Make region inference use a dirty list #47766

spastorino · 2018-01-26T05:17:25Z

nikomatsakis

Looks great. Left a small suggestion.

nikomatsakis · 2018-01-26T09:17:25Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

-                    },
-                );
+        let dependency_map = self.build_dependency_map();
+        let mut dirty_list = self.constraints.clone();


Nit: I think the dirty list ought to be either refs to constraints or indices of constraints. Since you used indices for the dependency-map, let's go with those. (It'll also avoid any borrowing issues.)

So something like:

let mut dirty_list: Vec<_> = (0..self.containts.len()).collect();

nikomatsakis · 2018-01-26T09:17:54Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

+        let mut dirty_list = self.constraints.clone();
+
+        debug!("propagate_constraints: --------------------");
+        while let Some(constraint) = dirty_list.pop() {


while let Some(constraint_index) = dirty_list.pop() { let constraint = &self.constraints[constraint_index]; ... }

nikomatsakis · 2018-01-26T09:21:20Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

+                debug!("propagate_constraints:   sup={:?}", constraint.sup);
+
+                if let Some(constraint_deps) = dependency_map.get(&constraint.sup) {
+                    for dep_idx in constraint_deps {


Now this can be

if let Some(constraint_deps) = dependency_map.get(&constraint.sup) { dirty_list.extend(constraint_deps); }

Not that it matters much, but generally invoking extend is better than for loop that calls push. This is because sometimes the iterator knows how long it will be, which lets dirty_list reallocate its backing array more efficiently.

Finally, you could simplify this further, like so:

dirty_list.extend( dependency_map.get(&constraint.sup).unwrap_or(&vec![]) );

but I'm not sure that's really clearer (arguably less so). The key here is that creating an empty vector (i.e., vec![]) is basically free -- it's just some null values, no allocation or anything. In the case where I have a "map to vec", I often do this style, under the idea that a missing entry is equivalent to an empty vector. But either way really.

spastorino · 2018-01-26T12:18:59Z

@nikomatsakis applied all the requested changes

leonardo-m · 2018-01-26T21:14:14Z

Have you run Clippy on your changes?

nikomatsakis · 2018-01-26T21:32:02Z

@leonardo-m why would we do that? is there some specific lint you are imagining?

fwiw, we don't usually run clippy on the compiler.

nikomatsakis

looking great! left 2 comments -- one is a nit, but also it's probably worth adding a bit vec to avoid duplicates as we discussed.

nikomatsakis · 2018-01-26T21:32:38Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

-                }
+        debug!("propagate_constraints: --------------------");
+        while let Some(constraint_idx) = dirty_list.pop() {
+            let constraint = self.constraints[constraint_idx];


this could be &self.constraints[constraint_idx] -- no reason to copy the value out

nikomatsakis · 2018-01-26T21:35:21Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

+                debug!("propagate_constraints:   sup={:?}", constraint.sup);
+
+                dirty_list.extend(
+                    dependency_map.get(&constraint.sup).unwrap_or(&vec![])


as you said, it might be nice to have a bitvec storing what is already present in the dirty_list, which you would then have to clear each time we pop something out of the dirty list.

spastorino · 2018-01-26T23:30:00Z

@nikomatsakis ready!

leonardo-m · 2018-01-26T23:43:41Z

we don't usually run clippy on the compiler.

I didn't know that Clippy is not used on the compiler. Clippy gives lot of false positives, but some advice is good, so perhaps you should use Clippy...

is there some specific lint you are imagining?

I think Clippy often suggests to replace or_insert with or_insert_with

nikomatsakis · 2018-01-27T00:01:15Z

I think Clippy often suggests to replace or_insert with or_insert_with

Ah. This case is better left as or_insert; or at least no worse, since creating an empty vector is just some constant bits.

nikomatsakis · 2018-01-27T00:52:14Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

-                    debug!("propagate_constraints:   sup={:?}", constraint.sup);
-                    changed = true;
+                for dep_idx in dependency_map.get(&constraint.sup).unwrap_or(&vec![]) {
+                    if !dirty_bit_vec.contains(*dep_idx) {


you can change this to:

if dirty_bit_vec.insert(*dep_idx) { dirty_list.push(*dep_idx); }

and save a bit of execution time.

nikomatsakis · 2018-01-27T00:52:28Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

