A large patchset has been submitted to mainline Linux for NVIDIA NVDLA AI accelerator Direct Rendering Manager (DRM) driver, accompanied by an open-source user mode driver.
The NVDLA (NVIDIA Deep Learning Accelerator) can be found in recent Jetson modules such as Jetson AGX Xavier and Jetson AGX Orin, and since NVDLA was made open-source hardware in 2017, it can also be integrated into third-party SoCs such as StarFive JH7100 Vision SoC and Allwinner V831 processor.
I actually assumed everything was open-source already since we were told that NVDLA was a “complete solution with Verilog and C-model for the chip, Linux drivers, test suites, kernel- and user-mode software, and software development tools all available on Github’s NVDLA account.” and the inference compiler was open-sourced in September 2019.
But apparently not, as developer Cai Huoqing submitted a patchset with 23 files changed, 13243 insertions, and the following short description:
The NVIDIA Deep Learning Accelerator (NVDLA) is an open source IP which is integrated into NVIDIA Jetson AGX Xavier, so add driver support for this accelerator.
NVDLA introduce: http://nvdla.org/primer.html
User mode driver: https://github.com/caihuoq/nvdla/tree/main/sw/umd
While the description only lists Jetson AGX Xavier, I’d assume it might be used by all SoCs with one or more NVDLA AI accelerators. It’s just been submitted for review, so we’ll have to see when/if gets merged into mainline.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
Is this actually posted by NVIDIA? The poster didn’t use a NV email address and someone I asked at NV wasn’t aware of the effort to get it upstream, nor what userspace they’re using for it. Please can you clarify the details in the article?
I just took the title of Phoronix article at face value without checking whether NVIDIA was actually involved or not.
I don’t actually know who “Cai Huoqing” is working for. Sorry about that. I’ve corrected the title and article in places where I implied NVIDIA was involved in the patch.