Auto merge of rust-lang#128265 - DianQK:instsimplify-before-inline, r=saethlin

 Perform instsimplify before inline to eliminate some trivial calls

I am currently working on rust-lang#128081. In the current pipeline, we can get the following clone statements ([godbolt](https://rust.godbolt.org/z/931316fhP)):

```
    bb0: {
        StorageLive(_2);
        _2 = ((*_1).0: i32);
        StorageLive(_3);
        _3 = ((*_1).1: u64);
        _0 = Foo { a: move _2, b: move _3 };
        StorageDead(_3);
        StorageDead(_2);
        return;
    }
```

Analyzing such statements will be simple and fast. We don't need to consider branches or some interfering statements. However, this requires us to run `InstSimplify`, `ReferencePropagation`, and `SimplifyCFG` at least once. I can introduce a new pass, but I think the best place for it would be within `InstSimplify`.

I put `InstSimplify` before `Inline`, which takes some of the burden away from `Inline`.

r? `@saethlin`

Author: bors

compiler/rustc_mir_transform/src/instsimplify.rs

@@ -13,9 +13,25 @@ use rustc_target::spec::abi::Abi;
 use crate::simplify::simplify_duplicate_switch_targets;
 use crate::take_array;
 
-pub struct InstSimplify;
+pub enum InstSimplify {
+    BeforeInline,
+    AfterSimplifyCfg,
+}
+
+impl InstSimplify {
+    pub fn name(&self) -> &'static str {
+        match self {
+            InstSimplify::BeforeInline => "InstSimplify-before-inline",
+            InstSimplify::AfterSimplifyCfg => "InstSimplify-after-simplifycfg",
+        }
+    }
+}
 
 impl<'tcx> MirPass<'tcx> for InstSimplify {
+    fn name(&self) -> &'static str {
+        self.name()
+    }
+
     fn is_enabled(&self, sess: &rustc_session::Session) -> bool {
         sess.mir_opt_level() > 0
     }

compiler/rustc_mir_transform/src/lib.rs

@@ -571,6 +571,8 @@ fn run_optimization_passes<'tcx>(tcx: TyCtxt<'tcx>, body: &mut Body<'tcx>) {
             // Has to be done before inlining, otherwise actual call will be almost always inlined.
             // Also simple, so can just do first
             &lower_slice_len::LowerSliceLenCalls,
+            // Perform instsimplify before inline to eliminate some trivial calls (like clone shims).
+            &instsimplify::InstSimplify::BeforeInline,
             // Perform inlining, which may add a lot of code.
             &inline::Inline,
             // Code from other crates may have storage markers, so this needs to happen after inlining.
@@ -590,7 +592,8 @@ fn run_optimization_passes<'tcx>(tcx: TyCtxt<'tcx>, body: &mut Body<'tcx>) {
             &match_branches::MatchBranchSimplification,
             // inst combine is after MatchBranchSimplification to clean up Ne(_1, false)
             &multiple_return_terminators::MultipleReturnTerminators,
-            &instsimplify::InstSimplify,
+            // After simplifycfg, it allows us to discover new opportunities for peephole optimizations.
+            &instsimplify::InstSimplify::AfterSimplifyCfg,
             &simplify::SimplifyLocals::BeforeConstProp,
             &dead_store_elimination::DeadStoreElimination::Initial,
             &gvn::GVN,

compiler/rustc_mir_transform/src/shim.rs

@@ -155,7 +155,7 @@ fn make_shim<'tcx>(tcx: TyCtxt<'tcx>, instance: ty::InstanceKind<'tcx>) -> Body<
             &deref_separator::Derefer,
             &remove_noop_landing_pads::RemoveNoopLandingPads,
             &simplify::SimplifyCfg::MakeShim,
-            &instsimplify::InstSimplify,
+            &instsimplify::InstSimplify::BeforeInline,
             &abort_unwinding_calls::AbortUnwindingCalls,
             &add_call_guards::CriticalCallEdges,
         ],

