speeding up the compile times (a lot)

[Trevor Blight]

Hi Scott,
Nice to see you're back to doing work on p5.
Over at p5c, I've just finished some code profiling and was able to obtain dramatic speed ups (10x!!) in compile time This was just by changing 20 or 30 lines of code, and it should be directly transferable to p5 and maybe p6.
Details here ---> https://sourceforge.net/p/pascal-p5c/blog/2017/07/how-to-use-profiling-to-speed-up-p5c-pascal-programs/

[Scott Franco]

Trevor,

Pretty cool. I also received a note from another user who claimed that the compile speed was increased by loading characters directly to strvs (string quanta) structures instead of buffering them, then transferring to strvsps. I haven't verified it.

Pascal-P5 has gone through a lot to increase its compliance with ISO 7185. I often wonder about the different branches of Pascal-P5, including p5c, which were taken from earlier revisions.

Anyways, Thanks for that, will take a look.

Scott Franco

[Trevor Blight]

Scott,
I'm sure you'll find the changes valuble for p5 & probably p6 as well.
The link shows how to analyse the code, so it could be repeated for pint as well, although the speedup might not be so dramatic.

I check into the p5 project every now and then to see what is new and track the changes to pcom.pas and the test suite. P5c passes all the p5 acceptance tests, and nearly all the rejection tests - the exceptions being things like new/dispose(ptr, tag).

-- Trevor//