Announcement

Collapse
No announcement yet.

Luoshuang's GPULightmass

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    Originally posted by Luoshuang View Post

    The source installation method is mentioned in the main post
    It should be in 4.20 anyway

    Comment


      This is so incredably awesome. Just want to say thank you so much!!!

      Comment


        I tryed to bake with swarm, my scene bake in 53s on my first PC, but when i active swarm with 4 PC, i exceed 2 min.... any idea? (my version is 4.19 VXGI version rebuild)

        Comment


          Apparently i have a large level that gives me an out of memory error after about 4 hours of rendering on swarm.. oddly enough it only takes 8-10 hours to render on default unreal stock renderer on production....

          How do i not get out of memory error?

          Comment


            Originally posted by kurylo3d View Post
            Apparently i have a large level that gives me an out of memory error after about 4 hours of rendering on swarm.. oddly enough it only takes 8-10 hours to render on default unreal stock renderer on production....

            How do i not get out of memory error?
            One way is to reduce the quality by setting MappingSurfaceCacheDownsampleFactor in BaseLightmass.ini to higher values like 4 or 8. Also I think you always want to use the fast preview quality preset firstly to check whether you will run into any errors so that you won't waste time .
            Last edited by Luoshuang; 05-23-2018, 03:21 AM.

            Comment


              Originally posted by Luoshuang View Post

              One way is to reduce the quality by setting MappingSurfaceCacheDownsampleFactor in BaseLightmass.ini to higher values like 4 or 8. Also I think you always want to use the fast preview quality preset firstly to check whether you will run into any errors so that you won't waste time .
              Thanks for the response ill give it a try. What does MappingSurfaceCacheDownsampleFactor actually do?

              Comment


                I'm running this error when I try to compile! When I changed to GPU it was OK, but today..I changed two things on my scene and it appears.
                Any idea how to repair?

                Thanks.

                <Nenhum> === Lightmass crashed: ===Unhandled Exception: 0x80000003
                Unhandled Exception: 0x80000003

                0x000000004322BA51 GPULightmassKernel.dll!UnknownFunction []
                0x000000004312EE02 GPULightmassKernel.dll!UnknownFunction []
                0x0000000043134A0D GPULightmassKernel.dll!UnknownFunction []
                0x000000004313671D GPULightmassKernel.dll!UnknownFunction []
                0x000000007883CFB2 UnrealLightmass.exe!UnknownFunction []
                0x0000000078809691 UnrealLightmass.exe!UnknownFunction []
                0x00000000787B1E41 UnrealLightmass.exe!UnknownFunction []
                0x00000000787DCF32 UnrealLightmass.exe!UnknownFunction []
                0x00000000787FED72 UnrealLightmass.exe!UnknownFunction []
                0x000000007888CF70 UnrealLightmass.exe!UnknownFunction []
                0x00000000B4E93034 KERNEL32.DLL!UnknownFunction []
                0x00000000B6E51551 ntdll.dll!UnknownFunction []
                0x00000000B6E51551 ntdll.dll!UnknownFunction []

                Comment


                  And this is what the Swarm Agent's Log shows:


