tech-toolchain archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Memory alignment not respected in thread-local objects



On 20.12.2018 23:15, Joerg Sonnenberger wrote:
> On Thu, Dec 20, 2018 at 10:21:30PM +0100, Michał Górny wrote:
>> How about forcing a larger default alignment?  64 bytes would be
>> sufficient for our compiler-rt, and I don't think we need to be sparing
>> with the address space here.
> 
> The only case I find in compiler-rt is a pure performance hack of
> questionable validity. Seems entirely optional.
> 
> Joerg
> 

It's used in xray for performance reasons (trying to store a struct in
cache line).

I've recommend to disable this alignas(64) for NetBSD unconditionally
and move on.

I'm also curious why handling alignment is supposed to be difficult.

Attachment: signature.asc
Description: OpenPGP digital signature



Home | Main Index | Thread Index | Old Index