You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
(Summarising DH/SB discussions today) element-wise binary operations i.e. operations combining two operands, are not trivial to implement with LAMA, unlike unary operations which are so & are already handled via LAMA in the codebase. We decided to think about & investigate efficient means to partition element-wise binary operations with LAMA towards adding logic to do so for a version after the imminent 3.2.0, ideally 3.3.0.
In particular, I recently implemented (fully except for LAMA management) the two-argument inverse tangent methodarctan2(y, x) which takes two data arrays as inputs and as a first step takes the element-wise quotient. As this needs such two-input LAMA logic, I have commented it out (tagged with AT2) pending that, so arctan2 i.e. division would be the natural case for experimentation.
We can then as appropriate add in further binary methods, though a different LAMA strategy could perhaps optimise (serially) any one case.
The text was updated successfully, but these errors were encountered:
sadielbartholomew
changed the title
LAMA partitioning for binary operations (e.g. arctan2 division)
Binary operation (e.g. arctan2 division) LAMA partitioning
Feb 28, 2020
sadielbartholomew
changed the title
Binary operation (e.g. arctan2 division) LAMA partitioning
Support binary operations e.g. arctan2 division
Nov 17, 2023
(Summarising DH/SB discussions today) element-wise binary operations i.e. operations combining two operands, are not trivial to implement with LAMA, unlike unary operations which are so & are already handled via LAMA in the codebase. We decided to think about & investigate efficient means to partition element-wise binary operations with LAMA towards adding logic to do so for a version after the imminent
3.2.0
, ideally3.3.0
.In particular, I recently implemented (fully except for LAMA management) the two-argument inverse tangent method
arctan2(y, x)
which takes two data arrays as inputs and as a first step takes the element-wise quotient. As this needs such two-input LAMA logic, I have commented it out (tagged withAT2
) pending that, soarctan2
i.e. division would be the natural case for experimentation.We can then as appropriate add in further binary methods, though a different LAMA strategy could perhaps optimise (serially) any one case.
The text was updated successfully, but these errors were encountered: