Merge commit '482e8540a1' into subtree-update_cg_gcc_2025-08-04

compiler/rustc_codegen_gcc/build_system/src/abi_test.rs

@@ -31,7 +31,7 @@ pub fn run() -> Result<(), String> {
         Some("clones/abi-cafe".as_ref()),
         true,
     )
-    .map_err(|err| (format!("Git clone failed with message: {err:?}!")))?;
+    .map_err(|err| format!("Git clone failed with message: {err:?}!"))?;
     // Configure abi-cafe to use the exact same rustc version we use - this is crucial.
     // Otherwise, the concept of ABI compatibility becomes meanignless.
     std::fs::copy("rust-toolchain", "clones/abi-cafe/rust-toolchain")

compiler/rustc_codegen_gcc/build_system/src/fuzz.rs

@@ -43,18 +43,18 @@ pub fn run() -> Result<(), String> {
             "--start" => {
                 start =
                     str::parse(&args.next().ok_or_else(|| "Fuzz start not provided!".to_string())?)
-                        .map_err(|err| (format!("Fuzz start not a number {err:?}!")))?;
+                        .map_err(|err| format!("Fuzz start not a number {err:?}!"))?;
             }
             "--count" => {
                 count =
                     str::parse(&args.next().ok_or_else(|| "Fuzz count not provided!".to_string())?)
-                        .map_err(|err| (format!("Fuzz count not a number {err:?}!")))?;
+                        .map_err(|err| format!("Fuzz count not a number {err:?}!"))?;
             }
             "-j" | "--jobs" => {
                 threads = str::parse(
                     &args.next().ok_or_else(|| "Fuzz thread count not provided!".to_string())?,
                 )
-                .map_err(|err| (format!("Fuzz thread count not a number {err:?}!")))?;
+                .map_err(|err| format!("Fuzz thread count not a number {err:?}!"))?;
             }
             _ => return Err(format!("Unknown option {arg}")),
         }
@@ -66,7 +66,7 @@ pub fn run() -> Result<(), String> {
         Some("clones/rustlantis".as_ref()),
         true,
     )
-    .map_err(|err| (format!("Git clone failed with message: {err:?}!")))?;
+    .map_err(|err| format!("Git clone failed with message: {err:?}!"))?;
 
     // Ensure that we are on the newest rustlantis commit.
     let cmd: &[&dyn AsRef<OsStr>] = &[&"git", &"pull", &"origin"];

compiler/rustc_codegen_gcc/patches/0001-Add-stdarch-Cargo.toml-for-testing.patch

@@ -1,29 +1,28 @@
-From b8f3eed3053c9333b5dfbeaeb2a6a65a4b3156df Mon Sep 17 00:00:00 2001
+From 190e26c9274b3c93a9ee3516b395590e6bd9213b Mon Sep 17 00:00:00 2001
-From: Antoni Boucher <bouanto@zoho.com>
+From: None <none@example.com>
-Date: Tue, 29 Aug 2023 13:06:34 -0400
+Date: Sun, 3 Aug 2025 19:54:56 -0400
 Subject: [PATCH] Patch 0001-Add-stdarch-Cargo.toml-for-testing.patch
 
 ---
- library/stdarch/Cargo.toml | 23 +++++++++++++++++++++++
+ library/stdarch/Cargo.toml | 20 ++++++++++++++++++++
- 1 file changed, 23 insertions(+)
+ 1 file changed, 20 insertions(+)
  create mode 100644 library/stdarch/Cargo.toml
 
 diff --git a/library/stdarch/Cargo.toml b/library/stdarch/Cargo.toml
 new file mode 100644
-index 0000000..4c63700
+index 0000000..bd6725c
 --- /dev/null
 +++ b/library/stdarch/Cargo.toml
-@@ -0,0 +1,21 @@
+@@ -0,0 +1,20 @@
 +[workspace]
 +resolver = "1"
 +members = [
-+  "crates/core_arch",
++  "crates/*",
-+  "crates/std_detect",
-+  "crates/stdarch-gen-arm",
 +  #"examples/"
 +]
 +exclude = [
-+  "crates/wasm-assert-instr-tests"
++  "crates/wasm-assert-instr-tests",
++  "rust_programs",
 +]
 +
 +[profile.release]
@@ -36,5 +35,5 @@ index 0000000..4c63700
 +opt-level = 3
 +incremental = true
 -- 
-2.42.0
+2.50.1
 

