rpcsx/rpcs3/Emu/RSX/GL/GLTextureCache.h

#pragma once

#include "stdafx.h"

#include <exception>
#include <string>
#include <functional>
#include <vector>
#include <memory>
#include <thread>
#include <condition_variable>
#include <chrono>

#include "Utilities/mutex.h"
#include "GLRenderTargets.h"
#include "GLOverlays.h"
#include "GLTexture.h"
#include "../Common/TextureUtils.h"
#include "../Common/texture_cache.h"

class GLGSRender;

namespace gl
{
	class blitter;

	extern GLenum get_sized_internal_format(u32);
	extern void copy_typeless(texture*, const texture*, const coord3u&, const coord3u&);
	extern blitter *g_hw_blitter;

	class cached_texture_section;
	class texture_cache;

	struct texture_cache_traits
	{
		using commandbuffer_type      = gl::command_context;
		using section_storage_type    = gl::cached_texture_section;
		using texture_cache_type      = gl::texture_cache;
		using texture_cache_base_type = rsx::texture_cache<texture_cache_type, texture_cache_traits>;
		using image_resource_type     = gl::texture*;
		using image_view_type         = gl::texture_view*;
		using image_storage_type      = gl::texture;
		using texture_format          = gl::texture::format;
	};

	class cached_texture_section : public rsx::cached_texture_section<gl::cached_texture_section, gl::texture_cache_traits>
	{
		using baseclass = rsx::cached_texture_section<gl::cached_texture_section, gl::texture_cache_traits>;
		friend baseclass;

		fence m_fence;
		buffer pbo;

		gl::viewable_image* vram_texture = nullptr;

		std::unique_ptr<gl::viewable_image> managed_texture;
		std::unique_ptr<gl::texture> scaled_texture;

		texture::format format = texture::format::rgba;
		texture::type type = texture::type::ubyte;

		void init_buffer(const gl::texture* src)
		{
			const u32 vram_size = src->pitch() * src->height();
			const u32 buffer_size = align(vram_size, 4096);

			if (pbo)
			{
				if (pbo.size() >= buffer_size)
					return;

				pbo.remove();
			}

			pbo.create(buffer::target::pixel_pack, buffer_size, nullptr, buffer::memory_type::host_visible, GL_STREAM_READ);
			glBindBuffer(GL_PIXEL_PACK_BUFFER, GL_NONE);
		}

	public:
		using baseclass::cached_texture_section;

		void create(u16 w, u16 h, u16 depth, u16 mipmaps, gl::texture* image, u32 rsx_pitch, bool read_only,
				gl::texture::format gl_format = gl::texture::format::rgba, gl::texture::type gl_type = gl::texture::type::ubyte, bool swap_bytes = false)
		{
			auto new_texture = static_cast<gl::viewable_image*>(image);
			ASSERT(!exists() || !is_managed() || vram_texture == new_texture);
			vram_texture = new_texture;

			if (read_only)
			{
				managed_texture.reset(vram_texture);
			}
			else
			{
				ASSERT(!managed_texture);
			}

			flushed = false;
			synchronized = false;
			sync_timestamp = 0ull;

			verify(HERE), rsx_pitch;

			this->rsx_pitch = rsx_pitch;
			this->width = w;
			this->height = h;
			this->real_pitch = 0;
			this->depth = depth;
			this->mipmaps = mipmaps;

			set_format(gl_format, gl_type, swap_bytes);

			// Notify baseclass
			baseclass::on_section_resources_created();
		}

		void set_dimensions(u32 width, u32 height, u32 /*depth*/, u32 pitch)
		{
			this->width = width;
			this->height = height;
			rsx_pitch = pitch;
		}

		void set_format(texture::format gl_format, texture::type gl_type, bool swap_bytes)
		{
			format = gl_format;
			type = gl_type;
			pack_unpack_swap_bytes = swap_bytes;

			if (format == gl::texture::format::rgba)
			{
				switch (type)
				{
				case gl::texture::type::f16:
					gcm_format = CELL_GCM_TEXTURE_W16_Z16_Y16_X16_FLOAT;
					break;
				case gl::texture::type::f32:
					gcm_format = CELL_GCM_TEXTURE_W32_Z32_Y32_X32_FLOAT;
					break;
				default:
					break;
				}
			}
		}

		void dma_transfer(gl::command_context& /*cmd*/, gl::texture* src, const areai& /*src_area*/, const utils::address_range& /*valid_range*/, u32 pitch)
		{
			init_buffer(src);

			glGetError();
			pbo.bind(buffer::target::pixel_pack);

			if (context == rsx::texture_upload_context::dma)
			{
				// Determine unpack config dynamically
				const auto format_info = gl::get_format_type(src->get_internal_format());
				format = static_cast<gl::texture::format>(format_info.format);
				type = static_cast<gl::texture::type>(format_info.type);

