Fix freezing because of multiplayer

README.md

@@ -8,7 +8,7 @@ A 3D, multiplayer racing game for casio fx-CG50 calculators
 ## Features
 - 3D graphics
 - Simple multiplayer (just connect two calculators)
-- Runs at about 16 FPS normally and 21 FPS overclocked, on the fx-CG50.
+- The multiplayer version runs at about 14 FPS and the singleplayer version runs at 24 FPS
 
 ## Controls
 - Press `up`/`8` to accelerate and `down`/`5` to brake
@@ -47,21 +47,18 @@ This version supports multiplayer, but it's slower than the gint version.
 This version doesn't support multiplayer, but it runs faster and doesn't have a border.
 #### Linux
 - Install gint ([https://gitea.planet-casio.com/Lephenixnoir/gint](https://gitea.planet-casio.com/Lephenixnoir/gint))
+- Install libprof ([https://gitea.planet-casio.com/Lephenixnoir/libprof](https://gitea.planet-casio.com/Lephenixnoir/libprof))
 - Run `make gint`
 
 ## Technical information
 ### 3D rendering
 - All the rendering code is in `src/rasterizer.h` and `src/rasterizer.cpp`
 - Every triangle is clipped to avoid drawing triangles outside the screen. If a triangle is only partially inside the screen, it's cut in one or two triangles. This doesn't happen with the cones and the car to improve performance.
-- The triangles are rasterized using the scan line algorithm with a depth buffer.
+- The triangles are split into two (one with a flat top and another with a flat bottom) and rasterized.
 - This renderer only supports diffuse directional lighting, because this way there is only one color per triangle, which increases performance.
 - Because the calculator doesn't have a floating point unit (FPU), everything related to rendering uses fixed point numbers (defined in src/fp.h). This caused some issues related to precision, most of which were solved by checking where the floating point calculations were overflowing.
 - To improve performance, the cones that are too far away from the camera are replaced with a simpler model and the ones even further away aren't drawn at all.
 
-#### Potential rendering performance improvements
-- Use DMA to clear the screen and draw the grass (in progress)
-- Clip models before clipping triangles
-
 ### Multiplayer
 All of the multiplayer code is in `src/main.cpp`
 - When the game starts, a second car is created outside the track.

gint/Makefile

@@ -1,7 +1,7 @@
 CC = sh-elf-g++
 CFLAGS += -Wall -Wextra -Ofast -funroll-loops -DGINT
 CFLAGS += -DFXCG50 -DTARGET_FXCG50 -m4-nofpu -mb -ffreestanding -nostdlib -fstrict-volatile-bitfields
-LDFLAGS = -m4-nofpu -mb -nostdlib -Wl,--no-warn-rwx-segments -T fxcg50.ld -lgint-cg -lc -lgcc -lgint-cg
+LDFLAGS = -m4-nofpu -mb -nostdlib -Wl,--no-warn-rwx-segments -T fxcg50.ld -lgint-cg -lc -lgcc -lgint-cg -lprof-cg
 
 INCLUDES =
 

src/main.cpp

@@ -28,13 +28,13 @@ vec3<float> cameraSpeed = {0, 0, 0};
 float cameraAngle = 0;
 
 #ifdef GINT
-static GALIGNED(32) fp depthBuffer[RENDER_WIDTH*RENDER_HEIGHT];
+static GALIGNED(32) unsigned char depthBuffer[RENDER_WIDTH*RENDER_HEIGHT];
 #include "models.h"
 #endif
 
 int main(){
 #ifndef GINT
-	fp depthBuffer[RENDER_WIDTH*RENDER_HEIGHT];
+	unsigned char depthBuffer[RENDER_WIDTH*RENDER_HEIGHT];
 #include "models.h"
 #endif
 	Rasterizer::depthBuffer = depthBuffer;
@@ -62,6 +62,8 @@ int main(){
 	Track::coneMesh = {22, cone_triangles};
 	Track::simpleConeMesh = {2, simpleConeTriangles};
 
+	Time::init();
+
 	Display::init();
 	Display::clear(newColor(70, 180, 220));
 	Display::show();
@@ -77,14 +79,14 @@ int main(){
 			{-1, -1, 0},
 			{-1, 1, 0},
 			{1, -1, 0},
-			{0, 1, 0},
+			{0, 0, -1},
 			newColor(255, 255, 0)
 		},
 		{
 			{-1, 1, 0},
 			{1, 1, 0},
 			{1, -1, 0},
-			{0, 1, 0},
+			{0, 0, -1},
 			newColor(255, 255, 0)
 		},
 	};
@@ -137,6 +139,7 @@ int main(){
 #endif
 
