Update SkLightingShader to support rotation

This also:
makes the SkLightingShader handle normal maps where the rects aren't aligned between the diffuse and normal maps.

adds a light aggregating class (Lights) to SkLightingShader (along with a Builder nested class).

Split out of https://codereview.chromium.org/1261433009/ (Add SkCanvas::drawLitAtlas call)

Committed: https://skia.googlesource.com/skia/+/45b59ed6e4e231814dbdb9f707b3d2a7ee50de84

Review URL: https://codereview.chromium.org/1291783003

1 files changed, 80 insertions, 0 deletions

diff --git a/src/core/SkPoint3.cpp b/src/core/SkPoint3.cpp
new file mode 100644
index 0000000000..3b5586b067
--- /dev/null
+++ b/src/core/SkPoint3.cpp
@@ -0,0 +1,80 @@
+/*
+ * Copyright 2015 Google Inc.
+ *
+ * Use of this source code is governed by a BSD-style license that can be
+ * found in the LICENSE file.
+ */
+
+#include "SkPoint3.h"
+
+// Returns the square of the Euclidian distance to (x,y,z).
+static inline float get_length_squared(float x, float y, float z) {
+    return x * x + y * y + z * z;
+}
+
+// Calculates the square of the Euclidian distance to (x,y,z) and stores it in
+// *lengthSquared.  Returns true if the distance is judged to be "nearly zero".
+//
+// This logic is encapsulated in a helper method to make it explicit that we
+// always perform this check in the same manner, to avoid inconsistencies
+// (see http://code.google.com/p/skia/issues/detail?id=560 ).
+static inline bool is_length_nearly_zero(float x, float y, float z, float *lengthSquared) {
+    *lengthSquared = get_length_squared(x, y, z);
+    return *lengthSquared <= (SK_ScalarNearlyZero * SK_ScalarNearlyZero);
+}
+
+SkScalar SkPoint3::Length(SkScalar x, SkScalar y, SkScalar z) {
+    float magSq = get_length_squared(x, y, z);
+    if (SkScalarIsFinite(magSq)) {
+        return sk_float_sqrt(magSq);
+    } else {
+        double xx = x;
+        double yy = y;
+        double zz = z;
+        return (float)sqrt(xx * xx + yy * yy + zz * zz);
+    }
+}
+
+/*
+ *  We have to worry about 2 tricky conditions:
+ *  1. underflow of magSq (compared against nearlyzero^2)
+ *  2. overflow of magSq (compared w/ isfinite)
+ *
+ *  If we underflow, we return false. If we overflow, we compute again using
+ *  doubles, which is much slower (3x in a desktop test) but will not overflow.
+ */
+bool SkPoint3::normalize() {
+    float magSq;
+    if (is_length_nearly_zero(fX, fY, fZ, &magSq)) {
+        this->set(0, 0, 0);
+        return false;
+    }
+
+    float scale;
+    if (SkScalarIsFinite(magSq)) {
+        scale = 1.0f / sk_float_sqrt(magSq);
+    } else {
+        // our magSq step overflowed to infinity, so use doubles instead.
+        // much slower, but needed when x, y or z is very large, otherwise we
+        // divide by inf. and return (0,0,0) vector.
+        double xx = fX;
+        double yy = fY;
+        double zz = fZ;
+#ifdef SK_CPU_FLUSH_TO_ZERO
+        // The iOS ARM processor discards small denormalized numbers to go faster.
+        // Casting this to a float would cause the scale to go to zero. Keeping it
+        // as a double for the multiply keeps the scale non-zero.
+        double dscale = 1.0f / sqrt(xx * xx + yy * yy + zz * zz);
+        fX = x * dscale;
+        fY = y * dscale;
+        fZ = z * dscale;
+        return true;
+#else
+        scale = (float)(1.0f / sqrt(xx * xx + yy * yy + zz * zz));
+#endif
+    }
+    fX *= scale;
+    fY *= scale;
+    fZ *= scale;
+    return true;
+}