				if ((src->aspect() & gl::image_aspect::stencil) == 0)
				{
					pack_unpack_swap_bytes = format_info.swap_bytes;
				}
				else
				{
					// Z24S8 decode is done on the CPU for now
					pack_unpack_swap_bytes = false;
				}
			}

			pixel_pack_settings pack_settings;
			pack_settings.alignment(1);
			pack_settings.swap_bytes(pack_unpack_swap_bytes);

			src->copy_to(nullptr, format, type, pack_settings);
			real_pitch = src->pitch();
			rsx_pitch = pitch;

			if (auto error = glGetError())
			{
				if (error == GL_OUT_OF_MEMORY && ::gl::get_driver_caps().vendor_AMD)
				{
					// AMD driver bug
					// Pixel transfer fails with GL_OUT_OF_MEMORY. Usually happens with float textures or operations attempting to swap endianness.
					// Failed operations also leak a large amount of memory
					rsx_log.error("Memory transfer failure (AMD bug). Please update your driver to Adrenalin 19.4.3 or newer. Format=0x%x, Type=0x%x, Swap=%d", static_cast<u32>(format), static_cast<u32>(type), pack_unpack_swap_bytes);
				}
				else
				{
					rsx_log.error("Memory transfer failed with error 0x%x. Format=0x%x, Type=0x%x", error, static_cast<u32>(format), static_cast<u32>(type));
				}
			}

			glBindBuffer(GL_PIXEL_PACK_BUFFER, GL_NONE);

			m_fence.reset();
			synchronized = true;
			sync_timestamp = get_system_time();
		}

		void copy_texture(gl::command_context& cmd, bool miss)
		{
			ASSERT(exists());

			if (!miss) [[likely]]
			{
				baseclass::on_speculative_flush();
			}
			else
			{
				baseclass::on_miss();
			}

			if (context == rsx::texture_upload_context::framebuffer_storage)
			{
				auto as_rtt = static_cast<gl::render_target*>(vram_texture);
				if (as_rtt->dirty()) as_rtt->read_barrier(cmd);
			}

			gl::texture* target_texture = vram_texture;
			if ((rsx::get_resolution_scale_percent() != 100 && context == rsx::texture_upload_context::framebuffer_storage) ||
				(vram_texture->pitch() != rsx_pitch))
			{
				u32 real_width = width;
				u32 real_height = height;

				if (context == rsx::texture_upload_context::framebuffer_storage)
				{
					auto surface = gl::as_rtt(vram_texture);
					real_width *= surface->samples_x;
					real_height *= surface->samples_y;
				}

				areai src_area = { 0, 0, 0, 0 };
				const areai dst_area = { 0, 0, static_cast<s32>(real_width), static_cast<s32>(real_height) };

				auto ifmt = vram_texture->get_internal_format();
				src_area.x2 = vram_texture->width();
				src_area.y2 = vram_texture->height();

				if (src_area.x2 != dst_area.x2 || src_area.y2 != dst_area.y2)
				{
					if (scaled_texture)
					{
						auto sfmt = scaled_texture->get_internal_format();
						if (scaled_texture->width() != real_width ||
							scaled_texture->height() != real_height ||
							sfmt != ifmt)
						{
							//Discard current scaled texture
							scaled_texture.reset();
						}
					}

					if (!scaled_texture)
					{
						scaled_texture = std::make_unique<gl::texture>(GL_TEXTURE_2D, real_width, real_height, 1, 1, static_cast<GLenum>(ifmt));
					}

					const bool linear_interp = is_depth_texture() ? false : true;
					g_hw_blitter->scale_image(cmd, vram_texture, scaled_texture.get(), src_area, dst_area, linear_interp, {});
					target_texture = scaled_texture.get();
				}
			}

			dma_transfer(cmd, target_texture, {}, {}, rsx_pitch);
		}

		/**
		 * Flush
		 */
		void* map_synchronized(u32 offset, u32 size)
		{
			AUDIT(synchronized && !m_fence.is_empty());

			m_fence.wait_for_signal();

			verify(HERE), (offset + size) <= pbo.size();
			pbo.bind(buffer::target::pixel_pack);

			return glMapBufferRange(GL_PIXEL_PACK_BUFFER, offset, size, GL_MAP_READ_BIT);
		}

		void finish_flush()
		{
			// Free resources
			glUnmapBuffer(GL_PIXEL_PACK_BUFFER);
			glBindBuffer(GL_PIXEL_PACK_BUFFER, GL_NONE);

			const auto valid_range = get_confirmed_range_delta();
			const u32 valid_offset = valid_range.first;
			const u32 valid_length = valid_range.second;
			void *dst = get_ptr(get_section_base() + valid_offset);

			if (pack_unpack_swap_bytes)
			{
				// Shuffle
				// TODO: Do this with a compute shader
				switch (type)
				{
				case gl::texture::type::sbyte:
				case gl::texture::type::ubyte:
				{
					if (pack_unpack_swap_bytes)
					{
						// byte swapping does not work on byte types, use uint_8_8_8_8 for rgba8 instead to avoid penalty
						rsx::shuffle_texel_data_wzyx<u8>(dst, rsx_pitch, width, align(valid_length, rsx_pitch) / rsx_pitch);
					}
					break;
				}
				case gl::texture::type::uint_24_8:
				{
					verify(HERE), pack_unpack_swap_bytes == false;
					verify(HERE), real_pitch == (width * 4);
					if (rsx_pitch == real_pitch) [[likely]]
					{
						rsx::convert_le_d24x8_to_be_d24x8(dst, dst, valid_length / 4, 1);
					}
					else
					{
						const u32 num_rows = align(valid_length, rsx_pitch) / rsx_pitch;
						u8* data = static_cast<u8*>(dst);
						for (u32 row = 0; row < num_rows; ++row)
						{
							rsx::convert_le_d24x8_to_be_d24x8(data, data, width, 1);
							data += rsx_pitch;
						}
					}
					break;
				}
				default:
					break;
				}
			}

			if (is_swizzled())
			{
				// This format is completely worthless to CPU processing algorithms where cache lines on die are linear.
				// If this is happening, usually it means it was not a planned readback (e.g shared pages situation)
				rsx_log.warning("[Performance warning] CPU readback of swizzled data");

				// Read-modify-write to avoid corrupting already resident memory outside texture region
				std::vector<u8> tmp_data(rsx_pitch * height);
				std::memcpy(tmp_data.data(), dst, tmp_data.size());

				switch (type)
				{
				case gl::texture::type::uint_8_8_8_8:
				case gl::texture::type::uint_24_8:
					rsx::convert_linear_swizzle<u32, false>(tmp_data.data(), dst, width, height, rsx_pitch);
					break;
				case gl::texture::type::ushort_5_6_5:
				case gl::texture::type::ushort:
					rsx::convert_linear_swizzle<u16, false>(tmp_data.data(), dst, width, height, rsx_pitch);
					break;
				default:
					rsx_log.error("Unexpected swizzled texture format 0x%x", static_cast<u32>(format));
				}
			}

			if (context == rsx::texture_upload_context::framebuffer_storage)
			{
				// Update memory tag
				static_cast<gl::render_target*>(vram_texture)->sync_tag();
			}
		}

		/**
		 * Misc
		 */
		void destroy()
		{
			if (!is_locked() && !pbo && vram_texture == nullptr && m_fence.is_empty() && !managed_texture)
				//Already destroyed
				return;

			if (pbo)
			{
				// Destroy pbo cache since vram texture is managed elsewhere
				pbo.remove();
				scaled_texture.reset();
			}

			managed_texture.reset();
			vram_texture = nullptr;

			if (!m_fence.is_empty())
			{
				m_fence.destroy();
			}

			baseclass::on_section_resources_destroyed();
		}

		bool exists() const
		{
			return (vram_texture != nullptr);
		}

		bool is_managed() const
		{
			return !exists() || managed_texture;
		}

		texture::format get_format() const
		{
			return format;
		}

		bool is_flushed() const
		{
			return flushed;
		}

		bool is_synchronized() const
		{
			return synchronized;
		}

		void set_flushed(bool state)
		{
			flushed = state;
		}

		bool is_empty() const
		{
			return vram_texture == nullptr;
		}

		gl::texture_view* get_view(u32 remap_encoding, const std::pair<std::array<u8, 4>, std::array<u8, 4>>& remap)
		{
			return vram_texture->get_view(remap_encoding, remap);
		}

		gl::texture* get_raw_texture() const
		{
			return managed_texture.get();
		}

		gl::texture_view* get_raw_view()
		{
			return vram_texture->get_view(0xAAE4, rsx::default_remap_vector);
		}

		bool is_depth_texture() const
		{
			switch (vram_texture->get_internal_format())
			{
			case gl::texture::internal_format::depth16:
			case gl::texture::internal_format::depth24_stencil8:
			case gl::texture::internal_format::depth32f_stencil8:
				return true;
			default:
				return false;
			}
		}

		bool has_compatible_format(gl::texture* tex) const
		{
			//TODO
			return (tex->get_internal_format() == vram_texture->get_internal_format());
		}
	};

	class texture_cache : public rsx::texture_cache<gl::texture_cache, gl::texture_cache_traits>
	{
	private:
		using baseclass = rsx::texture_cache<gl::texture_cache, gl::texture_cache_traits>;
		friend baseclass;

