engine/ac/draw_software.cpp

/* ScummVM - Graphic Adventure Engine
 *
 * ScummVM is the legal property of its developers, whose names
 * are too numerous to list here. Please refer to the COPYRIGHT
 * file distributed with this source distribution.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version 2
 * of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
 *
 */

//=============================================================================
//
// Software drawing component. Optimizes drawing for software renderer using
// dirty rectangles technique.
//
// TODO: do research/profiling to find out if this dirty rectangles thing
// is still giving ANY notable perfomance boost at all.
//
// TODO: would that give any benefit to reorganize the code and move dirty
// rectangles into SoftwareGraphicDriver?
// Alternatively: we could pass dirty rects struct pointer and room background
// DDB when calling BeginSpriteBatch(). Driver itself could be calling
// update_invalid_region(). That will keep gfx driver's changes to minimum.
//
// NOTE: this code, including structs and functions, has underwent several
// iterations of changes. Originally it was meant to perform full transform
// of dirty rects right away, but later I realized it won't work that way
// because a) Allegro does not support scaling bitmaps over destination with
// different colour depth (which may be a case when running 16-bit game),
// and b) Allegro does not support scaling and rotating of sprites with
// blending and lighting at the same time which means that room objects have
// to be drawn upon non-scaled background first. Possibly some of the code
// below may be therefore simplified.
//
//=============================================================================

#include "ags/lib/std/vector.h"
#include "ags/engine/ac/draw_software.h"
#include "ags/shared/gfx/bitmap.h"
#include "ags/shared/util/scaling.h"
#include "ags/globals.h"

namespace AGS3 {

using namespace AGS::Shared;
using namespace AGS::Engine;

IRSpan::IRSpan()
	: x1(0), x2(0) {
}

IRRow::IRRow()
	: numSpans(0) {
}

int IRSpan::mergeSpan(int tx1, int tx2) {
	if ((tx1 > x2) || (tx2 < x1))
		return 0;
	// overlapping, increase the span
	if (tx1 < x1)
		x1 = tx1;
	if (tx2 > x2)
		x2 = tx2;
	return 1;
}

DirtyRects::DirtyRects()
	: NumDirtyRegions(0) {
}

bool DirtyRects::IsInit() const {
	return DirtyRows.size() > 0;
}

void DirtyRects::Init(const Size &surf_size, const Rect &viewport) {
	int height = surf_size.Height;
	if (SurfaceSize != surf_size) {
		Destroy();
		SurfaceSize = surf_size;
		DirtyRows.resize(height);

		NumDirtyRegions = WHOLESCREENDIRTY;
		for (int i = 0; i < height; ++i)
			DirtyRows[i].numSpans = 0;
	}

	Viewport = viewport;
	Room2Screen.Init(surf_size, viewport);
	Screen2DirtySurf.Init(viewport, RectWH(0, 0, surf_size.Width, surf_size.Height));
}

void DirtyRects::SetSurfaceOffsets(int x, int y) {
	Room2Screen.SetSrcOffsets(x, y);
}

void DirtyRects::Destroy() {
	DirtyRows.clear();
	NumDirtyRegions = 0;
}

void DirtyRects::Reset() {
	NumDirtyRegions = 0;

	for (size_t i = 0; i < DirtyRows.size(); ++i)
		DirtyRows[i].numSpans = 0;
}

void dispose_invalid_regions(bool /* room_only */) {
	_GP(RoomCamRects).clear();
	_GP(RoomCamPositions).clear();
}

void set_invalidrects_globaloffs(int x, int y) {
	_GP(GlobalOffs) = Point(x, y);
}

void init_invalid_regions(int view_index, const Size &surf_size, const Rect &viewport) {
	if (view_index < 0) {
		_GP(BlackRects).Init(surf_size, viewport);
	} else {
		if (_GP(RoomCamRects).size() <= (size_t)view_index) {
			_GP(RoomCamRects).resize(view_index + 1);
			_GP(RoomCamPositions).resize(view_index + 1);
		}
		_GP(RoomCamRects)[view_index].Init(surf_size, viewport);
		_GP(RoomCamPositions)[view_index] = std::make_pair(-1000, -1000);
	}
}

