rpcs3/rpcs3/Emu/RSX/rsx_utils.h

#pragma once

#include "../system_config.h"
#include "Utilities/address_range.h"
#include "Utilities/geometry.h"
#include "gcm_enums.h"

#include <memory>
#include <bitset>
#include <chrono>

extern "C"
{
#include <libavutil/pixfmt.h>
}

#define RSX_SURFACE_DIMENSION_IGNORED 1

namespace rsx
{
	// Import address_range utilities
	using utils::address_range;
	using utils::address_range_vector;
	using utils::page_for;
	using utils::page_start;
	using utils::page_end;
	using utils::next_page;

	using flags64_t = u64;
	using flags32_t = u32;
	using flags16_t = u16;
	using flags8_t = u8;

	extern atomic_t<u64> g_rsx_shared_tag;

	enum class problem_severity : u8
	{
		low,
		moderate,
		severe,
		fatal
	};

	//Base for resources with reference counting
	class ref_counted
	{
	protected:
		atomic_t<s32> ref_count{ 0 }; // References held
		atomic_t<u8> idle_time{ 0 };  // Number of times the resource has been tagged idle

	public:
		void add_ref()
		{
			++ref_count;
			idle_time = 0;
		}

		void release()
		{
			--ref_count;
		}

		bool has_refs() const
		{
			return (ref_count > 0);
		}

		// Returns number of times the resource has been checked without being used in-between checks
		u8 unused_check_count()
		{
			if (ref_count)
			{
				return 0;
			}

			return idle_time++;
		}
	};

	namespace limits
	{
		enum
		{
			fragment_textures_count = 16,
			vertex_textures_count = 4,
			vertex_count = 16,
			fragment_count = 32,
			tiles_count = 15,
			zculls_count = 8,
			color_buffers_count = 4
		};
	}

	namespace constants
	{
		constexpr std::array<const char*, 16> fragment_texture_names =
		{
			"tex0", "tex1", "tex2", "tex3", "tex4", "tex5", "tex6", "tex7",
			"tex8", "tex9", "tex10", "tex11", "tex12", "tex13", "tex14", "tex15",
		};

		constexpr std::array<const char*, 4> vertex_texture_names =
		{
			"vtex0", "vtex1", "vtex2", "vtex3",
		};

		// Local RSX memory base (known as constant)
		constexpr u32 local_mem_base = 0xC0000000;
	}

	/**
	* Holds information about a framebuffer
	*/
	struct gcm_framebuffer_info
	{
		u32 address = 0;
		u32 pitch = 0;

		rsx::surface_color_format color_format;
		rsx::surface_depth_format2 depth_format;

		u16 width = 0;
		u16 height = 0;
		u8  bpp = 0;
		u8  samples = 0;

		address_range range{};

		gcm_framebuffer_info() = default;

		ENABLE_BITWISE_SERIALIZATION;

		void calculate_memory_range(u32 aa_factor_u, u32 aa_factor_v)
		{
			// Account for the last line of the block not reaching the end
			const u32 block_size = pitch * (height - 1) * aa_factor_v;
			const u32 line_size = width * aa_factor_u * bpp;
			range = address_range::start_length(address, block_size + line_size);
		}

		address_range get_memory_range(const u32* aa_factors)
		{
			calculate_memory_range(aa_factors[0], aa_factors[1]);
			return range;
		}

		address_range get_memory_range() const
		{
			ensure(range.start == address);
			return range;
		}
	};

	struct avconf
	{
		stereo_render_mode_options stereo_mode = stereo_render_mode_options::disabled;        // Stereo 3D display mode
		u8 format = 0;             // XRGB
		u8 aspect = 0;             // AUTO
		u8 resolution_id = 2;      // 720p
		u32 scanline_pitch = 0;    // PACKED
		atomic_t<f32> gamma = 1.f; // NO GAMMA CORRECTION
		u32 resolution_x = 1280;   // X RES
		u32 resolution_y = 720;    // Y RES
		atomic_t<u32> state = 0;   // 1 after cellVideoOutConfigure was called

