Closed steven-johnson closed 3 months ago
Looks like there are Vulkan failures -- any clues? Are there more fixes I missed?
Sorry! Yes, I saw this as well, but haven't been able to reproduce this on my Linux machine. Still investigating.
We should also grab: https://github.com/halide/Halide/pull/8130
PTAL -- lots of conflicts in the cherry pick so hopefully I did it right
It looks like a few things from main
got left behind in src/runtime/internal
. I'll add those to this PR.
I inspected buildbot-4 and it only had v1.2 of the Vulkan SDK installed, but the build target we're using is generating code for SDK v1.3. I've upgraded that machine, but the other linux workers will likely need to get upgraded too.
the other linux workers will likely need to get upgraded too.
Tell me the process to upgrade and I'll handle it
The packaging has changed since we originally setup the machines, and now everything is bundled under the vulkan-sdk for Ubuntu:
https://vulkan.lunarg.com/doc/view/latest/linux/getting_started_ubuntu.html
I just needed:
wget -qO - http://packages.lunarg.com/lunarg-signing-key-pub.asc | sudo apt-key add -
sudo wget -qO /etc/apt/sources.list.d/lunarg-vulkan-focal.list http://packages.lunarg.com/vulkan/lunarg-vulkan-focal.list
sudo apt update
sudo apt install vulkan-sdk
That should make vulkaninfo
report SDK v1.3.x
Fix Vulkan SIMT mappings for GPU loop vars. Previous refactoring accidentally used the fully qualified var name rather than the categorized vulkan intrinsic name.
Avoid formatting the GPU kernel to a string for Vulkan (since it's binary SPIR-V needs to remain intact).