|
From: | Paul Eggert |
Subject: | Re: emacs-26 9e59de9: Use GCALIGNED properly for GCC |
Date: | Fri, 10 Nov 2017 12:45:15 -0800 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.4.0 |
Stefan Monnier wrote:
Then I don't understand what you meant by "that does not work either".
I was confused (:-). I suppose I was worried about the admittedly unlikely possibility of a platform where alignof (int64_t) exceeds 8. This worry stemmed from an earlier iteration of this fix, where we ran into such a type on MS-Windows. Although perhaps I'm worrying too much, the point remains that because int64_t's alignment is not always a multiple of 8, using int64_t for alignment would not suffice for platforms where alignas does not work.
Also, there's at least one type (struct vectorlike_header) where alignment is convenient, and where the type's size is smaller than int64_t on some platforms, so using int64_t for alignment there would cause objects of this type (and of its containing types) to grow unnecessarily.
[Prev in Thread] | Current Thread | [Next in Thread] |