[haiku-commits] Re: BRANCH pdziepak-github.scheduler [5e78840] in src/system/kernel/arch/x86/paging: pae 32bit
- From: Pawel Dziepak <pdziepak@xxxxxxxxxxx>
- To: haiku-commits@xxxxxxxxxxxxx
- Date: Thu, 12 Dec 2013 01:49:44 +0100
2013/12/11 Ingo Weinhold <ingo_weinhold@xxxxxx>:
> On 12/11/2013 04:30 AM, pdziepak-github.scheduler wrote:
>>
>> 5e78840: x86/paging: Do not use array placement new
>>
>> Array placement new has implementation defined memory overhead what
>> makes it pretty useless.
>
>
> Oh, how much is it? I thought only a size_t was added to store the array
> size.
I haven't really investigated how compilers deal with this.
Apparently, when the type is not trivial gcc2 reserves a size_t at the
beginning of a given memory area (what also ruins natural alignment of
uint64). gcc4 and clang manges to deal with arrays without reserving
any additional memory (at least in this case).
Paweł
Other related posts:
