/*
==================
==================
*/
void Process_Fragment_4x4(
__int32 w_seed[2],
__int32 i_tile_in,
__int32 i_buffer_in,
const unsigned __int32 coverage_mask,
raster_output_& raster_output,
shader_input_& shader_input
) {
const __int32 i_buffer = i_buffer_in + (i_tile_in * 4 * 4);
__m128i bazza[3][4];
for (__int32 i_edge = 0; i_edge < 2; i_edge++) {
__m128i w_row = set_all(w_seed[i_edge]);
bazza[i_edge][0] = w_row + load_u(raster_output.reject_table[0][i_edge][0]);
bazza[i_edge][1] = w_row + load_u(raster_output.reject_table[0][i_edge][1]);
bazza[i_edge][2] = w_row + load_u(raster_output.reject_table[0][i_edge][2]);
bazza[i_edge][3] = w_row + load_u(raster_output.reject_table[0][i_edge][3]);
}
pixel_shader(i_buffer, coverage_mask, bazza, shader_input);
const __int32 i_buffer_depth_4x4 = i_buffer / (4 * 4);
const __int32 i_buffer_depth_16x16 = i_buffer / (16 * 16);
const __int32 i_buffer_depth_64x64 = i_buffer / (64 * 64);
shader_input.depth_tiles_4x4[i_buffer_depth_4x4] = shader_input.z_max;
shader_input.tile_mask_16x16 |= one_bit_64 << i_buffer_depth_16x16;
shader_input.tile_mask_64x64 |= one_bit_64 << i_buffer_depth_64x64;
}
void ShaderLoad::LoadData(const uint8_t* data, size_t size,
YCommon::YContainers::MemBuffer* buffer) {
YASSERT(YEngineData::VerifyShaderBuffer(flatbuffers::Verifier(data, size)),
"Invalid Shader Data.");
shader_ids = nullptr;
num_shader_ids = 0;
const YEngineData::Shader* shader_data = YEngineData::GetShader(data);
const char* shader_name = shader_data->name()->c_str();
const auto variants = shader_data->variants();
num_shader_ids = variants->size();
shader_ids = static_cast<YRenderer::ShaderID*>(
buffer->Allocate(sizeof(*shader_ids) * num_shader_ids));
YASSERT(shader_ids, "Out of memory - could not load shader data.");
YRenderer::ShaderID* shader_id_iter = shader_ids;
for (auto variant_iter = variants->begin();
variant_iter != variants->end();
++variant_iter) {
const char* variant_name = variant_iter->name()->c_str();
*shader_id_iter++ = YRenderer::Renderer::CreateShader(
shader_name, variant_name,
variant_iter->vertex_shader()->Data(),
variant_iter->vertex_shader()->size(),
variant_iter->pixel_shader()->Data(),
variant_iter->pixel_shader()->size());
}
}
/*
==================
==================
*/
void Process_Fragments(
raster_output_& raster_output,
shader_input_& shader_input
) {
const __m128 zero = set_all(0.0f);
shader_input.tile_mask_16x16 = 0x0;
shader_input.tile_mask_64x64 = 0x0;
//===============================================================================================
{
const __int32 n_fragments = raster_output.n_fragments[raster_output_::TRIVIAL_ACCEPT_64x64];
for (__int32 i_fragment = 0; i_fragment < n_fragments; i_fragment++) {
raster_fragment_& raster_fragment = raster_output.raster_fragment[raster_output_::TRIVIAL_ACCEPT_64x64][i_fragment];
const __int32 i_buffer = raster_fragment.buffer_mask_packed >> 16;
const unsigned __int32 coverage_mask = raster_fragment.buffer_mask_packed & 0xffff;
Process_Fragment_64x64(
raster_fragment.w,
i_buffer,
coverage_mask,
raster_output,
shader_input
);
}
}
//===============================================================================================
{
const __int32 n_fragments = raster_output.n_fragments[raster_output_::TRIVIAL_ACCEPT_16x16];
for (__int32 i_fragment = 0; i_fragment < n_fragments; i_fragment++) {
raster_fragment_& raster_fragment = raster_output.raster_fragment[raster_output_::TRIVIAL_ACCEPT_16x16][i_fragment];
const __int32 i_buffer = raster_fragment.buffer_mask_packed >> 16;
const unsigned __int32 coverage_mask = raster_fragment.buffer_mask_packed & 0xffff;
Process_Fragment_16x16(
raster_fragment.w,
0,
i_buffer,
coverage_mask,
raster_output,
shader_input
);
}
}
//===============================================================================================
{
const __int32 n_fragments = raster_output.n_fragments[raster_output_::TRIVIAL_ACCEPT_4x4];
for (__int32 i_fragment = 0; i_fragment < n_fragments; i_fragment++) {
raster_fragment_& raster_fragment = raster_output.raster_fragment[raster_output_::TRIVIAL_ACCEPT_4x4][i_fragment];
const __int32 i_buffer = raster_fragment.buffer_mask_packed >> 16;
