bug#24703: Store references in 8-byte chunks in compiled code

[Ludovic Courtès]

Leo Famulari <address@hidden> skribis:

> On Sat, Nov 05, 2016 at 12:15:25AM +0100, Ludovic Courtès wrote:

[...]

>> Instead, the knobs we have are (1) global flag to enable/disable each
>> built-in function (like -fno-builtin-… does), and (2) an x86-specific
>> knob to determine whether to use ‘movabs’ or not (‘-mmemcpy-strategy’
>> supposedly controls that, but ‘-mmemcpy-strategy=libcall:-1:noalign’
>> doesn’t seem to have any effect for instance.)
>
> Please correct me if I paraphrase the choices incorrectly:
> (1) Completely disable the strcpy optimization for all architectures
> (2) Ostensibly change how strcpy is optimized on x86, except the knob
> seems to have no effect

Correct.

>> These knobs are not great because that would lead us to disable the
>> optimization wholesale, which is not desirable.
>
> What are the costs of (1)? Should we report (2) upstream?

-fno-builtin-{strcpy,memcpy} disables not just use of ‘movabs’ but also
all the optimizations that GCC can make when the built-ins are being
used (see the doc for ‘-mmemcpy-strategy’).  It wouldn’t feel right to
disable that completely.

Ludo’.