Date: Wed, 10 Feb 2021 18:45:35 +0000
On Wed, 10 Feb 2021, Richard Smith via Liaison wrote:
> Agreed. The psABI documents still haven't been updated to properly address
> atomic size and alignment, years later. :-(
Perhaps that should be taken as an indication that WG14 needs to be very
careful about proposals with ABI implications (e.g. _BitInt in N2646) in
the absence of visible work with the relevant ABI groups for various
architectures on specifying what the ABI would be, in advance of accepting
such a proposal.
(In the case of atomics, ABIs need to consider the mapping to instructions
as discussed at <https://www.cl.cam.ac.uk/~pes20/cpp/cpp0xmappings.html>,
not just size and alignment.)
> Agreed. The psABI documents still haven't been updated to properly address
> atomic size and alignment, years later. :-(
Perhaps that should be taken as an indication that WG14 needs to be very
careful about proposals with ABI implications (e.g. _BitInt in N2646) in
the absence of visible work with the relevant ABI groups for various
architectures on specifying what the ABI would be, in advance of accepting
such a proposal.
(In the case of atomics, ABIs need to consider the mapping to instructions
as discussed at <https://www.cl.cam.ac.uk/~pes20/cpp/cpp0xmappings.html>,
not just size and alignment.)
-- Joseph S. Myers joseph_at_[hidden]
Received on 2021-02-10 12:45:44