wip

TODO.md

@@ -1,6 +1,49 @@
 
 ## todo
 
+- Settings
+    - Backlight support enabled 
+    - Volume SFX, Music
+
+- Save Game 
+    - Center the 2 slots 
+    - save text should represent what day your on, eg. day #1 
+    - back and action buttons at buttom of screen 
+
+- Manually created z layers 
+
+- GBA Z Depth Code (ASM)  -- 275,468
+    - employ the CPU mode switching
+    - load a of z span
+
+- Collect images of all scenes and characters 
+
+- pre-computed visiblity culling 
+    - encode the direction of a triangle as uint8_t
+    - sort triangles based on direction
+    - culling based on X and Z not Y
+
+- use triangles instead of quads, 
+- simplify character models 
+- LUT table at a given address, so we can simplfy the lookup
+
+- Enemy AI 
+    - Behaviour
+
+- Death 
+
+- Credits 
+
+- UI for picking up an item
+
+- Animated Environment
+
+- Content Completion ***
+    - Environments 
+    - Story 
+    - Enemy
+    - NPCs
+
 - State_Model 
     - better camera controls 
     - animation playback support 
@@ -14,7 +57,7 @@
 
 - switching to IRQ enabled gives up access to and alternative r8-r14 (I think r7 might be required for fiq_on / off)
 
-```
+```     8, 9, 10, 11, 12, 
 .macro fiq_on
     msr cpsr, #0x11     // switch r8-r14 to FIQ (IRQ enabled)
 .endm

source/common/graphics/graphics.c

@@ -13,7 +13,7 @@
 #include "image.h"
 
 //-----------------------------------------------------------------------------
-graphics_context_t g_graphics_context;			//IWRAM_DATA 
+graphics_context_t g_graphics_context;			//IWRAM_DATA -- put this at a fixed address or patch it in
 
 //	this could all be part of the graphics context 
 image_ptr boundimage = NULL;				//	think about bandwidth for texture sampling  // make this an array
@@ -212,6 +212,10 @@ edge_t _graphics_left_edge, _graphics_right_edge;
 //-----------------------------------------------------------------------------
 IWRAM_CODE void create_edge_and_calculate_deltas(edge_t* edge, vertex_t* v1, vertex_t* v2)
 {
+	//	this seems un-neccessarily heavy
+
+	//	lots of un-neccessary de-referencing
+
 	vertex_t* tv1 = v1;
 	vertex_t* tv2 = v2;
 
@@ -558,7 +562,7 @@ IWRAM_CODE void graphics_draw_model(model_t* model, uint16_t animation_id, uint1
 
 	//debug_printf(DEBUG_LOG_DEBUG, "graphics::draw_model - processed_vertices=0x%x", processed_vertices);
 
-	memory_copy(matrix_working_copy, m, sizeof(matrix4x4_t));
+	memory_copy(matrix_working_copy, m, sizeof(matrix4x4_t));	//	should be able to do a faster copy 
 	fixed16_t* p = (fixed16_t*)matrix_working_copy;
 	for (i = 0; i < 16; ++i)
 		p[i] >>= 6;
@@ -696,6 +700,8 @@ IWRAM_CODE void graphics_draw_model(model_t* model, uint16_t animation_id, uint1
 
 	//	sort and cull polygons	 move to it's own func 
 
+	//		reduce to one allocation, or better yet, none, just have scratch memory
+
 	polygon_t* polygons_to_draw = memory_allocate(sizeof(polygon_t) * 192, MEMORY_EWRAM);	//	linked list might make alot of sense here .
 	polygon_t* polygon_ptr = &polygons_to_draw[0];
 
@@ -746,6 +752,8 @@ IWRAM_CODE void graphics_draw_model(model_t* model, uint16_t animation_id, uint1
 		fixed16_t d = fixed16_mul((polygon_vertices[1].position.x - polygon_vertices[0].position.x), (polygon_vertices[2].position.y - polygon_vertices[0].position.y));
 		d -= fixed16_mul((polygon_vertices[1].position.y - polygon_vertices[0].position.y), (polygon_vertices[2].position.x - polygon_vertices[0].position.x));
 
