Wouldn't memory handling be outside of Mantle's API though? Even on PC Mantle would eventually need to work with different memory structures in GCN being on discrete cards and as part of APUs so I'd expect that part to be handled more abstracted on a higher level and be an implementation detail in the underlying driver. Or something like that.
Depends on the scope of the API, but if we're talking about something low level, I think there'd need to be exposure around what memory is being used for certain texture/buffer related operations in XB1, things that are taken for granted in a PC memory model.
Maybe they are just referring to it being compatible with DX's HLSL? Then using the above wording is a great way to save face. =D
When they talked about it, they said it was a case of starting with DX and taking things out and optimising the remainder for the chip.
That doesn't sound like shader syntax compatibility.
Like I said, I could see AMD licensing Mantle to MS, but couldn't see MS licensing something like the above for open cross platform use, under AMD's control. Of course it's possible MS isn't telling the whole truth about the genesis of their driver, and that they did do a license of Mantle, but it would be a modified version even in that case methinks rather than something the exact same as the PC version.
...
It's fairly academic though. It doesn't really matter, except that it would make PC versions of console games even easier to optimise for AMD hardware, if there was a direct analog in the console space. But I don't think Anandtech supports their speculation well, there's lots of other explanations for HLSL syntax compatibility.