Faster gammaHEADmain

1 files changed, 10 insertions, 9 deletions

diff --git a/src/swgfx.c b/src/swgfx.c
index 1901057..1e22ade 100644
--- a/src/swgfx.c
+++ b/src/swgfx.c
@@ -96,8 +96,10 @@ static inline sgVec3 sub3(sgVec3 a, sgVec3 b) { return (sgVec3){a.x - b.x, a.y -
 static inline sgVec3 mul3(sgVec3 a, sgVec3 b) { return (sgVec3){a.x * b.x, a.y * b.y, a.z * b.z}; }
 static inline sgVec3 div3(sgVec3 a, sgVec3 b) { return (sgVec3){a.x / b.x, a.y / b.y, a.z / b.z}; }
 static inline sgVec3 scale3(sgVec3 v, R s) { return (sgVec3){v.x * s, v.y * s, v.z * s}; }
-static inline sgVec3  lerp3(sgVec3 a, sgVec3 b, R t) { return add3(a, scale3(sub3(b,a), t)); }
+static inline sgVec3 lerp3(sgVec3 a, sgVec3 b, R t) { return add3(a, scale3(sub3(b,a), t)); }
 static inline sgVec3 exp3(sgVec3 v, R exp) { return (sgVec3){powf(v.x, exp), powf(v.y, exp), powf(v.z, exp)};}
+static inline sgVec3 square3(sgVec3 v) { return (sgVec3){v.x*v.x, v.y*v.y, v.z*v.z }; }
+static inline sgVec3 sqrt3  (sgVec3 v) { return (sgVec3){sqrt(v.x), sqrt(v.y), sqrt(v.z) }; }
 static inline R dot3(sgVec3 a, sgVec3 b) { return a.x * b.x + a.y * b.y + a.z * b.z; }
 static inline R normsq3(sgVec3 v) { return v.x * v.x + v.y * v.y + v.z * v.z; }
 static inline R norm3  (sgVec3 v) { return (R)sqrt(normsq3(v)); }
@@ -1058,23 +1060,22 @@ void sgNormals(swgfx* gfx) {
   }
 }
 
-void sgGamma(swgfx* gfx, sgTexel* texels, int width, int height) {
-  assert(gfx);
+void sgGamma(sgTexel* texels, int width, int height) {
   assert(texels);
-  constexpr R exp = 2.2f;
   for (int i = 0; i < width * height; ++i) {
     sgTexel* const p = &texels[i];
-    *p = Colour3ToTexel(exp3(TexelToColour3(*p), exp), p->a);
+    //*p = Colour3ToTexel(exp3(TexelToColour3(*p), 2.2f), p->a);
+    *p = Colour3ToTexel(square3(TexelToColour3(*p)), p->a);
   }
 }
 
-void sgGammaInv(swgfx* gfx, sgColour4* pixels, int width, int height) {
-  assert(gfx);
+void sgGammaInv(sgColour4* pixels, int width, int height) {
   assert(pixels);
-  constexpr R exp = 1.0f/2.2f;
   for (int i = 0; i < width * height; ++i) {
     sgColour4* const p = &pixels[i];
-    *p = Vec4FromVec3(exp3(Vec3FromVec4(*p), exp), p->a);
+    // sqrt() is faster (~500us at 320x240) than pow().
+    //*p = Vec4FromVec3(exp3(Vec3FromVec4(*p), 1.0f/2.2f), p->a);
+    *p = Vec4FromVec3(sqrt3(Vec3FromVec4(*p)), p->a);
   }
 }