		ENABLE_BITWISE_SERIALIZATION;
		SAVESTATE_INIT_POS(12);

		avconf() noexcept;
		~avconf() = default;
		avconf(utils::serial& ar);
		void save(utils::serial& ar);

		u32 get_compatible_gcm_format() const;
		u8 get_bpp() const;
		double get_aspect_ratio() const;

		areau aspect_convert_region(const size2u& image_dimensions, const size2u& output_dimensions) const;
		size2u aspect_convert_dimensions(const size2u& image_dimensions) const;
	};

	struct blit_src_info
	{
		blit_engine::transfer_source_format format;
		blit_engine::transfer_origin origin;
		u16 offset_x;
		u16 offset_y;
		u16 width;
		u16 height;
		u32 pitch;
		u32 rsx_address;
		void *pixels;
	};

	struct blit_dst_info
	{
		blit_engine::transfer_destination_format format;
		u16 offset_x;
		u16 offset_y;
		u16 width;
		u16 height;
		u16 clip_x;
		u16 clip_y;
		u16 clip_width;
		u16 clip_height;
		f32 scale_x;
		f32 scale_y;
		u32 pitch;
		u32 rsx_address;
		void *pixels;
		bool swizzled;
	};

	static const std::pair<std::array<u8, 4>, std::array<u8, 4>> default_remap_vector =
	{
		{ CELL_GCM_TEXTURE_REMAP_FROM_A, CELL_GCM_TEXTURE_REMAP_FROM_R, CELL_GCM_TEXTURE_REMAP_FROM_G, CELL_GCM_TEXTURE_REMAP_FROM_B },
		{ CELL_GCM_TEXTURE_REMAP_REMAP, CELL_GCM_TEXTURE_REMAP_REMAP, CELL_GCM_TEXTURE_REMAP_REMAP, CELL_GCM_TEXTURE_REMAP_REMAP }
	};

	static inline std::pair<std::array<u8, 4>, std::array<u8, 4>> decode_remap_encoding(u32 remap_ctl)
	{
		// Remapping tables; format is A-R-G-B
		// Remap input table. Contains channel index to read color from
		const std::array<u8, 4> remap_inputs =
		{
			static_cast<u8>(remap_ctl & 0x3),
			static_cast<u8>((remap_ctl >> 2) & 0x3),
			static_cast<u8>((remap_ctl >> 4) & 0x3),
			static_cast<u8>((remap_ctl >> 6) & 0x3),
		};

		// Remap control table. Controls whether the remap value is used, or force either 0 or 1
		const std::array<u8, 4> remap_lookup =
		{
			static_cast<u8>((remap_ctl >> 8) & 0x3),
			static_cast<u8>((remap_ctl >> 10) & 0x3),
			static_cast<u8>((remap_ctl >> 12) & 0x3),
			static_cast<u8>((remap_ctl >> 14) & 0x3),
		};

		return std::make_pair(remap_inputs, remap_lookup);
	}

	template <typename T>
	void pad_texture(void* input_pixels, void* output_pixels, u16 input_width, u16 input_height, u16 output_width, u16 /*output_height*/)
	{
		T *src = static_cast<T*>(input_pixels);
		T *dst = static_cast<T*>(output_pixels);

		for (u16 h = 0; h < input_height; ++h)
		{
			const u32 padded_pos = h * output_width;
			const u32 pos = h * input_width;
			for (u16 w = 0; w < input_width; ++w)
			{
				dst[padded_pos + w] = src[pos + w];
			}
		}
	}

	static constexpr u32 floor_log2(u32 value)
	{
		return value <= 1 ? 0 : std::countl_zero(value) ^ 31;
	}

	static constexpr u32 ceil_log2(u32 value)
	{
		return floor_log2(value) + u32{!!(value & (value - 1))};
	}

	static constexpr u32 next_pow2(u32 x)
	{
		if (x <= 2) return x;

		return static_cast<u32>((1ULL << 32) >> std::countl_zero(x - 1));
	}

	static inline bool fcmp(float a, float b, float epsilon = 0.000001f)
	{
		return fabsf(a - b) < epsilon;
	}

	// Returns an ever-increasing tag value
	static inline u64 get_shared_tag()
	{
		return g_rsx_shared_tag++;
	}

	static inline u32 get_location(u32 addr)
	{
		return (addr >= rsx::constants::local_mem_base) ?
			CELL_GCM_LOCATION_LOCAL :
			CELL_GCM_LOCATION_MAIN;
	}

	// General purpose alignment without power-of-2 constraint
	template <typename T, typename U>
	static inline T align2(T value, U alignment)
	{
		return ((value + alignment - 1) / alignment) * alignment;
	}

