pi benchmark on ldc and dmd

[Brad Roberts]

Ok.. I'm pretty sure that's a bug I discovered the other day in the 
initilization code of asm blocks.  I've already got a fix for it and will 
be sending a pull request shortly.

The asm semantic code calls the 32bit initialization code of the backend 
unconditionally, which is just wrong.

On Tue, 2 Aug 2011, Adam D. Ruppe wrote:

> Walter Bright wrote:
> > All I did with your example was replace BigInt with long.
> 
> hmm.... this is my error, but might be a bug too.
> 
> Take that same program and add some inline asm to it.
> 
> void main() {
>    asm { nop; }
> [... the rest is identical ...]
> }
> 
> 
> Now compile it and check the output. With the asm, I get the
> output I posted. If I cut it out, I get what you posted.
> 
> 
> My error here is when I did the obj2asm the first time, I added
> an instruction inline so I could confirm quickly that I was in
> the right place in the file. (I cut that out later but forgot to
> rerun obj2asm.)
>