Compiling with another compiler than gcc.

Hartmut Brandt hartmut.brandt at dlr.de
Mon Sep 17 14:36:56 PDT 2007


Anders Magnusson wrote:
> Hi,
> 
> Hartmut Brandt wrote:
>> Anders Magnusson wrote:
>>> It is not yet bug-free, but it can compile the i386 userspace.  The 
>>> big benefit of it
>>> (apart from that it's BSD licensed, for license geeks :-) is that it 
>>> is fast, 5-10 times
>>> faster than gcc, while still producing reasonable code.  The only 
>>
>> When reading the name pcc my first thought was: isn't that the 
>> compiler that was distributed on later Unix V7 tapes? And yes, the 
>> web-page says it is based on that one. I'm quite sure that the 
>> original code had no BSD copyright, so I wonder how it obtained one? 
>> For the rewritten code there is no question, but what for the 
>> remaining original code? Has it been relicensed by the original author?
> Caldera released it with BSD license some 6 years ago, as part of their 
> "ancient unix".

I've still one of these licenses that costed $100 in the beginning :-(
So that looks ok.

> 
>> It's interesting that the compiler is so much faster than gcc. I 
>> remember that it was around 3-5 times slower than the dmr compiler 
>> under V7. This tells a lot about gcc's speed :-(
>>
> Yes, and you are remembering correct.  And yes, it says something about 
> gcc.
> Even more interesting is that there are lots of quite slow sanity check 
> code in
> pcc, despite that it's really fast.

I suppose that the slowness when compared to Ritchie's compiler was 
because it was much larger (causing a lot of swapping) and it uses lex 
and yacc instead of hand-written parsers and lexers. Those where much 
slower on a PDP11. Nowadays this shouldn't make a big difference, 
though. The slowness of gcc comes probably from the machine-independent 
code representation in the backend. At least I remember a speed drop of 
a factor 3-4 when they inventend it going from gcc1.x to gcc2.x

Nice to hear, that you're working on this!

harti



More information about the freebsd-current mailing list