GCC 4.6

[Walter Bright]

retard wrote:
> Bearophile often compares DMD to world class products such as LLVM. Even 
> though Walter might be the best compiler writer in this world, he can't 
> compete with a motivated team of professionals.

What irritated me about bearophile's comparison of dmd to LLVM is he would say 
things like dmd didn't do data flow analysis (it does) and that it had poor 
integer arithmetic code gen (the integer arithmetic was the same) that it didn't 
do common loop optimizations (it did), etc.


> The Intel C++ compiler, 
> GCC, and LLVM teams are simply much bigger and well paid.

Yes, I know people just assume dmd doesn't do things because it is a small team. 
My favorite was the assumption that dmd didn't do the Named Return Value 
optimization, an optimization I invented back in 1990.

Back in the days of competitive compiler benchmarks, my compiler often won them, 
up against those well paid huge compiler teams from MS, Borland, etc. A friend 
told me once that the head of one large company used to yell down the halls how 
come Walter sitting over his garage can beat you guys. That made me feel good.


> Same things 
> with DMDScript vs Google's Javascript engine. It's almost impossible to 
> compete with these big teams. If you need performance, you probably use 
> these tools instead.

I have to laugh at that. When DMDScript was created, back in 2001, it was twice 
as fast as Microsoft's Jscript, and *twenty* times faster than Mozilla's 
javascript. All from little ole me. It passed Sun's javascript validation suite, 
and was sold by sun as part of their Chilicom software (the C++ version).

I haven't worked on it in 10 years, though, because of a lack of customer 
interest in it. It took the others 10 years to catch up.

Where the large teams pay off, though, is the breadth of the offering.