Auto merge of #45032 - alexcrichton:target-cfu, r=michaelwoerister

rustc: Allow target-specific default cgus

Some targets, like msp430 and nvptx, don't work with multiple codegen units
right now for bugs or fundamental reasons. To expose this allow targets to
express a default.

Closes #45000

Author: bors

src/librustc/session/config.rs

@@ -352,7 +352,7 @@ top_level_options!(
         actually_rustdoc: bool [TRACKED],
 
         // Number of object files/codegen units to produce on the backend
-        codegen_units: usize [UNTRACKED],
+        cli_forced_codegen_units: Option<usize> [UNTRACKED],
     }
 );
 
@@ -505,7 +505,7 @@ pub fn basic_options() -> Options {
         unstable_features: UnstableFeatures::Disallow,
         debug_assertions: true,
         actually_rustdoc: false,
-        codegen_units: 1,
+        cli_forced_codegen_units: None,
     }
 }
 
@@ -1711,48 +1711,6 @@ pub fn build_session_options_and_crate_config(matches: &getopts::Matches)
 
     let incremental = debugging_opts.incremental.as_ref().map(|m| PathBuf::from(m));
 
-    let codegen_units = codegen_units.unwrap_or_else(|| {
-        match opt_level {
-            // If we're compiling at `-O0` then default to 16 codegen units.
-            // The number here shouldn't matter too too much as debug mode
-            // builds don't rely on performance at all, meaning that lost
-            // opportunities for inlining through multiple codegen units is
-            // a non-issue.
-            //
-            // Note that the high number here doesn't mean that we'll be
-            // spawning a large number of threads in parallel. The backend
-            // of rustc contains global rate limiting through the
-            // `jobserver` crate so we'll never overload the system with too
-            // much work, but rather we'll only be optimizing when we're
-            // otherwise cooperating with other instances of rustc.
-            //
-            // Rather the high number here means that we should be able to
-            // keep a lot of idle cpus busy. By ensuring that no codegen
-            // unit takes *too* long to build we'll be guaranteed that all
-            // cpus will finish pretty closely to one another and we should
-            // make relatively optimal use of system resources
-            //
-            // Another note worth mentioning here, however, is that this number
-            // isn't *too* high. When codegen units are increased that means we
-            // currently have to codegen `#[inline]` functions into each codegen
-            // unit, which means the more codegen units we're using the more we
-            // may be generating. In other words, increasing codegen units may
-            // increase the overall work the compiler does. If we don't have
-            // enough cores to make up for this loss then increasing the number
-            // of codegen units could become an overall loss!
-            //
-            // As a result we choose a hopefully conservative value 16, which
-            // should be more than the number of cpus of most hardware compiling
-            // Rust but also not too much for 2-4 core machines to have too much
-            // loss of compile time.
-            OptLevel::No => 16,
-
-            // All other optimization levels default use one codegen unit,
-            // the historical default in Rust for a Long Time.
-            _ => 1,
-        }
-    });
-
     (Options {
         crate_types,
         optimize: opt_level,
@@ -1777,7 +1735,7 @@ pub fn build_session_options_and_crate_config(matches: &getopts::Matches)
         unstable_features: UnstableFeatures::from_environment(),
         debug_assertions,
         actually_rustdoc: false,
-        codegen_units,
+        cli_forced_codegen_units: codegen_units,
     },
     cfg)
 }

src/librustc/session/mod.rs

@@ -636,6 +636,43 @@ impl Session {
         }
         ret
     }
+
+    /// Returns the number of codegen units that should be used for this
+    /// compilation
+    pub fn codegen_units(&self) -> usize {
+        if let Some(n) = self.opts.cli_forced_codegen_units {
+            return n
+        }
+        if let Some(n) = self.target.target.options.default_codegen_units {
+            return n as usize
+        }
+
+        match self.opts.optimize {
+            // If we're compiling at `-O0` then default to 16 codegen units.
+            // The number here shouldn't matter too too much as debug mode
+            // builds don't rely on performance at all, meaning that lost
+            // opportunities for inlining through multiple codegen units is
+            // a non-issue.
+            //
+            // Note that the high number here doesn't mean that we'll be
+            // spawning a large number of threads in parallel. The backend
+            // of rustc contains global rate limiting through the
+            // `jobserver` crate so we'll never overload the system with too
+            // much work, but rather we'll only be optimizing when we're
+            // otherwise cooperating with other instances of rustc.
+            //
+            // Rather the high number here means that we should be able to
+            // keep a lot of idle cpus busy. By ensuring that no codegen
+            // unit takes *too* long to build we'll be guaranteed that all
+            // cpus will finish pretty closely to one another and we should
+            // make relatively optimal use of system resources
+            config::OptLevel::No => 16,
+
+            // All other optimization levels default use one codegen unit,
+            // the historical default in Rust for a Long Time.
+            _ => 1,
+        }
+    }
 }
 
