1 files changed, 75 insertions, 25 deletions

diff --git a/src/swgfx.c b/src/swgfx.c
index ddd64f7..6074eac 100644
--- a/src/swgfx.c
+++ b/src/swgfx.c
@@ -24,7 +24,8 @@ static constexpr size_t SWGFX_TEXTURE_REGISTER_SIZE = SWGFX_MAX_TEXTURES + 1;
 
 static constexpr R DepthClearValue = 1.0f;
 
-static constexpr sgVec3 Up3 = (sgVec3){0,1,0};
+static constexpr sgVec3 Up3   = (sgVec3){0,1,0};
+static constexpr sgVec3 Zero3 = (sgVec3){0,0,0};
 
 typedef struct sgViewport_t { int x0, y0, width, height; } sgViewport_t;
 typedef struct sgAABB2      { sgVec2 pmin, pmax; }         sgAABB2;
@@ -46,6 +47,7 @@ typedef struct swgfx {
   sgTextureId*  texture;   // Texture ID buffer.
   sgVec2*       texcoords; // Texture coords buffer.
   sgPixel*      albedo;    // Albedo buffer.
+  sgVec3*       normals;   // Normals buffer.
   sgViewport_t  viewport;
   sgMat4        model;     // Model matrix.
   sgMat4        view;      // View matrix.
@@ -87,12 +89,14 @@ static inline sgVec2  frac2(sgVec2 v) { return (sgVec2){frac(v.x), frac(v.y)}; }
 static inline sgVec2  lerp2(sgVec2 a, sgVec2 b, R t) { return add2(a, scale2(sub2(b,a), t)); }
 static inline sgVec2  mod2(sgVec2 v, R m) { return (sgVec2){mod1(v.x, m), mod1(v.y, m)}; }
 
+static inline sgVec3 max3(sgVec3 a, sgVec3 b) { return (sgVec3){fmaxf(a.x, b.x), fmaxf(a.y, b.y), fmaxf(a.z, b.z)}; }
 static inline sgVec3 add3(sgVec3 a, sgVec3 b) { return (sgVec3){a.x + b.x, a.y + b.y, a.z + b.z}; }
 static inline sgVec3 neg3(sgVec3 v) { return (sgVec3){-v.x, -v.y, -v.z}; }
 static inline sgVec3 sub3(sgVec3 a, sgVec3 b) { return (sgVec3){a.x - b.x, a.y - b.y, a.z - b.z}; }
 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 exp3(sgVec3 v, R exp) { return (sgVec3){powf(v.x, exp), powf(v.y, exp), powf(v.z, exp)};}
 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; }
@@ -326,11 +330,13 @@ static inline void SetPixelColour(swgfx* gfx, const sgVec2i p, sgPixel colour) {
 #endif // SWGFX_PROFILING
 }
 
-static inline void SetPixelDeferred(swgfx* gfx, const sgVec2i p, R depth, sgTextureId texid, sgVec2 uv) {
+static inline void SetPixelDeferred(swgfx* gfx, const sgVec2i p, R depth, sgTextureId texid, sgVec2 uv, sgVec3 normal) {
   assert(gfx);
-  gfx->depth[(p.y * gfx->dims.x) + p.x]     = depth;
-  gfx->texture[(p.y * gfx->dims.x) + p.x]   = texid;
-  gfx->texcoords[(p.y * gfx->dims.x) + p.x] = uv;
+  const int i = (p.y * gfx->dims.x) + p.x;
+  gfx->depth[i]     = depth;
+  gfx->texture[i]   = texid;
+  gfx->texcoords[i] = uv;
+  gfx->normals[i]   = normalize3(normal); // Non-unit after interpolation.
 #if SWGFX_PROFILING
   gfx->counters.pixels++;
 #endif // SWGFX_PROFILING
@@ -514,7 +520,7 @@ static inline R IntersectSegmentPlane(R near, const sgVec3* const a, const sgVec
   return t;
 }
 
-/// Interpolate depth and vertex attributes at the in/out vertex 'out'.
+/// Interpolate vertex attributes at the in/out vertex 'out'.
 static void InterpolateAttributes(const sgVert4* const a, const sgVert4* const b, R t, sgVert4* out) {
   assert(a);
   assert(b);
@@ -522,8 +528,9 @@ static void InterpolateAttributes(const sgVert4* const a, const sgVert4* const b
   assert(t >= 0.f);
   assert(t <= 1.f);
   const sgVec4 d = sub4(b->pos, a->pos); // Line direction.
-  out->pos = add4(a->pos, scale4(d, t));
-  out->uv  = lerp2(a->uv, b->uv, t);
+  out->pos    = add4(a->pos, scale4(d, t));
+  out->uv     = lerp2(a->uv, b->uv, t);
+  out->normal = lerp3(a->normal, b->normal, t);
 }
 
