Victor Do Nascimento <victor.donascime...@arm.com> writes:
> On 1/5/24 11:10, Richard Sandiford wrote:
>> Victor Do Nascimento <victor.donascime...@arm.com> writes:
>>> The introduction of further architectural-feature dependent ifuncs
>>> for AArch64 makes hard-coding ifunc `_i<n>' suffixes to functions
>>> cumbersome to work with. It is awkward to remember which ifunc maps
>>> onto which arch feature and makes the code harder to maintain when new
>>> ifuncs are added and their suffixes possibly altered.
>>>
>>> This patch uses pre-processor `#define' statements to map each suffix to
>>> a descriptive feature name macro, for example:
>>>
>>> #define LSE2 _i1
>>>
>>> and reconstructs function names with the pre-processor's token
>>> concatenation feature, such that for `MACRO(<name>_i<n>)', we would
>>> now have `MACRO_FEAT(name, feature)' and in the macro definition body
>>> we replace `name` with `name##feature`.
>>
>> FWIW, another way of doing this would be to have:
>>
>> #define CORE(NAME) NAME
>> #define LSE2(NAME) NAME##_i1
>>
>> and use feature(name) instead of name##feature. This has the slight
>> advantage of not using ## on empty tokens, and the maybe slightly
>> better advantage of not needing the extra forwarding step in:
>>
>> #define ENTRY_FEAT(name, feat) \
>> ENTRY_FEAT1(name, feat)
>>
>> #define ENTRY_FEAT1(name, feat) \
>>
>> WDYT?
>>
>> Richard
>>
>
> While from a strictly stylistic point of view, I'm not so keen on the
> resulting interface and its 'function call within a function call' look,
> e.g.
>
> ENTRY (LSE2 (libat_compare_exchange_16))
>
> and
>
> ALIAS (LSE128 (libat_compare_exchange_16), \
> LSE2 (libat_compare_exchange_16))
>
> on the implementation-side of things, I like the benefits this brings
> about. Namely allowing the use of the unaltered original
> implementations of the ENTRY, END and ALIAS macros with the
> aforementioned advantages of not having to use ## on empty tokens and
> abolishing the need for the extra forwarding step.
>
> I'm happy enough to go with this approach.
I was thinking that the invocations would stay the same. A C example is:
#define LSE2(NAME) NAME##_i2
#define ENTRY(NAME, FEAT) void FEAT (NAME) ()
ENTRY(foo, LSE2) {}
https://godbolt.org/z/rdn5dEMPM
Thanks,
Richard
