[PATCH] i386: Increase MALLOC_ALIGNMENT to 16 [BZ #21120]
H.J. Lu
hjl.tools@gmail.com
Fri Jun 30 13:35:00 GMT 2017
On Fri, Jun 30, 2017 at 6:05 AM, Carlos O'Donell <carlos@redhat.com> wrote:
> On 06/30/2017 07:54 AM, H.J. Lu wrote:
>> On Thu, Jun 29, 2017 at 1:10 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> Are you suggesting 3 i386 header files to define MALLOC_ALIGNMENT?
>>>
>>> 1. generic i386.
>>> 2. nptl i386.
>>> 3. hurd i386.
>>>
>>> That is very weird.
>
> I hope my explanation below makes the difference clearer.
>
>> I built and checked them for i686 with GCC 7. The biggest challenge is to
>> support GCC 7, not the header file itself.
>
> We have an existing malloc-machine.h for this purpose.
>
> The design question is:
>
> (a) Should we add a new file malloc-alignment.h to specify a new parameter.
> - Adds new file malloc-alignment.h.
> - 2 new files.
And define MALLOC_ALIGNMENT for i386 in one place.
>
> (b) Should we use existing malloc-machine.h to include the new parameter?
> - Extends existing malloc-machine.h to new sysdep dirs.
> - 3 new files.
And define MALLOC_ALIGNMENT for i386 in 3 difference places.
If we need to support a new OS or threading library, we need to
define MALLOC_ALIGNMENT for i386 in another place.
> You are arguing for (a) and I argue for (b).
>
> Use 1 less file but require a new header.
> Use 1 more file but use existing headers.
>
> I argue that it's simpler and easier for developers if we avoid additional headers
> and use existing headers.
>
Because your suggestion defines MALLOC_ALIGNMENT for i386 in
more than one place, I argue it is more complex and more error
prone than a single MALLOC_ALIGNMENT definition for i386. With
my approach, it is fixed once for all.
--
H.J.
More information about the Libc-alpha
mailing list