compiler/rustc_codegen_gcc/rust-toolchain

@@ -1,3 +1,3 @@
 [toolchain]
-channel = "nightly-2025-07-04"
+channel = "nightly-2025-08-03"
 components = ["rust-src", "rustc-dev", "llvm-tools-preview"]

compiler/rustc_codegen_gcc/src/allocator.rs

@@ -99,11 +99,15 @@ fn create_const_value_function(
     let func = context.new_function(None, FunctionType::Exported, output, &[], name, false);
 
     #[cfg(feature = "master")]
-    func.add_attribute(FnAttribute::Visibility(symbol_visibility_to_gcc(
+    {
-        tcx.sess.default_visibility(),
+        func.add_attribute(FnAttribute::Visibility(symbol_visibility_to_gcc(
-    )));
+            tcx.sess.default_visibility(),
+        )));
 
-    func.add_attribute(FnAttribute::AlwaysInline);
+        // FIXME(antoyo): cg_llvm sets AlwaysInline, but AlwaysInline is different in GCC and using
+        // it here will causes linking errors when using LTO.
+        func.add_attribute(FnAttribute::Inline);
+    }
 
     if tcx.sess.must_emit_unwind_tables() {
         // TODO(antoyo): emit unwind tables.

compiler/rustc_codegen_gcc/src/builder.rs

@@ -540,9 +540,15 @@ impl<'a, 'gcc, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'gcc, 'tcx> {
 
     fn ret(&mut self, mut value: RValue<'gcc>) {
         let expected_return_type = self.current_func().get_return_type();
-        if !expected_return_type.is_compatible_with(value.get_type()) {
+        let value_type = value.get_type();
-            // NOTE: due to opaque pointers now being used, we need to cast here.
+        if !expected_return_type.is_compatible_with(value_type) {
-            value = self.context.new_cast(self.location, value, expected_return_type);
+            // NOTE: due to opaque pointers now being used, we need to (bit)cast here.
+            if self.is_native_int_type(value_type) && self.is_native_int_type(expected_return_type)
+            {
+                value = self.context.new_cast(self.location, value, expected_return_type);
+            } else {
+                value = self.context.new_bitcast(self.location, value, expected_return_type);
+            }
         }
         self.llbb().end_with_return(self.location, value);
     }
@@ -1279,11 +1285,19 @@ impl<'a, 'gcc, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'gcc, 'tcx> {
 
     fn intcast(
         &mut self,
-        value: RValue<'gcc>,
+        mut value: RValue<'gcc>,
         dest_typ: Type<'gcc>,
-        _is_signed: bool,
+        is_signed: bool,
     ) -> RValue<'gcc> {
-        // NOTE: is_signed is for value, not dest_typ.
+        let value_type = value.get_type();
+        if is_signed && !value_type.is_signed(self.cx) {
+            let signed_type = value_type.to_signed(self.cx);
+            value = self.gcc_int_cast(value, signed_type);
+        } else if !is_signed && value_type.is_signed(self.cx) {
+            let unsigned_type = value_type.to_unsigned(self.cx);
+            value = self.gcc_int_cast(value, unsigned_type);
+        }
+
         self.gcc_int_cast(value, dest_typ)
     }
 

compiler/rustc_codegen_gcc/src/int.rs

@@ -4,12 +4,15 @@
 
 // cSpell:words cmpti divti modti mulodi muloti udivti umodti
 
-use gccjit::{BinaryOp, ComparisonOp, FunctionType, Location, RValue, ToRValue, Type, UnaryOp};
+use gccjit::{
+    BinaryOp, CType, ComparisonOp, FunctionType, Location, RValue, ToRValue, Type, UnaryOp,
+};
 use rustc_abi::{CanonAbi, Endian, ExternAbi};
 use rustc_codegen_ssa::common::{IntPredicate, TypeKind};
 use rustc_codegen_ssa::traits::{BackendTypes, BaseTypeCodegenMethods, BuilderMethods, OverflowOp};
 use rustc_middle::ty::{self, Ty};
 use rustc_target::callconv::{ArgAbi, ArgAttributes, FnAbi, PassMode};
+use rustc_type_ir::{Interner, TyKind};
 
