PENGUINLIONG
PENGUINLIONG
In the current implementation the device allocation ID is assigned by a problematic algo in multiple backends. For example: https://github.com/taichi-dev/taichi/blob/365e755190643c9f4504fbe7f990bd167762dfe0/taichi/rhi/cpu/cpu_device.cpp#L13-L28 In the CPU backend, freeing device allocations can lead to...
Profiling data shows that the current implementation only reaches 70% of the full ALU capacity (on Adreno 640), limited by the data dependency between instructions. The provided implementation can reach...
https://github.com/taichi-dev/taichi/pull/7329
https://github.com/taichi-dev/taichi/pull/7329
In the current implementation MoltenVK doesn't enforce a barrier when sampling a timestamp, in which case the timestamp may be sampled while some previous shaders are still running. As a...
Use `qlmanage -r` to reload the Quick Look service