/*
 * Copyright (c) 2020 Actions Technology Co., Ltd
 *
 * SPDX-License-Identifier: Apache-2.0
 */

#include <display/sw_draw.h>
#include <display/sw_rotate.h>
#include <assert.h>
#ifdef CONFIG_GUI_API_BROM
#  include <brom_interface.h>
#endif

#ifndef MAX
#define MAX(a, b) (((a) > (b)) ? (a) : (b))
#endif

#ifndef MIN
#define MIN(a, b) (((a) < (b)) ? (a) : (b))
#endif

/*
 * compute x range in pixels inside the src image
 *
 * The routine will update the range value [*x_min, *x_max].
 *
 * @param x_min address of minimum range value in pixels
 * @param x_max address of maximum range value in pixels
 * @param img_w source image width in pixels
 * @param img_h source image height in pixels
 * @param start_x X coord of start point in fixedpoint-16, corresponding to the original *x_min value
 * @param start_y Y coord of start point in fixedpoint-16, corresponding to the original *x_min value
 * @param dx_x X coord of point delta in x direction in fixedpoint-16
 * @param dx_y Y coord of point delta in x direction in fixedpoint-16
 *
 * @return N/A
 */
static inline void sw_transform_compoute_x_range(
		int32_t *x_min, int32_t *x_max, int16_t img_w, int16_t img_h,
		int32_t start_x, int32_t start_y, int32_t dx_x, int32_t dx_y)
{
	const int32_t img_w_m1 = FIXEDPOINT16(img_w - 1);
	const int32_t img_h_m1 = FIXEDPOINT16(img_h - 1);
	int x_1, x_2;

	/* FIXME: the compiler seems to divide towards to zero. */
	if (dx_x != 0) {
		/*
			* floor(Δx * dx_x + start_x) >= FIXEDPOINT16(0)
			* ceil(Δx * dx_x + start_x) <= img_w_m1
			*/
		if (dx_x > 0) {
			x_1 = (FIXEDPOINT16(0) - start_x + dx_x - 1) / dx_x;
			x_2 = (img_w_m1 - start_x) / dx_x;
		} else {
			x_2 = (FIXEDPOINT16(0) - start_x) / dx_x;
			x_1 = (img_w_m1 - start_x + dx_x + 1) / dx_x;
		}

		*x_min = MAX(*x_min, x_1);
		*x_max = MIN(*x_max, x_2);
	} else if (start_x < FIXEDPOINT16(0) || start_x > img_w_m1) {
		*x_max = *x_min - 1;
		return;
	}

	if (dx_y != 0) {
		/*
			* floor(Δy * dx_y + start_y) >= FIXEDPOINT16(0)
			* ceil(Δy * dx_y + start_y) <= img_h_m1
			*/
		if (dx_y > 0) {
			x_1 = (FIXEDPOINT16(0) - start_y + dx_y - 1) / dx_y;
			x_2 = (img_h_m1 - start_y) / dx_y;
		} else {
			x_2 = (FIXEDPOINT16(0) - start_y) / dx_y;
			x_1 = (img_h_m1 - start_y + dx_y + 1) / dx_y;
		}

		*x_min = MAX(*x_min, x_1);
		*x_max = MIN(*x_max, x_2);
	} else if (start_y < FIXEDPOINT16(0) || start_y > img_h_m1) {
		*x_max = *x_min - 1;
		return;
	}
}

void sw_transform_config(int16_t img_x, int16_t img_y,
		int16_t pivot_x, int16_t pivot_y, uint16_t angle,
		uint16_t scale_x, uint16_t scale_y, uint16_t scale_bits,
		sw_matrix_t *matrix)
{
	const uint16_t scale_factor = (1 << scale_bits);
	uint16_t revert_scale_x;
	uint16_t revert_scale_y;
	uint16_t revert_angle;

	assert(angle <= 3600 && scale_x > 0 && scale_y > 0);

	revert_scale_x = (scale_x > 0) ? (scale_factor * scale_factor / scale_x) : 1;
	revert_scale_y = (scale_x == scale_y) ? revert_scale_x :
			((scale_y > 0) ? (scale_factor * scale_factor / scale_y) : 1);

