(ob.buf);
    return NULL;
  }
 
  // SSE2 doesn't have _mm_shuffle_epi8 (introduced in SSSE3), so use scalar UTF-8 cache lookup
  // This is still much faster than the old approach since UTF-8 parsing is cached
 
  // Track current color state (copied from NEON)
  int curR = -1, curG = -1, curB = -1;
  int cur_color_idx = -1;
 
  for (int y = 0; y < height; y++) {
    const rgb_pixel_t *row = &((const rgb_pixel_t *)image->pixels)[y * width];
    int x = 0;
 
    // Process 16-pixel chunks with SSE2 (full 128-bit register capacity)
    while (x + 16 <= width) {
      // Manual deinterleave RGB components (SSE2 limitation vs NEON's vld3q_u8)
      uint8_t r_array[16], g_array[16], b_array[16];
      for (int j = 0; j < 16; j++) {
        r_array[j] = row[x + j].r;
        g_array[j] = row[x + j].g;
        b_array[j] = row[x + j].b;
      }
 
      // Load into SSE2 registers
      __m128i r_vec = _mm_loadl_epi64((__m128i *)r_array);
      __m128i g_vec = _mm_loadl_epi64((__m128i *)g_array);
      __m128i b_vec = _mm_loadl_epi64((__m128i *)b_array);
 
      // Convert to 16-bit for arithmetic
      __m128i r_16 = _mm_unpacklo_epi8(r_vec, _mm_setzero_si128());
      __m128i g_16 = _mm_unpacklo_epi8(g_vec, _mm_setzero_si128());
      __m128i b_16 = _mm_unpacklo_epi8(b_vec, _mm_setzero_si128());
 
      // Calculate luminance: (77*R + 150*G + 29*B + 128) >> 8
      __m128i luma_r = _mm_mullo_epi16(r_16, _mm_set1_epi16(LUMA_RED));
      __m128i luma_g = _mm_mullo_epi16(g_16, _mm_set1_epi16(LUMA_GREEN));
      __m128i luma_b = _mm_mullo_epi16(b_16, _mm_set1_epi16(LUMA_BLUE));
 
      __m128i luma_sum = _mm_add_epi16(luma_r, luma_g);
      luma_sum = _mm_add_epi16(luma_sum, luma_b);
      luma_sum = _mm_add_epi16(luma_sum, _mm_set1_epi16(LUMA_THRESHOLD));
      luma_sum = _mm_srli_epi16(luma_sum, 8);
 
      // Pack back to 8-bit and store
      __m128i luminance = _mm_packus_epi16(luma_sum, _mm_setzero_si128());
      uint8_t luma_array[8];
      _mm_storel_epi64((__m128i *)luma_array, luminance);
 
      // Convert to UTF-8 character indices using cached mappings
      uint8_t char_indices[8];
      for (int i = 0; i < 8; i++) {
        const uint8_t luma_idx = luma_array[i] >> 2; // 0-63 index
        char_indices[i] = luma_idx;                  // Direct index into cache64
      }
 
      if (use_256color) {
        // 256-color mode processing (copied from NEON logic)
        uint8_t color_indices[8];
        for (int i = 0; i < 8; i++) {
          color_indices[i] = rgb_to_256color_sse2(r_array[i], g_array[i], b_array[i]);
        }
 
        // Emit with RLE on (UTF-8 character, color) runs
        for (int i = 0; i < 8;) { // SSE2 processes 8 pixels, not 16
          const uint8_t char_idx = char_indices[i];
          const utf8_char_t *char_info = &utf8_cache->cache64[char_idx];
          const uint8_t color_idx = color_indices[i];
 
          int j = i + 1;
          while (j < 8 && char_indices[j] == char_idx && color_indices[j] == color_idx) {
            j++;
          }
          const uint32_t run = (uint32_t)(j - i);
 
          if (color_idx != cur_color_idx) {
            if (use_background) {
              emit_set_256_color_bg(&ob, color_idx);
            } else {
              emit_set_256_color_fg(&ob, color_idx);
            }
            cur_color_idx = color_idx;
          }
 
          // Emit UTF-8 character from cache
          ob_write(&ob, char_info->utf8_bytes, char_info->byte_len);
          if (rep_is_profitable(run)) {
            emit_rep(&ob, run - 1);
          } else {
            for (uint32_t k = 1; k < run; k++) {
              ob_write(&ob, char_info->utf8_bytes, char_info->byte_len);
            }
          }
          i = j;
        }
      } else {
        // Truecolor mode processing with UTF-8 characters
        for (int i = 0; i < 8;) { // SSE2 processes 8 pixels
          const uint8_t char_idx = char_indices[i];
          const utf8_char_t *char_info = &utf8_cache->cache64[char_idx];
          const uint8_t r = r_array[i];
          const uint8_t g = g_array[i];
          const uint8_t b = b_array[i];
 
          int j = i + 1;
          while (j < 8 && char_indices[j] == char_idx && r_array[j] == r && g_array[j] == g && b_array[j] == b) {
            j++;
          }
          const uint32_t run = (uint32_t)(j - i);
 
          if (r != curR || g != curG || b != curB) {
            if (use_background) {
              emit_set_truecolor_bg(&ob, r, g, b);
            } else {
              emit_set_truecolor_fg(&ob, r, g, b);
            }
            curR = r;
            curG = g;
            curB = b;
          }
 