+		//	triangles might sort better 
+
 		if (d > fixed16_zero)//(fixed16_one << 2))		//should be zero, but increasing this will only remove polygons that are at a sharp angle and probably don't take up many pixlelse
 		{
 			sort_index_t index = { depth_value, polygons_to_draw_idx };
@@ -794,7 +802,7 @@ IWRAM_CODE void graphics_draw_model(model_t* model, uint16_t animation_id, uint1
 
 		polygon_ptr = &polygons_to_draw[(index_ptr++)->idx];
 
-		draw_polygon(polygon_ptr->v, polygon_ptr->vertex_count);
+		draw_polygon(polygon_ptr->v, polygon_ptr->vertex_count);	//	jumping in and out of this function isn't ideal
 	}
 
 	profiler_end(phandle);

source/common/graphics/graphics.draw_spans.gba.s

@@ -78,7 +78,7 @@ draw_spans:
 	ldr r0, =_graphics_right_edge
 	ldmia r0, { r2-r6 }
 
-	mov r8, r2
+	mov r8, r2								@	these movs seem completely unneeded
 	mov r9, r4
 	mov r1, r6
 
@@ -98,28 +98,27 @@ draw_spans:
 												@ int32_t left_edge_floor_x_int = fixed16_to_int(_graphics_left_edge.x);
 	asr lr, r8, #16								@ int32_t right_edge_floor_x_int = fixed16_to_int(_graphics_right_edge.x);
 	subs lr, lr, r2, asr #16					@ int32_t width = right_edge_floor_x_int - left_edge_floor_x_int;
-	ble .draw_spans.scanline_end
-	@ if (width < 0 && dx is < 0) get out of the whole func 
+	ble .draw_spans.scanline_end				@ if (width < 0 && dx is < 0) get out of the whole func 
 
 
 	sub r3, r1, r6
 
-	ldr r0, =reciprocal_lut				@ fixed16_t[] reciprocal_lut			 
+	ldr r0, =reciprocal_lut						@ fixed16_t[] reciprocal_lut			 
 	lsl r1, lr, #2
-	ldr r0, [r0, r1]				@ fixed16_t overWidth = reciprocal_lut[width];
+	ldr r0, [r0, r1]							@ fixed16_t overWidth = reciprocal_lut[width];
 
 	asr r3, r3, #8
-	mul r1, r0, r3					@fixed16_t dv = fixed16_mul_approx2(right_edge_v - left_edge_v, overWidth);
+	mul r1, r0, r3								@fixed16_t dv = fixed16_mul_approx2(right_edge_v - left_edge_v, overWidth);
 
 
 	sub r3, r9, r4
 	asr r3, r3, #8
-	mul r0, r3, r0 					@fixed16_t du = fixed16_mul_approx2(right_edge_u - left_edge_u, overWidth);
+	mul r0, r3, r0 								@fixed16_t du = fixed16_mul_approx2(right_edge_u - left_edge_u, overWidth);
 
 
 	asr r3, r2, #16
 	asr r5, r3, #1
-	add r11, r11, r5, lsl #1		@uint16_t* output_span_ptr = output_scanline_ptr + (left_edge_floor_x_int >> 1);	
+	add r11, r11, r5, lsl #1					@uint16_t* output_span_ptr = output_scanline_ptr + (left_edge_floor_x_int >> 1);	
 