 use crate::builder::{Builder, ToGccComp};
 use crate::common::{SignType, TypeReflection};
@@ -167,9 +170,9 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
                 if a_type.is_vector() {
                     // Vector types need to be bitcast.
                     // TODO(antoyo): perhaps use __builtin_convertvector for vector casting.
-                    b = self.context.new_bitcast(self.location, b, a.get_type());
+                    b = self.context.new_bitcast(self.location, b, a_type);
                 } else {
-                    b = self.context.new_cast(self.location, b, a.get_type());
+                    b = self.context.new_cast(self.location, b, a_type);
                 }
             }
             self.context.new_binary_op(self.location, operation, a_type, a, b)
@@ -216,13 +219,22 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
         operation_name: &str,
         signed: bool,
         a: RValue<'gcc>,
-        b: RValue<'gcc>,
+        mut b: RValue<'gcc>,
     ) -> RValue<'gcc> {
         let a_type = a.get_type();
         let b_type = b.get_type();
         if (self.is_native_int_type_or_bool(a_type) && self.is_native_int_type_or_bool(b_type))
             || (a_type.is_vector() && b_type.is_vector())
         {
+            if !a_type.is_compatible_with(b_type) {
+                if a_type.is_vector() {
+                    // Vector types need to be bitcast.
+                    // TODO(antoyo): perhaps use __builtin_convertvector for vector casting.
+                    b = self.context.new_bitcast(self.location, b, a_type);
+                } else {
+                    b = self.context.new_cast(self.location, b, a_type);
+                }
+            }
             self.context.new_binary_op(self.location, operation, a_type, a, b)
         } else {
             debug_assert!(a_type.dyncast_array().is_some());
@@ -351,6 +363,11 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
             // TODO(antoyo): is it correct to use rhs type instead of the parameter typ?
             .new_local(self.location, rhs.get_type(), "binopResult")
             .get_address(self.location);
+        let new_type = type_kind_to_gcc_type(new_kind);
+        let new_type = self.context.new_c_type(new_type);
+        let lhs = self.context.new_cast(self.location, lhs, new_type);
+        let rhs = self.context.new_cast(self.location, rhs, new_type);
+        let res = self.context.new_cast(self.location, res, new_type.make_pointer());
         let overflow = self.overflow_call(intrinsic, &[lhs, rhs, res], None);
         (res.dereference(self.location).to_rvalue(), overflow)
     }
@@ -477,11 +494,27 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
             let lhs_low = self.context.new_cast(self.location, self.low(lhs), unsigned_type);
             let rhs_low = self.context.new_cast(self.location, self.low(rhs), unsigned_type);
 
+            let mut lhs_high = self.high(lhs);
+            let mut rhs_high = self.high(rhs);
+
+            match op {
+                IntPredicate::IntUGT
+                | IntPredicate::IntUGE
+                | IntPredicate::IntULT
+                | IntPredicate::IntULE => {
+                    lhs_high = self.context.new_cast(self.location, lhs_high, unsigned_type);
+                    rhs_high = self.context.new_cast(self.location, rhs_high, unsigned_type);
+                }
+                // TODO(antoyo): we probably need to handle signed comparison for unsigned
+                // integers.
+                _ => (),
+            }
+
             let condition = self.context.new_comparison(
                 self.location,
                 ComparisonOp::LessThan,
-                self.high(lhs),
+                lhs_high,
-                self.high(rhs),
+                rhs_high,
             );
             self.llbb().end_with_conditional(self.location, condition, block1, block2);
 
