Embed Notice

@mia@movsw.0x0.st @shibao@misskey.bubbletea.dev the fact that nvidia, amd, and intel, all have their own separate slightly janky and entirely incompatible toolchains for gpu compute is so annoying, i really wish there was a common standard that was as well supported as vulkan or something

especially with amd seemingly deciding which cards they wanna have support ROCm on with a dart board (its almost like they don't actually want to break apart the cuda monopoly or something)