[foreign-memaccess+abi] RFR: 8275644: Replace VMReg in shuffling code with something more fine grained.

[Maurizio Cimadamore]

On Fri, 5 Aug 2022 15:49:27 GMT, Jorn Vernee <jvernee at openjdk.org> wrote:

> This patch replaces `VMReg` with a new `VMStorage` class in the C++ code, which can be used to represent stack arguments of different sizes, as well as allowing finer grained stack addressing.
> 
> The old masm code that used `VMReg` for moving values is also replaced with new code that uses `VMStorage`. And `VMReg` is changed to `VMStorage` everywhere.
> 
> The `vmstorageBase` class was added in C++ to mirror the `jdk.internal.foreign.VMStorage`. That class now also has an additional field to encode the size of a stack argument, or a register mask, which can also be used to encode the size of a register argument. This also leads to some stubs needing to be updated on unsupported platforms.
> 
> Lastly, instead of encoding stack indexes as 64-bit slots as we did before, they are now encoded as byte offsets, which allows finer grained stack addressing. (this leads to a bunch of values being multiplied by 8 in the Java code as well).
> 
> ---
> 
> This change is needed to address https://bugs.openjdk.org/browse/JDK-8275584 as well as adding support for `long double` as well as more exotic ABIs that take arguments as vectors.

Java code looks good - left some minor comment.
I like the generality that this approach buys, I think it will be helpful when modelling more exotic ABIs/calling conventions.

src/java.base/share/classes/jdk/internal/foreign/abi/VMStorage.java line 31:

> 29: public class VMStorage {
> 30:     private final byte type;
> 31:     private final short segmentMaskOrSize;

We have a couple of fields here which do double duty - e.g. depending on whether the storage models a stack argument or a record, we end up with different meanings. One way is to use (as you did) names like `segmentMaskOrSize`.

But perhaps a better (?) way would be to make VMStorage an abstract class/interface and then have two implementations, for stack slots and registers. Although I realize this might impact the VM code in a bad way (as I think this class is mirrored there in some way).

-------------

Marked as reviewed by mcimadamore (Committer).

PR: https://git.openjdk.org/panama-foreign/pull/699