void delete_invalid_regions(int view_index) {
	if (view_index >= 0) {
		_GP(RoomCamRects).erase(_GP(RoomCamRects).begin() + view_index);
		_GP(RoomCamPositions).erase(_GP(RoomCamPositions).begin() + view_index);
	}
}

void set_invalidrects_cameraoffs(int view_index, int x, int y) {
	if (view_index < 0) {
		_GP(BlackRects).SetSurfaceOffsets(x, y);
		return;
	} else {
		_GP(RoomCamRects)[view_index].SetSurfaceOffsets(x, y);
	}

	int &posxwas = _GP(RoomCamPositions)[view_index].first;
	int &posywas = _GP(RoomCamPositions)[view_index].second;
	if ((x != posxwas) || (y != posywas)) {
		invalidate_all_camera_rects(view_index);
		posxwas = x;
		posywas = y;
	}
}

void invalidate_all_rects() {
	for (auto &rects : _GP(RoomCamRects)) {
		if (!IsRectInsideRect(rects.Viewport, _GP(BlackRects).Viewport))
			_GP(BlackRects).NumDirtyRegions = WHOLESCREENDIRTY;
		rects.NumDirtyRegions = WHOLESCREENDIRTY;
	}
}

void invalidate_all_camera_rects(int view_index) {
	if (view_index < 0)
		return;
	_GP(RoomCamRects)[view_index].NumDirtyRegions = WHOLESCREENDIRTY;
}

void invalidate_rect_on_surf(int x1, int y1, int x2, int y2, DirtyRects &rects) {
	if (rects.DirtyRows.size() == 0)
		return;
	if (rects.NumDirtyRegions >= MAXDIRTYREGIONS) {
		// too many invalid rectangles, just mark the whole thing dirty
		rects.NumDirtyRegions = WHOLESCREENDIRTY;
		return;
	}

	int a;

	const Size &surfsz = rects.SurfaceSize;
	if (x1 >= surfsz.Width) x1 = surfsz.Width - 1;
	if (y1 >= surfsz.Height) y1 = surfsz.Height - 1;
	if (x2 >= surfsz.Width) x2 = surfsz.Width - 1;
	if (y2 >= surfsz.Height) y2 = surfsz.Height - 1;
	if (x1 < 0) x1 = 0;
	if (y1 < 0) y1 = 0;
	if (x2 < 0) x2 = 0;
	if (y2 < 0) y2 = 0;
	rects.NumDirtyRegions++;

	// ** Span code
	std::vector<IRRow> &dirtyRow = rects.DirtyRows;
	int s, foundOne;
	// add this rect to the list for this row
	for (a = y1; a <= y2; a++) {
		foundOne = 0;
		for (s = 0; s < dirtyRow[a].numSpans; s++) {
			if (dirtyRow[a].span[s].mergeSpan(x1, x2)) {
				foundOne = 1;
				break;
			}
		}
		if (foundOne) {
			// we were merged into a span, so we're ok
			int t;
			// check whether now two of the spans overlap each other
			// in which case merge them
			for (s = 0; s < dirtyRow[a].numSpans; s++) {
				for (t = s + 1; t < dirtyRow[a].numSpans; t++) {
					if (dirtyRow[a].span[s].mergeSpan(dirtyRow[a].span[t].x1, dirtyRow[a].span[t].x2)) {
						dirtyRow[a].numSpans--;
						for (int u = t; u < dirtyRow[a].numSpans; u++)
							dirtyRow[a].span[u] = dirtyRow[a].span[u + 1];
						break;
					}
				}
			}
		} else if (dirtyRow[a].numSpans < MAX_SPANS_PER_ROW) {
			dirtyRow[a].span[dirtyRow[a].numSpans].x1 = x1;
			dirtyRow[a].span[dirtyRow[a].numSpans].x2 = x2;
			dirtyRow[a].numSpans++;
		} else {
			// didn't fit in an existing span, and there are none spare
			int nearestDist = 99999, nearestWas = -1, extendLeft = false;
			int tleft, tright;
			// find the nearest span, and enlarge that to include this rect
			for (s = 0; s < dirtyRow[a].numSpans; s++) {
				tleft = dirtyRow[a].span[s].x1 - x2;
				if ((tleft > 0) && (tleft < nearestDist)) {
					nearestDist = tleft;
					nearestWas = s;
					extendLeft = 1;
				}
				tright = x1 - dirtyRow[a].span[s].x2;
				if ((tright > 0) && (tright < nearestDist)) {
					nearestDist = tright;
					nearestWas = s;
					extendLeft = 0;
				}
			}
			if (extendLeft)
				dirtyRow[a].span[nearestWas].x1 = x1;
			else
				dirtyRow[a].span[nearestWas].x2 = x2;
		}
	}
	// ** End span code
	//}
}

