[<<][rtl][>>][..]

Fri May 17 17:20:16 EDT 2019

## accumulation

i:
Ai <- A(i-1) + Bi
This notation abstracts the need to initialize the accumulator before
the loop starts.
Representing accumulators is important, so maybe it is possible to
generalize:
i:
A_i <- A_f(i) + B_i
There based on what A_f(i) is, we can implement it as a single
accumulator, a finite set of accumulators (a shift register), or a
full array.
So that is all fluff. Let's use i' as the previous index, with
implicit initialization.
Actually I don't have any way to use the two types of references:
previous accumulator value and final, after the loop.
i:
A <- A Bi
i:
Di <- A Ci
So it appears that accumulators need some special notation. To make
it work for triangle patterns, use the triangle notation.
i:
Ai <- Ai1 Bi
i:
Di <- AI Ci
Where I is the last element that was stored in the array in its
constructing loop.

[Reply][About]

[<<][rtl][>>][..]