quicktex/quicktex/BlockEncoder.h

93 lines
3.5 KiB
C++

/* Python-rgbcx Texture Compression Library
Copyright (C) 2021 Andrew Cassidy <drewcassidy@me.com>
Partially derived from rgbcx.h written by Richard Geldreich 2020 <richgel99@gmail.com>
and licenced under the public domain
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#pragma once
#include <climits>
#include <cstdint>
#include <memory>
#include <string>
#include "BlockView.h"
namespace quicktex {
class BlockEncoder {
public:
using EncoderPtr = std::shared_ptr<BlockEncoder>;
virtual ~BlockEncoder() = default;
virtual void EncodeImage(uint8_t *encoded, Color *decoded, unsigned image_width, unsigned image_height) const = 0;
virtual size_t BlockSize() const = 0;
virtual size_t BlockWidth() const = 0;
virtual size_t BlockHeight() const = 0;
};
template <class B, size_t M, size_t N> class BlockEncoderTemplate : public BlockEncoder {
public:
using DecodedBlock = ColorBlockView<M, N>;
using EncodedBlock = B;
BlockEncoderTemplate() noexcept = default;
virtual ~BlockEncoderTemplate() noexcept = default;
virtual void EncodeBlock(DecodedBlock pixels, EncodedBlock *dest) const = 0;
virtual void EncodeImage(uint8_t *encoded, Color *decoded, unsigned image_width, unsigned image_height) const override {
assert(image_width % N == 0);
assert(image_width % M == 0);
unsigned block_width = image_width / N;
unsigned block_height = image_height / M;
auto blocks = reinterpret_cast<B *>(encoded);
// from experimentation, multithreading this using OpenMP sometimes actually makes decoding slower
// due to thread creation/teardown taking longer than the decoding process itself.
// As a result, this is sometimes left as a serial operation despite being embarassingly parallelizable
// threshold for number of blocks before multithreading is set by overriding MTThreshold()
#pragma omp parallel for if (block_width * block_height >= MTThreshold())
for (int y = 0; y < (int)block_height; y++) {
for (int x = 0; x < (int)block_width; x++) {
unsigned pixel_x = (unsigned)x * N;
unsigned pixel_y = (unsigned)y * M;
assert(pixel_x >= 0);
assert(pixel_y >= 0);
assert(pixel_y + M <= image_height);
assert(pixel_x + N <= image_width);
unsigned top_left = pixel_x + (pixel_y * image_width);
unsigned block_index = (unsigned)x + (block_width * (unsigned)y);
auto src = DecodedBlock(&decoded[top_left], (int)image_width);
EncodeBlock(src, &blocks[block_index]);
}
}
}
virtual size_t BlockSize() const override { return sizeof(B); }
virtual size_t BlockWidth() const override { return N; }
virtual size_t BlockHeight() const override { return M; }
virtual size_t MTThreshold() const { return SIZE_MAX; };
};
} // namespace quicktex