tests/incremental/hashes/call_expressions.rs

@@ -162,7 +162,7 @@ pub fn change_to_ufcs() {
 }
 
 #[cfg(not(any(cfail1,cfail4)))]
-#[rustc_clean(cfg="cfail2", except="opt_hir_owner_nodes,typeck")]
+#[rustc_clean(cfg="cfail2", except="opt_hir_owner_nodes,optimized_mir,typeck")]
 #[rustc_clean(cfg="cfail3")]
 #[rustc_clean(cfg="cfail5", except="opt_hir_owner_nodes,optimized_mir,typeck")]
 #[rustc_clean(cfg="cfail6")]

tests/mir-opt/const_prop/slice_len.rs

@@ -1,5 +1,5 @@
 //@ test-mir-pass: GVN
-//@ compile-flags: -Zmir-enable-passes=+InstSimplify -Zdump-mir-exclude-alloc-bytes
+//@ compile-flags: -Zmir-enable-passes=+InstSimplify-after-simplifycfg -Zdump-mir-exclude-alloc-bytes
 // EMIT_MIR_FOR_EACH_PANIC_STRATEGY
 // EMIT_MIR_FOR_EACH_BIT_WIDTH
 

tests/mir-opt/dataflow-const-prop/slice_len.rs

@@ -1,6 +1,6 @@
 // EMIT_MIR_FOR_EACH_PANIC_STRATEGY
 //@ test-mir-pass: DataflowConstProp
-//@ compile-flags: -Zmir-enable-passes=+InstSimplify
+//@ compile-flags: -Zmir-enable-passes=+InstSimplify-after-simplifycfg
 // EMIT_MIR_FOR_EACH_BIT_WIDTH
 
 // EMIT_MIR slice_len.main.DataflowConstProp.diff

tests/mir-opt/inline/dyn_trait.get_query.Inline.panic-abort.diff

@@ -21,14 +21,14 @@
       bb0: {
           StorageLive(_2);
           StorageLive(_3);
-          _3 = &(*_1);
+          _3 = _1;
           _2 = <Q as Query>::cache::<T>(move _3) -> [return: bb1, unwind unreachable];
       }
 
       bb1: {
           StorageDead(_3);
           StorageLive(_4);
-          _4 = &(*_2);
+          _4 = _2;
 -         _0 = try_execute_query::<<Q as Query>::C>(move _4) -> [return: bb2, unwind unreachable];
 +         StorageLive(_5);
 +         _5 = _4 as &dyn Cache<V = <Q as Query>::V> (PointerCoercion(Unsize));

tests/mir-opt/inline/dyn_trait.get_query.Inline.panic-unwind.diff

@@ -21,14 +21,14 @@
       bb0: {
           StorageLive(_2);
           StorageLive(_3);
-          _3 = &(*_1);
+          _3 = _1;
           _2 = <Q as Query>::cache::<T>(move _3) -> [return: bb1, unwind continue];
       }
 
       bb1: {
           StorageDead(_3);
           StorageLive(_4);
-          _4 = &(*_2);
+          _4 = _2;
 -         _0 = try_execute_query::<<Q as Query>::C>(move _4) -> [return: bb2, unwind continue];
 +         StorageLive(_5);
 +         _5 = _4 as &dyn Cache<V = <Q as Query>::V> (PointerCoercion(Unsize));

tests/mir-opt/inline/dyn_trait.mk_cycle.Inline.panic-abort.diff

@@ -8,7 +8,7 @@
 
       bb0: {
           StorageLive(_2);
-          _2 = &(*_1);
+          _2 = _1;
           _0 = <dyn Cache<V = V> as Cache>::store_nocache(move _2) -> [return: bb1, unwind unreachable];
       }
 

tests/mir-opt/inline/dyn_trait.mk_cycle.Inline.panic-unwind.diff

@@ -8,7 +8,7 @@
 
       bb0: {
           StorageLive(_2);
-          _2 = &(*_1);
+          _2 = _1;
           _0 = <dyn Cache<V = V> as Cache>::store_nocache(move _2) -> [return: bb1, unwind continue];
       }