 #ifdef PRIZM
+	Serial_Close(1);
 	while(Serial_IsOpen() != 1){
 		unsigned char mode[6] = {0, 5, 0, 0, 0, 0}; // 9600 bps 8n1
 		Serial_Open(mode);
@@ -189,6 +192,9 @@ int main(){
 #ifdef PRIZM
 			while(Input::keyDown(KEY_MENU))
 				Input::update();
+
+			Serial_Close(1);
+
 			timer = Timer_Install(0, []() {
 					Keyboard_PutKeycode(4, 9, 0);
 					Timer_Stop(timer);
@@ -199,6 +205,14 @@ int main(){
 			Bdisp_EnableColor(1);
 			GetKey(&k);
 
+			Serial_Close(1);
+			while(Serial_IsOpen() != 1){
+				unsigned char mode[6] = {0, 5, 0, 0, 0, 0}; // 9600 bps 8n1
+				Serial_Open(mode);
+			}
+			Serial_ClearTX();
+			Serial_ClearRX();
+
 			continue;
 #endif
 #ifdef SDL

src/rasterizer.cpp

@@ -37,7 +37,7 @@ inline int max(int a, int b){
 namespace Rasterizer {
 	Plane clippingPlanes[5];
 
-	fp *depthBuffer;
+	unsigned char *depthBuffer;
 
 	fp fov_d = 1;
 
@@ -50,14 +50,21 @@ namespace Rasterizer {
 	}
 
 	void reset(){
+		unsigned char value = -1;
+#if GINT || PRIZM
+		long v = value | (value << 8) | (value << 16) | (value << 24);
+#endif
 #if GINT && PIXEL_SIZE == 1
-		fp v = -1;
-		fp *depthBuffer_P1 = (fp*) mmu_translate_uram(depthBuffer);
-		cache_ocbp(depthBuffer, RENDER_WIDTH*RENDER_HEIGHT*sizeof(fp));
-		dma_memset(depthBuffer_P1, *((uint32_t*)&v), RENDER_WIDTH*RENDER_HEIGHT*sizeof(fp));
+		unsigned char *depthBuffer_P1 = (unsigned char*) mmu_translate_uram(depthBuffer);
+		cache_ocbp(depthBuffer, RENDER_WIDTH*RENDER_HEIGHT*sizeof(unsigned char));
+		dma_memset(depthBuffer_P1, *((uint32_t*)&v), RENDER_WIDTH*RENDER_HEIGHT*sizeof(unsigned char));
+#elif PRIZM
+		for(int i = 0; i < RENDER_WIDTH*RENDER_HEIGHT/4; i++){
+			*(((long*)depthBuffer) + i) = v;
+		}
 #else
 		for(int i = 0; i < RENDER_WIDTH*RENDER_HEIGHT; i++){
-			depthBuffer[i] = -1;
+			depthBuffer[i] = value;
 		}
 #endif
 	}
@@ -98,7 +105,7 @@ namespace Rasterizer {
 	}
 
 	// Draws a triangle which has a horizontal top or bottom
-	inline void _drawFlatSideTriangle(vec3<int> points[3], fp z, Color color, bool useDepth){
+	inline void _drawFlatSideTriangle(vec3<int> points[3], unsigned char z, Color color, bool useDepth){
 		if(points[0].y == points[1].y && points[1].y == points[2].y && points[2].y == points[0].y){
 			return;
 		}
@@ -141,7 +148,7 @@ namespace Rasterizer {
 #endif
 
 				for(int x = minX; x <= maxX; x++){
-					if(z < depthBuffer[p] || depthBuffer[p] == fp(-1) || !useDepth){
+					if(z < depthBuffer[p] || !useDepth){
 						if(useDepth){
 							depthBuffer[p] = z;
 						}
@@ -177,11 +184,7 @@ namespace Rasterizer {
 			toScreen(p2_d),
 		};
 
-		if(dot(mat4::toMat3(model->viewMatrix) * mat4::toMat3(model->modelMatrix) * triangle.normal, vec3<fp>(0, 0, 1)) > 0){
-			return;
-		}
-
-		fp z = (points[0].z + points[1].z + points[2].z) / 3;
+		unsigned char z = (points[0].z + points[1].z + points[2].z) / 3;
 
 		if(isShaded){
 			fp brightness = dot(mat4::toMat3(model->modelMatrix) * triangle.normal, vec3<fp>(I_SQRT_3, -I_SQRT_3, -I_SQRT_3)) * fp(0.6) + fp(0.4);
@@ -353,6 +356,10 @@ namespace Rasterizer {
 	}
 
