54 lines
1.1 KiB
Markdown
54 lines
1.1 KiB
Markdown
# """"Dynamic Programming""""
|
|
|
|
|
|
> Take a large problem
|
|
> Break it up
|
|
> Solve the pieces
|
|
|
|
|
|
|
|
# Definitions
|
|
|
|
_Recursive dynamic programming_
|
|
: Memoization a.k.a. top-down approach
|
|
: Go from big problems and build downward
|
|
|
|
_Tabulation_
|
|
: Bottom up approach
|
|
: Go from little problems and build up
|
|
|
|
|
|
# Recursive compiler tricks
|
|
|
|
Normal iterative:
|
|
```
|
|
start:
|
|
...
|
|
jcond start
|
|
```
|
|
|
|
Normal recursion(under non-priveleged procs):
|
|
|
|
Each "iteration" creates a new frame of whatever size:
|
|
hence we approach the hard stack limit
|
|
|
|
_Sick guess_:
|
|
```
|
|
At compile we should be able to predict an incoming iteration from a
|
|
recusive endpoint.
|
|
|
|
? is it just parameterized jumps ?
|
|
```
|
|
|
|
_Answer:_
|
|
```
|
|
Tail recusion optimization
|
|
|
|
Takes the last statement of function
|
|
Assuming its a recursive call, we can replace our current stack frame's arguments
|
|
The only state that changes between calls then is just the arguments
|
|
```
|
|
In this sense we are turning a recusive functino into an iterative one because the whole state is held in one frame & we only ever jump within our routines address space.
|
|
|
|
However, this tail optimization can only happen at the tail of a routine.
|