Zach Corse
Zach Corse
Fixed as of 1.4.0
Yes I have an MR in for this now.
Hi @jinz2014 . Warp doesn't support shared memory in kernels directly, but you are free to use shared memory in native function snippets: https://nvidia.github.io/warp/modules/differentiability.html#custom-native-functions
Yes, `wp.tid()` is a global ID. We don't have functions in Python for those, but you should be able to access eg `threadIdx.x` in a native snippet as you would...
No, I don't think templates would work in snippets currently.
This sort of problem isn't well suited for the Tile API I think. Could you use a SIMT approach instead? ```python n_pre = 1000 n_post = 1000 n_conn = 20...
I see, yeah. This is actually on our roadmap, so we'll be adding indexed load/stores.
@chaoming0625 , these API additions are all scheduled to be included in our 1.9 release (end of this month). You can also try them out now if you pull from...
Thanks for your post @supertan0204 . We have an autograd bug associated with complex types like vec3s that we're currently working on. Should be fixed in our next release, and...
In the example you provided, it should be `[9., 9., 9.]` . The adjoint for each element in x is the sum `adj_x[i] = 2.0 * adj_y[i].x + 3.0 *...