Differential D12888 Diff 43495 intern/cycles/kernel/bvh/bvh_nodes.h

Changeset View

Standalone View

intern/cycles/kernel/bvh/bvh_nodes.h

Show All 10 Lines
* distributed under the License is distributed on an "AS IS" BASIS,		* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.		* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and		* See the License for the specific language governing permissions and
* limitations under the License.		* limitations under the License.
*/		*/

// TODO(sergey): Look into avoid use of full Transform and use 3x3 matrix and		// TODO(sergey): Look into avoid use of full Transform and use 3x3 matrix and
// 3-vector which might be faster.		// 3-vector which might be faster.
ccl_device_forceinline Transform bvh_unaligned_node_fetch_space(ccl_global const KernelGlobals *kg,		ccl_device_forceinline Transform bvh_unaligned_node_fetch_space(KernelGlobals kg,
int node_addr,		int node_addr,
int child)		int child)
{		{
Transform space;		Transform space;
const int child_addr = node_addr + child * 3;		const int child_addr = node_addr + child * 3;
space.x = kernel_tex_fetch(__bvh_nodes, child_addr + 1);		space.x = kernel_tex_fetch(__bvh_nodes, child_addr + 1);
space.y = kernel_tex_fetch(__bvh_nodes, child_addr + 2);		space.y = kernel_tex_fetch(__bvh_nodes, child_addr + 2);
space.z = kernel_tex_fetch(__bvh_nodes, child_addr + 3);		space.z = kernel_tex_fetch(__bvh_nodes, child_addr + 3);
return space;		return space;
}		}

ccl_device_forceinline int bvh_aligned_node_intersect(ccl_global const KernelGlobals *kg,		ccl_device_forceinline int bvh_aligned_node_intersect(KernelGlobals kg,
const float3 P,		const float3 P,
const float3 idir,		const float3 idir,
const float t,		const float t,
const int node_addr,		const int node_addr,
const uint visibility,		const uint visibility,
float dist[2])		float dist[2])
{		{

Show All 31 Lines	#ifdef __VISIBILITY_FLAG__
/* this visibility test gives a 5% performance hit, how to solve? */		/* this visibility test gives a 5% performance hit, how to solve? */
return (((c0max >= c0min) && (__float_as_uint(cnodes.x) & visibility)) ? 1 : 0) \|		return (((c0max >= c0min) && (__float_as_uint(cnodes.x) & visibility)) ? 1 : 0) \|
(((c1max >= c1min) && (__float_as_uint(cnodes.y) & visibility)) ? 2 : 0);		(((c1max >= c1min) && (__float_as_uint(cnodes.y) & visibility)) ? 2 : 0);
#else		#else
return ((c0max >= c0min) ? 1 : 0) \| ((c1max >= c1min) ? 2 : 0);		return ((c0max >= c0min) ? 1 : 0) \| ((c1max >= c1min) ? 2 : 0);
#endif		#endif
}		}

ccl_device_forceinline bool bvh_unaligned_node_intersect_child(ccl_global const KernelGlobals *kg,		ccl_device_forceinline bool bvh_unaligned_node_intersect_child(KernelGlobals kg,
const float3 P,		const float3 P,
const float3 dir,		const float3 dir,
const float t,		const float t,
int node_addr,		int node_addr,
int child,		int child,
float dist[2])		float dist[2])
{		{
Transform space = bvh_unaligned_node_fetch_space(kg, node_addr, child);		Transform space = bvh_unaligned_node_fetch_space(kg, node_addr, child);
Show All 9 Lines	ccl_device_forceinline bool bvh_unaligned_node_intersect_child(KernelGlobals kg,
const float far_y = max(lower_xyz.y, upper_xyz.y);		const float far_y = max(lower_xyz.y, upper_xyz.y);
const float far_z = max(lower_xyz.z, upper_xyz.z);		const float far_z = max(lower_xyz.z, upper_xyz.z);
const float tnear = max4(0.0f, near_x, near_y, near_z);		const float tnear = max4(0.0f, near_x, near_y, near_z);
const float tfar = min4(t, far_x, far_y, far_z);		const float tfar = min4(t, far_x, far_y, far_z);
*dist = tnear;		*dist = tnear;
return tnear <= tfar;		return tnear <= tfar;
}		}

ccl_device_forceinline int bvh_unaligned_node_intersect(ccl_global const KernelGlobals *kg,		ccl_device_forceinline int bvh_unaligned_node_intersect(KernelGlobals kg,
const float3 P,		const float3 P,
const float3 dir,		const float3 dir,
const float3 idir,		const float3 idir,
const float t,		const float t,
const int node_addr,		const int node_addr,
const uint visibility,		const uint visibility,
float dist[2])		float dist[2])
{		{
Show All 15 Lines
#endif		#endif
{		{
mask \|= 2;		mask \|= 2;
}		}
}		}
return mask;		return mask;
}		}

ccl_device_forceinline int bvh_node_intersect(ccl_global const KernelGlobals *kg,		ccl_device_forceinline int bvh_node_intersect(KernelGlobals kg,
const float3 P,		const float3 P,
const float3 dir,		const float3 dir,
const float3 idir,		const float3 idir,
const float t,		const float t,
const int node_addr,		const int node_addr,
const uint visibility,		const uint visibility,
float dist[2])		float dist[2])
{		{
float4 node = kernel_tex_fetch(__bvh_nodes, node_addr);		float4 node = kernel_tex_fetch(__bvh_nodes, node_addr);
if (__float_as_uint(node.x) & PATH_RAY_NODE_UNALIGNED) {		if (__float_as_uint(node.x) & PATH_RAY_NODE_UNALIGNED) {
return bvh_unaligned_node_intersect(kg, P, dir, idir, t, node_addr, visibility, dist);		return bvh_unaligned_node_intersect(kg, P, dir, idir, t, node_addr, visibility, dist);
}		}
else {		else {
return bvh_aligned_node_intersect(kg, P, idir, t, node_addr, visibility, dist);		return bvh_aligned_node_intersect(kg, P, idir, t, node_addr, visibility, dist);
}		}
}		}