configure is missing

[keithbriggs]

Following the instructions in README:

kbriggs:~/Downloads/mpfa-0.0.0> ./configure bash: ./configure: No such file or directory

[jamesturner246]

Hi Keith. This applies to a tarball created with make dist.

Please run autoreconf -i -Wall to generate these files from the sources distributed via GitHub.

I should have made this clearer in the README. I will change it now. Thanks for pointing that out.

[keithbriggs]

Thanks, that works now.

Is there any documentation explaining the fundamentals of the method?

Is there a reason for using multiple mpfr_init and mpfr_clear function calls in most functions, instead of mpfr_inits and mpfr_clears?

[jamesturner246]

Some basic documentation is to come, when I get chance. Here are some posts on my (underused neglected) blog:

https://simulatingscience.wordpress.com/2016/12/27/bounding-the-error-of-a-simulation/ https://simulatingscience.wordpress.com/2016/12/31/linear-operations-for-affine-arithmetic/

The Wikipedia page is OK ish:

https://en.wikipedia.org/wiki/Affine_arithmetic

A better, more technical and longer introduction is given here:

https://pdfs.semanticscholar.org/b31d/7169c6aa63fbcab693d76ffab20d3621fd99.pdf

I tried to keep the functions as similar as I could with their MPFR counterparts, but there are a couple of new concepts, like symbol reduction, that might not be clear to users. Happy to answer specific questions for now.

[keithbriggs]

Ok, but I was asking e.g. why add.c has

mpfr_clear(alpha);
mpfr_clear(beta);
mpfr_clear(gamma);
mpfr_clear(delta);

rather than

mpfr_clears(alpha,beta.delta,gamma,NULL);

?

[jamesturner246]

Regarding the init/clear instead of inits/clears: I have no good reason. inits and clears are just variadic loops with init/clear internally, so the only difference is that it arguably looks nicer with single inits and clears calls. I'll look into changing them later.