ref: 8207a583af54b780757a278914e1deae8f2d1040
parent: 18949a3f82797cdde959037b56781e1437b56176
author: David <gek@katherine>
date: Fri Feb 19 11:53:15 EST 2021
Before ztriangle.h overhaul
--- a/include/zfeatures.h
+++ b/include/zfeatures.h
@@ -11,6 +11,8 @@
//This slows down the renderer so we don't usually do it, but
//it's part of the GL spec.
#define TGL_FEATURE_STRICT_OOM_CHECKS 0
+//Use Fast Inverse Square Root. Toggleable because it's actually slower on some systems, i've heard.
+#define TGL_FEATURE_FISR 1
//Clientside Arrays
#define TGL_FEATURE_ARRAYS 1
//This library has a super cool opcode system based on function pointers, it's pretty fast too.
--- a/src/get.c
+++ b/src/get.c
@@ -115,6 +115,9 @@
#if TGL_FEATURE_STRICT_OOM_CHECKS == 1
"TGL_FEATURE_STRICT_OOM_CHECKS "
#endif
+#if TGL_FEATURE_FISR == 1
+"TGL_FEATURE_FISR "
+#endif
#if TGL_FEATURE_ARRAYS == 1
"TGL_FEATURE_ARRAYS "
#endif
--- a/src/light.c
+++ b/src/light.c
@@ -274,15 +274,24 @@
d.X = l->position.v[0] - v->ec.v[0];
d.Y = l->position.v[1] - v->ec.v[1];
d.Z = l->position.v[2] - v->ec.v[2];
- tmp = fastInvSqrt(d.X * d.X + d.Y * d.Y + d.Z * d.Z);
- //dist = sq_rt(d.X * d.X + d.Y * d.Y + d.Z * d.Z);
+#if TGL_FEATURE_FISR == 1
+ tmp = fastInvSqrt(d.X * d.X + d.Y * d.Y + d.Z * d.Z); //FISR IMPL, MATCHED!
dist = 1.0f/tmp;
if (dist > 1E-3) {+ d.X *= tmp;
+ d.Y *= tmp;
+ d.Z *= tmp;
+ }
+#else
+ dist = sqrt(d.X * d.X + d.Y * d.Y + d.Z * d.Z);
+ //dist = 1.0f/tmp;
+ if (dist > 1E-3) {tmp = 1 / dist;
d.X *= tmp;
d.Y *= tmp;
d.Z *= tmp;
}
+#endif
att = 1.0f / (l->attenuation[0] + dist * (l->attenuation[1] + dist * l->attenuation[2]));
}
dot = d.X * n.X + d.Y * n.Y + d.Z * n.Z;
@@ -335,12 +344,19 @@
if (dot_spec > 0) {GLSpecBuf* specbuf;
GLint idx;
- //tmp = sqrt(s.X * s.X + s.Y * s.Y + s.Z * s.Z);
- tmp = fastInvSqrt(s.X * s.X + s.Y * s.Y + s.Z * s.Z);
+#if TGL_FEATURE_FISR == 1
+ 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;
dot_spec = dot_spec * tmp;
} else dot_spec = 0;
+#else
+ //reference implementation.
+ tmp=sqrt(s.X*s.X+s.Y*s.Y+s.Z*s.Z);
+ if (tmp > 1E-3) {+ dot_spec=dot_spec / tmp;
+ } else dot_spec = 0;
+#endif
+
/* TODO: optimize */
/* testing specular buffer code */
--- a/src/matrix.c
+++ b/src/matrix.c
@@ -134,15 +134,18 @@
GLfloat cost, sint;
/* normalize vector */
+
+#if TGL_FEATURE_FISR == 1
GLfloat len = u[0] + u[1] + u[2];
if (len == 0.0f)
return;
-/*OLD
-
-*/
-//NEW
- len = fastInvSqrt(len);
- //len = 1.0f / sqrt(len);
+ len = fastInvSqrt(len); //FISR
+#else
+ GLfloat len = u[0] * u[0] + u[1] * u[1] + u[2] * u[2];
+ if (len == 0.0f)
+ return;
+ len = 1.0f / sqrt(len);
+#endif
u[0] *= len;
u[1] *= len;
u[2] *= len;
--- a/src/zmath.c
+++ b/src/zmath.c
@@ -239,9 +239,16 @@
int gl_V3_Norm_Fast(V3* a) {GLfloat n;
- n = fastInvSqrt(a->X * a->X + a->Y * a->Y + a->Z * a->Z);
+#if TGL_FEATURE_FISR == 1
+ n = fastInvSqrt(a->X * a->X + a->Y * a->Y + a->Z * a->Z); //FISR
+ if(n>1E+3)
+ return 1;
+#else
+ n = sqrt(a->X * a->X + a->Y * a->Y + a->Z * a->Z); //NONFISR
if (n == 0)
return 1;
+ n = 1.0 / n;
+#endif
a->X *= n;
a->Y *= n;
a->Z *= n;
--- a/src/zmath.h
+++ b/src/zmath.h
@@ -52,7 +52,7 @@
int gl_Matrix_Inv(GLfloat* r, GLfloat* m, GLint n);
-
+/*
static inline GLfloat fastInvSqrt(float x){GLint i; GLfloat y;
memcpy(&i, &x, 4);
@@ -61,5 +61,15 @@
memcpy(&y, &i, 4);
return y * (1.5F - 0.5F * x * y * y);
}
+*/
+#if TGL_FEATURE_FISR == 1
+static inline GLfloat fastInvSqrt(float x){+ union{GLfloat f; GLint i;} conv;+ conv.f = x;
+ conv.i = 0x5F1FFFF9 - (conv.i>>1);
+ conv.f *= 0.703952253f * (2.38924456f - x * conv.f * conv.f);
+ return conv.f;
+}
+#endif
#endif
// __ZMATH__