mex or regexp boolean transitions

[JimHokanson]

identifying boolean transitions is apparently a bottleneck in running some code.

+sl+array\bool_transition_info.

obj.true_start_indices = find(~temp_logic(1:end-1) & temp_logic(2:end)); obj.false_start_indices = find(temp_logic(1:end-1) & ~temp_logic(2:end));

The goal is to make this faster.

I was thinking of using mex, but I wonder if regexp might work just as well ...

[JimHokanson]

I think we will need to use mex to avoid memory duplication. With find, I've always wondered if it is faster to look twice or to expand the memory as you go.

I think I've run into a similar problem before and considered using a goto based state machine:

States:

• 0_from_0
• 0_from_1
• 1_from_0
• 1_from_1
• 2_from_0
• 2_from_1

Also note that a logical is 1 byte, so we could use 2 as a sentinel.

Would probably also benefit from using jump tables.