 /// Clip a triangle, vertices in clip space. Return the number of output
@@ -597,10 +604,18 @@ static inline int TransformTri(const swgfx* gfx, const sgTri3* const tri, sgTri4
   const sgVec4 p0_clip = Mat4MulVec4(gfx->mvp, Vec4FromVec3(tri->p0.pos, 1));
   const sgVec4 p1_clip = Mat4MulVec4(gfx->mvp, Vec4FromVec3(tri->p1.pos, 1));
   const sgVec4 p2_clip = Mat4MulVec4(gfx->mvp, Vec4FromVec3(tri->p2.pos, 1));
+  // Model to world space for normals.
+  // This assumes the model matrix does not have non-uniform scaling.
+  // It seems more convenient to put the normals in world space instead of view
+  // space right now so that we don't have to transform lights to view space
+  // when lighting. But we might want to revisit this later.
+  const sgVec3 n0_view = Mat4MulVec3(gfx->model, tri->p0.normal, 0.f);
+  const sgVec3 n1_view = Mat4MulVec3(gfx->model, tri->p1.normal, 0.f);
+  const sgVec3 n2_view = Mat4MulVec3(gfx->model, tri->p2.normal, 0.f);
   const sgTri4 tri_clip = {
-    (sgVert4){ p0_clip, tri->p0.uv },
-    (sgVert4){ p1_clip, tri->p1.uv },
-    (sgVert4){ p2_clip, tri->p2.uv }};
+    (sgVert4){ p0_clip, tri->p0.uv, n0_view },
+    (sgVert4){ p1_clip, tri->p1.uv, n1_view },
+    (sgVert4){ p2_clip, tri->p2.uv, n2_view }};
   // Clip.
   // Our perspective matrix maps the near plane to z=-1 in clip space.
   constexpr R near_clip = -1.f;
@@ -673,20 +688,22 @@ static void DrawTriangle3PostClip(swgfx* gfx, const sgTri4* const tri) {
       if ((bar.x >= 0) && (bar.y >= 0) && (bar.z >= 0)) {
         assert((bar.x + bar.y + bar.z - 1e7) <= 1.f);
         const R p_one_over_z = dot3(bar, one_over_zs);
-        const R p_u_over_z   = dot3(bar, u_over_zs);
-        const R p_v_over_z   = dot3(bar, v_over_zs);
         const R p_depth      = dot3(bar, depths);
         const R z            = 1.f / p_one_over_z;
-        const sgVec2 uv      = (sgVec2){p_u_over_z * z, p_v_over_z * z};
         const R* depth = Depth(gfx, x, y);
         if ((0.f <= p_depth) && (p_depth <= 1.f) && (p_depth <= *depth)) {
-          // TODO: When doing lighting, need to tone-map here and apply inverse
-          //  gamma here.
+          const R p_u_over_z  = dot3(bar, u_over_zs);
+          const R p_v_over_z  = dot3(bar, v_over_zs);
+          const sgVec2 uv     = (sgVec2){p_u_over_z * z, p_v_over_z * z};
+          const sgVec3 n0     = scale3(tri->p0.normal, bar.x * one_over_zs.x);
+          const sgVec3 n1     = scale3(tri->p1.normal, bar.y * one_over_zs.y);
+          const sgVec3 n2     = scale3(tri->p2.normal, bar.z * one_over_zs.z);
+          const sgVec3 normal = add3(add3(n0,n1),n2);
           //const sgPixel colour = {(uint8_t)(bar.x*255.f), (uint8_t)(bar.y*255.f), (uint8_t)(bar.z*255.f), 255};
           //const sgPixel colour = {(int)(z*255.f), (int)(z*255.f), (int)(z*255.f), 255};
           //const sgPixel colour = {255, 0, 255, 255};
           //const sgPixel colour = {(int)(uv.x * 255.f), (int)(uv.y * 255.f), 255, 255};
-          SetPixelDeferred(gfx, (sgVec2i){x,y}, p_depth, gfx->activeTexture, uv);
+          SetPixelDeferred(gfx, (sgVec2i){x,y}, p_depth, gfx->activeTexture, uv, normal);
         }
       }
     }
@@ -742,6 +759,7 @@ size_t sgMem(int width, int height) {
          Align(N * sizeof(sgTextureId)) + // Texture ID buffer.
          Align(N * sizeof(sgVec2)) +      // Texture coords buffer.
          Align(N * sizeof(sgPixel)) +     // Albedo buffer.
+         Align(N * sizeof(sgVec3)) +      // Normals buffer.
          Align(SWGFX_TEXTURE_REGISTER_SIZE * sizeof(sgTexture)) + // Texture register.
          (SG_ALIGN - 1); // To make room to align allocations within the buffer.
 }
@@ -756,6 +774,7 @@ swgfx* sgNew(int width, int height, void* mem) {
   gfx->texture   = SG_ALLOC(&aligned, N, sgTextureId);
   gfx->texcoords = SG_ALLOC(&aligned, N, sgVec2);
   gfx->albedo    = SG_ALLOC(&aligned, N, sgPixel);
+  gfx->normals   = SG_ALLOC(&aligned, N, sgVec3);
   gfx->textureRegister = SG_ALLOC(&aligned, SWGFX_TEXTURE_REGISTER_SIZE, sgTexture);
   gfx->activeTexture = DefaultTextureId;
   gfx->defaultPixel  = (sgPixel){255, 255, 255, 255};
@@ -922,11 +941,12 @@ void sgTriangles(swgfx* gfx, size_t count, const sgTri3* tris, const sgNormal*)
   }
 }
 