	// Copy memory in inverse direction from source
	// Used to scale negatively x axis while transfering image data
	template <typename Ts = u8, typename Td = Ts>
	static void memcpy_r(void* dst, void* src, usz size)
	{
		for (u32 i = 0; i < size; i++)
		{
			*(static_cast<Td*>(dst) + i) = *(static_cast<Ts*>(src) - i);
		}
	}

	// Returns interleaved bits of X|Y|Z used as Z-order curve indices
	static inline u32 calculate_z_index(u32 x, u32 y, u32 z, u32 log2_width, u32 log2_height, u32 log2_depth)
	{
		AUDIT(x < (1u << log2_width) && y < (1u << log2_height) && z < (1u << log2_depth));

		// offset = X' | Y' | Z' which are x,y,z bits interleaved
		u32 offset = 0;
		u32 shift_count = 0;
		do
		{
			if (log2_width)
			{
				offset |= (x & 0x1) << shift_count++;
				x >>= 1;
				log2_width--;
			}

			if (log2_height)
			{
				offset |= (y & 0x1) << shift_count++;
				y >>= 1;
				log2_height--;
			}

			if (log2_depth)
			{
				offset |= (z & 0x1) << shift_count++;
				z >>= 1;
				log2_depth--;
			}
		}
		while (x | y | z);

		return offset;
	}

	/*   Note: What the ps3 calls swizzling in this case is actually z-ordering / morton ordering of pixels
	*       - Input can be swizzled or linear, bool flag handles conversion to and from
	*       - It will handle any width and height that are a power of 2, square or non square
	*    Restriction: It has mixed results if the height or width is not a power of 2
	*    Restriction: Only works with 2D surfaces
	*/
	template <typename T, bool input_is_swizzled>
	void convert_linear_swizzle(const void* input_pixels, void* output_pixels, u16 width, u16 height, u32 pitch)
	{
		u32 log2width = ceil_log2(width);
		u32 log2height = ceil_log2(height);

		// Max mask possible for square texture
		u32 x_mask = 0x55555555;
		u32 y_mask = 0xAAAAAAAA;

		// We have to limit the masks to the lower of the two dimensions to allow for non-square textures
		u32 limit_mask = (log2width < log2height) ? log2width : log2height;
		// double the limit mask to account for bits in both x and y
		limit_mask = 1 << (limit_mask << 1);

		//x_mask, bits above limit are 1's for x-carry
		x_mask = (x_mask | ~(limit_mask - 1));
		//y_mask. bits above limit are 0'd, as we use a different method for y-carry over
		y_mask = (y_mask & (limit_mask - 1));

		u32 offs_y = 0;
		u32 offs_x = 0;
		u32 offs_x0 = 0; //total y-carry offset for x
		u32 y_incr = limit_mask;

		u32 adv = pitch / sizeof(T);

		if constexpr (!input_is_swizzled)
		{
			for (int y = 0; y < height; ++y)
			{
				auto src = static_cast<const T*>(input_pixels) + y * adv;
				auto dst = static_cast<T*>(output_pixels) + offs_y;
				offs_x = offs_x0;

				for (int x = 0; x < width; ++x)
				{
					dst[offs_x] = src[x];
					offs_x = (offs_x - x_mask) & x_mask;
				}

				offs_y = (offs_y - y_mask) & y_mask;

				if (offs_y == 0)
				{
					offs_x0 += y_incr;
				}
			}
		}
		else
		{
			for (int y = 0; y < height; ++y)
			{
				auto src = static_cast<const T*>(input_pixels) + offs_y;
				auto dst = static_cast<T*>(output_pixels) + y * adv;
				offs_x = offs_x0;

				for (int x = 0; x < width; ++x)
				{
					dst[x] = src[offs_x];
					offs_x = (offs_x - x_mask) & x_mask;
				}

				offs_y = (offs_y - y_mask) & y_mask;

				if (offs_y == 0)
				{
					offs_x0 += y_incr;
				}
			}
		}
	}

	/**
	 * Write swizzled data to linear memory with support for 3 dimensions
	 * Z ordering is done in all 3 planes independently with a unit being a 2x2 block per-plane
	 * A unit in 3d textures is a group of 2x2x2 texels advancing towards depth in units of 2x2x1 blocks
	 * i.e 32 texels per "unit"
	 */
	template <typename T>
	void convert_linear_swizzle_3d(const void* input_pixels, void* output_pixels, u16 width, u16 height, u16 depth)
	{
		if (depth == 1)
		{
			convert_linear_swizzle<T, true>(input_pixels, output_pixels, width, height, width * sizeof(T));
			return;
		}