@@ -495,8 +528,8 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
             let condition = self.context.new_comparison(
                 self.location,
                 ComparisonOp::GreaterThan,
-                self.high(lhs),
+                lhs_high,
-                self.high(rhs),
+                rhs_high,
             );
             block2.end_with_conditional(self.location, condition, block3, block4);
 
@@ -620,7 +653,7 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
         }
     }
 
-    pub fn gcc_xor(&self, a: RValue<'gcc>, b: RValue<'gcc>) -> RValue<'gcc> {
+    pub fn gcc_xor(&self, a: RValue<'gcc>, mut b: RValue<'gcc>) -> RValue<'gcc> {
         let a_type = a.get_type();
         let b_type = b.get_type();
         if a_type.is_vector() && b_type.is_vector() {
@@ -628,6 +661,9 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
             a ^ b
         } else if self.is_native_int_type_or_bool(a_type) && self.is_native_int_type_or_bool(b_type)
         {
+            if !a_type.is_compatible_with(b_type) {
+                b = self.context.new_cast(self.location, b, a_type);
+            }
             a ^ b
         } else {
             self.concat_low_high_rvalues(
@@ -1042,3 +1078,25 @@ impl<'gcc, 'tcx> CodegenCx<'gcc, 'tcx> {
         self.context.new_array_constructor(None, typ, &values)
     }
 }
+
+fn type_kind_to_gcc_type<I: Interner>(kind: TyKind<I>) -> CType {
+    use rustc_middle::ty::IntTy::*;
+    use rustc_middle::ty::UintTy::*;
+    use rustc_middle::ty::{Int, Uint};
+
+    match kind {
+        Int(I8) => CType::Int8t,
+        Int(I16) => CType::Int16t,
+        Int(I32) => CType::Int32t,
+        Int(I64) => CType::Int64t,
+        Int(I128) => CType::Int128t,
+
+        Uint(U8) => CType::UInt8t,
+        Uint(U16) => CType::UInt16t,
+        Uint(U32) => CType::UInt32t,
+        Uint(U64) => CType::UInt64t,
+        Uint(U128) => CType::UInt128t,
+
+        _ => unimplemented!("Kind: {:?}", kind),
+    }
+}

