cacheing compiled integrand in quadpack functions



Camm Maguire <camm at enhanced.com> wrote:

> We had a compiler speed discussion some time ago, and the input then
> was that people wanted fast executables even if at the expense of some
> compile time.  That said, there are some obvious things we could do if
> there is interest, such as leaving a gcc process open and speaking to
> it via a pipe, etc.  I won't be able to look into this for a month,
> but ideas/suggestions are always most welcome.

a couple of ideas --

(1) can a "warm start" of the compiler be arranged?
talking to gcc via a pipe might address that.

(2) can we arrange it so the compiler doesn't touch the file system?
that would have 2 nice effects -- the compiler could
run faster, and gcl could run on read-only file systems.

for what it's worth,
robert dodier