		auto src = static_cast<const T*>(input_pixels);
		auto dst = static_cast<T*>(output_pixels);

		const u32 log2_w = ceil_log2(width);
		const u32 log2_h = ceil_log2(height);
		const u32 log2_d = ceil_log2(depth);

		for (u32 z = 0; z < depth; ++z)
		{
			for (u32 y = 0; y < height; ++y)
			{
				for (u32 x = 0; x < width; ++x)
				{
					*dst++ = src[calculate_z_index(x, y, z, log2_w, log2_h, log2_d)];
				}
			}
		}
	}

	void convert_scale_image(u8 *dst, AVPixelFormat dst_format, int dst_width, int dst_height, int dst_pitch,
		const u8 *src, AVPixelFormat src_format, int src_width, int src_height, int src_pitch, int src_slice_h, bool bilinear);

	void clip_image(u8 *dst, const u8 *src, int clip_x, int clip_y, int clip_w, int clip_h, int bpp, int src_pitch, int dst_pitch);
	void clip_image_may_overlap(u8 *dst, const u8 *src, int clip_x, int clip_y, int clip_w, int clip_h, int bpp, int src_pitch, int dst_pitch, u8* buffer);

	std::array<float, 4> get_constant_blend_colors();

	/**
	 * Shuffle texel layout from xyzw to wzyx
	 * TODO: Variable src/dst and optional se conversion
	 */
	template <typename T>
	void shuffle_texel_data_wzyx(void* data, u32 row_pitch_in_bytes, u16 row_length_in_texels, u16 num_rows)
	{
		char* raw_src = static_cast<char*>(data);
		T tmp[4];

		for (u16 n = 0; n < num_rows; ++n)
		{
			T* src = reinterpret_cast<T*>(raw_src);
			raw_src += row_pitch_in_bytes;

			for (u16 m = 0; m < row_length_in_texels; ++m)
			{
				tmp[0] = src[3];
				tmp[1] = src[2];
				tmp[2] = src[1];
				tmp[3] = src[0];

				src[0] = tmp[0];
				src[1] = tmp[1];
				src[2] = tmp[2];
				src[3] = tmp[3];

				src += 4;
			}
		}
	}

	/**
	 * Clips a rect so that it never falls outside the parent region
	 * attempt_fit: allows resizing of the requested region. If false, failure to fit will result in the child rect being pinned to (0, 0)
	 */
	template <typename T>
	std::tuple<T, T, T, T> clip_region(T parent_width, T parent_height, T clip_x, T clip_y, T clip_width, T clip_height, bool attempt_fit)
	{
		T x = clip_x;
		T y = clip_y;
		T width = clip_width;
		T height = clip_height;

		if ((clip_x + clip_width) > parent_width)
		{
			if (clip_x >= parent_width)
			{
				if (clip_width < parent_width)
					width = clip_width;
				else
					width = parent_width;

				x = static_cast<T>(0);
			}
			else
			{
				if (attempt_fit)
					width = parent_width - clip_x;
				else
					width = std::min(clip_width, parent_width);
			}
		}

		if ((clip_y + clip_height) > parent_height)
		{
			if (clip_y >= parent_height)
			{
				if (clip_height < parent_height)
					height = clip_height;
				else
					height = parent_height;

				y = static_cast<T>(0);
			}
			else
			{
				if (attempt_fit)
					height = parent_height - clip_y;
				else
					height = std::min(clip_height, parent_height);
			}
		}

		return std::make_tuple(x, y, width, height);
	}

	/**
	 * Extracts from 'parent' a region that fits in 'child'
	 */
	static inline std::tuple<position2u, position2u, size2u> intersect_region(
		u32 parent_address, u16 parent_w, u16 parent_h,
		u32 child_address, u16 child_w, u16 child_h,
		u32 pitch)
	{
		if (child_address < parent_address)
		{
			const auto offset = parent_address - child_address;
			const auto src_x = 0u;
			const auto src_y = 0u;
			const auto dst_y = (offset / pitch);
			const auto dst_x = (offset % pitch);
			const auto w = std::min<u32>(parent_w, std::max<u32>(child_w, dst_x) - dst_x); // Clamp negatives to 0!
			const auto h = std::min<u32>(parent_h, std::max<u32>(child_h, dst_y) - dst_y);