                  17:41:43: [Interface:TryOpenConnection] Local connection established
                  17:41:43: [Job] Accepted Job 3BAC86BB-42462A52-42C79395-84FFFC22
                  17:42:00: [Job] Launched Job UnrealLightmass_2018-04-14_15-00-09_1391616-64bit.exe
                  17:42:00: [Job] PID is 14288
                  17:42:00: [Job] GUID is "3BAC86BB-42462A52-42C79395-84FFFC22"
                  17:42:00: LogLightmass: Display: Lightmass Win64 started on: DESKTOP-SNFBN4F. Command-line: 3BAC86BB42462A5242C7939584FFFC22
                  17:42:00: Failed to determine engine directory: Defaulting to ../../../Engine/
                  17:42:00: Failed to load 'aqProf.dll' (GetLastError=126)
                  17:42:00: File 'aqProf.dll' does not exist
                  17:42:00: Failed to load 'VtuneApi.dll' (GetLastError=126)
                  17:42:00: File 'VtuneApi.dll' does not exist
                  17:42:00: Failed to load 'VtuneApi32e.dll' (GetLastError=126)
                  17:42:00: File 'VtuneApi32e.dll' does not exist
                  17:42:00: Lightmass Win64 started on: DESKTOP-SNFBN4F. Command-line: 3BAC86BB42462A5242C7939584FFFC22
                  17:42:00: LogLightmass: Display: Processing scene GUID: 3BAC86BB42462A5242C7939584FFFC22 with 6 threads
                  17:42:00: Processing scene GUID: 3BAC86BB42462A5242C7939584FFFC22 with 6 threads
                  17:42:00: Building static lighting...
                  17:42:01: [OpenConnection] Registering TCP channel ...
                  17:42:01: [OpenConnection] Connecting to agent ...
                  17:42:01: [TryOpenConnection] Testing the Agent
                  17:42:01: [TryOpenConnection] Opening Connection to Agent
                  17:42:01: [TryOpenConnection] Local Process ID is 14288
                  17:42:01: [Job] Found a parent connection for PID 14288
                  17:42:01: [Job] 02D7CA36 -> 5D2B620A
                  17:42:01: [Interface:TryOpenConnection] Local connection established
                  17:42:01: Skylight import processing 0.006s with CubemapSize 128
                  17:42:02: Measured CPU frequency: 3.60 GHz
                  17:42:02: FStaticLightingSystem started using GKDOPMaxTrisPerLeaf: 4
                  17:42:02: Number of texture mappings: 301
                  17:42:02: Number of fluid mappings: 0
                  17:42:02: Number of landscape mappings: 64
                  17:42:02: Number of BSP mappings: 18
                  17:42:02: Number of static mesh instance mappings: 301
                  17:42:03: Reserving memory for 383 meshes, 886983 vertices, 1257695 triangles
                  17:42:03: Reserving memory for 64 meshes, 97344 vertices, 184832 triangles
                  17:42:03: Scene surface area calculated at 2298.464 million units (53.518% of the estimated 4294.758 million units)
                  17:42:03: Importance volume surface area calculated at 1064.637 million units (102.888% of the estimated 1034.754 million units)
                  17:42:03: Embree Build 0.1s
                  17:42:03:
                  17:42:03:
                  17:42:03: Collision Mesh Overview:
                  17:42:03: Num Triangles : 1253727
                  17:42:03: MeshInfos : 0.0Mb
                  17:42:03: UVs : 9.3Mb
                  17:42:03: LightmapUVs : 9.3Mb
                  17:42:03: Embree Used Memory : 92.0Mb
                  17:42:03:
                  17:42:03:
                  17:42:03: Preallocated 0.2Gb for kDOP nodes and triangles
                  17:42:04: Building kDOP took 0.89 seconds.
                  17:42:04: Static lighting kDOP: 816964 nodes, 408483 leaves, 1633932 triangles, 884775 vertices
                  17:42:04: Static lighting kDOP: 23.305% wasted space in leaves
                  17:42:04: Preallocated 0.0Gb for kDOP nodes and triangles
                  17:42:04: Building kDOP took 0.11 seconds.
                  17:42:04: Static lighting kDOP: 931264 nodes, 465634 leaves, 1862536 triangles, 97344 vertices
                  17:42:04: Static lighting kDOP: 90.076% wasted space in leaves
                  17:42:04: Processing...
                  17:42:05: GPULightmass Kernel: Importing mesh: 884775 vertices, 1253320 triangles
                  17:42:06: GPULightmass Kernel: Embree SBVH Construction finished, 465MS
                  17:42:06: GPULightmass Kernel: Convert to CudaBVH finished, 501MS
                  17:42:07: GPULightmass Kernel: GPU BVH video memory size: 36.64MB, triangle payload size: 87.84MB
                  17:42:07: GPULightmass Kernel: Bind CudaBVH finished, 407MS
                  17:42:07: GPULightmass Kernel: GPU triangle parameterization data size: 32.62MB
                  17:42:07: GPULightmass Kernel: 64 masked collision maps imported
                  17:42:07: GPULightmass Kernel: 2 point lights and 3 spot lights imported
                  17:42:07: Started task graph with 4 named threads and 10 total threads with 1 sets of task threads.
                  17:42:13: GPULightmass Kernel: Total surface cache video memory data: 2009.42MB
                  17:42:13: GPULightmass Kernel: RadiosityPass 0
                  17:42:38: GPULightmass Kernel: Line 475: cudaCheckError: unspecified launch failure