-void sgTrianglesIndexed(swgfx* gfx, size_t numIndices, const sgIdx* indices, const sgVec3* positions, const sgVec2* texcoords) {
+void sgTrianglesIndexed(swgfx* gfx, size_t numIndices, const sgIdx* indices, const sgVec3* positions, const sgVec2* texcoords, const sgVec3* normals) {
   assert(gfx);
   assert(indices);
   assert(positions);
   assert(texcoords);
+  assert(normals);
   for (size_t i = 0; i < numIndices; i+=3) {
     const sgIdx i0 = indices[i];
     const sgIdx i1 = indices[i+1];
@@ -937,25 +957,29 @@ void sgTrianglesIndexed(swgfx* gfx, size_t numIndices, const sgIdx* indices, con
     const sgVec2 uv0 = texcoords[i0];
     const sgVec2 uv1 = texcoords[i1];
     const sgVec2 uv2 = texcoords[i2];
+    const sgVec3 n0  = normals[i0];
+    const sgVec3 n1  = normals[i1];
+    const sgVec3 n2  = normals[i2];
     const sgTri3 tri = (sgTri3){
-      (sgVert3){p0, uv0},
-      (sgVert3){p1, uv1},
-      (sgVert3){p2, uv2}};
+      (sgVert3){p0, uv0, n0},
+      (sgVert3){p1, uv1, n1},
+      (sgVert3){p2, uv2, n2}};
     DrawTriangle3(gfx, &tri);
   }
 }
 
-void sgTrianglesIndexedNonUniform(swgfx* gfx, size_t numTris, const sgTriIdx* tris, const sgVec3* positions, const sgVec2* texcoords) {
+void sgTrianglesIndexedNonUniform(swgfx* gfx, size_t numTris, const sgTriIdx* tris, const sgVec3* positions, const sgVec2* texcoords, const sgVec3* normals) {
   assert(gfx);
   assert(tris);
   assert(positions);
   assert(texcoords);
+  assert(normals);
   for (size_t t = 0; t < numTris; ++t) {
     const sgTriIdx* triIdx = &tris[t];
     const sgTri3 tri = (sgTri3){
-      (sgVert3){positions[triIdx->v0.pos], texcoords[triIdx->v0.uv]},
-      (sgVert3){positions[triIdx->v1.pos], texcoords[triIdx->v1.uv]},
-      (sgVert3){positions[triIdx->v2.pos], texcoords[triIdx->v2.uv]}};
+      (sgVert3){positions[triIdx->v0.pos], texcoords[triIdx->v0.uv], normals[triIdx->v0.normal]},
+      (sgVert3){positions[triIdx->v1.pos], texcoords[triIdx->v1.uv], normals[triIdx->v1.normal]},
+      (sgVert3){positions[triIdx->v2.pos], texcoords[triIdx->v2.uv], normals[triIdx->v2.normal]}};
     DrawTriangle3(gfx, &tri);
   }
 }
@@ -996,6 +1020,32 @@ void sgAmbient(swgfx* gfx, sgVec3 ambient) {
   }
 }
 
+void sgDepth(swgfx* gfx) {
+  assert(gfx);
+  const int N = gfx->dims.x * gfx->dims.y;
+  for (int i = 0; i < N; ++i) {
+    const R depth = gfx->depth[i];
+    if (depth != DepthClearValue) {
+      sgPixel* colour = &gfx->colour[i];
+      const R d       = gfx->depth[i];
+      *colour = Vec3ToPixel((sgVec3){d,d,d}, 1.f);
+    }
+  }
+}
+
+void sgNormals(swgfx* gfx) {
+  assert(gfx);
+  const int N = gfx->dims.x * gfx->dims.y;
+  for (int i = 0; i < N; ++i) {
+    const R depth = gfx->depth[i];
+    if (depth != DepthClearValue) {
+      sgPixel* colour      = &gfx->colour[i];
+      const sgVec3* normal = &gfx->normals[i];
+      *colour = Vec3ToPixel(max3(Zero3, *normal), 1.f);
+    }
+  }
+}
+
 void sgGamma(swgfx* gfx, sgPixel* pixels, int width, int height) {
   assert(gfx);
   assert(pixels);