#ifdef CONFIG_GUI_API_BROM_LEOPARD
	if (revert_scale_x == revert_scale_y) {
		p_brom_libgui_api->p_sw_transform_config(img_x, img_y,
				pivot_x, pivot_y, angle, revert_scale_x, revert_scale_y, scale_bits, matrix);
		return;
	}
#endif

	revert_angle = 3600 - angle;

	/* coordinates in the destination coordinate system */
	matrix->tx = PX_FIXEDPOINT16(0);
	matrix->ty = PX_FIXEDPOINT16(0);
	matrix->sx = FIXEDPOINT16(1);
	matrix->shy = FIXEDPOINT16(0);
	matrix->shx = FIXEDPOINT16(0);
	matrix->sy = FIXEDPOINT16(1);

	/* transform back to the source coordinate system (rotate -> scaling) */
	sw_transform_point32_rot_first(&matrix->tx, &matrix->ty, matrix->tx, matrix->ty,
			PX_FIXEDPOINT16(img_x + pivot_x), PX_FIXEDPOINT16(img_y + pivot_y),
			revert_angle, revert_scale_x, revert_scale_y, scale_bits);
	sw_transform_point32_rot_first(&matrix->sx, &matrix->shy, matrix->sx, matrix->shy,
			FIXEDPOINT16(0), FIXEDPOINT16(0), revert_angle,
			revert_scale_x, revert_scale_y, scale_bits);
	sw_transform_point32_rot_first(&matrix->shx, &matrix->sy, matrix->shx, matrix->sy,
			FIXEDPOINT16(0), FIXEDPOINT16(0), revert_angle,
			revert_scale_x, revert_scale_y, scale_bits);

	/* map to the source pixel coordinate system */
	matrix->tx -= PX_FIXEDPOINT16(img_x);
	matrix->ty -= PX_FIXEDPOINT16(img_y);
}

void sw_transform_config_with_mode(uint16_t img_w, uint16_t img_h,
		uint8_t mode, sw_matrix_t *matrix)
{
	switch (mode) {
	case SW_ROT_90:
		*matrix = (sw_matrix_t) {
			.sx = 0, .shx = FIXEDPOINT16(1), .tx = 0,
			.shy = FIXEDPOINT16(-1), .sy = 0, .ty = FIXEDPOINT16(img_h - 1),
		};
		break;
	case SW_ROT_180:
		*matrix = (sw_matrix_t) {
			.sx = FIXEDPOINT16(-1), .shx = 0, .tx = FIXEDPOINT16(img_w - 1),
			.shy = 0, .sy = FIXEDPOINT16(-1), .ty = FIXEDPOINT16(img_h - 1),
		};
		break;
	case SW_ROT_270:
		*matrix = (sw_matrix_t) {
			.sx = 0, .shx = FIXEDPOINT16(-1), .tx = FIXEDPOINT16(img_w - 1),
			.shy = FIXEDPOINT16(1), .sy = 0, .ty = 0,
		};
		break;
	case SW_FLIP_H:
		*matrix = (sw_matrix_t) {
			.sx = FIXEDPOINT16(-1), .shx = 0, .tx = FIXEDPOINT16(img_w - 1),
			.shy = 0, .sy = FIXEDPOINT16(1), .ty = 0,
		};
		break;
	case SW_FLIP_V:
		*matrix = (sw_matrix_t) {
			.sx = FIXEDPOINT16(1), .shx = 0, .tx = 0,
			.shy = 0, .sy = FIXEDPOINT16(-1), .ty = FIXEDPOINT16(img_h - 1),
		};
		break;
	default:
		*matrix = (sw_matrix_t) {
			.sx = FIXEDPOINT16(1), .shx = 0, .tx = 0,
			.shy = 0, .sy = FIXEDPOINT16(1), .ty = 0,
		};
		break;
	}
}

void sw_transform_rgb565_over_rgb565(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
#ifdef CONFIG_GUI_API_BROM_LARK
	if (src_pitch == src_w * 2) {
		p_brom_libgui_api->p_sw_rotate_rgb565_over_rgb565(
				dst, src, dst_pitch / 2, src_w, src_h, x, y, w, h, matrix);
		return;
	}
#endif