void invalidate_rect_ds(DirtyRects &rects, int x1, int y1, int x2, int y2, bool in_room) {
	if (!in_room) {
		// TODO: for most opimisation (esp. with multiple viewports) should perhaps
		// split/cut parts of the original rectangle which overlap room viewport(s).
		Rect r(x1, y1, x2, y2);
		// If overlay is NOT completely over the room, then invalidate black rect
		if (!IsRectInsideRect(rects.Viewport, r))
			invalidate_rect_on_surf(x1, y1, x2, y2, _GP(BlackRects));
		// If overlay is NOT intersecting room viewport at all, then stop
		if (!AreRectsIntersecting(rects.Viewport, r))
			return;

		// Transform from screen to room coordinates through the known viewport
		x1 = rects.Screen2DirtySurf.X.ScalePt(x1);
		x2 = rects.Screen2DirtySurf.X.ScalePt(x2);
		y1 = rects.Screen2DirtySurf.Y.ScalePt(y1);
		y2 = rects.Screen2DirtySurf.Y.ScalePt(y2);
	} else {
		// Transform only from camera pos to room background
		x1 -= rects.Room2Screen.X.GetSrcOffset();
		y1 -= rects.Room2Screen.Y.GetSrcOffset();
		x2 -= rects.Room2Screen.X.GetSrcOffset();
		y2 -= rects.Room2Screen.Y.GetSrcOffset();
	}

	invalidate_rect_on_surf(x1, y1, x2, y2, rects);
}

void invalidate_rect_ds(int x1, int y1, int x2, int y2, bool in_room) {
	if (!in_room) { // convert from game viewport to global screen coords
		x1 += _GP(GlobalOffs).X;
		x2 += _GP(GlobalOffs).X;
		y1 += _GP(GlobalOffs).Y;
		y2 += _GP(GlobalOffs).Y;
	}

	for (auto &rects : _GP(RoomCamRects))
		invalidate_rect_ds(rects, x1, y1, x2, y2, in_room);
}

void invalidate_rect_global(int x1, int y1, int x2, int y2) {
	for (auto &rects : _GP(RoomCamRects))
		invalidate_rect_ds(rects, x1, y1, x2, y2, false);
}

// Note that this function is denied to perform any kind of scaling or other transformation
// other than blitting with offset. This is mainly because destination could be a 32-bit virtual screen
// while room background was 16-bit and Allegro lib does not support stretching between colour depths.
// The no_transform flag here means essentially "no offset", and indicates that the function
// must blit src on ds at 0;0. Otherwise, actual Viewport offset is used.
void update_invalid_region(Bitmap *ds, Bitmap *src, const DirtyRects &rects, bool no_transform) {
	if (rects.NumDirtyRegions == 0)
		return;

	if (!no_transform)
		ds->SetClip(rects.Viewport);

	const int src_x = rects.Room2Screen.X.GetSrcOffset();
	const int src_y = rects.Room2Screen.Y.GetSrcOffset();
	const int dst_x = no_transform ? 0 : rects.Viewport.Left;
	const int dst_y = no_transform ? 0 : rects.Viewport.Top;