          // Emit UTF-8 character from cache
          ob_write(&ob, char_info->utf8_bytes, char_info->byte_len);
          if (rep_is_profitable(run)) {
            emit_rep(&ob, run - 1);
          } else {
            for (uint32_t k = 1; k < run; k++) {
              ob_write(&ob, char_info->utf8_bytes, char_info->byte_len);
            }
          }
          i = j;
        }
      }
      x += 16;
    }
 
    // Scalar tail for remaining pixels (copied from NEON logic)
    for (; x < width;) {
      const rgb_pixel_t *p = &row[x];
      uint32_t R = p->r, G = p->g, B = p->b;
      uint8_t Y = (uint8_t)((LUMA_RED * R + LUMA_GREEN * G + LUMA_BLUE * B + LUMA_THRESHOLD) >> 8);
      uint8_t luma_idx = Y >> 2;
      const utf8_char_t *char_info = &utf8_cache->cache64[luma_idx];
 
      if (use_256color) {
        // 256-color scalar tail with UTF-8
        uint8_t color_idx = rgb_to_256color_sse2((uint8_t)R, (uint8_t)G, (uint8_t)B);
 
        int j = x + 1;
        while (j < width) {
          const rgb_pixel_t *q = &row[j];
          uint32_t R2 = q->r, G2 = q->g, B2 = q->b;
          uint8_t Y2 = (uint8_t)((LUMA_RED * R2 + LUMA_GREEN * G2 + LUMA_BLUE * B2 + LUMA_THRESHOLD) >> 8);
          uint8_t luma_idx2 = Y2 >> 2;
          uint8_t color_idx2 = rgb_to_256color_sse2((uint8_t)R2, (uint8_t)G2, (uint8_t)B2);
          if (luma_idx2 != luma_idx || color_idx2 != color_idx)
            break;
          j++;
        }
        uint32_t run = (uint32_t)(j - x);
 
        if (color_idx != cur_color_idx) {
          if (use_background) {
            emit_set_256_color_bg(&ob, color_idx);
          } else {
            emit_set_256_color_fg(&ob, color_idx);
          }
          cur_color_idx = color_idx;
        }
 
        // Emit UTF-8 character from cache
        ob_write(&ob, char_info->utf8_bytes, char_info->byte_len);
        if (rep_is_profitable(run)) {
          emit_rep(&ob, run - 1);
        } else {
          for (uint32_t k = 1; k < run; k++) {
            ob_write(&ob, char_info->utf8_bytes, char_info->byte_len);
          }
        }
        x = j;
      } else {
        // Truecolor scalar tail with UTF-8
        int j = x + 1;
        while (j < width) {
          const rgb_pixel_t *q = &row[j];
          uint32_t R2 = q->r, G2 = q->g, B2 = q->b;
          uint8_t Y2 = (uint8_t)((LUMA_RED * R2 + LUMA_GREEN * G2 + LUMA_BLUE * B2 + LUMA_THRESHOLD) >> 8);
          uint8_t luma_idx2 = Y2 >> 2;
          if (luma_idx2 != luma_idx || R2 != R || G2 != G || B2 != B)
            break;
          j++;
        }
        uint32_t run = (uint32_t)(j - x);
 
        if ((int)R != curR || (int)G != curG || (int)B != curB) {
          if (use_background) {
            emit_set_truecolor_bg(&ob, (uint8_t)R, (uint8_t)G, (uint8_t)B);
          } else {
            emit_set_truecolor_fg(&ob, (uint8_t)R, (uint8_t)G, (uint8_t)B);
          }
          curR = (int)R;
          curG = (int)G;
          curB = (int)B;
        }
 
        // Emit UTF-8 character from cache
        ob_write(&ob, char_info->utf8_bytes, char_info->byte_len);
        if (rep_is_profitable(run)) {
          emit_rep(&ob, run - 1);
        } else {
          for (uint32_t k = 1; k < run; k++) {
            ob_write(&ob, char_info->utf8_bytes, char_info->byte_len);
          }
        }
        x = j;
      }
    }
 
    // End row: clear to EOL, reset SGR, add newline (except for last row) (copied from NEON)
    ob_putc(&ob, '\033');
    ob_putc(&ob, '[');
    ob_putc(&ob, 'K');
    emit_reset(&ob);
    if (y < height - 1) {
      ob_putc(&ob, '\n');
    }
    curR = curG = curB = -1;
    cur_color_idx = -1;
  }
 
  ob_term(&ob);
  return ob.buf;
}
 
// Destroy SSE2 cache resources (called at program shutdown)
void sse2_caches_destroy(void) {
  // SSE2 currently uses shared caches from common.c, so no specific cleanup needed
  log_debug("SSE2_CACHE: SSE2 caches cleaned up");
}
 
#endif /* SIMD_SUPPORT_SSE2 */