#ifdef CONFIG_GUI_API_BROM_LEOPARD
	p_brom_libgui_api->p_sw_transform_rgb565_over_rgb565(
			dst, src, dst_pitch, src_pitch, src_w, src_h, x, y, w, h, matrix);
#else
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 2;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			*tmp_dst = bilinear_rgb565_fast_m6(*(uint16_t*)src1,
					*(uint16_t*)src2, *(uint16_t*)src3, *(uint16_t*)src4,
					x_frac >> 10, y_frac >> 10, 6);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
#endif /* CONFIG_GUI_API_BROM_LEOPARD */
}

void sw_transform_rgb565_over_rgb888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	uint16_t src_bytes_per_pixel = 2;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint16_t c16;

			c16 = bilinear_rgb565_fast_m6(*(uint16_t*)src1,
					*(uint16_t*)src2, *(uint16_t*)src3, *(uint16_t*)src4,
					x_frac >> 10, y_frac >> 10, 6);

			*tmp_dst++ = ((c16 & 0x1f) << 3) | (c16 & 0x07);
			*tmp_dst++ = ((c16 & 0x07e0) >> 3) | ((c16 & 0x60) >> 5);
			*tmp_dst++ = ((c16 & 0xf800) >> 8) | ((c16 & 0x3800) >> 11);

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_rgb565_over_argb8888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
//#ifdef CONFIG_GUI_API_BROM_LEOPARD
//	p_brom_libgui_api->p_sw_transform_rgb565_over_argb8888(
//			dst, src, dst_pitch, src_pitch, src_w, src_h, x, y, w, h, matrix);
//#else
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 2;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint16_t c16;

			c16 = bilinear_rgb565_fast_m6(*(uint16_t*)src1,
					*(uint16_t*)src2, *(uint16_t*)src3, *(uint16_t*)src4,
					x_frac >> 10, y_frac >> 10, 6);

			*tmp_dst = ((c16 & 0x1f) << 3) | (c16 & 0x07) |
					((c16 & 0x07e0) << 5) | ((c16 & 0x60) << 3) |
					((c16 & 0xf800) << 8) | ((c16 & 0x3800) << 5) | 0xFF000000;

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
//#endif /* CONFIG_GUI_API_BROM_LEOPARD */
}

void sw_transform_rgb565a8_over_rgb565(void *dst, const void *src, const void *src_opa,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_opa_pitch,
		uint16_t src_w, uint16_t src_h, int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 2;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint8_t *src_a1 = (uint8_t *)src_opa + y * src_opa_pitch + x;
			uint8_t *src_a2 = src1 + 1;
			uint8_t *src_a3 = src1 + src_opa_pitch;
			uint8_t *src_a4 = src2 + src_opa_pitch;
			sw_color16a8_t col24;

			col24.rgb = bilinear_rgb565_fast_m6(*(uint16_t*)src1,
					*(uint16_t*)src2, *(uint16_t*)src3, *(uint16_t*)src4,
					x_frac >> 10, y_frac >> 10, 6);
			col24.a = bilinear_a8_fast_m8(*src_a1, *src_a2, *src_a3, *src_a4,
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = blend_argb8565_over_rgb565(*tmp_dst, (uint8_t *)&col24);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_rgb565a8_over_rgb888(void *dst, const void *src, const void *src_opa,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_opa_pitch,
		uint16_t src_w, uint16_t src_h, int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	uint16_t src_bytes_per_pixel = 2;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint8_t *src_a1 = (uint8_t *)src_opa + y * src_opa_pitch + x;
			uint8_t *src_a2 = src1 + 1;
			uint8_t *src_a3 = src1 + src_opa_pitch;
			uint8_t *src_a4 = src2 + src_opa_pitch;
			sw_color32_t col32 = {
				.a = 255,
				.r = tmp_dst[2],
				.g = tmp_dst[1],
				.b = tmp_dst[0],
			};
			sw_color16a8_t col24;

			col24.rgb = bilinear_rgb565_fast_m6(*(uint16_t*)src1,
					*(uint16_t*)src2, *(uint16_t*)src3, *(uint16_t*)src4,
					x_frac >> 10, y_frac >> 10, 6);
			col24.a = bilinear_a8_fast_m8(*src_a1, *src_a2, *src_a3, *src_a4,
					x_frac >> 8, y_frac >> 8, 8);

			col32.full = blend_argb8565_over_argb8888(col32.full, (uint8_t *)&col24);
			*tmp_dst++ = col32.b;
			*tmp_dst++ = col32.g;
			*tmp_dst++ = col32.r;

