Cycles OpenCL hardware raytracing intersection for AMD Rx 6000 series #82557

New Issue

Brian Savery (AMD) · 2020-11-10T01:58:13+01:00

2020-11-10 01:58:13 +01:00

Overview

AMD Radeon RDNA2 architecture adds support for hardware accelerated ray intersection and BVH traversal. This is exposed in OpenCL via a “hardware intrinsic”. We will use this to modify the existing OpenCL path in Cycles to use a separate hardware accelerated code path when device support is present.

From a user point of view, they continue using the OpenCL version of cycles, and we automatically enable hw raytracing if available.

Design
Enablement:

A flag in the make file

WITH_CYCLES_DEVICE_AMD_HW

If the cmake flag is set, device capabilities are assessed for HW raytracing and if available, HW RT is enabled

Host side, BVH

Addition of a new BVH layout mask (bvh_amd)
Addition of a new bvh class to generate hw compatible bvh and pack instances accordingly
- QBVH based with top level bounding boxes, up to depth 5, with full precision and the rest with half precision

Host side, Device

more details to come...

Detecting device compatibility

When AMD HW raytracing flag is set, the hardware ray tracing capability is queried in the OpenCLDevice constructor. The result of the query is checked when assigning the BVH layout and when kernels are compiled. The BVH layout routes the BVH creation to BVHAMD class and later in the BVH building, methods of this class are used to pack nodes and generate the hardware compatible BVH. The compiler options routes the intersection calls to the appropriate BVH function where AMD hardware supported intersection functionality is used.

(Patch to be public soon)

**Overview** AMD Radeon RDNA2 architecture adds support for hardware accelerated ray intersection and BVH traversal. This is exposed in OpenCL via a “hardware intrinsic”. We will use this to modify the existing OpenCL path in Cycles to use a separate hardware accelerated code path when device support is present. From a user point of view, they continue using the OpenCL version of cycles, and we automatically enable hw raytracing if available. **Design** Enablement: - A flag in the make file ``` WITH_CYCLES_DEVICE_AMD_HW ``` - If the cmake flag is set, device capabilities are assessed for HW raytracing and if available, HW RT is enabled Host side, BVH - Addition of a new BVH layout mask (bvh_amd) - Addition of a new bvh class to generate hw compatible bvh and pack instances accordingly - QBVH based with top level bounding boxes, up to depth 5, with full precision and the rest with half precision Host side, Device more details to come... **Detecting device compatibility** When AMD HW raytracing flag is set, the hardware ray tracing capability is queried in the OpenCLDevice constructor. The result of the query is checked when assigning the BVH layout and when kernels are compiled. The BVH layout routes the BVH creation to BVHAMD class and later in the BVH building, methods of this class are used to pack nodes and generate the hardware compatible BVH. The compiler options routes the intersection calls to the appropriate BVH function where AMD hardware supported intersection functionality is used. (Patch to be public soon)

Brian Savery (AMD) self-assigned this 2020-11-10 01:58:13 +01:00

Brian Savery (AMD) commented

2020-11-10 01:58:13 +01:00

Changed status from 'Needs Triage' to: 'Confirmed'

Brian Savery (AMD) commented

2020-11-10 01:58:13 +01:00

Added subscribers: @BrianSavery, @brecht

blender-admin commented

2020-11-10 01:58:13 +01:00

#82543 was marked as duplicate of this issue

Paul Larson commented

2020-11-10 05:57:52 +01:00

Added subscriber: @GeorgiaPacific

Janne Aliu commented

2020-11-10 14:02:25 +01:00

Added subscriber: @Jaydead

Jeroen Bakker commented

2020-11-10 14:10:59 +01:00

Added subscriber: @Jeroen-Bakker

Stefan Werner commented

2020-11-10 15:40:03 +01:00

Added subscriber: @Stefan_Werner

Pietro Elia Appeddu commented

2020-11-10 18:00:05 +01:00

Added subscriber: @PietroEliaAppeddu

Ghost commented

2020-11-11 02:23:53 +01:00

Added subscriber: @dabuxian

Andrea Monzini commented

2020-11-18 16:32:34 +01:00

Added subscriber: @AndreaMonzini

Alex Fuller commented

2020-11-20 00:04:27 +01:00

Added subscriber: @boberfly

Andrea Monzini commented

2020-11-20 14:27:13 +01:00

Hello, i would gently ask info about GNU/Linux design support for the hardware raytracing in Blender for the Rx 6000 :

support the AMDGPU-Pro ?
support the AMDGPU + OpenCL part from AMDGPU-Pro ?
support the new ROCm 4.0 ?

The Rx 6000 seem to be supported in ROCm :
https://www.phoronix.com/scan.php?page=article&item=amd-rx6800-opencl&num=1

Personally i hope for a working solution with free and open source drivers.

EDIT: it appears that from Mesa 20.3 ( AMDGPU ) we will have support for "Clover" OpenCL 1.2 !

Anyway thank you for your work !

Andrea

Hello, i would gently ask info about GNU/Linux design support for the hardware raytracing in Blender for the Rx 6000 : - support the AMDGPU-Pro ? - support the AMDGPU + OpenCL part from AMDGPU-Pro ? - support the new ROCm 4.0 ? The Rx 6000 seem to be supported in ROCm : https://www.phoronix.com/scan.php?page=article&item=amd-rx6800-opencl&num=1 Personally i hope for a working solution with free and open source drivers. EDIT: it appears that from Mesa 20.3 ( AMDGPU ) we will have support for "Clover" OpenCL 1.2 ! Anyway thank you for your work ! Andrea

Makram commented

2020-12-05 13:13:20 +01:00

Added subscriber: @Pliou