Vortex 2.2: New Open GPGPU on RISC-V Released

The project Vortex 2.2 has been published, focusing on developing open GPGPU based on the RISC-V architecture. This project is designed for parallel calculations using the OpenCl API and SIMT (Single Instruction, Multiple Threads). Vortex 2.2 can be utilized for 3D graphics research and new GPU architecture development. Schemes, Verilog hardware block descriptions, Simulator, Drivers, and other design documentation are available under the Apache 2.0 license.

The GPGPU foundation is the standard ISA RISC-V, expanded with additional instructions necessary for GPU support and flow management functions. The developed GPGPU supports 32- and 64-bit architectures of the RISC-V RV32IMF and RV64IMafd command set. Optional components like separated memory, L1, L2, and L3 cache, varying numbers of cores, tasks, and flows can be included. Custom features like FPU, LSU, and SFU can also be added for each core.

For developing applications, an instrumentation process is proposed, which includes tools specifically adapted for working with Vortex. This includes pocl (compiler and Runtime OpenCl), LLVM/Clang, GCC, and Binutils. The project complies with the OpenCl 1.2 specification and implements support for the intermediate SPIR-V shader representation.

Additionally, the project develops the Open GPU Skybox technology, supporting the

/Reports, release notes, official announcements.