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_rgb565a8_over_argb8888(void *dst, const void *src, const void *src_opa,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_opa_pitch,
		uint16_t src_w, uint16_t src_h, int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 2;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint8_t *src_a1 = (uint8_t *)src_opa + y * src_opa_pitch + x;
			uint8_t *src_a2 = src1 + 1;
			uint8_t *src_a3 = src1 + src_opa_pitch;
			uint8_t *src_a4 = src2 + src_opa_pitch;
			sw_color16a8_t col24;

			col24.rgb = bilinear_rgb565_fast_m6(*(uint16_t*)src1,
					*(uint16_t*)src2, *(uint16_t*)src3, *(uint16_t*)src4,
					x_frac >> 10, y_frac >> 10, 6);
			col24.a = bilinear_a8_fast_m8(*src_a1, *src_a2, *src_a3, *src_a4,
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = blend_argb8565_over_argb8888(*tmp_dst, (uint8_t *)&col24);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_argb8565_over_rgb565(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
#ifdef CONFIG_GUI_API_BROM_LEOPARD
	p_brom_libgui_api->p_sw_transform_argb8565_over_rgb565(
			dst, src, dst_pitch, src_pitch, src_w, src_h, x, y, w, h, matrix);
#else
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint8_t result[3];

			bilinear_argb8565_fast_m6(result, src1, src2, src3, src4,
					x_frac >> 10, y_frac >> 10, 6);

			*tmp_dst = blend_argb8565_over_rgb565(*tmp_dst, result);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
#endif /* CONFIG_GUI_API_BROM_LEOPARD */
}

void sw_transform_argb8565_over_rgb888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	uint16_t src_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint8_t result[3];
			sw_color32_t col32 = {
				.a = 255,
				.r = tmp_dst[2],
				.g = tmp_dst[1],
				.b = tmp_dst[0],
			};

			bilinear_argb8565_fast_m6(result, src1, src2, src3, src4,
					x_frac >> 10, y_frac >> 10, 6);

			col32.full = blend_argb8565_over_argb8888(col32.full, result);
			*tmp_dst++ = col32.b;
			*tmp_dst++ = col32.g;
			*tmp_dst++ = col32.r;

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_argb8565_over_argb8888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
#ifdef CONFIG_GUI_API_BROM_LEOPARD
	p_brom_libgui_api->p_sw_transform_argb8565_over_argb8888(
			dst, src, dst_pitch, src_pitch, src_w, src_h, x, y, w, h, matrix);
#else
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint8_t result[3];

			bilinear_argb8565_fast_m6(result, src1, src2, src3, src4,
					x_frac >> 10, y_frac >> 10, 6);

			*tmp_dst = blend_argb8565_over_argb8888(*tmp_dst, result);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
#endif /* CONFIG_GUI_API_BROM_LEOPARD */
}

void sw_transform_argb6666_over_rgb565(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
#ifdef CONFIG_GUI_API_BROM_LEOPARD
	p_brom_libgui_api->p_sw_transform_argb6666_over_rgb565(
			dst, src, dst_pitch, src_pitch, src_w, src_h, x, y, w, h, matrix);
#else
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint8_t result[3];

			bilinear_argb6666_fast_m6(result, src1, src2, src3, src4,
					x_frac >> 10, y_frac >> 10, 6);

			*tmp_dst = blend_argb6666_over_rgb565(*tmp_dst, result);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
#endif /* CONFIG_GUI_API_BROM_LEOPARD */
}

void sw_transform_argb6666_over_rgb888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	uint16_t src_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint8_t result[3];
			sw_color32_t col32 = {
				.a = 255,
				.r = tmp_dst[2],
				.g = tmp_dst[1],
				.b = tmp_dst[0],
			};

			bilinear_argb6666_fast_m6(result, src1, src2, src3, src4,
					x_frac >> 10, y_frac >> 10, 6);

			col32.full = blend_argb6666_over_argb8888(col32.full, result);
			*tmp_dst++ = col32.b;
			*tmp_dst++ = col32.g;
			*tmp_dst++ = col32.r;