compiler/rustc_codegen_gcc/src/intrinsic/archs.rs

@@ -95,8 +95,11 @@ fn map_arch_intrinsic(full_name: &str) -> &'static str {
                     "cubema" => "__builtin_amdgcn_cubema",
                     "cubesc" => "__builtin_amdgcn_cubesc",
                     "cubetc" => "__builtin_amdgcn_cubetc",
+                    "cvt.f16.bf8" => "__builtin_amdgcn_cvt_f16_bf8",
+                    "cvt.f16.fp8" => "__builtin_amdgcn_cvt_f16_fp8",
                     "cvt.f32.bf8" => "__builtin_amdgcn_cvt_f32_bf8",
                     "cvt.f32.fp8" => "__builtin_amdgcn_cvt_f32_fp8",
+                    "cvt.f32.fp8.e5m3" => "__builtin_amdgcn_cvt_f32_fp8_e5m3",
                     "cvt.off.f32.i4" => "__builtin_amdgcn_cvt_off_f32_i4",
                     "cvt.pk.bf8.f32" => "__builtin_amdgcn_cvt_pk_bf8_f32",
                     "cvt.pk.f16.bf8" => "__builtin_amdgcn_cvt_pk_f16_bf8",
@@ -181,6 +184,12 @@ fn map_arch_intrinsic(full_name: &str) -> &'static str {
                     "dot4.f32.fp8.bf8" => "__builtin_amdgcn_dot4_f32_fp8_bf8",
                     "dot4.f32.fp8.fp8" => "__builtin_amdgcn_dot4_f32_fp8_fp8",
                     "ds.add.gs.reg.rtn" => "__builtin_amdgcn_ds_add_gs_reg_rtn",
+                    "ds.atomic.async.barrier.arrive.b64" => {
+                        "__builtin_amdgcn_ds_atomic_async_barrier_arrive_b64"
+                    }
+                    "ds.atomic.barrier.arrive.rtn.b64" => {
+                        "__builtin_amdgcn_ds_atomic_barrier_arrive_rtn_b64"
+                    }
                     "ds.bpermute" => "__builtin_amdgcn_ds_bpermute",
                     "ds.bpermute.fi.b32" => "__builtin_amdgcn_ds_bpermute_fi_b32",
                     "ds.gws.barrier" => "__builtin_amdgcn_ds_gws_barrier",
@@ -198,8 +207,32 @@ fn map_arch_intrinsic(full_name: &str) -> &'static str {
                     "fdot2.f16.f16" => "__builtin_amdgcn_fdot2_f16_f16",
                     "fdot2.f32.bf16" => "__builtin_amdgcn_fdot2_f32_bf16",
                     "fdot2c.f32.bf16" => "__builtin_amdgcn_fdot2c_f32_bf16",
+                    "flat.prefetch" => "__builtin_amdgcn_flat_prefetch",
                     "fmul.legacy" => "__builtin_amdgcn_fmul_legacy",
+                    "global.load.async.to.lds.b128" => {
+                        "__builtin_amdgcn_global_load_async_to_lds_b128"
+                    }
+                    "global.load.async.to.lds.b32" => {
+                        "__builtin_amdgcn_global_load_async_to_lds_b32"
+                    }
+                    "global.load.async.to.lds.b64" => {
+                        "__builtin_amdgcn_global_load_async_to_lds_b64"
+                    }
+                    "global.load.async.to.lds.b8" => "__builtin_amdgcn_global_load_async_to_lds_b8",
                     "global.load.lds" => "__builtin_amdgcn_global_load_lds",
+                    "global.prefetch" => "__builtin_amdgcn_global_prefetch",
+                    "global.store.async.from.lds.b128" => {
+                        "__builtin_amdgcn_global_store_async_from_lds_b128"
+                    }
+                    "global.store.async.from.lds.b32" => {
+                        "__builtin_amdgcn_global_store_async_from_lds_b32"
+                    }
+                    "global.store.async.from.lds.b64" => {
+                        "__builtin_amdgcn_global_store_async_from_lds_b64"
+                    }
+                    "global.store.async.from.lds.b8" => {
+                        "__builtin_amdgcn_global_store_async_from_lds_b8"
+                    }
                     "groupstaticsize" => "__builtin_amdgcn_groupstaticsize",
                     "iglp.opt" => "__builtin_amdgcn_iglp_opt",
                     "implicit.buffer.ptr" => "__builtin_amdgcn_implicit_buffer_ptr",
@@ -291,6 +324,7 @@ fn map_arch_intrinsic(full_name: &str) -> &'static str {
