@jservon may appreciate the addition of a cache for scalars, which he asked for some years ago. This is a cache only for results of function cvalue(). I became convinced when practising by myself the computation of myriads of indices . See new.rst for details. The increase in performance is quite convincing on some use cases
Also, in the same context (multiple indices base on multi-model analysis), I was impressed by the size of CRS expressions (some Mega bytes), their memory footprint, and by the time taken re-computing it. Digging a bit, there was numerous re-computation of the CRS occurring for each object (each time it was involved in an expression). I fixed that. But I cannot guarantee that macros still work fine
@jservon may appreciate the addition of a cache for scalars, which he asked for some years ago. This is a cache only for results of function cvalue(). I became convinced when practising by myself the computation of myriads of indices . See new.rst for details. The increase in performance is quite convincing on some use cases
Also, in the same context (multiple indices base on multi-model analysis), I was impressed by the size of CRS expressions (some Mega bytes), their memory footprint, and by the time taken re-computing it. Digging a bit, there was numerous re-computation of the CRS occurring for each object (each time it was involved in an expression). I fixed that. But I cannot guarantee that macros still work fine