	private:
		struct discardable_storage
		{
			std::unique_ptr<gl::texture> image;
			std::unique_ptr<gl::texture_view> view;

			discardable_storage() = default;

			discardable_storage(std::unique_ptr<gl::texture>& tex)
			{
				image = std::move(tex);
			}

			discardable_storage(std::unique_ptr<gl::texture_view>& _view)
			{
				view = std::move(_view);
			}

			discardable_storage(std::unique_ptr<gl::texture>& tex, std::unique_ptr<gl::texture_view>& _view)
			{
				image = std::move(tex);
				view = std::move(_view);
			}
		};

	private:

		blitter m_hw_blitter;
		std::vector<discardable_storage> m_temporary_surfaces;

		void clear()
		{
			baseclass::clear();
			clear_temporary_subresources();
		}

		void clear_temporary_subresources()
		{
			m_temporary_surfaces.clear();
		}

		gl::texture_view* create_temporary_subresource_impl(gl::command_context& cmd, gl::texture* src, GLenum sized_internal_fmt, GLenum dst_type, u32 gcm_format,
				u16 x, u16 y, u16 width, u16 height, const rsx::texture_channel_remap_t& remap, bool copy)
		{
			if (sized_internal_fmt == GL_NONE)
			{
				sized_internal_fmt = gl::get_sized_internal_format(gcm_format);
			}

			std::unique_ptr<gl::texture> dst = std::make_unique<gl::viewable_image>(dst_type, width, height, 1, 1, sized_internal_fmt);

			if (copy)
			{
				std::vector<copy_region_descriptor> region =
				{{
					src,
					rsx::surface_transform::coordinate_transform,
					0,
					x, y, 0, 0, 0,
					width, height, width, height
				}};

				copy_transfer_regions_impl(cmd, dst.get(), region);
			}

			std::array<GLenum, 4> swizzle;
			if (!src || static_cast<GLenum>(src->get_internal_format()) != sized_internal_fmt)
			{
				// Apply base component map onto the new texture if a data cast has been done
				swizzle = get_component_mapping(gcm_format, rsx::texture_create_flags::default_component_order);
			}
			else
			{
				swizzle = src->get_native_component_layout();
			}

			if (memcmp(remap.first.data(), rsx::default_remap_vector.first.data(), 4) ||
				memcmp(remap.second.data(), rsx::default_remap_vector.second.data(), 4))
				swizzle = apply_swizzle_remap(swizzle, remap);

			auto view = std::make_unique<gl::texture_view>(dst.get(), dst_type, sized_internal_fmt, swizzle.data());
			auto result = view.get();

			m_temporary_surfaces.emplace_back(dst, view);
			return result;
		}

		std::array<GLenum, 4> get_component_mapping(u32 gcm_format, rsx::texture_create_flags flags) const
		{
			switch (gcm_format)
			{
			case CELL_GCM_TEXTURE_DEPTH24_D8:
			case CELL_GCM_TEXTURE_DEPTH24_D8_FLOAT:
			case CELL_GCM_TEXTURE_DEPTH16:
			case CELL_GCM_TEXTURE_DEPTH16_FLOAT:
				//Dont bother letting this propagate
				return{ GL_RED, GL_RED, GL_RED, GL_RED };
			default:
				break;
			}

			switch (flags)
			{
			case rsx::texture_create_flags::default_component_order:
			{
				return gl::get_swizzle_remap(gcm_format);
			}
			case rsx::texture_create_flags::native_component_order:
			{
				return{ GL_ALPHA, GL_RED, GL_GREEN, GL_BLUE };
			}
			case rsx::texture_create_flags::swapped_native_component_order:
			{
				return{ GL_BLUE, GL_ALPHA, GL_RED, GL_GREEN };
			}
			default:
				fmt::throw_exception("Unknown texture create flags" HERE);
			}
		}

		void copy_transfer_regions_impl(gl::command_context& cmd, gl::texture* dst_image, const std::vector<copy_region_descriptor>& sources) const
		{
			const auto dst_bpp = dst_image->pitch() / dst_image->width();
			const auto dst_aspect = dst_image->aspect();

			for (const auto &slice : sources)
			{
				if (!slice.src)
					continue;

				const bool typeless = dst_aspect != slice.src->aspect() ||
					!formats_are_bitcast_compatible(static_cast<GLenum>(slice.src->get_internal_format()), static_cast<GLenum>(dst_image->get_internal_format()));

				std::unique_ptr<gl::texture> tmp;
				auto src_image = slice.src;
				auto src_x = slice.src_x;
				auto src_y = slice.src_y;
				auto src_w = slice.src_w;
				auto src_h = slice.src_h;

