Re: [BDW-GC] Static cell/string/symbol allocation

[Ken Raeburn]

On Jan 12, 2009, at 19:04, Ludovic Courtès wrote:
> > union {
> >  scm_t_cell cell[2];
> >  double d_for_alignment;
> >  long long ll_for_alignment;
> > }
>
> The issue with this is that there's nothing telling us how compilers
> should behave when encountering this.  Even if the underlying hardware
> has a preferred alignment for these types, the compiler doesn't have  
> to
> honor it (on some RISC architectures the alignment can be mandated,  
> and
> failing to honor them would lead to SIGBUS).  So that appears to be
> quite unreliable.

That's true, there's no guarantee at all.  Though I do think Guile is  
probably making a bunch of other assumptions about what the compiler  
or OS will do, especially when it comes to garbage collection,  
equality checks, stuff like that.  There's a difference between what's  
guaranteed by the language specs and what's reliable in the types of  
platforms we'd care about.  For many of them, I think the above will  
cause the desired alignment; I don't know about all of them.

> ... with code that just keeps using dynamic allocation when
> `SCM_ALIGNED' is undefined.

That works too.  The compilers and platforms of greatest interest for  
the biggest part of the user community (GCC and Visual Studio, and  
then...what else?) are likely to provide some facility for this, so  
the performance gain probably only fails to be realized for a small  
part of the audience.

Ken