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_argb6666_over_argb8888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
#ifdef CONFIG_GUI_API_BROM_LEOPARD
	p_brom_libgui_api->p_sw_transform_argb6666_over_argb8888(
			dst, src, dst_pitch, src_pitch, src_w, src_h, x, y, w, h, matrix);
#else
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			uint8_t result[3];

			bilinear_argb6666_fast_m6(result, src1, src2, src3, src4,
					x_frac >> 10, y_frac >> 10, 6);

			*tmp_dst = blend_argb6666_over_argb8888(*tmp_dst, result);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
#endif /* CONFIG_GUI_API_BROM_LEOPARD */
}

void sw_transform_argb8888_over_rgb565(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
#ifdef CONFIG_GUI_API_BROM_LARK
	if (src_pitch == src_w * 4) {
		p_brom_libgui_api->p_sw_rotate_argb8888_over_rgb565(
				dst, src, dst_pitch / 2, src_w, src_h, x, y, w, h, matrix);
		return;
	}
#endif

#ifdef CONFIG_GUI_API_BROM_LEOPARD
	p_brom_libgui_api->p_sw_transform_argb8888_over_rgb565(
			dst, src, dst_pitch, src_pitch, src_w, src_h, x, y, w, h, matrix);
#else
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 4;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			uint32_t color = bilinear_argb8888_fast_m8(*(uint32_t*)src1,
					*(uint32_t*)src2, *(uint32_t*)src3, *(uint32_t*)src4,
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = blend_argb8888_over_rgb565(*tmp_dst, color);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
#endif /* CONFIG_GUI_API_BROM_LEOPARD */
}

void sw_transform_argb8888_over_rgb888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	uint16_t src_bytes_per_pixel = 4;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			sw_color32_t col32 = {
				.a = 255,
				.r = tmp_dst[2],
				.g = tmp_dst[1],
				.b = tmp_dst[0],
			};

			uint32_t color = bilinear_argb8888_fast_m8(*(uint32_t*)src1,
					*(uint32_t*)src2, *(uint32_t*)src3, *(uint32_t*)src4,
					x_frac >> 8, y_frac >> 8, 8);

			col32.full = blend_argb8888_over_argb8888(col32.full, color);
			*tmp_dst++ = col32.b;
			*tmp_dst++ = col32.g;
			*tmp_dst++ = col32.r;

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_argb8888_over_argb8888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
#ifdef CONFIG_GUI_API_BROM_LARK
	if (src_pitch == src_w * 4) {
		p_brom_libgui_api->p_sw_rotate_argb8888_over_argb8888(
				dst, src, dst_pitch / 4, src_w, src_h, x, y, w, h, matrix);
		return;
	}
#endif

#ifdef CONFIG_GUI_API_BROM_LEOPARD
	p_brom_libgui_api->p_sw_transform_argb8888_over_argb8888(
			dst, src, dst_pitch, src_pitch, src_w, src_h, x, y, w, h, matrix);
#else
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 4;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			uint32_t color = bilinear_argb8888_fast_m8(*(uint32_t*)src1,
					*(uint32_t*)src2, *(uint32_t*)src3, *(uint32_t*)src4,
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = blend_argb8888_over_argb8888(*tmp_dst, color);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
#endif /* CONFIG_GUI_API_BROM_LEOPARD */
}

void sw_transform_xrgb8888_over_rgb565(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 4;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			uint32_t color = bilinear_argb8888_fast_m8(*(uint32_t*)src1,
					*(uint32_t*)src2, *(uint32_t*)src3, *(uint32_t*)src4,
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = ((color & 0xf80000) >> 8) | ((color & 0x00fc00) >> 5) |
					((color & 0x0000f8) >> 3);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_xrgb8888_over_rgb888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	uint16_t src_bytes_per_pixel = 4;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			sw_color32_t col32;

			col32.full = bilinear_argb8888_fast_m8(*(uint32_t*)src1,
					*(uint32_t*)src2, *(uint32_t*)src3, *(uint32_t*)src4,
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst++ = col32.b;
			*tmp_dst++ = col32.g;
			*tmp_dst++ = col32.r;

