shithub: tinygl

--- a/include/zfeatures.h

+++ b/include/zfeatures.h

@@ -75,7 +75,11 @@

 #define TGL_FEATURE_MULTITHREADED_ZB_COPYBUFFER 0

 //Enable stdalign

 #define TGL_FEATURE_ALIGNAS 1

+//Optimization hint- cost of branching.

+//0- branching has zero cost, avoid extraneous code.

+//1- Branching has some cost, allow some extraneous code

+//2- Branching has extreme cost, allow a lot of extraneous code. Modern processors work best on this setting.

+#define TGL_OPTIMIZATION_HINT_BRANCH_COST 2

 //Disable it for TinyC

 #ifdef __TINYC__

--- a/src/clip.c

+++ b/src/clip.c

@@ -20,13 +20,16 @@

 		v->zp.z = (GLint)(v->pc.Z * winv * c->viewport.scale.Z + c->viewport.trans.Z);

 	/* color */

-	v->zp.r = (GLint)(v->color.v[0] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

-	v->zp.g = (GLint)(v->color.v[1] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

-	v->zp.b = (GLint)(v->color.v[2] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

+	v->zp.r = (GLuint)(v->color.v[0] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

+	v->zp.g = (GLuint)(v->color.v[1] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

+	v->zp.b = (GLuint)(v->color.v[2] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

 	/* texture */

-	if (c->texture_2d_enabled) {

+#if TGL_OPTIMIZATION_HINT_BRANCH_COST < 1

+	if (c->texture_2d_enabled)

+#endif

+	{

 		v->zp.s = (GLint)(v->tex_coord.X * (ZB_POINT_S_MAX - ZB_POINT_S_MIN) + ZB_POINT_S_MIN); //MARKED

 		v->zp.t = (GLint)(v->tex_coord.Y * (ZB_POINT_T_MAX - ZB_POINT_T_MIN) + ZB_POINT_T_MIN); //MARKED

@@ -138,7 +141,7 @@

 	q->pc.Y = p0->pc.Y + (p1->pc.Y - p0->pc.Y) * t;

 	q->pc.Z = p0->pc.Z + (p1->pc.Z - p0->pc.Z) * t;

 	q->pc.W = p0->pc.W + (p1->pc.W - p0->pc.W) * t;

+#pragma omp simd

 	for(int i = 0; i < 3; i++)

 		q->color.v[i] = p0->color.v[i] + (p1->color.v[i] - p0->color.v[i]) * t;

--- a/src/get.c

+++ b/src/get.c

@@ -56,6 +56,7 @@

 const GLubyte* extensions_string = (const GLubyte*)"TGL_TEXTURE "

 "TGL_SMOOTHSHADING "

 "TGL_LIGHTING "

+"TGL_OPTIMIZATION_HINT_BRANCH_COST=" xstr(TGL_OPTIMIZATION_HINT_BRANCH_COST) " "

 #if TGL_FEATURE_ERROR_CHECK == 1

 "TGL_FEATURE_ERROR_CHECK "

 #endif

--- a/src/light.c

+++ b/src/light.c

@@ -280,7 +280,7 @@

 			d.Y = l->position.v[1] - v->ec.v[1];

 			d.Z = l->position.v[2] - v->ec.v[2];

 #if TGL_FEATURE_FISR == 1

-			tmp = clampf(fastInvSqrt(d.X * d.X + d.Y * d.Y + d.Z * d.Z),0,1); //FISR IMPL, MATCHED!

+			tmp = fastInvSqrt(d.X * d.X + d.Y * d.Y + d.Z * d.Z); //FISR IMPL, MATCHED!

 				d.X *= tmp;

 				d.Y *= tmp;

@@ -354,7 +354,7 @@

 #endif

 					dot_spec = clampf(dot_spec, 0, 1);

 #if TGL_FEATURE_FISR == 1

-					tmp = clampf(fastInvSqrt(s.X * s.X + s.Y * s.Y + s.Z * s.Z),0,1); //FISR IMPL, MATCHED!

+					tmp = fastInvSqrt(s.X * s.X + s.Y * s.Y + s.Z * s.Z); //FISR IMPL, MATCHED!

