axis_merge

`CartesianAxisMerge`

Bases: ScheduleNodeTransformer

Merge a cartesian axis if they are contiguous in code-flow.

Can do

merge a given axis with the next maps at the same recursion level
can overcompute to allow for more merging at the cost of an if

It expects

All Maps and ForLoop are on a single axis - but doesn't check for it.

This pass has been designed to work in coordination with OffGridConditional and OffGridTasklet passes to allow to align maps for merge

Parameters:

Name	Type	Description	Default
`axis`	`AxisIterator`	AxisIterator to be merged	required
`overcompute`	`bool`	merge at the cost of an if statement.	required

`visit_ScheduleTreeRoot(node)`

Merge consecutive maps of the same cartesian axis.

Dev NOTE: this pass has been implemented in coordination with the rest of the passes deployed in CartesianMerge. Some of the other passes aim to surface maps in a way to maximize the efficiency of this pass.

The algorithm works as follows

Try to merge the surfaced maps
When done, count the number of actual merges
If NO merges - restore the previous children (undo potential changes that didn't lead to map merge)
If merges - go again since we have modified the tree Then exit.

ToDo

ForLoop are not merge at the moment, only Maps.
Non-cartesian ForLoop could be merged down if the maps below are unique (e.g. if everything has been merged) and the hint is PARALLEL. This is relevant for linear solvers and other iteration-dependent algorithmics

`InsertOvercomputationGuard`

Bases: ScheduleNodeTransformer

Recurse down the cartesian-block of maps before patching the inner maps with the if-guard.

Recursing down allows to surface the maps for more subsequent merging. Recursing down also ensures that maximum parallelization is conserved.