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_xrgb8888_over_argb8888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 4;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			uint32_t color = bilinear_argb8888_fast_m8(*(uint32_t*)src1,
					*(uint32_t*)src2, *(uint32_t*)src3, *(uint32_t*)src4,
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = color | 0xff000000;

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_rgb888_over_rgb565(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			uint32_t color = bilinear_rgb888_fast_m8(src1,
					src2, src3, src4, x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = ((color & 0xf80000) >> 8) | ((color & 0x00fc00) >> 5) |
					((color & 0x0000f8) >> 3);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_rgb888_over_rgb888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	uint16_t src_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			sw_color32_t col32;

			col32.full = bilinear_rgb888_fast_m8(src1,
					src2, src3, src4, x_frac >> 8, y_frac >> 8, 8);
			*tmp_dst++ = col32.b;
			*tmp_dst++ = col32.g;
			*tmp_dst++ = col32.r;

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_rgb888_over_argb8888(void *dst, const void *src,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			uint32_t color = bilinear_rgb888_fast_m8(src1,
					src2, src3, src4, x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = color | 0xff000000;

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_a8_over_rgb565(void *dst, const void *src, uint32_t src_color,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint8_t src_opa = src_color >> 24;
	uint16_t src_bytes_per_pixel = 1;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			uint32_t opa = bilinear_a8_fast_m8(*src1,
					*src2, *src3, *src4, x_frac >> 8, y_frac >> 8, 8);

			if (src_opa < 255)
				opa = (opa * src_opa) >> 8;

			src_color = (src_color & 0xffffff) | (opa << 24);

			*tmp_dst = blend_argb8888_over_rgb565(*tmp_dst, src_color);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_a8_over_rgb888(void *dst, const void *src, uint32_t src_color,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	uint8_t src_opa = src_color >> 24;
	uint16_t src_bytes_per_pixel = 1;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			sw_color32_t col32 = {
				.a = 255,
				.r = tmp_dst[2],
				.g = tmp_dst[1],
				.b = tmp_dst[0],
			};

            uint32_t opa = bilinear_a8_fast_m8(*src1,
                *src2, *src3, *src4, x_frac >> 8, y_frac >> 8, 8);

			if (src_opa < 255)
				opa = (opa * src_opa) >> 8;

			src_color = (src_color & 0xffffff) | (opa << 24);

			col32.full = blend_argb8888_over_argb8888(col32.full, src_color);
			*tmp_dst++ = col32.b;
			*tmp_dst++ = col32.g;
			*tmp_dst++ = col32.r;

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_a8_over_argb8888(void *dst, const void *src, uint32_t src_color,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint8_t src_opa = src_color >> 24;
	uint16_t src_bytes_per_pixel = 1;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

            uint32_t opa = bilinear_a8_fast_m8(*src1,
                *src2, *src3, *src4, x_frac >> 8, y_frac >> 8, 8);

			if (src_opa < 255)
				opa = (opa * src_opa) >> 8;

			src_color = (src_color & 0xffffff) | (opa << 24);

			*tmp_dst = blend_argb8888_over_argb8888(*tmp_dst, src_color);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_index8_over_rgb565(void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 1;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			uint32_t color = bilinear_argb8888_fast_m8(src_clut[*src1],
					src_clut[*src2], src_clut[*src3], src_clut[*src4],
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = blend_argb8888_over_rgb565(*tmp_dst, color);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_index8_over_rgb888(void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	uint16_t src_bytes_per_pixel = 1;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;
			sw_color32_t col32 = {
				.a = 255,
				.r = tmp_dst[2],
				.g = tmp_dst[1],
				.b = tmp_dst[0],
			};

			uint32_t color = bilinear_argb8888_fast_m8(src_clut[*src1],
					src_clut[*src2], src_clut[*src3], src_clut[*src4],
					x_frac >> 8, y_frac >> 8, 8);

			col32.full = blend_argb8888_over_argb8888(col32.full, color);
			*tmp_dst++ = col32.b;
			*tmp_dst++ = col32.g;
			*tmp_dst++ = col32.r;

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_index8_over_argb8888(void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t src_bytes_per_pixel = 1;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + x * src_bytes_per_pixel;
			uint8_t *src2 = src1 + src_bytes_per_pixel;
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			uint32_t color = bilinear_argb8888_fast_m8(src_clut[*src1],
					src_clut[*src2], src_clut[*src3], src_clut[*src4],
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = blend_argb8888_over_argb8888(*tmp_dst, color);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

static void sw_transform_index124_over_rgb565(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		uint8_t src_bpp, int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;
	const uint8_t src_bmask = (1 << src_bpp) - 1;
	const uint8_t src_bofs_max = 8 - src_bpp;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint16_t *tmp_dst = (uint16_t *)dst8;

		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint16_t x_ofs1 = x * src_bpp;
			uint16_t x_ofs2 = x * src_bpp + src_bpp;
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + (x_ofs1 >> 3);
			uint8_t *src2 = (uint8_t *)src + y * src_pitch + (x_ofs2 >> 3);
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			x_ofs1 = src_bofs_max - (x_ofs1 & 0x7);
			x_ofs2 = src_bofs_max - (x_ofs2 & 0x7);

			uint8_t src1_idx = (*src1 >> x_ofs1) & src_bmask;
			uint8_t src2_idx = (*src2 >> x_ofs2) & src_bmask;
			uint8_t src3_idx = (*src3 >> x_ofs1) & src_bmask;
			uint8_t src4_idx = (*src4 >> x_ofs2) & src_bmask;

			uint32_t color = bilinear_argb8888_fast_m8(src_clut[src1_idx],
					src_clut[src2_idx], src_clut[src3_idx], src_clut[src4_idx],
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = blend_argb8888_over_rgb565(*tmp_dst, color);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

static void sw_transform_index124_over_rgb888(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		uint8_t src_bpp, int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	uint16_t dst_bytes_per_pixel = 3;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;
	const uint8_t src_bmask = (1 << src_bpp) - 1;
	const uint8_t src_bofs_max = 8 - src_bpp;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint8_t *tmp_dst = (uint8_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1 * dst_bytes_per_pixel;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint16_t x_ofs1 = x * src_bpp;
			uint16_t x_ofs2 = x * src_bpp + src_bpp;
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + (x_ofs1 >> 3);
			uint8_t *src2 = (uint8_t *)src + y * src_pitch + (x_ofs2 >> 3);
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			x_ofs1 = src_bofs_max - (x_ofs1 & 0x7);
			x_ofs2 = src_bofs_max - (x_ofs2 & 0x7);

			uint8_t src1_idx = (*src1 >> x_ofs1) & src_bmask;
			uint8_t src2_idx = (*src2 >> x_ofs2) & src_bmask;
			uint8_t src3_idx = (*src3 >> x_ofs1) & src_bmask;
			uint8_t src4_idx = (*src4 >> x_ofs2) & src_bmask;

			sw_color32_t col32 = {
				.a = 255,
				.r = tmp_dst[2],
				.g = tmp_dst[1],
				.b = tmp_dst[0],
			};

			uint32_t color = bilinear_argb8888_fast_m8(src_clut[src1_idx],
					src_clut[src2_idx], src_clut[src3_idx], src_clut[src4_idx],
					x_frac >> 8, y_frac >> 8, 8);

			col32.full = blend_argb8888_over_argb8888(col32.full, color);
			*tmp_dst++ = col32.b;
			*tmp_dst++ = col32.g;
			*tmp_dst++ = col32.r;

			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

static void sw_transform_index124_over_argb8888(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		uint8_t src_bpp, int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	uint8_t * dst8 = dst;
	int32_t src_coord_x = matrix->tx +
			y * matrix->shx + x * matrix->sx;
	int32_t src_coord_y = matrix->ty +
			y * matrix->sy + x * matrix->shy;
	const uint8_t src_bmask = (1 << src_bpp) - 1;
	const uint8_t src_bofs_max = 8 - src_bpp;

	for (int j = h; j > 0; j--) {
		int32_t p_x = src_coord_x;
		int32_t p_y = src_coord_y;
		uint32_t *tmp_dst = (uint32_t *)dst8;
		int x1 = 0, x2 = w - 1;

		sw_transform_compoute_x_range(&x1, &x2, src_w, src_h,
				p_x, p_y, matrix->sx, matrix->shy);
		if (x1 > x2) {
			goto next_line;
		} else if (x1 > 0) {
			p_x += matrix->sx * x1;
			p_y += matrix->shy * x1;
			tmp_dst += x1;
		}

