🚀 go-pugleaf

In article <dmurdoch.253.721061938@mast.queensu.ca> dmurdoch@mast.queensu.ca (Duncan Murdoch) writes:
>Why do people accept such slow compile times?  Turbo Pascal demonstrates 
>that fast compiler technology exists.  Is C++ five times harder to compile 
>than TP?

Someone from Borland pointed out to me that C++ is indeed a much 
more complex language than TP, but the main thing is that the C++ language 
is not under Borland's control.  Borland can pick and choose features for 
Turbo Pascal with an eye to compiler efficiency, while someone else chooses 
what goes into C++.  Another consideration that comes from the size of C++ 
is that it's hard to fit the compiler in real mode memory; DPMI slows it 
down.  I'd tend to discount this; TP for Windows runs in protected 
mode nearly as fast as TP for DOS runs.  Finally, C++ includes more 
optimizations than TP - but even with those turned off, it runs much slower.

One consequence of this that wasn't mentioned, but which is probably really 
the reason for TP's superior speed, is that the TP compiler is written in 
hand-optimized assembler.  Because C++ is so big, and because it's liable to 
change under someone else's control, it has to be written in a high level 
language.  (I'd assume C, with low level service routines in assembler; can 
anyone confirm or deny that?)  


Duncan Murdoch
dmurdoch@mast.queensu.ca