 					//if (tmp < 1E+3)

 						dot_spec = dot_spec * tmp;

--- a/src/vertex.c

+++ b/src/vertex.c

@@ -68,15 +68,15 @@

 			gl_M4_Inv(&tmp, c->matrix_stack_ptr[0]);

 			gl_M4_Transpose(&c->matrix_model_view_inv, &tmp);

 		} else {

-			//GLfloat* m = &c->matrix_model_projection.m[0][0];

+			GLfloat* m = &c->matrix_model_projection.m[0][0];

 			/* precompute projection matrix */

 			gl_M4_Mul(&c->matrix_model_projection, c->matrix_stack_ptr[1], c->matrix_stack_ptr[0]);

 			/* test to accelerate computation */

 			c->matrix_model_projection_no_w_transform = 0;

-			//if (m[12] == 0.0 && m[13] == 0.0 && m[14] == 0.0)

-			if(c->matrix_model_projection.m[3][0] == 0.0 &&

-				c->matrix_model_projection.m[3][1] == 0.0 &&

-				c->matrix_model_projection.m[3][2] == 0.0)

+			if (m[12] == 0.0 && m[13] == 0.0 && m[14] == 0.0)

+			//if(c->matrix_model_projection.m[3][0] == 0.0 &&

+			//	c->matrix_model_projection.m[3][1] == 0.0 &&

+			//	c->matrix_model_projection.m[3][2] == 0.0)

 				c->matrix_model_projection_no_w_transform = 1;

@@ -137,13 +137,15 @@

 		v->zp.z = (GLint)(v->pc.Z * winv * c->viewport.scale.Z + c->viewport.trans.Z);

-	v->zp.r = (GLint)(v->color.v[0] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

-	v->zp.g = (GLint)(v->color.v[1] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

-	v->zp.b = (GLint)(v->color.v[2] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

+	v->zp.r = (GLuint)(v->color.v[0] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

+	v->zp.g = (GLuint)(v->color.v[1] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

+	v->zp.b = (GLuint)(v->color.v[2] * COLOR_CORRECTED_MULT_MASK + COLOR_MIN_MULT) & COLOR_MASK;

-	if (c->texture_2d_enabled) {

+//#if TGL_OPTIMIZATION_HINT_BRANCH_COST < 1

+	if (c->texture_2d_enabled)

+//#endif

+	{

 		v->zp.s = (GLint)(v->tex_coord.X * (ZB_POINT_S_MAX - ZB_POINT_S_MIN) + ZB_POINT_S_MIN); //MARKED

 		v->zp.t = (GLint)(v->tex_coord.Y * (ZB_POINT_T_MAX - ZB_POINT_T_MIN) + ZB_POINT_T_MIN); //MARKED

@@ -153,7 +155,10 @@

 	GLfloat* m;

 	GLContext* c = gl_get_context();

-	if (c->lighting_enabled) {

+//#if TGL_OPTIMIZATION_HINT_BRANCH_COST < 2

+	if (c->lighting_enabled)

+//#endif

+	{

 		/* eye coordinates needed for lighting */

 		V4* n;

 		m = &c->matrix_stack_ptr[0]->m[0][0];

@@ -179,7 +184,9 @@

 		if (c->normalize_enabled) {

 			gl_V3_Norm_Fast(&v->normal);

-	} else {

+	}

+//#if TGL_OPTIMIZATION_HINT_BRANCH_COST < 2

+	else {

 		/* no eye coordinates needed, no normal */

 		/* NOTE: W = 1 is assumed */

 		m = &c->matrix_model_projection.m[0][0];

@@ -193,7 +200,7 @@

 			v->pc.W = (v->coord.X * m[12] + v->coord.Y * m[13] + v->coord.Z * m[14] + m[15]);

+//#endif

 	v->clip_code = gl_clipcode(v->pc.X, v->pc.Y, v->pc.Z, v->pc.W);

--- a/src/zmath.c

+++ b/src/zmath.c

@@ -9,6 +9,7 @@

 /* ******* Gestion des matrices 4x4 ****** */

 void gl_M4_Id(M4* a) {

+/*

 	GLint i, j;

 #pragma omp simd collapse(2)

 	for (i = 0; i < 4; i++)

@@ -17,6 +18,14 @@

 				a->m[i][j] = 1.0;

 			else

 				a->m[i][j] = 0.0;

+*/

+const M4 c = (M4){{

+	{1,0,0,0},

+	{0,1,0,0},

+	{0,0,1,0},

+	{0,0,0,1},}

+	};

+*a = c;

 GLint gl_M4_IsId(M4* a) {

@@ -211,7 +220,7 @@

 void gl_M4_Inv(M4* a, M4* b) {

 	M4 tmp;

-	memcpy(&tmp, b, 16 * sizeof(GLfloat));

+	memcpy(&tmp, b, sizeof(M4));

 	/*tmp=*b;*/

 	Matrix_Inv(&a->m[0][0], &tmp.m[0][0], 4);

--- a/src/zmath.h

+++ b/src/zmath.h

@@ -88,7 +88,7 @@

 #endif

-static inline GLint gl_V3_Norm_Fast(V3* a) {

+static inline int gl_V3_Norm_Fast(V3* a) {

 	GLfloat n;

 #if TGL_FEATURE_FISR == 1

 	n = fastInvSqrt(a->X * a->X + a->Y * a->Y + a->Z * a->Z); //FISR