-                    debug!("propagate_constraints:   sub={:?}", constraint.sub);
-                    debug!("propagate_constraints:   sup={:?}", constraint.sup);
-                    changed = true;
+                for dep_idx in dependency_map.get(&constraint.sup).unwrap_or(&vec![]) {


Nit: I'd usually write for &dep_idx in ... just so you don't need the *dep_idx below

nikomatsakis · 2018-01-27T00:52:55Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

            debug!("\n");
        }

        self.inferred_values = Some(inferred_values);
    }

+    fn build_dependency_map(&self) -> HashMap<RegionVid, Vec<usize>> {


Maybe a comment here would be nice, actually. You could steal the stuff I wrote in the issue description (i.e., the example). But something like this would suffice:

Builds up a map from each region variable X to a vector with the indices of constraints that need to be re-evaluated when X changes. These are constraints like Y: X @ P -- so if X changed, we may need to grow Y.

Fixes rust-lang#47602

I suspect the lines would be long for tidy.

nikomatsakis

Since I know you just stepped out, I might make these changes :)

nikomatsakis · 2018-01-27T01:01:06Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

-                );
+        let dependency_map = self.build_dependency_map();
+        let mut dirty_list: Vec<_> = (0..self.constraints.len()).collect();
+        let mut dirty_bit_vec = BitVector::new(dirty_list.len());


Nit: this starts as all zeroes, but we really want it to start as all ones I suppose. But it won't be wrong this way, just mildly less efficient. Actually, we could "invert the sense" of the vector, and have it store a 1 if the thing is not present....call it clean_bit_vec...

nikomatsakis · 2018-01-27T01:01:29Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

+
+        debug!("propagate_constraints: --------------------");
+        while let Some(constraint_idx) = dirty_list.pop() {
+            dirty_bit_vec.remove(constraint_idx);


...in that case, this would be clean_bit_vec.insert(constraint_idx);...

nikomatsakis · 2018-01-27T01:01:43Z

src/librustc_mir/borrow_check/nll/region_infer/mod.rs

-                    debug!("propagate_constraints:   sup={:?}", constraint.sup);
-                    changed = true;
+                for &dep_idx in dependency_map.get(&constraint.sup).unwrap_or(&vec![]) {
+                    if dirty_bit_vec.insert(dep_idx) {


...and this would be if clean_bit_vec.remove(dep_idx) {

Otherwise the vector is initially out of sync

nikomatsakis · 2018-01-27T09:44:26Z

My measurements here locally using the testcase from #47267:

No changes, no NLL: ~405 ms
No changes, with NLL: ~608 ms, so NLL adds around 203ms
This PR, with NLL: ~518ms, so NLL adds around 113ms

So this looks like a solid win for NLL to me! :)

These are averages of 6 runs. The results were fairly stable but jumped around some. The test case is kind of small. So take them with a grain of salt. Here is a gist with my full measurements in case you are curious, each one is labeled with the SHA1 commit hash and whether it had NLL enabled or disabled.

nikomatsakis · 2018-01-27T09:44:33Z

@bors r+

bors · 2018-01-27T09:44:33Z

📌 Commit 205eba8 has been approved by nikomatsakis

leonardo-m · 2018-01-27T09:48:16Z

These are averages of 6 runs. The results were fairly stable but jumped around some. The test case is kind of small.

The next Nightly I'll test this with a much larger amount of code.

nikomatsakis · 2018-01-27T10:41:09Z

@leonardo-m if you have a test, I can run it now.

bors · 2018-01-27T16:40:58Z

⌛ Testing commit 205eba8 with merge 4349bc51453d0c74495589c8036622a9b1a05726...

bors · 2018-01-27T19:41:38Z

💔 Test failed - status-appveyor

kennytm · 2018-01-28T18:53:57Z

@bors retry #46903

bors · 2018-01-29T02:20:21Z

⌛ Testing commit 205eba8 with merge b7d4de9edf202b3a9cd889bf038ee663aecaf573...

bors · 2018-01-29T04:55:44Z

💔 Test failed - status-travis

kennytm · 2018-01-29T09:11:19Z

@bors retry #43283

bors · 2018-01-29T10:51:14Z

⌛ Testing commit 205eba8 with merge 70f7d58...

