taichi
taichi copied to clipboard
taichi-dev
Productive, portable, and performant GPU programming in Python.
Dear Taichi Development Team, I've been trying to use the CUPTI profiler on MPM samples to analyze the bottlenecks of the generated sub kernels. And I encountered and solved several...
From forum: https://forum.taichi.graphics/t/quantaichi-main/2803 Some quantaichi examples are broken due to the use of deprecated Taichi interfaces. Let's fix quantaichi examples then add a simple CI to monitor whether they're in...
**To reproduce** ```py import taichi as ti ti.init() vec2 = ti.types.vector(2, ti.f32) @ti.kernel def test(a: vec2): print(a) test(a=[1, 2]) ``` Output: ``` [Taichi] version 1.0.4, llvm 10.0.0, commit 2827db2c, linux,...
**Describe the bug** The problem arises when I try to let a `ti.func` and a `ti.kernel` share the same function body. I try to do this through the following code:...
Since we are doing frontend type inference now in C++ and it's fully under our control, I would suggest we adopt a stronger & simpler type promotion system. 1. No...
**Concisely describe the proposed feature** I would like to run struct-for partially (with some dims fixed) on a SNode field. **Describe the solution you'd like (if any)** One possible solution...
Related issue = #2906 Related PR = https://github.com/taichi-dev/taichi/pull/4904 The sparse matrix on GPU is based on the `cuSparse` library. In this pr, the user can only build the sparse matrix...
**Concisely describe the proposed feature** I would like to serialize/deserialize sparse data structure with to a buffer or to file, this will benefit online transmission and data logging/replay **Describe the...
It would be nice if we can come up with a list of designed use cases of the C-API interfaces, and make sure they are fully covered by the CI.
Metadata
Owner
Metadata
Productive, portable, and performant GPU programming in Python.