From d4fcd32cdc54bae93c53b6646739eb9d650251b8 Mon Sep 17 00:00:00 2001
From: Alex Crichton <alex@alexcrichton.com>
Date: Mon, 24 Feb 2020 15:18:09 -0600
Subject: [PATCH] Optimize generated code via the CLI by default (#973)

* Optimize generated code via the CLI by default

This commit updates the behavior of the CLI and adds a new flag. It
first enables the `--optimize` flag by default, ensuring that usage of
the `wasmtime` CLI will enable cranelift optimizations by default. Next
it also adds a `--opt-level` flag which is similar to Rust's
`-Copt-level` where it takes a string argument of how to optimize. This
is updates to support 0/1/2/s, where 1 is currently the same as 2 but
added for consistency with other compilers. The default setting is
`--opt-level=2`.

When the `-O` flag is not passed the `--opt-level` flag is used,
otherwise `-O` takes precedent in the sense that it implies
`--opt-level=2` which is the highest optimization level. The thinking is
that these flags will in general select the highest optimization level
specified as the final optimization level.

* Add inline docs

* fix a test
---
 src/commands/wasm2obj.rs |  2 +-
 src/lib.rs               | 35 +++++++++++++++++++++++++++++++----
 src/obj.rs               | 13 ++++++++++---
 tests/debug/obj.rs       |  2 +-
 4 files changed, 43 insertions(+), 9 deletions(-)

diff --git a/src/commands/wasm2obj.rs b/src/commands/wasm2obj.rs
index 4d144d3d7d..494857b2a2 100644
--- a/src/commands/wasm2obj.rs
+++ b/src/commands/wasm2obj.rs
@@ -75,7 +75,7 @@ impl WasmToObjCommand {
             self.target.as_ref(),
             strategy,
             self.common.enable_simd,
-            self.common.optimize,
+            self.common.opt_level(),
             self.common.debug_info,
             self.output.clone(),
             &cache_config,
diff --git a/src/lib.rs b/src/lib.rs
index 1e64ecb98f..dc9b1bfca2 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -132,9 +132,18 @@ struct CommonOptions {
     #[structopt(long)]
     jitdump: bool,
 
-    /// Run optimization passes on translated functions
+    /// Run optimization passes on translated functions, on by default
     #[structopt(short = "O", long)]
     optimize: bool,
+
+    /// Optimization level for generated functions (0 (none), 1, 2 (most), or s
+    /// (size))
+    #[structopt(
+        long,
+        parse(try_from_str = parse_opt_level),
+        default_value = "2",
+    )]
+    opt_level: wasmtime::OptLevel,
 }
 
 impl CommonOptions {
@@ -148,11 +157,9 @@ impl CommonOptions {
             .wasm_reference_types(self.enable_reference_types || self.enable_all)
             .wasm_multi_value(self.enable_multi_value || self.enable_all)
             .wasm_threads(self.enable_threads || self.enable_all)
+            .cranelift_opt_level(self.opt_level())
             .strategy(pick_compilation_strategy(self.cranelift, self.lightbeam)?)?
             .profiler(pick_profiling_strategy(self.jitdump)?)?;
-        if self.optimize {
-            config.cranelift_opt_level(wasmtime::OptLevel::Speed);
-        }
         if !self.disable_cache {
             match &self.config {
                 Some(path) => {
@@ -165,4 +172,24 @@ impl CommonOptions {
         }
         Ok(config)
     }
+
+    fn opt_level(&self) -> wasmtime::OptLevel {
+        match (self.optimize, self.opt_level.clone()) {
+            (true, _) => wasmtime::OptLevel::Speed,
+            (false, other) => other,
+        }
+    }
+}
+
+fn parse_opt_level(opt_level: &str) -> Result<wasmtime::OptLevel> {
+    match opt_level {
+        "s" => Ok(wasmtime::OptLevel::SpeedAndSize),
+        "0" => Ok(wasmtime::OptLevel::None),
+        "1" => Ok(wasmtime::OptLevel::Speed),
+        "2" => Ok(wasmtime::OptLevel::Speed),
+        other => bail!(
+            "unknown optimization level `{}`, only 0,1,2,s accepted",
+            other
+        ),
+    }
 }
diff --git a/src/obj.rs b/src/obj.rs
index 20c466cfe8..87c8951b24 100644
--- a/src/obj.rs
+++ b/src/obj.rs
@@ -19,7 +19,7 @@ pub fn compile_to_obj(
     target: Option<&Triple>,
     strategy: Strategy,
     enable_simd: bool,
-    optimize: bool,
+    opt_level: wasmtime::OptLevel,
     debug_info: bool,
     artifact_name: String,
     cache_config: &CacheConfig,
@@ -38,8 +38,15 @@ pub fn compile_to_obj(
         flag_builder.enable("enable_simd").unwrap();
     }
 
-    if optimize {
-        flag_builder.set("opt_level", "speed").unwrap();
+    match opt_level {
+        wasmtime::OptLevel::None => {}
+        wasmtime::OptLevel::Speed => {
+            flag_builder.set("opt_level", "speed").unwrap();
+        }
+        wasmtime::OptLevel::SpeedAndSize => {
+            flag_builder.set("opt_level", "speed_and_size").unwrap();
+        }
+        other => bail!("unknown optimization level {:?}", other),
     }
 
     let isa = isa_builder.finish(settings::Flags::new(flag_builder));
diff --git a/tests/debug/obj.rs b/tests/debug/obj.rs
index f49cdf0def..a800373514 100644
--- a/tests/debug/obj.rs
+++ b/tests/debug/obj.rs
@@ -16,7 +16,7 @@ pub fn compile_cranelift(
         target.as_ref(),
         Strategy::Cranelift,
         false,
-        false,
+        wasmtime::OptLevel::None,
         true,
         output
             .as_ref()