 
 .draw_spans.uv_mask_assignment:
@@ -143,8 +142,8 @@ draw_spans:
 	@ r5	= u_bitmask
 	@ r6	= v
 	@ r7	= v_bitmask
-	@ r8	= * - not used
-	@ r9	= *	- not used 
+	@ r8	= * - scratchd
+	@ r9	= *	- scratch 
 	@ r10	= iheight
 	@ r11	= output_span_ptr 
 	@ r12	= texture_data_ptr

source/common/graphics/graphics.draw_spans.win32.c

@@ -149,6 +149,17 @@ void draw_spans(void)
 //-----------------------------------------------------------------------------
 void draw_spans_depth(void)
 {
+	/*
+		This is the wrong approach, we don't need per pixel checks in this manner 
+
+		array of spans instead of ???
+
+		could it be a 1-bit depth map and we just bind the appropriate one
+		this would mean, only need to read depth value every 32 pixel stride 
+			
+		consider making depth test branchless (use a conditional mov or something)
+	*/
+
 	fixed16_t fheight = fixed16_min(_graphics_left_edge.height, _graphics_right_edge.height);	//	one of the edges might continue to be used for the next draw_spans, so heights might be different 
 
 	uint16_t* output_scanline_ptr = g_graphics_context.frame_buffer + (fixed16_to_int(_graphics_left_edge.y) * g_graphics_context.width);		//	output_scanline_ptr is 16-bit wide as VRAM requires writes of 16-bit wide

source/common/states/debug/state_model.c

@@ -6,6 +6,7 @@
 #include "common/math/trigonometry.h"
 #include "common/math/point.h"
 #include "common/graphics/graphics.h"
+#include "common/graphics/overlay.h"
 #include "common/graphics/image.h"
 #include "common/graphics/camera.h"
 #include "common/input/input.h"

source/common/utils/bitstream.h

@@ -18,9 +18,9 @@ typedef struct bitstream_s
 	const uint32_t* buffer;
 	uint32_t size;
 
-	uint32_t buffer_idx;
+	uint32_t buffer_idx;	//	byte_idx
 
-	uint8_t bit_idx;
+	uint8_t bit_idx;		//	might as well be an usize 
 	uint8_t reserved[3];
 
 	uint32_t value;	//	cache value

source/common/video/smacker.c

@@ -556,6 +556,14 @@ IWRAM_CODE static char smk_render_video(struct smk_video_t* s, const uint16_t* f
 	smk_huff16_v5_reset(s->tree[2]);
 	smk_huff16_v5_reset(s->tree[3]);
 
+	//	possibly, don't use huff16, pre-decompress, 
+
+	//	tree to IWRAM ?? 
+
+	//	lz77 compress chunks, 
+	//		decompress these to EWRAM 
+	//		playback video 
+
 	while (row < s->h)
 	{
 		unpack = smk_huff16_v5_lookup(&bs, s->tree[SMK_TREE_TYPE]);		//	decompress this ahead of time???
@@ -587,7 +595,7 @@ IWRAM_CODE static char smk_render_video(struct smk_video_t* s, const uint16_t* f
 		{
 			skip = (row * pitch) + half_col;
 
-			switch (type)	
+			switch (type)	//	lots of branching assembly code ... 
 			{
 				case 0:	
 					unpack = smk_huff16_v5_lookup(&bs, s->tree[SMK_TREE_MCLR]);

source/games/7days/entities/animation.c

@@ -2,6 +2,8 @@
 #include "games/7days/entities/entity.h"
 
 #include "common/debug/debug.h"
+#include "common/graphics/graphics.h"
+#include "common/resources/resources.h"
 
 //	movement has already been applied, this is just a presentation layer 
 
@@ -73,10 +75,10 @@ void animation_update(entity_ptr entity)		//	apply this to player and npc
 
 		animation_t* animation = model_find_animation(model, entity->animation_id);
 
-		debug_assert(animation, "animation could not be found");
+		debug_assert(animation != NULL, "animation could not be found");
 
 		entity->frame_precise = fixed16_zero;
-		entity->frame_count = animation->frame_count;
+		entity->frame_count = (int16_t)animation->frame_count;
 	}
 	else
 	{

source/games/7days/states/state_savegame.c

@@ -101,6 +101,7 @@ void st_savegame_generate_palette_remap(st_savegame_context_ptr context)
 
 		palette_remap[idx] = sum + 1;
 	}
+	palette_delete(graphics_pal);
 }
 
 //-----------------------------------------------------------------------------

source/games/7days/world.c

@@ -8,6 +8,8 @@
 #include "games/7days/sequence.h"
 #include "games/7days/states/states.h"
 
