Overview / Usage

Pixar's development team is continuing their journey to develop RenderMan XPU™, their next-generation photorealistic production rendering technology to deliver the animation and film visual effects of tomorrow. We'll update you on our techniques leveraging heterogeneous compute across CPUs and GPUs. Pixar and Intel® are working together to refine practical techniques to achieve coherent execution and improved data access over this massively parallel problem space. Combined with Intel's SIMD version of OSL (Open Shading Language), RenderMan XPU™ aims to achieve high SIMD utilization and scale up with Intel® Advanced Vector Extensions 512 (Intel® AVX-512).

Methodology / Approach

Using Intel® C++ Compiler with OpenMP SIMD directives we vectorize kernels in Pixar's RenderMan XPU to target Intel® Advanced Vector Extensions 512 (Intel® AVX-512). We then use compiler optimization reports to identify data layout and other issues, after some refactoring scaling of the kernels can be improved. We also explore improving data coherency to limit number of kernel launches and increase SIMD lane utilization. Such reorganization also improves batch utilization of Intel's SIMD enabled Open Shading Language.

Technologies Used

Open Shading Language (OSL)
Intel® Parallel Studio XE
Intel® C++ Compiler
C++14
OpenMP
Intel® Threading Building Blocks (TBB)
Intel® VTune™ Profiler
Intel® Advisor
GDB*

Other links

• OpenMP SIMD Directives

• Open Shading Language (https://github.com/imageworks/OpenShadingLanguage)

• Pixar's RenderMan
• Intel Parallel Studio XE
• Intel® oneAPI Toolkits
• Slashing Open Shading Language Render Times with a SIMD Scalable Architecture
• From RenderMan 22.0(r) to Next-Gen RenderMan XPU and Beyond: Role of Open Shadding Language (OSL) with Intel(r) Advanced Vector Extensions (Intel(r) AVX-512)
• Unleashing Intel® Advanced Vector Extensions 512 Inside the Open Shading Language