				if (slice.xform == rsx::surface_transform::coordinate_transform)
				{
					// Dimensions were given in 'dst' space. Work out the real source coordinates
					const auto src_bpp = slice.src->pitch() / slice.src->width();
					src_x = (src_x * dst_bpp) / src_bpp;
					src_w = ::aligned_div<u16>(src_w * dst_bpp, src_bpp);
				}

				if (auto surface = dynamic_cast<gl::render_target*>(slice.src))
				{
					surface->transform_samples_to_pixels(src_x, src_w, src_y, src_h);
				}

				if (typeless) [[unlikely]]
				{
					const auto src_bpp = slice.src->pitch() / slice.src->width();
					const u16 convert_w = u16(slice.src->width() * src_bpp) / dst_bpp;
					tmp = std::make_unique<texture>(GL_TEXTURE_2D, convert_w, slice.src->height(), 1, 1, static_cast<GLenum>(dst_image->get_internal_format()));

					src_image = tmp.get();

					// Compute src region in dst format layout
					const u16 src_w2 = u16(src_w * src_bpp) / dst_bpp;
					const u16 src_x2 = u16(src_x * src_bpp) / dst_bpp;

					if (src_w2 == slice.dst_w && src_h == slice.dst_h && slice.level == 0)
					{
						// Optimization, avoid typeless copy to tmp followed by data copy to dst
						// Combine the two transfers into one
						const coord3u src_region = { { src_x, src_y, 0 }, { src_w, src_h, 1 } };
						const coord3u dst_region = { { slice.dst_x, slice.dst_y, slice.dst_z }, { slice.dst_w, slice.dst_h, 1 } };
						gl::copy_typeless(dst_image, slice.src, dst_region, src_region);

						continue;
					}

					const coord3u src_region = { { src_x, src_y, 0 }, { src_w, src_h, 1 } };
					const coord3u dst_region = { { src_x2, src_y, 0 }, { src_w2, src_h, 1 } };
					gl::copy_typeless(src_image, slice.src, dst_region, src_region);

					src_x = src_x2;
					src_w = src_w2;
				}

				if (src_w == slice.dst_w && src_h == slice.dst_h)
				{
					glCopyImageSubData(src_image->id(), GL_TEXTURE_2D, 0, src_x, src_y, 0,
						dst_image->id(), static_cast<GLenum>(dst_image->get_target()), slice.level, slice.dst_x, slice.dst_y, slice.dst_z, src_w, src_h, 1);
				}
				else
				{
					verify(HERE), dst_image->get_target() == gl::texture::target::texture2D;

					auto _blitter = gl::g_hw_blitter;
					const areai src_rect = { src_x, src_y, src_x + src_w, src_y + src_h };
					const areai dst_rect = { slice.dst_x, slice.dst_y, slice.dst_x + slice.dst_w, slice.dst_y + slice.dst_h };

					gl::texture* _dst;
					if (src_image->get_internal_format() == dst_image->get_internal_format() && slice.level == 0)
					{
						_dst = dst_image;
					}
					else
					{
						tmp = std::make_unique<texture>(GL_TEXTURE_2D, dst_rect.x2, dst_rect.y2, 1, 1, static_cast<GLenum>(slice.src->get_internal_format()));
						_dst = tmp.get();
					}

					_blitter->scale_image(cmd, src_image, _dst,
						src_rect, dst_rect, false, {});

					if (_dst != dst_image)
					{
						// Data cast comes after scaling
						glCopyImageSubData(tmp->id(), GL_TEXTURE_2D, 0, slice.dst_x, slice.dst_y, 0,
							dst_image->id(), static_cast<GLenum>(dst_image->get_target()), slice.level, slice.dst_x, slice.dst_y, slice.dst_z, slice.dst_w, slice.dst_h, 1);
					}
				}
			}
		}

		gl::texture* get_template_from_collection_impl(const std::vector<copy_region_descriptor>& sections_to_transfer) const
		{
			if (sections_to_transfer.size() == 1) [[likely]]
			{
				return sections_to_transfer.front().src;
			}

			gl::texture* result = nullptr;
			for (const auto &section : sections_to_transfer)
			{
				if (!section.src)
					continue;

				if (!result)
				{
					result = section.src;
				}
				else
				{
					const auto set1 = result->get_native_component_layout();
					const auto set2 = section.src->get_native_component_layout();

					if (set1[0] != set2[0] ||
						set1[1] != set2[1] ||
						set1[2] != set2[2] ||
						set1[3] != set2[3])
					{
						// TODO
						// This requires a far more complex setup as its not always possible to mix and match without compute assistance
						return nullptr;
					}
				}
			}

			return result;
		}

	protected:

