Lukasz Wawrzyniak

CUDA synchronization can be a little tricky, especially when launching work using multiple frameworks that use different scheduling mechanisms under the hood. **Short answer**: If you're not explicitly creating and...

Hi @cadop, I agree that the output of the timers is a little confusing/misleading. The "step" timer only measures the time to schedule the CUDA work, not the time taken...

Hi @ricetwice, thanks for reporting this issue. Looks like it's related to state swaps in the `Example.simulate()` method: ```python # swap states (self.state_0, self.state_1) = (self.state_1, self.state_0) ``` To support...

Looks like there's a C++ exception thrown by NanoVDB which goes unhandled and crashes the program. We can do a better job there and convert it to a Python exception...

@dongwoonhyun this should be fixed with the support for in-out arguments we added in https://github.com/NVIDIA/warp/issues/815. You can include the `a_sum` argument in the `in_out_argnames` list and drop the `dummy` arg:...

Thanks @ramenguy99, this looks like a potentially useful contribution. My main concern is that I'm not aware of any immediate use for it, so we don't have good test cases...

Fixed in 01777f92872f90afa78da5801d43c5a7e2e01aa7

Good news: We have a working implementation of this that doesn't require fiddling with any native bits. The basic usage looks like this: ```python import warp as wp import jax...

Hi @itk22, we're planning to add JAX backward pass interoperability in an upcoming release (#515). This will include optimization examples similar to the PyTorch ones.