Re: [9front] Bounty: OCaml 4.11.1

[McKay Marston <mckay.marston@greymanlabs.com>]

> On Nov 20, 2020, at 10:46 PM, ori@eigenstate.org wrote:
> 
> Quoth McKay Marston <mckay.marston@greymanlabs.com>:
>> I tried porting it myself, but it relies heavily
>> on aligned memory blocks that are generated in
>> preprocessor macros, and I gave up trying to
>> figure out how that could be done in plan9.
> 
> I don't have the time to do the full port myself,
> but do you have any specific examples of
> the kind of thing that's giving trouble?

Well, specifically this:

runtime/caml/misc.h:
#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201112L
#define CAMLalign(n) _Alignas(n)
#elif defined(SUPPORTS_ALIGNED_ATTRIBUTE)
#define CAMLalign(n) __attribute__((aligned(n)))
#elif _MSC_VER >= 1500
#define CAMLalign(n) __declspec(align(n))
#else
#error "How do I align values on this platform?”   <- this is where it bails on plan9
#endif

which is used by:

runtime/caml/domain_state.h
#define DOMAIN_STATE(type, name) CAMLalign(8) type name;
#define DOMAIN_STATE(type, name) CAMLalign(8) type _##name;

which is used by a bunch of stuff like this:

DOMAIN_STATE(intnat, stat_compactions)
DOMAIN_STATE(intnat, stat_heap_chunks)

And my issue is that the only way that I can see to align memory in plan9 is with mallocalign, which isn’t easily translated (by me, at least) into something that works with the “set an aligned attribute at declaration" method used here.

If I’m missing something, though, I’d appreciate any guidance.

—
zgasma
Mckay Marston