		gl::texture_view* create_temporary_subresource_view(gl::command_context &cmd, gl::texture** src, u32 gcm_format, u16 x, u16 y, u16 w, u16 h,
				const rsx::texture_channel_remap_t& remap_vector) override
		{
			return create_temporary_subresource_impl(cmd, *src, GL_NONE, GL_TEXTURE_2D, gcm_format, x, y, w, h, remap_vector, true);
		}

		gl::texture_view* create_temporary_subresource_view(gl::command_context &cmd, gl::texture* src, u32 gcm_format, u16 x, u16 y, u16 w, u16 h,
				const rsx::texture_channel_remap_t& remap_vector) override
		{
			return create_temporary_subresource_impl(cmd, src, static_cast<GLenum>(src->get_internal_format()),
					GL_TEXTURE_2D, gcm_format, x, y, w, h, remap_vector, true);
		}

		gl::texture_view* generate_cubemap_from_images(gl::command_context& cmd, u32 gcm_format, u16 size, const std::vector<copy_region_descriptor>& sources, const rsx::texture_channel_remap_t& /*remap_vector*/) override
		{
			const GLenum ifmt = gl::get_sized_internal_format(gcm_format);
			std::unique_ptr<gl::texture> dst_image = std::make_unique<gl::viewable_image>(GL_TEXTURE_CUBE_MAP, size, size, 1, 1, ifmt);
			auto view = std::make_unique<gl::texture_view>(dst_image.get(), GL_TEXTURE_CUBE_MAP, ifmt);

			//Empty GL_ERROR
			glGetError();

			copy_transfer_regions_impl(cmd, dst_image.get(), sources);

			if (GLenum err = glGetError())
			{
				rsx_log.warning("Failed to copy image subresource with GL error 0x%X", err);
				return nullptr;
			}

			auto result = view.get();
			m_temporary_surfaces.emplace_back(dst_image, view);
			return result;
		}

		gl::texture_view* generate_3d_from_2d_images(gl::command_context& cmd, u32 gcm_format, u16 width, u16 height, u16 depth, const std::vector<copy_region_descriptor>& sources, const rsx::texture_channel_remap_t& /*remap_vector*/) override
		{
			const GLenum ifmt = gl::get_sized_internal_format(gcm_format);
			std::unique_ptr<gl::texture> dst_image = std::make_unique<gl::viewable_image>(GL_TEXTURE_3D, width, height, depth, 1, ifmt);
			auto view = std::make_unique<gl::texture_view>(dst_image.get(), GL_TEXTURE_3D, ifmt);

			//Empty GL_ERROR
			glGetError();

			copy_transfer_regions_impl(cmd, dst_image.get(), sources);

			if (GLenum err = glGetError())
			{
				rsx_log.warning("Failed to copy image subresource with GL error 0x%X", err);
				return nullptr;
			}

			auto result = view.get();
			m_temporary_surfaces.emplace_back(dst_image, view);
			return result;
		}

		gl::texture_view* generate_atlas_from_images(gl::command_context& cmd, u32 gcm_format, u16 width, u16 height, const std::vector<copy_region_descriptor>& sections_to_copy,
				const rsx::texture_channel_remap_t& remap_vector) override
		{
			auto _template = get_template_from_collection_impl(sections_to_copy);
			auto result = create_temporary_subresource_impl(cmd, _template, GL_NONE, GL_TEXTURE_2D, gcm_format, 0, 0, width, height, remap_vector, false);

			copy_transfer_regions_impl(cmd, result->image(), sections_to_copy);
			return result;
		}

		gl::texture_view* generate_2d_mipmaps_from_images(gl::command_context& cmd, u32 /*gcm_format*/, u16 width, u16 height, const std::vector<copy_region_descriptor>& sections_to_copy,
			const rsx::texture_channel_remap_t& remap_vector) override
		{
			const auto _template = sections_to_copy.front().src;
			const GLenum ifmt = static_cast<GLenum>(_template->get_internal_format());
			const u8 mipmaps = ::narrow<u8>(sections_to_copy.size());
			const auto swizzle = _template->get_native_component_layout();

			auto image_ptr = new gl::viewable_image(GL_TEXTURE_2D, width, height, 1, mipmaps, ifmt);
			image_ptr->set_native_component_layout(swizzle);

			copy_transfer_regions_impl(cmd, image_ptr, sections_to_copy);

			auto view = image_ptr->get_view(get_remap_encoding(remap_vector), remap_vector);

			std::unique_ptr<gl::texture> dst_image(image_ptr);
			m_temporary_surfaces.emplace_back(dst_image);
			return view;
		}

		void release_temporary_subresource(gl::texture_view* view) override
		{
			for (auto& e : m_temporary_surfaces)
			{
				if (e.image.get() == view->image())
				{
					e.view.reset();
					e.image.reset();
					return;
				}
			}
		}

		void update_image_contents(gl::command_context& cmd, gl::texture_view* dst, gl::texture* src, u16 width, u16 height) override
		{
			std::vector<copy_region_descriptor> region =
			{{
				src,
				rsx::surface_transform::identity,
				0,
				0, 0, 0, 0, 0,
				width, height, width, height
			}};

			copy_transfer_regions_impl(cmd, dst->image(), region);
		}

		cached_texture_section* create_new_texture(gl::command_context&, const utils::address_range &rsx_range, u16 width, u16 height, u16 depth, u16 mipmaps, u16 pitch,
			u32 gcm_format, rsx::texture_upload_context context, rsx::texture_dimension_extended type, bool swizzled, rsx::texture_create_flags flags) override
		{
			auto image = gl::create_texture(gcm_format, width, height, depth, mipmaps, type);

			const auto swizzle = get_component_mapping(gcm_format, flags);
			image->set_native_component_layout(swizzle);

			auto& cached = *find_cached_texture(rsx_range, gcm_format, true, true, width, height, depth, mipmaps);
			ASSERT(!cached.is_locked());

			// Prepare section
			cached.reset(rsx_range);
			cached.set_view_flags(flags);
			cached.set_context(context);
			cached.set_image_type(type);
			cached.set_gcm_format(gcm_format);
			cached.set_swizzled(swizzled);

			cached.create(width, height, depth, mipmaps, image, pitch, true);
			cached.set_dirty(false);

			if (context != rsx::texture_upload_context::blit_engine_dst)
			{
				AUDIT(cached.get_memory_read_flags() != rsx::memory_read_flags::flush_always);
				read_only_range = cached.get_min_max(read_only_range, rsx::section_bounds::locked_range); // TODO ruipin: This was outside the if, but is inside the if in Vulkan. Ask kd-11
				cached.protect(utils::protection::ro);
			}
			else
			{
				//TODO: More tests on byte order
				//ARGB8+native+unswizzled is confirmed with Dark Souls II character preview
				switch (gcm_format)
				{
				case CELL_GCM_TEXTURE_A8R8G8B8:
				{
					cached.set_format(gl::texture::format::bgra, gl::texture::type::uint_8_8_8_8, false);
					break;
				}
				case CELL_GCM_TEXTURE_R5G6B5:
				{
					cached.set_format(gl::texture::format::rgb, gl::texture::type::ushort_5_6_5, true);
					break;
				}
				case CELL_GCM_TEXTURE_DEPTH24_D8:
				{
					cached.set_format(gl::texture::format::depth_stencil, gl::texture::type::uint_24_8, false);
					break;
				}
				case CELL_GCM_TEXTURE_DEPTH16:
				{
					cached.set_format(gl::texture::format::depth, gl::texture::type::ushort, true);
					break;
				}
				default:
					fmt::throw_exception("Unexpected gcm format 0x%X" HERE, gcm_format);
				}

				//NOTE: Protection is handled by the caller
				cached.set_dimensions(width, height, depth, (rsx_range.length() / height));
				no_access_range = cached.get_min_max(no_access_range, rsx::section_bounds::locked_range);
			}

			update_cache_tag();
			return &cached;
		}

		cached_texture_section* create_nul_section(gl::command_context& /*cmd*/, const utils::address_range& rsx_range, bool /*memory_load*/) override
		{
			auto& cached = *find_cached_texture(rsx_range, RSX_GCM_FORMAT_IGNORED, true, false);
			ASSERT(!cached.is_locked());

			// Prepare section
			cached.reset(rsx_range);
			cached.set_context(rsx::texture_upload_context::dma);
			cached.set_dirty(false);

			no_access_range = cached.get_min_max(no_access_range, rsx::section_bounds::locked_range);
			update_cache_tag();
			return &cached;
		}

		cached_texture_section* upload_image_from_cpu(gl::command_context &cmd, const utils::address_range& rsx_range, u16 width, u16 height, u16 depth, u16 mipmaps, u16 pitch, u32 gcm_format,
			rsx::texture_upload_context context, const std::vector<rsx_subresource_layout>& subresource_layout, rsx::texture_dimension_extended type, bool input_swizzled) override
		{
			auto section = create_new_texture(cmd, rsx_range, width, height, depth, mipmaps, pitch, gcm_format, context, type, input_swizzled,
				rsx::texture_create_flags::default_component_order);

			gl::upload_texture(section->get_raw_texture()->id(), gcm_format, width, height, depth, mipmaps,
					input_swizzled, type, subresource_layout);

			section->last_write_tag = rsx::get_shared_tag();
			return section;
		}

		void enforce_surface_creation_type(cached_texture_section& section, u32 gcm_format, rsx::texture_create_flags flags) override
		{
			if (flags == section.get_view_flags())
				return;

			const auto swizzle = get_component_mapping(gcm_format, flags);
			auto image = static_cast<gl::viewable_image*>(section.get_raw_texture());

			verify(HERE), image != nullptr;
			image->set_native_component_layout(swizzle);

			section.set_view_flags(flags);
		}

		void insert_texture_barrier(gl::command_context&, gl::texture*) override
		{
			auto &caps = gl::get_driver_caps();

			if (caps.ARB_texture_barrier_supported)
				glTextureBarrier();
			else if (caps.NV_texture_barrier_supported)
				glTextureBarrierNV();
		}

		bool render_target_format_is_compatible(gl::texture* tex, u32 gcm_format) override
		{
			auto ifmt = tex->get_internal_format();
			switch (gcm_format)
			{
			default:
				//TODO
				warn_once("Format incompatibility detected, reporting failure to force data copy (GL_INTERNAL_FORMAT=0x%X, GCM_FORMAT=0x%X)", static_cast<u32>(ifmt), gcm_format);
				return false;
			case CELL_GCM_TEXTURE_W16_Z16_Y16_X16_FLOAT:
				return (ifmt == gl::texture::internal_format::rgba16f);
			case CELL_GCM_TEXTURE_W32_Z32_Y32_X32_FLOAT:
				return (ifmt == gl::texture::internal_format::rgba32f);
			case CELL_GCM_TEXTURE_X32_FLOAT:
				return (ifmt == gl::texture::internal_format::r32f);
			case CELL_GCM_TEXTURE_R5G6B5:
				return (ifmt == gl::texture::internal_format::rgb565);
			case CELL_GCM_TEXTURE_A8R8G8B8:
			case CELL_GCM_TEXTURE_D8R8G8B8:
				return (ifmt == gl::texture::internal_format::rgba8 ||
						ifmt == gl::texture::internal_format::depth24_stencil8 ||
						ifmt == gl::texture::internal_format::depth32f_stencil8);
			case CELL_GCM_TEXTURE_B8:
				return (ifmt == gl::texture::internal_format::r8);
			case CELL_GCM_TEXTURE_G8B8:
				return (ifmt == gl::texture::internal_format::rg8);
			case CELL_GCM_TEXTURE_DEPTH24_D8:
			case CELL_GCM_TEXTURE_DEPTH24_D8_FLOAT:
				return (ifmt == gl::texture::internal_format::depth24_stencil8 ||
						ifmt == gl::texture::internal_format::depth32f_stencil8 ||
						ifmt == gl::texture::internal_format::depth_stencil);
			case CELL_GCM_TEXTURE_X16:
			case CELL_GCM_TEXTURE_DEPTH16:
			case CELL_GCM_TEXTURE_DEPTH16_FLOAT:
				return (ifmt == gl::texture::internal_format::depth16 ||
						ifmt == gl::texture::internal_format::depth);
			}
		}

		void prepare_for_dma_transfers(gl::command_context&) override
		{}

		void cleanup_after_dma_transfers(gl::command_context&) override
		{}

	public:

		using baseclass::texture_cache;

		void initialize()
		{
			m_hw_blitter.init();
			g_hw_blitter = &m_hw_blitter;
		}

		void destroy() override
		{
			clear();
			g_hw_blitter = nullptr;
			m_hw_blitter.destroy();
		}

		bool is_depth_texture(u32 rsx_address, u32 rsx_size) override
		{
			reader_lock lock(m_cache_mutex);

			auto &block = m_storage.block_for(rsx_address);

			if (block.get_locked_count() == 0)
				return false;

			for (auto& tex : block)
			{
				if (tex.is_dirty())
					continue;

				if (!tex.overlaps(rsx_address, rsx::section_bounds::full_range))
					continue;

				if ((rsx_address + rsx_size - tex.get_section_base()) <= tex.get_section_size())
					return tex.is_depth_texture();
			}

			return false;
		}

		void on_frame_end() override
		{
			if (m_storage.m_unreleased_texture_objects >= m_max_zombie_objects)
			{
				purge_unreleased_sections();
			}

			clear_temporary_subresources();

			baseclass::on_frame_end();
		}

		bool blit(gl::command_context &cmd, rsx::blit_src_info& src, rsx::blit_dst_info& dst, bool linear_interpolate, gl_render_targets& m_rtts)
		{
			auto result = upload_scaled_image(src, dst, linear_interpolate, cmd, m_rtts, m_hw_blitter);

			if (result.succeeded)
			{
				if (result.real_dst_size)
				{
					flush_if_cache_miss_likely(cmd, result.to_address_range());
				}

				return true;
			}

			return false;
		}
	};
}