		for (int i = x2 - x1; i >= 0; i--) {
			int x = FLOOR_FIXEDPOINT16(p_x);
			int y = FLOOR_FIXEDPOINT16(p_y);
			int x_frac = p_x - FIXEDPOINT16(x);
			int y_frac = p_y - FIXEDPOINT16(y);
			uint16_t x_ofs1 = x * src_bpp;
			uint16_t x_ofs2 = x * src_bpp + src_bpp;
			uint8_t *src1 = (uint8_t *)src + y * src_pitch + (x_ofs1 >> 3);
			uint8_t *src2 = (uint8_t *)src + y * src_pitch + (x_ofs2 >> 3);
			uint8_t *src3 = src1 + src_pitch;
			uint8_t *src4 = src2 + src_pitch;

			x_ofs1 = src_bofs_max - (x_ofs1 & 0x7);
			x_ofs2 = src_bofs_max - (x_ofs2 & 0x7);

			uint8_t src1_idx = (*src1 >> x_ofs1) & src_bmask;
			uint8_t src2_idx = (*src2 >> x_ofs2) & src_bmask;
			uint8_t src3_idx = (*src3 >> x_ofs1) & src_bmask;
			uint8_t src4_idx = (*src4 >> x_ofs2) & src_bmask;

			uint32_t color = bilinear_argb8888_fast_m8(src_clut[src1_idx],
					src_clut[src2_idx], src_clut[src3_idx], src_clut[src4_idx],
					x_frac >> 8, y_frac >> 8, 8);

			*tmp_dst = blend_argb8888_over_argb8888(*tmp_dst, color);

			tmp_dst += 1;
			p_x += matrix->sx;
			p_y += matrix->shy;
		}

next_line:
		src_coord_x += matrix->shx;
		src_coord_y += matrix->sy;
		dst8 += dst_pitch;
	}
}

void sw_transform_index4_over_rgb565(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	sw_transform_index124_over_rgb565(dst, src, src_clut, dst_pitch, src_pitch,
			src_w, src_h, 4, x, y, w, h, matrix);
}

void sw_transform_index4_over_rgb888(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	sw_transform_index124_over_rgb888(dst, src, src_clut, dst_pitch, src_pitch,
			src_w, src_h, 4, x, y, w, h, matrix);
}

void sw_transform_index4_over_argb8888(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	sw_transform_index124_over_argb8888(dst, src, src_clut, dst_pitch, src_pitch,
			src_w, src_h, 4, x, y, w, h, matrix);
}

void sw_transform_index2_over_rgb565(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	sw_transform_index124_over_rgb565(dst, src, src_clut, dst_pitch, src_pitch,
			src_w, src_h, 2, x, y, w, h, matrix);
}

void sw_transform_index2_over_rgb888(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	sw_transform_index124_over_rgb888(dst, src, src_clut, dst_pitch, src_pitch,
			src_w, src_h, 2, x, y, w, h, matrix);
}

void sw_transform_index2_over_argb8888(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	sw_transform_index124_over_argb8888(dst, src, src_clut, dst_pitch, src_pitch,
			src_w, src_h, 2, x, y, w, h, matrix);
}

void sw_transform_index1_over_rgb565(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	sw_transform_index124_over_rgb565(dst, src, src_clut, dst_pitch, src_pitch,
			src_w, src_h, 1, x, y, w, h, matrix);
}

void sw_transform_index1_over_rgb888(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	sw_transform_index124_over_rgb888(dst, src, src_clut, dst_pitch, src_pitch,
			src_w, src_h, 1, x, y, w, h, matrix);
}

void sw_transform_index1_over_argb8888(
		void *dst, const void *src, const uint32_t *src_clut,
		uint16_t dst_pitch, uint16_t src_pitch, uint16_t src_w, uint16_t src_h,
		int16_t x, int16_t y, uint16_t w, uint16_t h,
		const sw_matrix_t *matrix)
{
	sw_transform_index124_over_argb8888(dst, src, src_clut, dst_pitch, src_pitch,
			src_w, src_h, 1, x, y, w, h, matrix);
}