                  Comment


                    Originally posted by Chassiloko View Post

                    17:42:38: GPULightmass Kernel: Line 475: cudaCheckError: unspecified launch failure
                    Follow the instructions in the main post to change your TDR settings.

                    Comment


                      "Follow the instructions in the main post to change your TDR settings."

                      you mean the first page? What is the name of the main post in case if its not the first page, please?

                      Comment


                        Originally posted by Chassiloko View Post
                        "Follow the instructions in the main post to change your TDR settings."

                        you mean the first page? What is the name of the main post in case if its not the first page, please?
                        You can easily find that thing by pressing Ctrl+F on the first page and type "TDR".

                        Comment


                          Thank you very much!!!

                          Comment


                            Originally posted by Luoshuang View Post
                            [*]Current the number of samples (quality settings) is hardcoded in the program and you cannot change it since I haven't found an reliable way to expose them to the user
                            Not even through .ini file setting ?

                            Comment


                              Originally posted by motorsep View Post

                              Not even through .ini file setting ?
                              The total number of samples taken by the Brute Force GI Engine is fixed:

                              Fast Preview: 16 * 16 * 2 = 512
                              Medium Quality: 32 * 32 * 2 (Fixed) + 32 * 32 * 2 (Adaptive) = 4096
                              Ultra High: 64 * 64 * 2 (Fixed) + 64 * 64 * 2 (Adaptive) = 16384

                              However, the secondary radiosity caching GI engine does accept a parameter to reduce/increase the quality: MappingSurfaceCacheDownsampleFactor (minimum 1) in BaseLightmass.ini. Basically the engine uses your lightmap resolution as a reference and divides it by the downsample factor to deduce the final indirect lighting resolution. Raising the downsample factor can reduce video memory consumption but can also introduce leaks. A workaround is to split the lightmap UV at where the lighting leaks.

                              Comment


                                Originally posted by Luoshuang View Post

                                The total number of samples taken by the Brute Force GI Engine is fixed:

                                Fast Preview: 16 * 16 * 2 = 512
                                Medium Quality: 32 * 32 * 2 (Fixed) + 32 * 32 * 2 (Adaptive) = 4096
                                Ultra High: 64 * 64 * 2 (Fixed) + 64 * 64 * 2 (Adaptive) = 16384

                                However, the secondary radiosity caching GI engine does accept a parameter to reduce/increase the quality: MappingSurfaceCacheDownsampleFactor (minimum 1) in BaseLightmass.ini. Basically the engine uses your lightmap resolution as a reference and divides it by the downsample factor to deduce the final indirect lighting resolution. Raising the downsample factor can reduce video memory consumption but can also introduce leaks. A workaround is to split the lightmap UV at where the lighting leaks.
                                Right, but I don't really want to have 3 different custom engines built. That's why I wonder why not to put total number of samples into .ini so that people (including myself) only need to build 1 copy of the engine and then restart it (after changing .ini settings) to be able to build either Fast Preview or Ultra High quality.

                                Comment

                                Working...
                                X