Operations a pair rule merged into another.
Color stages the operations collapsed into.
Extent the decode produced.
Operations an identity or annihilation rule removed.
What running this plan is expected to cost in milliseconds, before any encoder.
An estimate and not a promise: the rates behind it were measured on one machine, and against real transforms it lands within about 20%. That is the accuracy the question needs when the question is whether work will fit inside a CPU limit, and it is not a substitute for measuring, which TinyImgModule.measure does.
0 when the source header could not be read.
Special cases the planner took, by name.
Operations left after the rewrites.
Extent of the final image, and how many channels it carries.
Channels per pixel, 1 through 4.
Fused passes the run will make.
The region of the source the decoder was asked for.
The subsampling denominator the decoder was asked for: 1, 2, 4 or 8.
What the planner decided, before any pixel was touched.