			return std::make_tuple<position2u, position2u, size2u>({ src_x, src_y }, { dst_x, dst_y }, { w, h });
		}
		else
		{
			const auto offset = child_address - parent_address;
			const auto src_y = (offset / pitch);
			const auto src_x = (offset % pitch);
			const auto dst_x = 0u;
			const auto dst_y = 0u;
			const auto w = std::min<u32>(child_w, std::max<u32>(parent_w, src_x) - src_x);
			const auto h = std::min<u32>(child_h, std::max<u32>(parent_h, src_y) - src_y);

			return std::make_tuple<position2u, position2u, size2u>({ src_x, src_y }, { dst_x, dst_y }, { w, h });
		}
	}

	static inline f32 get_resolution_scale()
	{
		return g_cfg.video.strict_rendering_mode ? 1.f : (g_cfg.video.resolution_scale_percent / 100.f);
	}

	static inline int get_resolution_scale_percent()
	{
		return g_cfg.video.strict_rendering_mode ? 100 : g_cfg.video.resolution_scale_percent;
	}

	template <bool clamp = false>
	static inline const std::pair<u16, u16> apply_resolution_scale(u16 width, u16 height, u16 ref_width = 0, u16 ref_height = 0)
	{
		ref_width = (ref_width)? ref_width : width;
		ref_height = (ref_height)? ref_height : height;
		const u16 ref = std::max(ref_width, ref_height);

		if (ref > g_cfg.video.min_scalable_dimension)
		{
			// Upscale both width and height
			width = (get_resolution_scale_percent() * width) / 100;
			height = (get_resolution_scale_percent() * height) / 100;

			if constexpr (clamp)
			{
				width = std::max<u16>(width, 1);
				height = std::max<u16>(height, 1);
			}
		}

		return { width, height };
	}

	template <bool clamp = false>
	static inline const std::pair<u16, u16> apply_inverse_resolution_scale(u16 width, u16 height)
	{
		// Inverse scale
		auto width_ = (width * 100) / get_resolution_scale_percent();
		auto height_ = (height * 100) / get_resolution_scale_percent();

		if constexpr (clamp)
		{
			width_ = std::max<u16>(width_, 1);
			height_ = std::max<u16>(height_, 1);
		}

		if (std::max(width_, height_) > g_cfg.video.min_scalable_dimension)
		{
			return { width_, height_ };
		}

		return { width, height };
	}

	/**
	 * Calculates the regions used for memory transfer between rendertargets on succession events
	 * Returns <src_w, src_h, dst_w, dst_h>
	 */
	template <typename SurfaceType>
	std::tuple<u16, u16, u16, u16> get_transferable_region(const SurfaceType* surface)
	{
		auto src = static_cast<const SurfaceType*>(surface->old_contents[0].source);
		auto area1 = src->get_normalized_memory_area();
		auto area2 = surface->get_normalized_memory_area();

		auto w = std::min(area1.x2, area2.x2);
		auto h = std::min(area1.y2, area2.y2);

		const auto src_scale_x = src->get_bpp() * src->samples_x;
		const auto src_scale_y = src->samples_y;
		const auto dst_scale_x = surface->get_bpp() * surface->samples_x;
		const auto dst_scale_y = surface->samples_y;

		const u16 src_w = u16(w / src_scale_x);
		const u16 src_h = u16(h / src_scale_y);
		const u16 dst_w = u16(w / dst_scale_x);
		const u16 dst_h = u16(h / dst_scale_y);

		return std::make_tuple(src_w, src_h, dst_w, dst_h);
	}

	template <typename SurfaceType>
	inline bool pitch_compatible(const SurfaceType* a, const SurfaceType* b)
	{
		if (a->get_surface_height() == 1 || b->get_surface_height() == 1)
			return true;

		return (a->get_rsx_pitch() == b->get_rsx_pitch());
	}

	template <bool __is_surface = true, typename SurfaceType>
	inline bool pitch_compatible(const SurfaceType* surface, u32 pitch_required, u16 height_required)
	{
		if constexpr (__is_surface)
		{
			if (height_required == 1 || surface->get_surface_height() == 1)
				return true;
		}
		else
		{
			if (height_required == 1 || surface->get_height() == 1)
				return true;
		}

