Nikita Sirgienko Sirgienko
1 Followers · 0 Following
  • Joined on 2020-12-10
Repositories
1
 Projects Public Activity Starred Repositories
Nikita Sirgienko commented on issue blender/blender#107414 2023-05-11 20:10:20 +02:00
Cycles Embree HWRT shading differences

I am plan to investigate this issue and resolve it during next few weeks (after begin of Bcon3)

Nikita Sirgienko commented on issue blender/blender#107414 2023-05-11 20:03:18 +02:00
Cycles Embree HWRT shading differences

Thanks @sentharn for reporting and the attached scene as well. I have able to reproduce this problem on a machine with Intel® Arc™ A770 for RT HW (Embree4) path and I have also can confirm that…

Nikita Sirgienko pushed to main at blender/blender 2023-05-03 12:06:37 +02:00
1dcc8e6ffa Fix #107356: Cycles: improve oneAPI error handling
Nikita Sirgienko closed issue blender/blender#107356 2023-05-03 12:06:36 +02:00
oneAPI Embree HWRT crash when using Ashikhmin-Shirley Glossy BSDF node
Nikita Sirgienko commented on issue blender/blender#107356 2023-05-02 11:02:51 +02:00
oneAPI Embree HWRT crash when using Ashikhmin-Shirley Glossy BSDF node

@PratikPB2123 > Nvidia GPUs are reporting this error only in 3.6 (maybe we need a separate report to handle this?) Something wrong on Blender 3.6 (and 3.6 only) side with execution of this…

Nikita Sirgienko commented on issue blender/blender#107356 2023-04-26 18:41:19 +02:00
oneAPI Embree HWRT crash when using Ashikhmin-Shirley Glossy BSDF node

@sentharn I can confirm that I can observe this crash with my Intel® Arc™ A770 Graphics - but for me, a crash are happening regardless of Embree4/RTHW usage and it is related to denoising -…

Nikita Sirgienko pushed to main at blender/blender 2023-04-20 21:22:00 +02:00
4babb7c02e Cycles: oneAPI: Fix volume intersection for Embree GPU execution
0d9fa73b42 Cycles: oneAPI: Fix motion blur rendering for Embree GPU execution
7e92fb92ec Cycles: oneAPI: Fix kernels preloading in case of incompatible AoT binaries
Compare 3 commits »
Nikita Sirgienko pushed to main at blender/blender 2023-04-20 11:10:18 +02:00
7ce10ebbbf Cycles: oneAPI: Remove excess quotes in a capabilities output
Nikita Sirgienko commented on pull request blender/blender#106266 2023-04-17 15:45:55 +02:00
Cycles: add RT HW support through Embree for oneAPI backend

I don' think that Metal and Optix are using this file at all - Metal have own version of scene_intersect* functions in "kernel\device\metal\bvh.h" and Optix also have own version in `kernel\devic…

Nikita Sirgienko pushed to main at blender/blender 2023-03-15 22:04:13 +01:00
7ee0bf671e Cycles: use 8-bit type for number of ray hits when possible
f9922b7074 Cycles: Use ray->tfar in Embree filter functions
1a580dbfdd Cycles: Use IntegratorShadowState directly in Embree filter functions
b97a6daa9a Cycles: Use geometryUserPtr from Embree filter functions arguments
Compare 4 commits »
Nikita Sirgienko merged pull request blender/blender#105800 2023-03-15 22:04:12 +01:00
Cycles: Few optimizations in Embree filter functions
Nikita Sirgienko pushed to cycles_embree4_support at blender/blender 2023-03-15 22:02:09 +01:00
7ee0bf671e Cycles: use 8-bit type for number of ray hits when possible
f9922b7074 Cycles: Use ray->tfar in Embree filter functions
1a580dbfdd Cycles: Use IntegratorShadowState directly in Embree filter functions
b97a6daa9a Cycles: Use geometryUserPtr from Embree filter functions arguments
59a083e948 Cleanup: format
Compare 34 commits »
Nikita Sirgienko commented on pull request blender/blender#105800 2023-03-15 21:55:35 +01:00
Cycles: Few optimizations in Embree filter functions

As I see buildbot have successfully build this PR for all platforms and all tests have beeen successfuly finished - considering this as success, I will rebase this changes and will merge them into…

Nikita Sirgienko commented on pull request blender/blender#105800 2023-03-15 18:59:29 +01:00
Cycles: Few optimizations in Embree filter functions

In case of approval of this pull request I would prefer to keep initial 4 commit separated (if there are no objections from Blender side) - so when review part of the process will finished, I will…

Nikita Sirgienko pushed to cycles_embree4_support at blender/blender 2023-03-15 18:55:23 +01:00
e5ebb351fa Add requested changes
Nikita Sirgienko commented on pull request blender/blender#105800 2023-03-15 18:29:33 +01:00
Cycles: Few optimizations in Embree filter functions

Well, an idea here what when we have free slots in INTEGRATOR_SHADOW_ARRAY tfar won't be decreased basically from base value, but as soon as we reach this condition if (num_recorded_hits + 1 >= max_record_hits) {, then we will find maximumum t value among all founded intersection and set it to tfar (which is supported, according to documentation "The filter function is further allowed to change the hit and decrease the tfar value of the ray but it should not modify other ray data nor any inactive components of the ray or hit." - as a result Embree will try to find any other intersection which will be closer to the ray origin that our the most distant hit - and if it will be possible, then we will got a hit, and in this intersection filter function we will replace this the most distant hit by new one (according to logic of "max_t" filtering).

Nikita Sirgienko created pull request blender/blender#105800 2023-03-15 16:53:01 +01:00
Cycles: Few optimizations in Embree filter functions
Nikita Sirgienko created branch cycles_embree4_support in blender/blender 2023-03-15 16:24:57 +01:00
Nikita Sirgienko pushed to cycles_embree4_support at blender/blender 2023-03-15 16:24:57 +01:00
633a4d36ab Cycles: use 8-bit type for number of ray hits when possible
5af80f3fe5 Cycles: Use ray->tfar in Embree filter functions
d4ebfd2812 Cycles: Use IntegratorShadowState directly in Embree filter functions
c2a3ad368d Cycles: Use geometryUserPtr from Embree filter functions arguments
Compare 4 commits »
Nikita Sirgienko commented on issue blender/blender#101726 2023-03-13 12:10:53 +01:00
Cycles does not generate the exact same images when a scene is rendered twice

@brech Considering that this problem making check of functional changes really hard (which is a problem, if you working on some CPU optimistaions) - is it possible at least add preprocess…