Cleanup: use float3 SSE instead of ssef for voronoi texture.

3 files changed, 40 insertions, 0 deletions

diff --git a/intern/cycles/util/util_math.h b/intern/cycles/util/util_math.h
index fd3199f209f..85cbd18b7ba 100644
--- a/intern/cycles/util/util_math.h
+++ b/intern/cycles/util/util_math.h
@@ -266,6 +266,11 @@ ccl_device_inline int floor_to_int(float f)
 	return float_to_int(floorf(f));
 }
 
+ccl_device_inline int quick_floor_to_int(float x)
+{
+	return float_to_int(x) - ((x < 0) ? 1 : 0);
+}
+
 ccl_device_inline int ceil_to_int(float f)
 {
 	return float_to_int(ceilf(f));
@@ -550,6 +555,11 @@ ccl_device_inline float xor_signmask(float x, int y)
 	return __int_as_float(__float_as_int(x) ^ y);
 }
 
+ccl_device float bits_to_01(uint bits)
+{
+	return bits * (1.0f/(float)0xFFFFFFFF);
+}
+
 /* projections */
 ccl_device_inline float2 map_to_tube(const float3 co)
 {
diff --git a/intern/cycles/util/util_math_float3.h b/intern/cycles/util/util_math_float3.h
index f5149fe13ed..e42ded76c75 100644
--- a/intern/cycles/util/util_math_float3.h
+++ b/intern/cycles/util/util_math_float3.h
@@ -377,6 +377,18 @@ ccl_device_inline bool isequal_float3(const float3 a, const float3 b)
 #endif
 }
 
+ccl_device_inline int3 quick_floor_to_int3(const float3 a)
+{
+#ifdef __KERNEL_SSE__
+	int3 b = int3(_mm_cvttps_epi32(a.m128));
+	int3 isneg = int3(_mm_castps_si128(_mm_cmplt_ps(a.m128, _mm_set_ps1(0.0f))));
+	/* Unsaturated add 0xffffffff is the same as subtract -1. */
+	return b + isneg;
+#else
+	return make_int3(quick_floor_to_int(a.x), quick_floor_to_int(a.y), quick_floor_to_int(a.z));
+#endif
+}
+
 ccl_device_inline bool isfinite3_safe(float3 v)
 {
 	return isfinite_safe(v.x) && isfinite_safe(v.y) && isfinite_safe(v.z);
diff --git a/intern/cycles/util/util_math_int3.h b/intern/cycles/util/util_math_int3.h
index 6eef8517665..81b10f31f4a 100644
--- a/intern/cycles/util/util_math_int3.h
+++ b/intern/cycles/util/util_math_int3.h
@@ -91,6 +91,24 @@ ccl_device_inline bool operator<(const int3 &a, const int3 &b)
 {
 	return a.x < b.x && a.y < b.y && a.z < b.z;
 }
+
+ccl_device_inline int3 operator+(const int3 &a, const int3 &b)
+{
+#ifdef __KERNEL_SSE__
+	return int3(_mm_add_epi32(a.m128, b.m128));
+#else
+	return make_int3(a.x + b.x, a.y + b.y, a.z + b.z);
+#endif
+}
+
+ccl_device_inline int3 operator-(const int3 &a, const int3 &b)
+{
+#ifdef __KERNEL_SSE__
+	return int3(_mm_sub_epi32(a.m128, b.m128));
+#else
+	return make_int3(a.x - b.x, a.y - b.y, a.z - b.z);
+#endif
+}
 #endif  /* !__KERNEL_OPENCL__ */
 
 CCL_NAMESPACE_END