	if (rects.NumDirtyRegions == WHOLESCREENDIRTY) {
		ds->Blit(src, src_x, src_y, dst_x, dst_y, rects.SurfaceSize.Width, rects.SurfaceSize.Height);
	} else {
		const std::vector<IRRow> &dirtyRow = rects.DirtyRows;
		const int surf_height = rects.SurfaceSize.Height;
		// TODO: is this IsMemoryBitmap check is still relevant?
		// If bitmaps properties match and no transform required other than linear offset
		if (src->GetColorDepth() == ds->GetColorDepth()) {
			const int bypp = src->GetBPP();
			// do the fast memory copy
			for (int i = 0; i < surf_height; i++) {
				const uint8_t *src_scanline = src->GetScanLine(i + src_y);
				uint8_t *dst_scanline = ds->GetScanLineForWriting(i + dst_y);
				const IRRow &dirty_row = dirtyRow[i];
				for (int k = 0; k < dirty_row.numSpans; k++) {
					int tx1 = dirty_row.span[k].x1;
					int tx2 = dirty_row.span[k].x2;
					memcpy(&dst_scanline[(tx1 + dst_x) * bypp], &src_scanline[(tx1 + src_x) * bypp], ((tx2 - tx1) + 1) * bypp);
				}
			}
		}
		// If has to use Blit, but still must draw with no transform but offset
		else {
			// do fast copy without transform
			for (int i = 0, rowsInOne = 1; i < surf_height; i += rowsInOne, rowsInOne = 1) {
				// if there are rows with identical masks, do them all in one go
				// TODO: what is this for? may this be done at the invalidate_rect merge step?
				while ((i + rowsInOne < surf_height) && (memcmp(&dirtyRow[i], &dirtyRow[i + rowsInOne], sizeof(IRRow)) == 0))
					rowsInOne++;

				const IRRow &dirty_row = dirtyRow[i];
				for (int k = 0; k < dirty_row.numSpans; k++) {
					int tx1 = dirty_row.span[k].x1;
					int tx2 = dirty_row.span[k].x2;
					ds->Blit(src, tx1 + src_x, i + src_y, tx1 + dst_x, i + dst_y, (tx2 - tx1) + 1, rowsInOne);
				}
			}
		}
	}
}

void update_invalid_region(Bitmap *ds, color_t fill_color, const DirtyRects &rects) {
	ds->SetClip(rects.Viewport);

	if (rects.NumDirtyRegions == WHOLESCREENDIRTY) {
		ds->FillRect(rects.Viewport, fill_color);
	} else {
		const std::vector<IRRow> &dirtyRow = rects.DirtyRows;
		const int surf_height = rects.SurfaceSize.Height;
		{
			const AGS::Shared::PlaneScaling &tf = rects.Room2Screen;
			for (int i = 0, rowsInOne = 1; i < surf_height; i += rowsInOne, rowsInOne = 1) {
				// if there are rows with identical masks, do them all in one go
				// TODO: what is this for? may this be done at the invalidate_rect merge step?
				while ((i + rowsInOne < surf_height) && (memcmp(&dirtyRow[i], &dirtyRow[i + rowsInOne], sizeof(IRRow)) == 0))
					rowsInOne++;

				const IRRow &dirty_row = dirtyRow[i];
				for (int k = 0; k < dirty_row.numSpans; k++) {
					Rect src_r(dirty_row.span[k].x1, i, dirty_row.span[k].x2, i + rowsInOne - 1);
					Rect dst_r = tf.ScaleRange(src_r);
					ds->FillRect(dst_r, fill_color);
				}
			}
		}
	}
}

void update_black_invreg_and_reset(Bitmap *ds) {
	if (!_GP(BlackRects).IsInit())
		return;
	update_invalid_region(ds, (color_t)0, _GP(BlackRects));
	_GP(BlackRects).Reset();
}

void update_room_invreg_and_reset(int view_index, Bitmap *ds, Bitmap *src, bool no_transform) {
	if (view_index < 0 || _GP(RoomCamRects).size() == 0)
		return;

	update_invalid_region(ds, src, _GP(RoomCamRects)[view_index], no_transform);
	_GP(RoomCamRects)[view_index].Reset();
}

} // namespace AGS3