Re: mongodb4 on aarch64

[Paul Ripke <stix%stix.id.au@localhost>]

On Tue, Sep 24, 2024 at 09:38:56AM -0400, Greg Troxel wrote:
> Martin Husemann <martin%duskware.de@localhost> writes:
> 
> > I think the flag is new in gcc12.
> >
> > It makes the compiler change the value of __GCC_DESTRUCTIVE_SIZE, which
> > gets it default from the CPU selected by -mtune=... (but only in C++17
> > or newer mode).
> >
> > You can check the value like this:
> >
> > 	touch /tmp/empty.cpp
> > 	c++ -std=c++17 -dM -E /tmp/empty.cpp | fgrep STRUCTIVE
> >
> > and if the compiler does not define that macro, you could skip the option.
> 
> Sure, but that requires writing and patching a configure test into the
> upstream code.  I left at todo
> 
> > The C++ library uses that macro for a stupid feature in the C++17 standard,
> > std::hardware_destructive_interference_size.
> 
> THanks for the explanation - stuck in comments to help the next person.

One idly wonders how this is supposed to work on machines that have
asymmetric cores (e.g. big.LITTLE) that happen to have differing cache
line sizes...

Yes, they exist. Yes, there are (partial) solutions. Yes, this is
also all horrible.

-- 
Paul Ripke
"Great minds discuss ideas, average minds discuss events, small minds
 discuss people."
-- Disputed: Often attributed to Eleanor Roosevelt. 1948.