gdaviet

Note that in the meantime it is also possible to implement said decompositions directly in Warp. For instance, here are some QR and Cholesky decomposition routines that I have been...

Note that since Warp 1.7.x this code should no longer throw an error (the vec should get automatically reshaped). The req about getting a more informative message when reshaping is...

Hi @yuhan-zh , user-friendly bindings to cuDSS is indeed something we want to have from warp.sparse eventually. In the meantime, depending on your problem the [iterative linear solvers](https://nvidia.github.io/warp/modules/sparse.html#iterative-linear-solvers) from `warp.optim.linear`...

Thanks @shi-eric. I've asked cursor to generate an example using this API and it seems to work fine. We could potentially add a `csr_to_cupy` helper to streamline this, though the...

There are a couple of dynamic kernels on __main__ that can should moved to their own modules, I'll do that. But the compile time is surprising -- I get about...

For reference the same kernel takes about 5s to compile on my machine (CUDA 13.0 as well, but sm_86) ``` Module __main__._test_cell_lookup__itp_domainDefGeo_NodalField_Tet_0f54a1c3 2b967c1 load on device 'cuda:0' took 5536.38 ms...

Hi @andrinr, I think the short answer is yes -- I am not sure we have an existing example that does exactly that, but happy to provide one. I would...

Cool, I'll try to setup a shape optimization example, I agree that it would be good to have one. Probably with a linear model to keep it simple

Thanks @Cucchi01 for this thorough PR! I 'll have a closer look a the changes. In the meantime: here's a branch https://github.com/gdaviet/warp/tree/fix/temporary_self_reference that marks the `.array` access as deprecated and...

> Whenever we assign self.attr = fem.borrow_temporary(...), in my changes the old buffer is released first. > Even in situation in which the temporary could be returned externally, e.g. Yes,...