Microsoft Doubles Their Commits To Mesa This Week
More than a dozen patches were merged by a Microsoft engineer into Mesa yesterday.
Yes, it may be surprising there is any commits to Mesa by Microsoft engineers, but in recent months there have been patches from at least two Microsoft employees. No, they are not porting the Mesa drivers for usage on Windows, but rather part of their play for getting GPU acceleration -- compute in particular -- up and running within Windows Subsystem for Linux 2 (WSL2).
Microsoft partnered with Collabora as part of their efforts for Direct3D 12 on Linux/WSL2 and getting OpenCL/OpenGL working within WSL2 too. While Collabora engineers have done a lot of the heavy lifting, Microsoft engineers have been involved as well with the likes of these Mesa patches and other areas like their DirectX kernel driver.
Mesa is now up to just over two dozen patches from Microsoft.com developers, most of which pertain to the NIR intermediate representation. This newest batch of patches is for implementing some missing OpenCL floating point opcodes for NIR/SPIR-V.
It will certainly be interesting to see how this GPU acceleration within WSL2 ultimately works out and how well it's received by the developer community.
Yes, it may be surprising there is any commits to Mesa by Microsoft engineers, but in recent months there have been patches from at least two Microsoft employees. No, they are not porting the Mesa drivers for usage on Windows, but rather part of their play for getting GPU acceleration -- compute in particular -- up and running within Windows Subsystem for Linux 2 (WSL2).
Microsoft partnered with Collabora as part of their efforts for Direct3D 12 on Linux/WSL2 and getting OpenCL/OpenGL working within WSL2 too. While Collabora engineers have done a lot of the heavy lifting, Microsoft engineers have been involved as well with the likes of these Mesa patches and other areas like their DirectX kernel driver.
Mesa is now up to just over two dozen patches from Microsoft.com developers, most of which pertain to the NIR intermediate representation. This newest batch of patches is for implementing some missing OpenCL floating point opcodes for NIR/SPIR-V.
It will certainly be interesting to see how this GPU acceleration within WSL2 ultimately works out and how well it's received by the developer community.
31 Comments