Repository navigation
Implement matrix_cl overloads for prim & rev functions #1854
Description
Activity
- changed the title
[-]Implement matrix_cl overloads for remaining /prim functions[/-][+]Implement `matrix_cl<T>`/`var_value<matrix_cl<T>>` overloads for all prim & rev functions[/+]on Dec 7, 2020 Rewrote this issue to now be a TODO list for non-distributions functions (_lpdf/_lpmf,_lcdf, _lccdf, etc are excluded). For now, I went through /prim again, will update /rev tomorrow.
I am also attaching a test file with which I double-checked all of them compile.
prim_sigs.cpp.txt- changed the title
[-]Implement `matrix_cl<T>`/`var_value<matrix_cl<T>>` overloads for all prim & rev functions[/-][+]Implement matrix_cl<T> && var_value<matrix_cl<T>> overloads for all prim & rev functions[/+]on Dec 14, 2020 - changed the title
[-]Implement matrix_cl<T> && var_value<matrix_cl<T>> overloads for all prim & rev functions[/-][+]Implement matrix_cl<T> & var_value<matrix_cl<T>> overloads for all prim & rev functions[/+]on Dec 14, 2020 - changed the title
[-]Implement matrix_cl<T> & var_value<matrix_cl<T>> overloads for all prim & rev functions[/-][+]Implement matrix_cl overloads for prim & rev functions[/+]on Dec 14, 2020
Description
For a list of functions that have OpenCL support for both prim and rev for all exposed signatures see http://mc-stan.org/math/opencl_support.html
Functions with partial or no support that we are targeting to support in the immediate future:
Legend:
xdone*not all signatures are supportedWe also have implementations of rep_vector/row_vector but those can not be used in Stan with OpenCL because the inputs are scalars. Maybe only if we give them a suffix and stanc3 generates a suffixed call to rep_vector/row_vector. It's the same for
rep_matrix(real, int, int).Other functions not in our immediate plans: