Multiple return values...

[Timon Gehr]

On 03/11/2012 09:30 PM, Robert Jacques wrote:
> On Sun, 11 Mar 2012 12:39:56 -0500, Manu <turkeyman at gmail.com> wrote:
>> On 11 March 2012 18:50, Robert Jacques <sandford at jhu.edu> wrote:
>>> On Sun, 11 Mar 2012 05:57:05 -0500, Manu <turkeyman at gmail.com> wrote:
>>> On 11 March 2012 04:35, Sean Cavanaugh <WorksOnMyMachine at gmail.com>
>>>> wrote:
>>>> On 3/10/2012 8:08 PM, Mantis wrote:
>
> [snip]
>
>>> And Walter has talked about using the XMM registers to return floating
>>> point data for exactly this reason. But those optimization apply to all
>>> structs and all data types. There's nothing special about MRV. It has to
>>> return a set of data in a structured manner; this is identical to the
>>> case
>>> of returning a struct.
>>>
>>
>> Well you can't build these sorts of tightly packed structs in XMM
>> registers... and even for the GPR's, it's very inefficient to do so. I
>> haven't done tests, but I suspect this feature is probably a
>> de-optimisation in all cases other than an int-pair struct (ranges,
>> delegates? I suspect that's why this was invented). Structure
>> packing/unpacking code will likely be slower than a store/load.
>>
>> We just need proper multiple return values, then this can go away. And
>> the
>> earlier any ABI breaking changes are implemented, the better, while there
>> are still few(/none?) closed source, binary distributed D libraries.
>
> Manu, why are you assuming that the struct is returned in any sort of
> fixed memory layout?

Because that is what the ABI says.

* 1, 2 and 4 byte structs are returned in EAX.
* 8 byte structs are returned in EDX,EAX, where EDX gets the most 
significant half.
* For other struct sizes, the return value is stored through a hidden 
pointer passed as an argument to the function.

> You can _not_ take the memory address, etc of a
> register so maintaining the layout of a struct is a) impossible and b)
> not necessary, if you're returning it via registers. The compiler is
> free to use XMM registers for floats/doubles and GPRs for pointers/ints.
> And there's no need for the caller to place those struct register values
> on the stack if they're never used. Furthermore, even on the stack, the
> struct's memory layout is only ever valid immediately before (and
> forever after due to pointer escapes) a non-inlined function call or
> address taking. At a low level, MRV is a structured set of value, ergo
> it's a struct.

Multiple function arguments are a structured set of values too.