 pub fn build_session(sopts: config::Options,

src/librustc_back/target/mod.rs

@@ -430,6 +430,9 @@ pub struct TargetOptions {
 
     /// The minimum alignment for global symbols.
     pub min_global_align: Option<u64>,
+
+    /// Default number of codegen units to use in debug mode
+    pub default_codegen_units: Option<u64>,
 }
 
 impl Default for TargetOptions {
@@ -492,6 +495,7 @@ impl Default for TargetOptions {
             crt_static_respected: false,
             stack_probes: false,
             min_global_align: None,
+            default_codegen_units: None,
         }
     }
 }
@@ -732,6 +736,7 @@ impl Target {
         key!(crt_static_respected, bool);
         key!(stack_probes, bool);
         key!(min_global_align, Option<u64>);
+        key!(default_codegen_units, Option<u64>);
 
         if let Some(array) = obj.find("abi-blacklist").and_then(Json::as_array) {
             for name in array.iter().filter_map(|abi| abi.as_string()) {
@@ -924,6 +929,7 @@ impl ToJson for Target {
         target_option_val!(crt_static_respected);
         target_option_val!(stack_probes);
         target_option_val!(min_global_align);
+        target_option_val!(default_codegen_units);
 
         if default.abi_blacklist != self.options.abi_blacklist {
             d.insert("abi-blacklist".to_string(), self.options.abi_blacklist.iter()

src/librustc_back/target/msp430_none_elf.rs

@@ -48,6 +48,11 @@ pub fn target() -> TargetResult {
             // code because of the extra costs it involves.
             relocation_model: "static".to_string(),
 
+            // Right now we invoke an external assembler and this isn't
+            // compatible with multiple codegen units, and plus we probably
+            // don't want to invoke that many gcc instances.
+            default_codegen_units: Some(1),
+
             .. Default::default( )
         }
     })

src/librustc_trans/back/link.rs

@@ -467,7 +467,7 @@ fn link_rlib<'a>(sess: &'a Session,
                 // of when we do and don't keep .#module-name#.bc files around.
                 let user_wants_numbered_bitcode =
                         sess.opts.output_types.contains_key(&OutputType::Bitcode) &&
-                        sess.opts.codegen_units > 1;
+                        sess.codegen_units() > 1;
                 if !sess.opts.cg.save_temps && !user_wants_numbered_bitcode {
                     remove(sess, &bc_filename);
                 }

src/librustc_trans/back/write.rs

@@ -1037,10 +1037,10 @@ fn produce_final_output_artifacts(sess: &Session,
         let needs_crate_object = crate_output.outputs.contains_key(&OutputType::Exe);
 
         let keep_numbered_bitcode = needs_crate_bitcode ||
-                (user_wants_bitcode && sess.opts.codegen_units > 1);
+                (user_wants_bitcode && sess.codegen_units() > 1);
 
         let keep_numbered_objects = needs_crate_object ||
-                (user_wants_objects && sess.opts.codegen_units > 1);
+                (user_wants_objects && sess.codegen_units() > 1);
 
         for module in compiled_modules.modules.iter() {
             let module_name = Some(&module.name[..]);
@@ -2052,7 +2052,7 @@ impl OngoingCrateTranslation {
 
         // FIXME: time_llvm_passes support - does this use a global context or
         // something?
-        if sess.opts.codegen_units == 1 && sess.time_llvm_passes() {
+        if sess.codegen_units() == 1 && sess.time_llvm_passes() {
             unsafe { llvm::LLVMRustPrintPassTimings(); }
         }
 

src/librustc_trans/base.rs

@@ -1218,7 +1218,7 @@ fn collect_and_partition_translation_items<'a, 'tcx>(
     let strategy = if tcx.sess.opts.debugging_opts.incremental.is_some() {
         PartitioningStrategy::PerModule
     } else {
-        PartitioningStrategy::FixedUnitCount(tcx.sess.opts.codegen_units)
+        PartitioningStrategy::FixedUnitCount(tcx.sess.codegen_units())
     };
 
     let codegen_units = time(time_passes, "codegen unit partitioning", || {