Rollup merge of #131586 - taiki-e:s390x-vector-abi, r=compiler-errors,uweigand

Support s390x z13 vector ABI

cc #130869

This resolves the following fixmes:
- https://github.com/rust-lang/rust/blob/58420a065b68ecb3eec03b942740c761cdadd5c4/compiler/rustc_target/src/abi/call/s390x.rs#L1-L2
- https://github.com/rust-lang/rust/blob/58420a065b68ecb3eec03b942740c761cdadd5c4/compiler/rustc_target/src/spec/targets/s390x_unknown_linux_gnu.rs#L9-L11

Refs: Section 1.2.3 "Parameter Passing" and section 1.2.5 "Return Values" in ELF Application Binary Interface s390x Supplement, Version 1.6.1 (lzsabi_s390x.pdf in https://github.com/IBM/s390x-abi/releases/tag/v1.6.1)

This PR extends ~~#127731 #132173 (merged) 's ABI check to handle cases where `vector` target feature is disabled.
If we do not do ABI check, we run into the ABI problems as described in #116558 and #130869 (comment), and the problem of the compiler generating strange code (#131586 (comment)).

cc `@uweigand`

`@rustbot` label +O-SystemZ +A-ABI

Author: matthiaskrgr

Diff for: compiler/rustc_abi/src/layout/ty.rs

@@ -209,6 +209,24 @@ impl<'a, Ty> TyAndLayout<'a, Ty> {
         }
     }
 
+    pub fn is_single_vector_element<C>(self, cx: &C, expected_size: Size) -> bool
+    where
+        Ty: TyAbiInterface<'a, C>,
+        C: HasDataLayout,
+    {
+        match self.backend_repr {
+            BackendRepr::Vector { .. } => self.size == expected_size,
+            BackendRepr::Memory { .. } => {
+                if self.fields.count() == 1 && self.fields.offset(0).bytes() == 0 {
+                    self.field(cx, 0).is_single_vector_element(cx, expected_size)
+                } else {
+                    false
+                }
+            }
+            _ => false,
+        }
+    }
+
     pub fn is_adt<C>(self) -> bool
     where
         Ty: TyAbiInterface<'a, C>,

Diff for: compiler/rustc_target/src/callconv/s390x.rs

@@ -1,12 +1,16 @@
-// FIXME: The assumes we're using the non-vector ABI, i.e., compiling
-// for a pre-z13 machine or using -mno-vx.
+// Reference: ELF Application Binary Interface s390x Supplement
+// https://github.com/IBM/s390x-abi
 
-use crate::abi::call::{ArgAbi, FnAbi, Reg};
-use crate::abi::{HasDataLayout, TyAbiInterface};
+use crate::abi::call::{ArgAbi, FnAbi, Reg, RegKind};
+use crate::abi::{BackendRepr, HasDataLayout, TyAbiInterface};
 use crate::spec::HasTargetSpec;
 
 fn classify_ret<Ty>(ret: &mut ArgAbi<'_, Ty>) {
-    if !ret.layout.is_aggregate() && ret.layout.size.bits() <= 64 {
+    let size = ret.layout.size;
+    if size.bits() <= 128 && matches!(ret.layout.backend_repr, BackendRepr::Vector { .. }) {
+        return;
+    }
+    if !ret.layout.is_aggregate() && size.bits() <= 64 {
         ret.extend_integer_width_to(64);
     } else {
         ret.make_indirect();
@@ -32,19 +36,25 @@ where
         }
         return;
     }
-    if !arg.layout.is_aggregate() && arg.layout.size.bits() <= 64 {
+
+    let size = arg.layout.size;
+    if size.bits() <= 128 && arg.layout.is_single_vector_element(cx, size) {
+        arg.cast_to(Reg { kind: RegKind::Vector, size });
+        return;
+    }
+    if !arg.layout.is_aggregate() && size.bits() <= 64 {
         arg.extend_integer_width_to(64);
         return;
     }
 
     if arg.layout.is_single_fp_element(cx) {
-        match arg.layout.size.bytes() {
+        match size.bytes() {
             4 => arg.cast_to(Reg::f32()),
             8 => arg.cast_to(Reg::f64()),
             _ => arg.make_indirect(),
         }
     } else {
-        match arg.layout.size.bytes() {
+        match size.bytes() {
             1 => arg.cast_to(Reg::i8()),
             2 => arg.cast_to(Reg::i16()),
             4 => arg.cast_to(Reg::i32()),

Diff for: compiler/rustc_target/src/spec/targets/s390x_unknown_linux_gnu.rs

@@ -6,9 +6,6 @@ pub(crate) fn target() -> Target {
     base.endian = Endian::Big;
     // z10 is the oldest CPU supported by LLVM
     base.cpu = "z10".into();
-    // FIXME: The ABI implementation in abi/call/s390x.rs is for now hard-coded to assume the no-vector
-    // ABI. Pass the -vector feature string to LLVM to respect this assumption.
-    base.features = "-vector".into();
     base.max_atomic_width = Some(128);
     base.min_global_align = Some(16);
     base.stack_probes = StackProbeType::Inline;

Diff for: compiler/rustc_target/src/spec/targets/s390x_unknown_linux_musl.rs

@@ -6,9 +6,6 @@ pub(crate) fn target() -> Target {
     base.endian = Endian::Big;
     // z10 is the oldest CPU supported by LLVM
     base.cpu = "z10".into();
-    // FIXME: The ABI implementation in abi/call/s390x.rs is for now hard-coded to assume the no-vector
-    // ABI. Pass the -vector feature string to LLVM to respect this assumption.
-    base.features = "-vector".into();
     base.max_atomic_width = Some(128);
     base.min_global_align = Some(16);
     base.static_position_independent_executables = true;