/**
 ******************************************************************************
 * Xenia : Xbox 360 Emulator Research Project                                 *
 ******************************************************************************
 * Copyright 2022 Ben Vanik. All rights reserved.                             *
 * Released under the BSD license - see LICENSE in the root for more details. *
 ******************************************************************************
 */

#ifndef XENIA_GPU_SHADERS_EDRAM_XESLI_
#define XENIA_GPU_SHADERS_EDRAM_XESLI_

#include "pixel_formats.xesli"

#define kXenosMsaaSamples_1X 0u
#define kXenosMsaaSamples_2X 1u
#define kXenosMsaaSamples_4X 2u

// `wrap = false` can be used if it's known that the resulting tile indices
// can't exceed 11 bits, and the modulo operator doesn't need to be performed to
// access the data in the render targets that are located in both ends of the
// EDRAM at the same time.
uint XeEdramOffsetInts(xesl_uint2 pixel_index, uint base_tiles, bool wrap,
                       uint pitch_tiles, uint msaa_samples, bool is_depth,
                       uint format_ints_log2, uint pixel_sample_index,
                       xesl_uint2 resolution_scale) {
  xesl_uint2 rt_sample_index =
      pixel_index <<
      xesl_uint2(xesl_greaterThanEqual(
          xesl_uint_x2(msaa_samples),
          xesl_uint2(kXenosMsaaSamples_4X, kXenosMsaaSamples_2X)));
  rt_sample_index +=
      (xesl_uint_x2(pixel_sample_index) >> xesl_uint2(1u, 0u)) & 1u;
  // For now, while the actual storage of 64bpp render targets in comparison to
  // 32bpp is not known, storing 40x16 64bpp samples per tile for simplicity of
  // addressing in different scenarios.
  xesl_uint2 tile_size_at_32bpp = xesl_uint2(80u, 16u) * resolution_scale;
  xesl_uint2 tile_size_samples =
      tile_size_at_32bpp >> xesl_uint2(format_ints_log2, 0u);
  xesl_uint2 tile_offset_xy = rt_sample_index / tile_size_samples;
  base_tiles += tile_offset_xy.y * pitch_tiles + tile_offset_xy.x;
  rt_sample_index -= tile_offset_xy * tile_size_samples;
  if (is_depth) {
    uint tile_width_half = tile_size_samples.x >> 1u;
    rt_sample_index.x =
        uint(int(rt_sample_index.x) +
             ((rt_sample_index.x >= tile_width_half) ? -int(tile_width_half)
                                                     : int(tile_width_half)));
  }
  uint address =
      base_tiles * (tile_size_at_32bpp.x * tile_size_at_32bpp.y) +
      ((rt_sample_index.y * tile_size_samples.x + rt_sample_index.x) <<
       format_ints_log2);
  if (wrap) {
    // EDRAM addressing is periodic (modulo the EDRAM size).
    address %= tile_size_at_32bpp.x * tile_size_at_32bpp.y * 2048u;
  }
  return address;
}

#endif  // XENIA_GPU_SHADERS_EDRAM_XESLI_