Raye Kimmerer
Raye Kimmerer
Currently the semantics of `cholmod_*_xtype` are not great since it mutates the data type of an object. It would be nicer in most cases to have `cholmod_*_copy_xtype` which returns a...
There's a note within about zero length VectorTypes. Seems to me that it is impossible by construction. fixes #856
``` [100%] Linking C shared library libgraphblas.so - %80 = tail call @__mulsc3(float noundef %60, float noundef %59, float noundef %67, float noundef %69) #69 Unknown type for libm UNREACHABLE...
Could an option be added to `GrB_get` the equivalent of `GxB_fprint` for all objects as a string? I may make a PR if I get some time.
Dense `apply` seems to allocate when it shouldn't. `GrB_apply(C, NULL, NULL, GrB_SIN_FP64, A, NULL)` shouldn't allocate if `C` and `A` are both non-iso full-dense. I believe `T` is being written...
Many functions now allow a Scalar argument in addition to the existing `GrB_Function_` functions that exist today. Several are missing though, including Monoid. These would be nice to have, and...
Right now compilation is controlled in one of two ways: globally with `GxB_JIT_C_CONTROL` and quasi-local control in the form of type / operator definitions (providing a definition enables JIT, not...
I fully expect this to die, but it's too large to test locally.
1. When do we map and broadcast in the Matrix way vs the Graph way? How do we expose this to the user succinctly. 2. Backup methods for when cfunc...
I believe this is correct, I haven't run it though. This should fix the issue we faced with `xtrue` being different when the number of ranks are > 1.