                     "s.incperflevel" => "__builtin_amdgcn_s_incperflevel",
                     "s.memrealtime" => "__builtin_amdgcn_s_memrealtime",
                     "s.memtime" => "__builtin_amdgcn_s_memtime",
+                    "s.monitor.sleep" => "__builtin_amdgcn_s_monitor_sleep",
                     "s.sendmsg" => "__builtin_amdgcn_s_sendmsg",
                     "s.sendmsghalt" => "__builtin_amdgcn_s_sendmsghalt",
                     "s.setprio" => "__builtin_amdgcn_s_setprio",
@@ -300,11 +334,15 @@ fn map_arch_intrinsic(full_name: &str) -> &'static str {
                     "s.sleep.var" => "__builtin_amdgcn_s_sleep_var",
                     "s.ttracedata" => "__builtin_amdgcn_s_ttracedata",
                     "s.ttracedata.imm" => "__builtin_amdgcn_s_ttracedata_imm",
+                    "s.wait.asynccnt" => "__builtin_amdgcn_s_wait_asynccnt",
                     "s.wait.event.export.ready" => "__builtin_amdgcn_s_wait_event_export_ready",
+                    "s.wait.tensorcnt" => "__builtin_amdgcn_s_wait_tensorcnt",
                     "s.waitcnt" => "__builtin_amdgcn_s_waitcnt",
                     "sad.hi.u8" => "__builtin_amdgcn_sad_hi_u8",
                     "sad.u16" => "__builtin_amdgcn_sad_u16",
                     "sad.u8" => "__builtin_amdgcn_sad_u8",
+                    "sat.pk4.i4.i8" => "__builtin_amdgcn_sat_pk4_i4_i8",
+                    "sat.pk4.u4.u8" => "__builtin_amdgcn_sat_pk4_u4_u8",
                     "sched.barrier" => "__builtin_amdgcn_sched_barrier",
                     "sched.group.barrier" => "__builtin_amdgcn_sched_group_barrier",
                     "sdot2" => "__builtin_amdgcn_sdot2",
@@ -346,8 +384,13 @@ fn map_arch_intrinsic(full_name: &str) -> &'static str {
                     "smfmac.i32.16x16x64.i8" => "__builtin_amdgcn_smfmac_i32_16x16x64_i8",
                     "smfmac.i32.32x32x32.i8" => "__builtin_amdgcn_smfmac_i32_32x32x32_i8",
                     "smfmac.i32.32x32x64.i8" => "__builtin_amdgcn_smfmac_i32_32x32x64_i8",
+                    "struct.ptr.buffer.load.lds" => "__builtin_amdgcn_struct_ptr_buffer_load_lds",
                     "sudot4" => "__builtin_amdgcn_sudot4",
                     "sudot8" => "__builtin_amdgcn_sudot8",
+                    "tensor.load.to.lds" => "__builtin_amdgcn_tensor_load_to_lds",
+                    "tensor.load.to.lds.d2" => "__builtin_amdgcn_tensor_load_to_lds_d2",
+                    "tensor.store.from.lds" => "__builtin_amdgcn_tensor_store_from_lds",
+                    "tensor.store.from.lds.d2" => "__builtin_amdgcn_tensor_store_from_lds_d2",
                     "udot2" => "__builtin_amdgcn_udot2",
                     "udot4" => "__builtin_amdgcn_udot4",
                     "udot8" => "__builtin_amdgcn_udot8",
@@ -6326,6 +6369,23 @@ fn map_arch_intrinsic(full_name: &str) -> &'static str {
             }
             s390(name, full_name)
         }
+        "spv" => {
+            #[allow(non_snake_case)]
+            fn spv(name: &str, full_name: &str) -> &'static str {
+                match name {
+                    // spv
+                    "num.subgroups" => "__builtin_spirv_num_subgroups",
+                    "subgroup.id" => "__builtin_spirv_subgroup_id",
+                    "subgroup.local.invocation.id" => {
+                        "__builtin_spirv_subgroup_local_invocation_id"
+                    }
+                    "subgroup.max.size" => "__builtin_spirv_subgroup_max_size",
+                    "subgroup.size" => "__builtin_spirv_subgroup_size",
+                    _ => unimplemented!("***** unsupported LLVM intrinsic {full_name}"),
+                }
+            }
+            spv(name, full_name)
+        }
         "ve" => {
             #[allow(non_snake_case)]
             fn ve(name: &str, full_name: &str) -> &'static str {

