On 29 Dec 2025 14:50, Thiago Macieira via Std-Proposals wrote:
> On Sunday, 28 December 2025 22:35:31 Brasilia Standard Time Jonathan Wakely
> via Std-Proposals wrote:
>>> Amazingly, this wasn't lockfree either. I'm dealing with the latest
>>> 'trunk' GNU g++ compiler on x86_64 Linux, which has a CPU instruction
>>> (compare and swap 16 bytes) to make this possible.
>>
>> You need -mcx16 to let GCC use that instruction, and even with that option
>> atomic ops will call into libatomic which decides at runtime whether to use
>> cmpxchg16b or not.
>
> Indeed. Do note that a lockfree CAS isn't enough: there needs to be an atomic
> way of loading and storing 128 bits, which is lacking a dedicated instruction
> on x86-64. Without that, one must lock and unlock to perform loads and stores.
>
> This is possible but undocumented: all machines capable of AVX have 128-bit
> atomic data paths. Using the VMOVDQA instruction will also have a much higher
> performance penalty due to the register file transfers.

It is documented in Intel SDM Volume 3, 10.1.1 Guaranteed Atomic Operations.

But this is only needed if you place an atomic in read-only memory,
which is a pretty pointless use case.