Re: LALR parser question

["Randy Brukardt" <randy@rrsoftware.com>]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1884 bytes --]

"Yannick Duchêne (Hibou57)" <yannick_duchene@yahoo.fr> wrote in message 
news:op.wwfqzzi0ule2fv@cardamome...
Le Thu, 02 May 2013 03:49:53 +0200, Randy Brukardt <randy@rrsoftware.com>
a écrit:
>> I tend to agree that the speed of parsing doesn't matter that much, but 
>> you
>> would be surprised at how slow some compilers were at parsing and syntax
>> analysis, especially before GNAT came out. We actually had a pretty 
>> decent
>> business from companies that bought our fast compiler just to do
>> pre-compiles on before using their main development slug to build the 
>> actual
>> system.
>
>More precisely, is there an average picture? What is considered fast and 
>what is considered slow? Which amount of sample Ada sources parsed in 
>which amount of time? How slow was their "main development slug"? How fast 
>is Janus Ada compared to theirs?

I couldn't answer these questions numerically; we didn't own those other 
compilers that were too slow. And it was a long time ago. I'd heard that in 
some cases it was a factor of 10. But the numbers are hardly relevant today. 
Obviously, the developers thought that they saved enough time doing 
check-out compilations using Janus/Ada to justify paying for a license, 
which was not free. So that suggests that there the main system was pretty 
slow (probably on the order of hours).

After all, in those days it took 4 or so hours to do a full recompile of 
Janus/Ada. (It's under 5 minutes today.) We spent a lot of effort trying not 
to make changes to the symboltable in order to avoid those long compiles, 
which lead to all kinds of misuses of existing components. It's always a 
pain when one of those turns up (these days, I change them to have 
meaningful names and fewer tricks, of course, although its not always worth 
the effort to change).

                                                           Randy.