compiler/rustc_codegen_gcc/src/intrinsic/mod.rs

@@ -925,10 +925,17 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
         // TODO(antoyo): use width?
         let arg_type = arg.get_type();
         let result_type = self.u32_type;
+        let arg = if arg_type.is_signed(self.cx) {
+            let new_type = arg_type.to_unsigned(self.cx);
+            self.gcc_int_cast(arg, new_type)
+        } else {
+            arg
+        };
+        let arg_type = arg.get_type();
         let count_leading_zeroes =
             // TODO(antoyo): write a new function Type::is_compatible_with(&Type) and use it here
             // instead of using is_uint().
-            if arg_type.is_uint(self.cx) {
+            if arg_type.is_uchar(self.cx) || arg_type.is_ushort(self.cx) || arg_type.is_uint(self.cx) {
                 "__builtin_clz"
             }
             else if arg_type.is_ulong(self.cx) {

compiler/rustc_codegen_gcc/src/intrinsic/simd.rs

@@ -206,6 +206,28 @@ pub fn generic_simd_intrinsic<'a, 'gcc, 'tcx>(
         );
     }
 
+    #[cfg(feature = "master")]
+    if name == sym::simd_funnel_shl {
+        return Ok(simd_funnel_shift(
+            bx,
+            args[0].immediate(),
+            args[1].immediate(),
+            args[2].immediate(),
+            true,
+        ));
+    }
+
+    #[cfg(feature = "master")]
+    if name == sym::simd_funnel_shr {
+        return Ok(simd_funnel_shift(
+            bx,
+            args[0].immediate(),
+            args[1].immediate(),
+            args[2].immediate(),
+            false,
+        ));
+    }
+
     if name == sym::simd_bswap {
         return Ok(simd_bswap(bx, args[0].immediate()));
     }
@@ -1434,3 +1456,62 @@ pub fn generic_simd_intrinsic<'a, 'gcc, 'tcx>(
 
     unimplemented!("simd {}", name);
 }
+
+#[cfg(feature = "master")]
+fn simd_funnel_shift<'a, 'gcc, 'tcx>(
+    bx: &mut Builder<'a, 'gcc, 'tcx>,
+    a: RValue<'gcc>,
+    b: RValue<'gcc>,
+    shift: RValue<'gcc>,
+    shift_left: bool,
+) -> RValue<'gcc> {
+    use crate::common::SignType;
+
+    let a_type = a.get_type();
+    let vector_type = a_type.unqualified().dyncast_vector().expect("vector type");
+    let num_units = vector_type.get_num_units();
+    let elem_type = vector_type.get_element_type();
+
+    let (new_int_type, int_shift_val, int_mask) = if elem_type.is_compatible_with(bx.u8_type)
+        || elem_type.is_compatible_with(bx.i8_type)
+    {
+        (bx.u16_type, 8, u8::MAX as u64)
+    } else if elem_type.is_compatible_with(bx.u16_type) || elem_type.is_compatible_with(bx.i16_type)
+    {
+        (bx.u32_type, 16, u16::MAX as u64)
+    } else if elem_type.is_compatible_with(bx.u32_type) || elem_type.is_compatible_with(bx.i32_type)
+    {
+        (bx.u64_type, 32, u32::MAX as u64)
+    } else if elem_type.is_compatible_with(bx.u64_type) || elem_type.is_compatible_with(bx.i64_type)
+    {
+        (bx.u128_type, 64, u64::MAX)
+    } else {
+        unimplemented!("funnel shift on {:?}", elem_type);
+    };
+
+    let int_mask = bx.context.new_rvalue_from_long(new_int_type, int_mask as i64);
+    let int_shift_val = bx.context.new_rvalue_from_int(new_int_type, int_shift_val);
+    let mut elements = vec![];
+    let unsigned_type = elem_type.to_unsigned(bx);
+    for i in 0..num_units {
+        let index = bx.context.new_rvalue_from_int(bx.int_type, i as i32);
+        let a_val = bx.context.new_vector_access(None, a, index).to_rvalue();
+        let a_val = bx.context.new_bitcast(None, a_val, unsigned_type);
+        // TODO: we probably need to use gcc_int_cast instead.
+        let a_val = bx.gcc_int_cast(a_val, new_int_type);
+        let b_val = bx.context.new_vector_access(None, b, index).to_rvalue();
+        let b_val = bx.context.new_bitcast(None, b_val, unsigned_type);
+        let b_val = bx.gcc_int_cast(b_val, new_int_type);
+        let shift_val = bx.context.new_vector_access(None, shift, index).to_rvalue();
+        let shift_val = bx.gcc_int_cast(shift_val, new_int_type);
+        let mut val = a_val << int_shift_val | b_val;
+        if shift_left {
+            val = (val << shift_val) >> int_shift_val;
+        } else {
+            val = (val >> shift_val) & int_mask;
+        }
+        let val = bx.gcc_int_cast(val, elem_type);
+        elements.push(val);
+    }
+    bx.context.new_rvalue_from_vector(None, a_type, &elements)
+}

compiler/rustc_codegen_gcc/src/lib.rs

@@ -50,6 +50,7 @@ extern crate rustc_session;
 extern crate rustc_span;
 extern crate rustc_symbol_mangling;
 extern crate rustc_target;
+extern crate rustc_type_ir;
 
