Render Options
Software version : 10.0.148 (Jun 17 2013)
* Render Options :
Render used the 'raytrace' hider
Image resolution: 640 x 480
Pixel oversampling: 2 x 2
Total number of displays (AOVs) : 2
Shared instances (for massive instancing) : on
The information included in these statistics give some general information about render settings that can affect performance. The first line is always the software version so it is easy to see what 3Delight version was used to render the frame.
| Statistic | Description |
|---|---|
Render used the 'raytrace' hider | This says wheter REYES or ray tracing was used to render the images. |
Image resolution: 640x480 | Image resolution. |
Pixel oversampling: 2 x 2 | The oversampling used for this image. |
Total number of displays (AOVs) | Says how many AOVs are output in the current render. |
Shared instances (for massive instancing) : on | This says if 3Delight is using a special optimization to render large amounts of object instances. |
System
* System :
Total physical memory : 8.6GB
CPU Physical cores count: 2, logical cores count: 2
Render threads started by user : 2
Network reads : 1.48 Mb total. Transfer rate: 137.6 Kb/s
Network writes : 1.48 Mb total. Transfer rate: 137.7 Kb/s
Load average: 1.2 0.9 1.1
Time to first pixel : 1 seconds.
This statistics block gives information about the system that was used to render.
| Statistic | Description |
|---|---|
Total physical memory : 8.6GB | Gives the total amount of memory installed on the system. |
CPU Physical cores count: 2, logical cores count: 2 | Gives physical and logical core count. Logical core count includes "hyperthreading" processors. |
Render threads started by user : 2 | Says how many threads have been started by the user. |
Network reads : 1.48 Mb total. Transfer rate: 137.6 Kb/s | This interesting statistics gives the total amount of data that has been read from the network while rendering the image. Note that this count all the network data and not only the data that is used for rendering. This statistic is helpful to debug network performance problems. |
Network writes : 1.48 Mb total. Transfer rate: 137.7 Kb/s | Similar to the statistic above but for network writes. |
Load average: 1.2 0.9 1.1 | Load average is a gauge of how many processes are on average, concurrently demanding CPU attention. The three numbers are averages for 1, 5 and 15 minutes. Very detailed description of load averages can be obtained on Linux manpages of uptime. A good technical overview can be found here and good vulgarization can be found here. |
Time to first pixel : 1 seconds. | This somewhat misplaced statistic shows the time it took before the first pixel has been sent to the user. It is always a very important goal in 3Delight to minimize this time. |
Profiling
* Profiling :
task time
raytracer init 0.0
raytracing 10.3
surface shaders 0.0
imager shaders 0.0
display drivers 0.2
screen filtering 0.1
raytrace hider 1.0
Total : 11.6
Total process times (user/sys/real) : 19.56 / 0.27 / 11.66
CPU use : 167.8% user + 2.3% sys = 170.1%
This statistic block gives several timings about the render. Each task shows the wall clock time it took in the total render time. Follows a description for all possible tasks.
| Task | Description |
|---|---|
raytracer init | This task includes all the work that needs to be done before starting the space parition. Basically this means iterating through all scene objects and prepare structures for the upcoming render. |
ray tracer instance init | This task counts the time that is spent on optimizing instances (when "shared instances" are enabled). |
space partition init | This task counts the time it takes 3Delight to build the space partitioner (a BVH-like structure). The time spent here will go proportionally with the scene size. |
surface shaders | Counts the amount of time the renderer is spending inside surface shaders. Complex shaders (as generated, for example, by comlexe shading networks in Maya or Softimage) will lead to higher times in this task. |
light shaders | Counts the time spent in light shaders. |
volume shaders | Counts the time spent in volume shaders. |
displacement shaders | Counts the time spent in displacement shaders. |
imager shaders | Counts the time spent in imager shaders. |
RSL plugins | Counts the time spent in custom RSL plug-ins. |
shadowmap filtering | Counts the time spent filtering shadow maps to render shadows. |
dsm filtering | Same as for shadow map filtering but for deep shadow maps. |
texture filtering | Counts the time spent filtering texture maps. If this times seems too high compared to total render time (>10%) it could be one of the following reasons:
|
point cloud filtering | Counts the time spent filtering point clouds for point-based color bleeding. |
brick map filtering | Counts the time spent filtering brick maps. |
texture loading | Counts the time spent loading texture tiles. Since 3Delight has a caching system for texture loading this time should be marginal compared to the total render time. If not, this probably means that the texture cache is not large enough (read below). |
shadow map loading | Counts the time spent loading shadow map tiles. |
Texture Cache
The texture cache plays an important part in the overall performance of 3Delight. The cache is needed since available physical memory is usually much less than the space taken by all the textures for a certain render. The role of this cache is to keep and manage a limited amount of texture data that is needed for rendering.