		return (surface->get_rsx_pitch() == pitch_required);
	}

	/**
	 * Remove restart index and emulate using degenerate triangles
	 * Can be used as a workaround when restart_index doesnt work too well
	 * dst should be able to hold at least 2xcount entries
	 */
	template <typename T>
	u32 remove_restart_index(T* dst, T* src, int count, T restart_index)
	{
		// Converts a stream e.g [1, 2, 3, -1, 4, 5, 6] to a stream with degenerate splits
		// Output is e.g [1, 2, 3, 3, 3, 4, 4, 5, 6] (5 bogus triangles)
		T last_index{}, index;
		u32 dst_index = 0;
		for (int n = 0; n < count;)
		{
			index = src[n];
			if (index == restart_index)
			{
				for (; n < count; ++n)
				{
					if (src[n] != restart_index)
						break;
				}

				if (n == count)
					return dst_index;

				dst[dst_index++] = last_index; //Duplicate last

				if ((dst_index & 1) == 0)
					//Duplicate last again to fix face winding
					dst[dst_index++] = last_index;

				last_index = src[n];
				dst[dst_index++] = last_index; //Duplicate next
			}
			else
			{
				dst[dst_index++] = index;
				last_index = index;
				++n;
			}
		}

		return dst_index;
	}

	// The rsx internally adds the 'data_base_offset' and the 'vert_offset' and masks it
	// before actually attempting to translate to the internal address. Seen happening heavily in R&C games
	static inline u32 get_vertex_offset_from_base(u32 vert_data_base_offset, u32 vert_base_offset)
	{
		return (vert_data_base_offset + vert_base_offset) & 0xFFFFFFF;
	}

	// Similar to vertex_offset_base calculation, the rsx internally adds and masks index
	// before using
	static inline u32 get_index_from_base(u32 index, u32 index_base)
	{
		return (index + index_base) & 0x000FFFFF;
	}

	// Convert color write mask for G8B8 to R8G8
	static inline u32 get_g8b8_r8g8_clearmask(u32 mask)
	{
		u32 result = 0;
		if (mask & RSX_GCM_CLEAR_GREEN_BIT) result |= RSX_GCM_CLEAR_GREEN_BIT;
		if (mask & RSX_GCM_CLEAR_BLUE_BIT) result |= RSX_GCM_CLEAR_RED_BIT;

		return result;
	}

	static inline void get_g8b8_r8g8_colormask(bool &red, bool &/*green*/, bool &blue, bool &alpha)
	{
		red = blue;
		blue = false;
		alpha = false;
	}

	static inline void get_g8b8_clear_color(u8& red, u8& /*green*/, u8& blue, u8& /*alpha*/)
	{
		red = blue;
	}

	static inline u32 get_abgr8_clearmask(u32 mask)
	{
		u32 result = 0;
		if (mask & RSX_GCM_CLEAR_RED_BIT) result |= RSX_GCM_CLEAR_BLUE_BIT;
		if (mask & RSX_GCM_CLEAR_GREEN_BIT) result |= RSX_GCM_CLEAR_GREEN_BIT;
		if (mask & RSX_GCM_CLEAR_BLUE_BIT) result |= RSX_GCM_CLEAR_RED_BIT;
		if (mask & RSX_GCM_CLEAR_ALPHA_BIT) result |= RSX_GCM_CLEAR_ALPHA_BIT;
		return result;
	}

	static inline void get_abgr8_colormask(bool& red, bool& /*green*/, bool& blue, bool& /*alpha*/)
	{
		std::swap(red, blue);
	}

	static inline void get_abgr8_clear_color(u8& red, u8& /*green*/, u8& blue, u8& /*alpha*/)
	{
		std::swap(red, blue);
	}

	template <typename T, typename U>
		requires std::is_integral_v<T> && std::is_integral_v<U>
	u8 renormalize_color8(T input, U base)
	{
		// Base will be some POT-1 value
		const int value = static_cast<u8>(input & base);
		return static_cast<u8>((value * 255) / base);
	}