 	inline void drawTriangle(Model *model, Triangle triangle, bool useDepth, bool isShaded, bool clipTriangles){
+		if(dot(mat4::toMat3(model->viewMatrix) * mat4::toMat3(model->modelMatrix) * triangle.normal, vec3<fp>(0, 0, 1)) > 0){
+			return;
+		}
+
 		triangle.p0 = model->viewMatrix * model->modelMatrix * triangle.p0;
 		triangle.p1 = model->viewMatrix * model->modelMatrix * triangle.p1;
 		triangle.p2 = model->viewMatrix * model->modelMatrix * triangle.p2;
@@ -362,12 +369,14 @@ namespace Rasterizer {
 		}
 
 		int inside = 5;
-		for(int i = 0; i < 5; i++){
-			if(dot(clippingPlanes[i].n, triangle.p0) + clippingPlanes[i].d < 0
-					|| dot(clippingPlanes[i].n, triangle.p1) + clippingPlanes[i].d < 0
-					|| dot(clippingPlanes[i].n, triangle.p2) + clippingPlanes[i].d < 0){
-				inside--;
-				break;
+		if(clipTriangles){
+			for(int i = 0; i < 5; i++){
+				if(dot(clippingPlanes[i].n, triangle.p0) + clippingPlanes[i].d < 0
+						|| dot(clippingPlanes[i].n, triangle.p1) + clippingPlanes[i].d < 0
+						|| dot(clippingPlanes[i].n, triangle.p2) + clippingPlanes[i].d < 0){
+					inside--;
+					break;
+				}
 			}
 		}
 
@@ -389,8 +398,28 @@ Model::Model(){
 }
 Model::Model(Mesh mesh){
 	this->mesh = mesh;
+	fp radius2 = 0;
+	for(int i = 0; i < mesh.numTriangles; i++){
+		fp d0 = mesh.triangles[i].p0.length2();
+		fp d1 = mesh.triangles[i].p1.length2();
+		fp d2 = mesh.triangles[i].p2.length2();
+		radius2 = max(radius2, d0);
+		radius2 = max(radius2, d1);
+		radius2 = max(radius2, d2);
+	}
+	float i_radius = _isqrt(radius2);
+	radius = 1.0f/i_radius;
 }
 void Model::draw(bool useDepth, bool isShaded, bool clipTriangles){
+	if(!clipTriangles){
+		vec3<fp> center = viewMatrix * modelMatrix * vec3<fp>(0, 0, 0);
+		for(int i = 0; i < 5; i++){
+			if(dot(Rasterizer::clippingPlanes[i].n, center) + Rasterizer::clippingPlanes[i].d < radius){
+				return;
+			}
+		}
+	}
+
 	for(int i = 0; i < mesh.numTriangles; i++){
 		Rasterizer::drawTriangle(this, mesh.triangles[i], useDepth, isShaded, clipTriangles);
 	}

src/rasterizer.h

@@ -26,6 +26,7 @@ struct Mesh {
 };
 
 class Model {
+	fp radius;
 public:
 	Mesh mesh;
 	mat4 modelMatrix;
@@ -38,7 +39,7 @@ public:
 namespace Rasterizer {
 	void init();
 	void reset();
-	extern fp *depthBuffer;
+	extern unsigned char *depthBuffer;
 
 	extern fp fov_d;
 	void setFOV(int fov);

src/time-gint.cpp

@@ -1,12 +1,23 @@
 #ifdef GINT
 #include "time.h"
 #include <gint/rtc.h>
+#include <libprof.h>
+
 namespace Time {
-	void init(){}
+	prof_t prof;
+	void init(){
+		prof_init();
+		prof = prof_make();
+	}
 	void update(){
+		prof_leave(prof);
+
 		const float lastTime = time;
 		time = rtc_ticks();
-		delta = time - lastTime;
+		delta = prof_time(prof) / 1000.0f / (1000.0f / 128.0f);
+
+		prof = prof_make();
+		prof_enter(prof);
 	}
 };
 #endif

src/time-prizm.cpp

@@ -3,6 +3,8 @@
 #include <fxcg/rtc.h>
 
 namespace Time {
+	void init(){
+	}
 	void update(){
 		const float lastTime = time;
 		time = RTC_GetTicks();

src/time-sdl.cpp

@@ -3,6 +3,8 @@
 #include <SDL2/SDL.h>
 
 namespace Time {
+	void init(){
+	}
 	void update(){
 		const float lastTime = time;
 		time = ((float)(SDL_GetTicks()) / (1000.0/128.0));

src/time.h

@@ -3,5 +3,6 @@
 namespace Time {
 	extern float time;
 	extern float delta;
+	void init();
 	void update();
 };