Project

General

Profile

Slug #884

DistrMPolyInlPP::myPushFront and DistrMPolyInlPP::myPushBack inefficient if arg is a PP

Added by John Abbott almost 8 years ago. Updated almost 8 years ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
Improving
Target version:
CoCoALib-1.0
Start date:
24 May 2016
Due date:
% Done:

0%

Estimated time:
Spent time:
0.50 h

Description

In the class DistrMPolyInlPP the mem fns myPushFront and myPushBack@ can accept raw coeff and raw pp, BUT the impl then "explodes" the PP into a vector of exponents then calls the myPushFront or myPushBack mem fns on the vector. This is rather inefficient!

Find a way to copy the aleady encoded PP!

Related issues

Related to CoCoALib - Slug #881: ReadExpr is too slow on large polysClosed2016-05-09

History

#1 Updated by John Abbott almost 8 years ago

I wrote a special reading fn for polynomials (using a "private" linearization). It was not as fast as I hoped, and profiling showed that a lot of time was spent in DistrMPolyInlPP::myPushFront.

Looking at the code I saw that it "explodes" the PP into an expv, and then recompresses it into an ordv. This is a perceptible waste of time!

The source code for myPushFront contained a commented-out line calling a straight assignment; activating the line triggered a compilation error. The problem is that the PP which is passed in is of type PPMonoidElemConstRawPtr but the assignment wants a const OrdvElem*. How to do what I want reasonably cleanly and safely?

#2 Updated by John Abbott almost 8 years ago

  • Related to Slug #881: ReadExpr is too slow on large polys added

Also available in: Atom PDF