Making life (even) harder for proprietary modules
Making life (even) harder for proprietary modules
Posted Aug 4, 2023 15:42 UTC (Fri) by jgg (subscriber, #55211)In reply to: Making life (even) harder for proprietary modules by rahulsundaram
Parent article: Making life (even) harder for proprietary modules
NVIDIA provided an open GPL-compatible licensed driver and a closed FW. Just like the majority of server HW out there these days. This new FW is NVIDIA's primary FW for all new HW. It is is not some special crippled open source release.
If you are familiar with the NVIDIA ecosystem you can see that the new FW and open source driver support the entire propritary userspace NVIDIA software stack. Today. This means everything you can do with CUDA/etc can be done through this FW.
So was "critical functionality" blocked off somehow? I guess not.
I think the more productive way to look at the situation is to ask what can be done on the open source side with this FW interface.
DRM maintainers have long held that an open userspace and open compiler are essential parts they need to consider something open.
It appears this FW is good enough that nouveau can make an open vulkan graphics stack, running at full performance, using the open userspace compiler that already exists. This is a big improvement.
If this FW release made nouveau, and its fully open stack, dramatically better - is it more or less open source friendly?