* Texture Cache :
Cache size: 0.133 G
Overall texture cache efficiency: 0 %
cache efficiency (2D texture): 0 %
cache efficiency (misc texture): 0 %
2D texture footprint: 0.000 G
misc texture footprint: 0.000 G
In these stats, there are three important concepts: size, footprint and efficiency. Both are explained below.
| Statistic | Description |
|---|---|
Cache size | Total cache size as specified by the user. |
texture footprint | The texture footprint is the total amount of data seen by the cache during a render. It's the size the cache should be made to hold everything needed in memory. This is a good metric of how much data a scene is using. Note that the cache does not have to be that large to be fully efficient in most cases. |
cache efficiency | The cache efficiency is related to this: it is the ratio between the footprint and the amount of data actually loaded into the cache, counting data which is loaded more than once because it was flushed. For example, at 50% each texture tile is loaded on average twice, which is actually not that bad. At 10%, we're loading each tile 10 times, which is not that good anymore. |
Point Based Occlusion
* Point Based Occlusion :
Number of evaluations : 0
Considered elements per evaluation : 0
These two statistic lines only concern point-based global illumination and are actually just used for debugging by the 3Delight team. The "Number of Evaluation" is simply the number of calls that were made to the point-based version of indirectdiffuse() and the "Considered elements" is related to the total number of octree nodes accessed for each evaluation, on average.
Map Manager
The map manager gives information about the files maps used during the render. For each map it shows additional information that might be helpful for TDs when debugging scene performance.
* Map Manager :
Total number of maps used : 0
Total number of texture maps : 0
Total number of shadow maps : 0
Total number of environment maps : 0
Total number of deep shadow maps : 0
Total number of 'invalid' maps : 0
Total number of file open operations : 0
Total number of map requests : 0
Texture IDs
Texture stats
ID Refs Same Search Faults Rate Mipmap (min) Mipmap (max)
Scene Manager
* Scene Manager :
Number of objects in the scene : 1
Number of attributes in the scene : 2
Number of transforms in the scene : 3
Primvar compression ratio : 1.8
Saved memory due to compression : 5.0 Mb
This statistics block gives some basic information about the current scene.
| Statistics | Description |
|---|---|
Number of objects in the scene | Counts the total number of objects in the scene. |
Number of attributes in the scene | Counts the total number of RenderMan attributes in the scene. |
Number of transforms in the scene | Counts the total number of transforms in the scene. |
Primvar compression ratio | 3Delight uses a compression algorithm to minimize the memory used by the geometry and its auxiliary data. This number gives an idea of the performance of that compression algorithm. |
Saved memory due to compression | Gives the total number of megabytes saved thanks to compression. |
Memory Profile
Memory profiling is an important tool when it comes to analyze 3Delight's memory usage. A rapid look at this stat can give a very good idea on where the memory is going.
* Memory Profiling :
item current peak category peak
unprocessed gprim 211 K 211 K 418 K
gprim 0 0 0
gprim (polygons) 0 0 0
gprim (subdiv) 0 0 0
gprim (curves) 0 0 0
gprim variables 26 K 617 K 617 K
particles 0 0 4536
topology data 0 0 0
crack elimination geo 0 0 0
grid 0 0 0
shaded grid 0 0 0
grid AOV 0 0 0
texture objects 0 0 0
texture cache 0 0 0
hider and visible points 0 0 90 K
image data 0 251 K 313 K
buffered display data 0 31 K 61 K
shader code 4033 4033 4033
shader instance 2332 2332 2332
shader variables 0 814 K 814 K
attributes 796 796 796
transform 1628 1628 1776
subsurface 0 0 0
point based occlusion 0 0 0
point cloud 0 0 0
photon map 0 0 0
inline archive 0 0 0
ray tracer 289 K 289 K 289 K
ray tracer cache 0 0 0
space partition 90 K 90 K 90 K
diffuse ray cache 0 0 0
misc 1756 K 1756 K 1756 K
Total : 2381 K 4068 K 4462 K
Peak VM size : 0 K
Peak resident : 35623 K
Peak from malloc : 12899 K