From 955cdd5f83bd694b1c87d7c69e25cbbee06d675e Mon Sep 17 00:00:00 2001
From: Julian Seward <jseward@acm.org>
Date: Sat, 7 Sep 2019 17:34:08 +0200
Subject: [PATCH] VirtRegs::find: use SmallVec instead of Vec for val_stack.

Pushing on the `val_stack` vector is CL's biggest source of calls to
malloc/realloc/free, by some margin.  It accounts for about 27.7% of all heap
blocks allocated when compiling wasm_lua_binarytrees.  This change removes
pretty much all dynamic allocation by changing to a SmallVec<[Value; 8]>
instead.  A fixed size of 4 gets all the gains to be had, in testing, so 8
gives some safety margin and is harmless from a stack-use perspective: 8
Values will occupy 32 bytes.

As a bonus, this change also reduces the compiler's dynamic instruction count
by about 0.5%.
---
 cranelift/codegen/Cargo.toml               | 1 +
 cranelift/codegen/src/regalloc/virtregs.rs | 3 ++-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/cranelift/codegen/Cargo.toml b/cranelift/codegen/Cargo.toml
index 2c5ccb4d2e..2d81a660e8 100644
--- a/cranelift/codegen/Cargo.toml
+++ b/cranelift/codegen/Cargo.toml
@@ -21,6 +21,7 @@ hashmap_core = { version = "0.1.9", optional = true }
 target-lexicon = "0.8.1"
 log = { version = "0.4.6", default-features = false }
 serde = { version = "1.0.94", features = ["derive"], optional = true }
+smallvec = { version = "0.6.10" }
 # It is a goal of the cranelift-codegen crate to have minimal external dependencies.
 # Please don't add any unless they are essential to the task of creating binary
 # machine code. Integration tests that need external dependencies can be
diff --git a/cranelift/codegen/src/regalloc/virtregs.rs b/cranelift/codegen/src/regalloc/virtregs.rs
index 584ad9c53a..fc267c3f61 100644
--- a/cranelift/codegen/src/regalloc/virtregs.rs
+++ b/cranelift/codegen/src/regalloc/virtregs.rs
@@ -21,6 +21,7 @@ use crate::packed_option::PackedOption;
 use crate::ref_slice::ref_slice;
 use core::cmp::Ordering;
 use core::fmt;
+use smallvec::SmallVec;
 use std::vec::Vec;
 
 /// A virtual register reference.
@@ -292,7 +293,7 @@ impl VirtRegs {
     /// Find the leader value and rank of the set containing `v`.
     /// Compress the path if needed.
     fn find(&mut self, mut val: Value) -> (Value, u32) {
-        let mut val_stack = vec![];
+        let mut val_stack = SmallVec::<[Value; 8]>::new();
         let found = loop {
             match UFEntry::decode(self.union_find[val]) {
                 UFEntry::Rank(rank) => break (val, rank),