+//#define WORLD_DEBUG_DRAW
+
 world_t g_world;	// put this on the heap
 
 world_ptr g_main_world = &g_world;
@@ -63,7 +65,7 @@ void world_initialize(world_ptr world)
 		}
 	}
 
-	world->ephermeral.highlight_timer = timer_start(20, TIMER_MODE_LOOP);
+	world->ephermeral.highlight_timer = timer_start(40, TIMER_MODE_LOOP);
 	world->ephermeral.active_highlight = NULL;
 }
 
@@ -203,9 +205,11 @@ vector2_t world_closestpointonpath(vector2_t point, uint32_t mask)
 	vector2_t currentClosestPoint = { 0, 0 }, tmpDist = { 0, 0 };
 	fixed16_t currentClosestDist = fixed16_maximum;
 
+	//	this is abit brute force, at least limit based on view or something...
+
 	for (uint8_t i = 0; i < g_main_world->ephermeral.numCollision; ++i)
 	{
-		//	use collision mask 
+		//	use collision mask to optimize
 
 		vector2_t closestPoint = collisionClosestPointOBB(point, g_main_world->ephermeral.collision[i].box);
 		if (collisionCheckPointInsideOBB(point, g_main_world->ephermeral.collision[i].box))
@@ -613,6 +617,8 @@ void world_draw(world_ptr world)
 		I don't like this code for entity draw order sorting, it feels messy 
 	*/
 
+	//	can use the existing sorting code 
+
 	for (idx = 1; idx < draw_list_count; ++idx)
 	{
 		if (entity_distance_list[idx - 1] < entity_distance_list[idx])
@@ -659,24 +665,7 @@ void world_draw(world_ptr world)
 
 	profiler_end(phandle);
 
-
-	phandle = profiler_begin("w:d_entites");
-
-	//	model_entity_draw
-	for (idx = 0; idx < draw_list_count; ++idx)
-	{
-		entity_draw(
-			entity_draw_list[idx],
-			&world->ephermeral.views[world->persistent.view_idx].wvp);
-	}
-
-	if (g_graphics_context.frame_dirty_scanline_span.start < g_graphics_context.frame_dirty_scanline_span.end)
-		world->ephermeral.dirty_scanline_span[g_graphics_context.page_flip] = g_graphics_context.frame_dirty_scanline_span;
-
-	profiler_end(phandle);
-
-
-#if WORLD_DEBUG_DRAW
+#ifdef WORLD_DEBUG_DRAW
 
 	fixed16_t ground_y = entity_draw_list[0]->position.y;
 
@@ -698,14 +687,30 @@ void world_draw(world_ptr world)
 		if ((trigger->layer & current_level->layer) > 0)
 		{
 			graphicsDrawWireframeOBB(
-				&trigger->area, 
+				&trigger->area,
 				ground_y,
 				&world->ephermeral.views[world->persistent.view_idx].wvp);
 		}
 	}
 
 #endif
 
+	phandle = profiler_begin("w:d_entites");
+
+	//	model_entity_draw
+	for (idx = 0; idx < draw_list_count; ++idx)
+	{
+		entity_draw(
+			entity_draw_list[idx],
+			&world->ephermeral.views[world->persistent.view_idx].wvp);
+	}
+
+	if (g_graphics_context.frame_dirty_scanline_span.start < g_graphics_context.frame_dirty_scanline_span.end)
+		world->ephermeral.dirty_scanline_span[g_graphics_context.page_flip] = g_graphics_context.frame_dirty_scanline_span;
+
+	profiler_end(phandle);
+
+
 }
 
 //-----------------------------------------------------------------------------