 // This prevents duplicating functions and statics that are already part of the host rustc process.
 #[allow(unused_extern_crates)]
@@ -362,9 +363,9 @@ impl WriteBackendMethods for GccCodegenBackend {
         _exported_symbols_for_lto: &[String],
         each_linked_rlib_for_lto: &[PathBuf],
         modules: Vec<FatLtoInput<Self>>,
-        diff_fncs: Vec<AutoDiffItem>,
+        diff_functions: Vec<AutoDiffItem>,
     ) -> Result<ModuleCodegen<Self::Module>, FatalError> {
-        if !diff_fncs.is_empty() {
+        if !diff_functions.is_empty() {
             unimplemented!();
         }
 

compiler/rustc_codegen_gcc/tests/failing-lto-tests.txt

@@ -28,6 +28,6 @@ tests/ui/macros/macro-comma-behavior-rpass.rs
 tests/ui/macros/rfc-2011-nicer-assert-messages/assert-with-custom-errors-does-not-create-unnecessary-code.rs
 tests/ui/macros/rfc-2011-nicer-assert-messages/feature-gate-generic_assert.rs
 tests/ui/macros/stringify.rs
-tests/ui/reexport-test-harness-main.rs
 tests/ui/rfcs/rfc-1937-termination-trait/termination-trait-in-test.rs
 tests/ui/binding/fn-arg-incomplete-pattern-drop-order.rs
+tests/ui/lto/debuginfo-lto-alloc.rs

compiler/rustc_codegen_gcc/tests/failing-run-make-tests.txt

@@ -6,7 +6,6 @@ tests/run-make/doctests-keep-binaries/
 tests/run-make/doctests-runtool/
 tests/run-make/emit-shared-files/
 tests/run-make/exit-code/
-tests/run-make/issue-22131/
 tests/run-make/issue-64153/
 tests/run-make/llvm-ident/
 tests/run-make/native-link-modifier-bundle/

compiler/rustc_codegen_gcc/tests/failing-ui-tests.txt

@@ -10,11 +10,10 @@ tests/ui/iterators/iter-sum-overflow-overflow-checks.rs
 tests/ui/mir/mir_drop_order.rs
 tests/ui/mir/mir_let_chains_drop_order.rs
 tests/ui/mir/mir_match_guard_let_chains_drop_order.rs
-tests/ui/oom_unwind.rs
+tests/ui/panics/oom-panic-unwind.rs
 tests/ui/panic-runtime/abort-link-to-unwinding-crates.rs
 tests/ui/panic-runtime/abort.rs
 tests/ui/panic-runtime/link-to-abort.rs
-tests/ui/unwind-no-uwtable.rs
 tests/ui/parser/unclosed-delimiter-in-dep.rs
 tests/ui/consts/missing_span_in_backtrace.rs
 tests/ui/drop/dynamic-drop.rs
@@ -82,3 +81,8 @@ tests/ui/coroutine/panic-drops.rs
 tests/ui/coroutine/panic-safe.rs
 tests/ui/process/nofile-limit.rs
 tests/ui/simd/intrinsic/generic-arithmetic-pass.rs
+tests/ui/linking/no-gc-encapsulation-symbols.rs
+tests/ui/panics/unwind-force-no-unwind-tables.rs
+tests/ui/attributes/fn-align-dyn.rs
+tests/ui/linkage-attr/raw-dylib/elf/glibc-x86_64.rs
+tests/ui/explicit-tail-calls/recursion-etc.rs

compiler/rustc_codegen_gcc/tools/cspell_dicts/rustc_codegen_gcc.txt

@@ -8,6 +8,7 @@ clzll
 cmse
 codegened
 csky
+ctfe
 ctlz
 ctpop
 cttz
@@ -25,6 +26,7 @@ fwrapv
 gimple
 hrtb
 immediates
+interner
 liblto
 llbb
 llcx
@@ -47,6 +49,7 @@ mavx
 mcmodel
 minimumf
 minnumf
+miri
 monomorphization
 monomorphizations
 monomorphized