	static inline void get_rgb565_clear_color(u8& red, u8& green, u8& blue, u8& /*alpha*/)
	{
		// RSX clear color is just a memcpy, so in this case the input is ARGB8 so only BG have the 16-bit input
		const u16 raw_value = static_cast<u16>(green) << 8 | blue;
		blue = renormalize_color8(raw_value, 0x1f);
		green = renormalize_color8(raw_value >> 5, 0x3f);
		red = renormalize_color8(raw_value >> 11, 0x1f);
	}

	static inline void get_a1rgb555_clear_color(u8& red, u8& green, u8& blue, u8& alpha, u8 alpha_override)
	{
		// RSX clear color is just a memcpy, so in this case the input is ARGB8 so only BG have the 16-bit input
		const u16 raw_value = static_cast<u16>(green) << 8 | blue;
		blue = renormalize_color8(raw_value, 0x1f);
		green = renormalize_color8(raw_value >> 5, 0x1f);
		red = renormalize_color8(raw_value >> 10, 0x1f);

		// Alpha can technically be encoded into the clear but the format normally just injects constants.
		// Will require hardware tests when possible to determine which approach makes more sense.
		// alpha = static_cast<u8>((raw_value & (1 << 15)) ? 255 : 0);
		alpha = alpha_override;
	}

	static inline u32 get_b8_clearmask(u32 mask)
	{
		u32 result = 0;
		if (mask & RSX_GCM_CLEAR_BLUE_BIT) result |= RSX_GCM_CLEAR_RED_BIT;
		return result;
	}

	static inline void get_b8_colormask(bool& red, bool& green, bool& blue, bool& alpha)
	{
		red = blue;
		green = false;
		blue = false;
		alpha = false;
	}

	static inline void get_b8_clear_color(u8& red, u8& /*green*/, u8& blue, u8& /*alpha*/)
	{
		std::swap(red, blue);
	}

	static inline color4f decode_border_color(u32 colorref)
	{
		color4f result;
		result.b = (colorref & 0xFF) / 255.f;
		result.g = ((colorref >> 8) & 0xFF) / 255.f;
		result.r = ((colorref >> 16) & 0xFF) / 255.f;
		result.a = ((colorref >> 24) & 0xFF) / 255.f;
		return result;
	}

	template <uint integer, uint frac, bool sign = true, typename To = f32>
	static inline To decode_fxp(u32 bits)
	{
		static_assert(u64{sign} + integer + frac <= 32, "Invalid decode_fxp range");

		// Classic fixed point, see PGRAPH section of nouveau docs for TEX_FILTER (lod_bias) and TEX_CONTROL (min_lod, max_lod)
		// Technically min/max lod are fixed 4.8 but a 5.8 decoder should work just as well since sign bit is 0

		if constexpr (sign) if (bits & (1 << (integer + frac)))
		{
			bits = (0 - bits) & (~0u >> (31 - (integer + frac)));
			return bits / (-To(1u << frac));
		}

		return bits / To(1u << frac);
	}

	static inline f32 decode_fp16(u16 bits)
	{
		if (bits == 0)
		{
			return 0.f;
		}

		// Extract components
		unsigned int sign = (bits >> 15) & 1;
		unsigned int exp = (bits >> 10) & 0x1f;
		unsigned int mantissa = bits & 0x3ff;

		float base = (sign != 0) ? -1.f : 1.f;
		float scale;

		if (exp == 0x1F)
		{
			// specials (nan, inf)
			u32 nan = 0x7F800000 | mantissa;
			nan |= (sign << 31);
			return std::bit_cast<f32>(nan);
		}
		else if (exp > 0)
		{
			// normal number, borrows a '1' from the hidden mantissa bit
			base *= std::exp2f(f32(exp) - 15.f);
			scale = (float(mantissa) / 1024.f) + 1.f;
		}
		else
		{
			// subnormal number, borrows a '0' from the hidden mantissa bit
			base *= std::exp2f(1.f - 15.f);
			scale = float(mantissa) / 1024.f;
		}

		return base * scale;
	}

	template<bool _signed = false>
	u16 encode_fx12(f32 value)
	{
		u16 raw = u16(std::abs(value) * 256.);

		if constexpr (!_signed)
		{
			return raw;
		}
		else
		{
			if (value >= 0.f) [[likely]]
			{
				return raw;
			}
			else
			{
				return u16(0 - raw) & 0x1fff;
			}
		}
	}
}