cranelift: introduce a new WasmtimeAppleAarch64 calling convention

The previous choice to use the WasmtimeSystemV calling convention for
apple-aarch64 devices was incorrect: padding of arguments was
incorrectly computed. So we have to use some flavor of the apple-aarch64
ABI there.

Since we want to support the wasmtime custom convention for multiple
returns on apple-aarch64 too, a new custom Wasmtime calling convention
was introduced to support this.

cranelift/codegen/src/isa/aarch64/abi.rs

@@ -183,7 +183,7 @@ impl ABIMachineSpec for AArch64MachineDeps {
         args_or_rets: ArgsOrRets,
         add_ret_area_ptr: bool,
     ) -> CodegenResult<(Vec<ABIArg>, i64, Option<usize>)> {
-        let is_apple_cc = call_conv == isa::CallConv::AppleAarch64;
+        let is_apple_cc = call_conv.extends_apple_aarch64();
         let is_baldrdash = call_conv.extends_baldrdash();
         let has_baldrdash_tls = call_conv == isa::CallConv::Baldrdash2020;
 

cranelift/codegen/src/isa/call_conv.rs

@@ -18,7 +18,7 @@ pub enum CallConv {
     SystemV,
     /// Windows "fastcall" convention, also used for x64 and ARM.
     WindowsFastcall,
-    /// Mac aarch64 calling convention, which is a tweak aarch64 ABI.
+    /// Mac aarch64 calling convention, which is a tweaked aarch64 ABI.
     AppleAarch64,
     /// SpiderMonkey WebAssembly convention on systems using natively SystemV.
     BaldrdashSystemV,
@@ -39,6 +39,10 @@ pub enum CallConv {
     ///
     /// Differs from fastcall in the same way as `WasmtimeSystemV`.
     WasmtimeFastcall,
+    /// Wasmtime equivalent of AppleAarch64, not ABI-stable.
+    ///
+    /// Differs from apple-aarch64 in the same way as `WasmtimeSystemV`.
+    WasmtimeAppleAarch64,
 }
 
 impl CallConv {
@@ -78,6 +82,14 @@ impl CallConv {
         }
     }
 
+    /// Is the calling convention extending the Apple aarch64 ABI?
+    pub fn extends_apple_aarch64(self) -> bool {
+        match self {
+            Self::AppleAarch64 | Self::WasmtimeAppleAarch64 => true,
+            _ => false,
+        }
+    }
+
     /// Is the calling convention extending the Baldrdash ABI?
     pub fn extends_baldrdash(self) -> bool {
         match self {
@@ -89,7 +101,7 @@ impl CallConv {
     /// Is the calling convention extending the Wasmtime ABI?
     pub fn extends_wasmtime(self) -> bool {
         match self {
-            Self::WasmtimeSystemV | Self::WasmtimeFastcall => true,
+            Self::WasmtimeSystemV | Self::WasmtimeFastcall | Self::WasmtimeAppleAarch64 => true,
             _ => false,
         }
     }
@@ -109,6 +121,7 @@ impl fmt::Display for CallConv {
             Self::Probestack => "probestack",
             Self::WasmtimeSystemV => "wasmtime_system_v",
             Self::WasmtimeFastcall => "wasmtime_fastcall",
+            Self::WasmtimeAppleAarch64 => "wasmtime_apple_aarch64",
         })
     }
 }
@@ -128,6 +141,7 @@ impl str::FromStr for CallConv {
             "probestack" => Ok(Self::Probestack),
             "wasmtime_system_v" => Ok(Self::WasmtimeSystemV),
             "wasmtime_fastcall" => Ok(Self::WasmtimeFastcall),
+            "wasmtime_apple_aarch64" => Ok(Self::WasmtimeAppleAarch64),
             _ => Err(()),
         }
     }

cranelift/codegen/src/isa/x64/abi.rs

@@ -906,7 +906,7 @@ fn get_intreg_for_retval(
             _ => None,
         },
         CallConv::BaldrdashWindows | CallConv::Probestack => todo!(),
-        CallConv::AppleAarch64 => unreachable!(),
+        CallConv::AppleAarch64 | CallConv::WasmtimeAppleAarch64 => unreachable!(),
     }
 }
 
@@ -936,7 +936,7 @@ fn get_fltreg_for_retval(
             _ => None,
         },
         CallConv::BaldrdashWindows | CallConv::Probestack => todo!(),
-        CallConv::AppleAarch64 => unreachable!(),
+        CallConv::AppleAarch64 | CallConv::WasmtimeAppleAarch64 => unreachable!(),
     }
 }
 
@@ -1005,7 +1005,7 @@ fn get_callee_saves(call_conv: &CallConv, regs: &Set<Writable<RealReg>>) -> Vec<
             .filter(|r| is_callee_save_fastcall(r.to_reg()))
             .collect(),
         CallConv::Probestack => todo!("probestack?"),
-        CallConv::AppleAarch64 => unreachable!(),
+        CallConv::AppleAarch64 | CallConv::WasmtimeAppleAarch64 => unreachable!(),
     };
     // Sort registers for deterministic code output. We can do an unstable sort because the
     // registers will be unique (there are no dups).

cranelift/codegen/src/isa/x86/abi.rs

@@ -514,7 +514,7 @@ pub fn prologue_epilogue(func: &mut ir::Function, isa: &dyn TargetIsa) -> Codege
         }
         CallConv::Probestack => unimplemented!("probestack calling convention"),
         CallConv::Baldrdash2020 => unimplemented!("Baldrdash ABI 2020"),
-        CallConv::AppleAarch64 => unreachable!(),
+        CallConv::AppleAarch64 | CallConv::WasmtimeAppleAarch64 => unreachable!(),
     }
 }