I want to integrate K different scalar-valued functions over the same range (from 0 to 1). Option A) is to loop from k=1:1:K, and call integral on each function with ArrayValued=False (the default). Option B) is to create a cell-array of the K functions and call integral once with ArrayValued=True. From my reading (in particular this answer: ) it would seem that performing Option A) with ArrayValued=True is going to be slower than Option A) with ArrayValued=False, since ArrayValue=True will turn off vectorization for univariate integrals. This makes sense to me. However, what about Option B) vs. Option A)? Will Option B) also be slower?

Will the ArrayValued option for integral improve or diminish performance relative to looping when integrating a vector-valued function?

17 Giu 2016

0 Risposte

3 Visualizzazioni (30 giorni)

0 voti

I want to integrate K different scalar-valued functions over the same range (from 0 to 1).

Option A) is to loop from k=1:1:K, and call integral on each function with ArrayValued=False (the default).

Option B) is to create a cell-array of the K functions and call integral once with ArrayValued=True.

From my reading (in particular this answer: http://www.mathworks.com/matlabcentral/answers/147338-integrate-function-that-calls-a-function) it would seem that performing Option A) with ArrayValued=True is going to be slower than Option A) with ArrayValued=False, since ArrayValue=True will turn off vectorization for univariate integrals. This makes sense to me.

However, what about Option B) vs. Option A)? Will Option B) also be slower?