machinst: Avoid a full instructions traversal of all the blocks when computing the final block ordering;

cranelift/codegen/src/machinst/blockorder.rs

@@ -1,7 +1,6 @@
 //! Computation of basic block order in emitted code.
 
 use crate::machinst::*;
-use regalloc::{BlockIx, Function};
 
 /// Simple reverse postorder-based block order emission.
 ///
@@ -10,15 +9,13 @@ use regalloc::{BlockIx, Function};
 struct BlockRPO {
     visited: Vec<bool>,
     postorder: Vec<BlockIndex>,
-    deferred_last: Option<BlockIndex>,
 }
 
 impl BlockRPO {
     fn new<I: VCodeInst>(vcode: &VCode<I>) -> BlockRPO {
         BlockRPO {
             visited: vec![false; vcode.num_blocks()],
-            postorder: vec![],
-            deferred_last: None,
+            postorder: Vec::with_capacity(vcode.num_blocks()),
         }
     }
 
@@ -29,22 +26,15 @@ impl BlockRPO {
                 self.visit(vcode, *succ);
             }
         }
-
-        for i in vcode.block_insns(BlockIx::new(block)) {
-            if vcode.get_insn(i).is_epilogue_placeholder() {
-                debug_assert!(self.deferred_last.is_none());
-                self.deferred_last = Some(block);
-                return;
-            }
-        }
-
+        if Some(block) != vcode.fallthrough_return_block {
             self.postorder.push(block);
         }
+    }
 
-    fn rpo(self) -> Vec<BlockIndex> {
+    fn rpo<I: VCodeInst>(self, vcode: &VCode<I>) -> Vec<BlockIndex> {
         let mut rpo = self.postorder;
         rpo.reverse();
-        if let Some(block) = self.deferred_last {
+        if let Some(block) = vcode.fallthrough_return_block {
             rpo.push(block);
         }
         rpo
@@ -55,5 +45,5 @@ impl BlockRPO {
 pub fn compute_final_block_order<I: VCodeInst>(vcode: &VCode<I>) -> Vec<BlockIndex> {
     let mut rpo = BlockRPO::new(vcode);
     rpo.visit(vcode, vcode.entry());
-    rpo.rpo()
+    rpo.rpo(vcode)
 }

cranelift/codegen/src/machinst/lower.rs

@@ -332,6 +332,7 @@ impl<'func, I: VCodeInst> Lower<'func, I> {
                     GenerateReturn::Yes
                 } else {
                     debug_assert!(last_insn_opcode == Opcode::FallthroughReturn);
+                    self.vcode.set_fallthrough_return_block(*bb);
                     GenerateReturn::No
                 };
                 self.gen_retval_setup(gen_ret);

cranelift/codegen/src/machinst/vcode.rs

@@ -18,8 +18,7 @@
 //! backend pipeline.
 
 use crate::entity::SecondaryMap;
-use crate::ir;
-use crate::ir::SourceLoc;
+use crate::ir::{self, Block, SourceLoc};
 use crate::machinst::*;
 use crate::settings;
 
@@ -101,6 +100,9 @@ pub struct VCode<I: VCodeInst> {
 
     /// ABI object.
     abi: Box<dyn ABIBody<I = I>>,
+
+    /// The block targeted by fallthrough_returns, if there's one.
+    pub fallthrough_return_block: Option<BlockIndex>,
 }
 
 /// A builder for a VCode function body. This builder is designed for the
@@ -153,6 +155,16 @@ impl<I: VCodeInst> VCodeBuilder<I> {
         &mut *self.vcode.abi
     }
 
+    /// Set the fallthrough_return target block for this function. There must be at most once per
+    /// function.
+    pub fn set_fallthrough_return_block(&mut self, bb: Block) {
+        debug_assert!(
+            self.vcode.fallthrough_return_block.is_none(),
+            "a function must have at most one fallthrough-return instruction"
+        );
+        self.vcode.fallthrough_return_block = Some(self.bb_to_bindex(bb));
+    }
+
     /// Set the type of a VReg.
     pub fn set_vreg_type(&mut self, vreg: VirtualReg, ty: Type) {
         while self.vcode.vreg_types.len() <= vreg.get_index() {
@@ -315,6 +327,7 @@ impl<I: VCodeInst> VCode<I> {
             final_block_offsets: vec![],
             code_size: 0,
             abi,
+            fallthrough_return_block: None,
         }
     }