const unsigned __int32 coverage_mask = raster_fragment.buffer_mask_packed & 0xffff;
Process_Fragment_4x4(raster_fragment.w, 0, i_buffer, coverage_mask, raster_output, shader_input);
}
}
//===============================================================================================
{
//const __int32 start = raster_output_::MAX_FRAGMENTS - 1;
//const __int32 end = raster_output.n_fragments[raster_output_::PARTIAL_ACCEPT_4x4];
//for (__int32 i_fragment = start; i_fragment > end; i_fragment--) {
// raster_fragment_& raster_fragment = raster_output.raster_fragment[raster_output_::PARTIAL_ACCEPT_4x4][i_fragment];
// const __int32 i_buffer = raster_fragment.buffer_mask_packed >> 16;
// const unsigned __int32 coverage_mask = raster_fragment.buffer_mask_packed & 0xffff;
// Process_Fragment_4x4(raster_fragment.w, 0, i_buffer, coverage_mask, raster_output, shader_input);
//}
}
//===============================================================================================
{
const __int32 n_fragments = raster_output.n_fragments_COMPLETE;
__int32 n_depth_fragments = 0;
for (__int32 i_fragment = 0; i_fragment < n_fragments; i_fragment++) {
raster_fragment_complete_& raster_fragment = raster_output.raster_fragment_complete[i_fragment];
const __int32 i_buffer = raster_fragment.buffer_mask_packed >> 16;
const unsigned __int32 coverage_mask = raster_fragment.buffer_mask_packed & 0xffff;
pixel_shader(i_buffer, coverage_mask, raster_fragment.bazza, shader_input);
const __int32 i_buffer_depth_4x4 = i_buffer / (4 * 4);
const __int32 i_buffer_depth_16x16 = i_buffer / (16 * 16);
const __int32 i_buffer_depth_64x64 = i_buffer / (64 * 64);
shader_input.depth_tiles_4x4[i_buffer_depth_4x4] = shader_input.z_max;
shader_input.tile_mask_16x16 |= one_bit_64 << i_buffer_depth_16x16;
shader_input.tile_mask_64x64 |= one_bit_64 << i_buffer_depth_64x64;
}
}
//===============================================================================================
{
//printf_s(" %llu ", shader_input.tile_mask_16x16);
__int64 n_tiles = _mm_popcnt_u64(shader_input.tile_mask_16x16);
for (__int32 i_bit = 0; i_bit < n_tiles; i_bit++) {
unsigned long i_tile_16x16;
_BitScanForward64(&i_tile_16x16, shader_input.tile_mask_16x16);
shader_input.tile_mask_16x16 ^= one_bit_64 << i_tile_16x16;
const __int32 i_tile_4x4 = i_tile_16x16 * (4 * 4);
__m128 depth_4x4[4];
depth_4x4[0] = load_u(shader_input.depth_tiles_4x4 + i_tile_4x4 + (0 * 4));
depth_4x4[1] = load_u(shader_input.depth_tiles_4x4 + i_tile_4x4 + (1 * 4));
depth_4x4[2] = load_u(shader_input.depth_tiles_4x4 + i_tile_4x4 + (2 * 4));
depth_4x4[3] = load_u(shader_input.depth_tiles_4x4 + i_tile_4x4 + (3 * 4));
__m128 z_max;
z_max = depth_4x4[0];
z_max = min_vec(depth_4x4[1], z_max);
z_max = min_vec(depth_4x4[2], z_max);
z_max = min_vec(depth_4x4[3], z_max);
__m128 z_out = z_max;
z_max = rotate_left(z_max);
z_out = min_vec(z_max, z_out);
z_max = rotate_left(z_max);
z_out = min_vec(z_max, z_out);
z_max = rotate_left(z_max);
z_out = min_vec(z_max, z_out);
shader_input.depth_tiles_16x16[i_tile_16x16] = store_s(z_out);
}
}
{
__int64 n_tiles = _mm_popcnt_u64(shader_input.tile_mask_64x64);
//printf_s(" %llu ", n_tiles);
for (__int32 i_bit = 0; i_bit < n_tiles; i_bit++) {
unsigned long i_tile_64x64;
_BitScanForward64(&i_tile_64x64, shader_input.tile_mask_64x64);
shader_input.tile_mask_64x64 ^= one_bit_64 << i_tile_64x64;
const __int32 i_tile_16x16 = i_tile_64x64 * (4 * 4);
__m128 depth_16x16[4];
depth_16x16[0] = load_u(shader_input.depth_tiles_16x16 + i_tile_16x16 + (0 * 4));
depth_16x16[1] = load_u(shader_input.depth_tiles_16x16 + i_tile_16x16 + (1 * 4));
depth_16x16[2] = load_u(shader_input.depth_tiles_16x16 + i_tile_16x16 + (2 * 4));
depth_16x16[3] = load_u(shader_input.depth_tiles_16x16 + i_tile_16x16 + (3 * 4));
__m128 z_max;
z_max = depth_16x16[0];
z_max = min_vec(depth_16x16[1], z_max);
z_max = min_vec(depth_16x16[2], z_max);
z_max = min_vec(depth_16x16[3], z_max);
__m128 z_out = z_max;
z_max = rotate_left(z_max);
z_out = min_vec(z_max, z_out);
z_max = rotate_left(z_max);
z_out = min_vec(z_max, z_out);
z_max = rotate_left(z_max);
z_out = min_vec(z_max, z_out);
shader_input.depth_tiles_64x64[i_tile_64x64] = store_s(z_out);
}
}
}
