Counting an initialised array, and segments

Cecil Ward cecil at cecilward.com
Sun Jun 25 22:08:19 UTC 2023


I recently had some problems

dchar[] arr = [ ‘ ‘, TAB, CR, LF … ];

and I got errors from the compiler which led to me having to 
count the elements in the initialiser and declare the array with 
an explicit size. I don’t want the array to be mutable so I later 
added immutable to it, but that didn’t help matters. At one 
point, because the array was quite long, I got the arr[ 
n_elements ] number wrong, it was too small and the remainder of 
the array was full of 0xffs (or something), which was good, 
helped me spot the bug.

Is there any way to get the compiler to count the number of 
elements in the initialiser and set the array to that size ? And 
it’s immutable.

The only reason that I’m giving it a name is that I want the 
object to be used in several places and I don’t want multiple 
copies of it in the code/readonly initialised data segment.

Another couple of unrelated questions: is there such a thing as a 
no-execute initialised readonly data segment? I’m seeing 
immutables going into the code segment, I think, with x86 LDC at 
least, can’t remember about GDC. Anyway on x86-64 immutables are 
addressed as [rip + displ] which is very pleasing as it’s vastly 
more efficient than accessing statics in TLS which seems to be a 
nightmare in Linux at least.

In MS Windows, isn’t TLS dealt with using FS: ( or GS: ?) 
prefixes? Shame this doesn’t seem to be exploited in Linux, or am 
I wrong?

I’d like to deal with the overhead of retrieving the static base 
address all the time in the Linux situation (if I have got the 
right end of the stick) but having an ‘application object’ which 
contains all the statics in a struct in an alloc cell or 
something, and passing a pointer to this static base app object 
everywhere seems a nightmare too as it eats a register and worse 
eats one of the limited number of precious function argument 
registers which are in short supply in eg x86-64, where there are 
less than half a dozen argument registers allowed. I realise that 
one can deal with that limited number by rolling some passed 
arguments up into a passed struct, but that’s introducing a level 
of indirection and other overhead, that or just live with the 
fact that the extra args are going into the stack, which isn’t 
the worst thing in the world. I wonder what others do about 
statics in TLS?


More information about the Digitalmars-d-learn mailing list