@nikomatsakis

Make region inference use a dirty list r? @nikomatsakis

bors · 2018-01-29T13:45:03Z

☀️ Test successful - status-appveyor, status-travis
Approved by: nikomatsakis
Pushing 70f7d58 to master...

rust-highfive assigned nikomatsakis Jan 26, 2018

nikomatsakis requested changes Jan 26, 2018

View reviewed changes

spastorino force-pushed the inference-dirty-list branch from b1ed025 to 50280ea Compare January 26, 2018 12:18

kennytm added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Jan 26, 2018

spastorino mentioned this pull request Jan 26, 2018

Significantly slower --emit=metadata #47267

Closed

nikomatsakis requested changes Jan 26, 2018

View reviewed changes

spastorino force-pushed the inference-dirty-list branch from 50280ea to 6b01707 Compare January 26, 2018 23:29

nikomatsakis requested changes Jan 27, 2018

View reviewed changes

Make region inference use a dirty list

da545ce

Fixes rust-lang#47602

spastorino force-pushed the inference-dirty-list branch from 6b01707 to da545ce Compare January 27, 2018 00:57

Adjust comment spacing

44b6668

I suspect the lines would be long for tidy.

nikomatsakis requested changes Jan 27, 2018

View reviewed changes

nikomatsakis added 2 commits January 26, 2018 20:13

change from dirty_bit_vec to clean_bit_vec

b57ca9d

Otherwise the vector is initially out of sync

pacify the mercilous tidy

205eba8

nikomatsakis approved these changes Jan 27, 2018

View reviewed changes

bors mentioned this pull request Jan 27, 2018

Fix off-by-one spans in MIR borrowck errors #47420

Merged

kennytm added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jan 29, 2018

bors added a commit that referenced this pull request Jan 29, 2018

Auto merge of #47766 - spastorino:inference-dirty-list, r=nikomatsakis

70f7d58

Make region inference use a dirty list r? @nikomatsakis

bors merged commit 205eba8 into rust-lang:master Jan 29, 2018

nikomatsakis mentioned this pull request Jan 30, 2018

NLL performance tracking issue #47879

Closed

6 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Make region inference use a dirty list #47766

Make region inference use a dirty list #47766

spastorino commented Jan 26, 2018

nikomatsakis left a comment

nikomatsakis Jan 26, 2018

nikomatsakis Jan 26, 2018

nikomatsakis Jan 26, 2018

spastorino commented Jan 26, 2018

leonardo-m commented Jan 26, 2018

nikomatsakis commented Jan 26, 2018

nikomatsakis left a comment

nikomatsakis Jan 26, 2018

nikomatsakis Jan 26, 2018

spastorino commented Jan 26, 2018

leonardo-m commented Jan 26, 2018

nikomatsakis commented Jan 27, 2018

nikomatsakis Jan 27, 2018

nikomatsakis Jan 27, 2018

nikomatsakis Jan 27, 2018 •

edited

Loading

nikomatsakis left a comment

nikomatsakis Jan 27, 2018

nikomatsakis Jan 27, 2018

nikomatsakis Jan 27, 2018

nikomatsakis commented Jan 27, 2018

nikomatsakis commented Jan 27, 2018

bors commented Jan 27, 2018

leonardo-m commented Jan 27, 2018

nikomatsakis commented Jan 27, 2018

bors commented Jan 27, 2018

bors commented Jan 27, 2018

kennytm commented Jan 28, 2018

bors commented Jan 29, 2018

bors commented Jan 29, 2018

kennytm commented Jan 29, 2018

bors commented Jan 29, 2018

bors commented Jan 29, 2018

Make region inference use a dirty list #47766

Make region inference use a dirty list #47766

Conversation

spastorino commented Jan 26, 2018

nikomatsakis left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

spastorino commented Jan 26, 2018

leonardo-m commented Jan 26, 2018

nikomatsakis commented Jan 26, 2018

nikomatsakis left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

spastorino commented Jan 26, 2018

leonardo-m commented Jan 26, 2018

nikomatsakis commented Jan 27, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nikomatsakis Jan 27, 2018 • edited Loading

Choose a reason for hiding this comment

nikomatsakis left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nikomatsakis commented Jan 27, 2018

nikomatsakis commented Jan 27, 2018

bors commented Jan 27, 2018

leonardo-m commented Jan 27, 2018

nikomatsakis commented Jan 27, 2018

bors commented Jan 27, 2018

bors commented Jan 27, 2018

kennytm commented Jan 28, 2018

bors commented Jan 29, 2018

bors commented Jan 29, 2018

kennytm commented Jan 29, 2018

bors commented Jan 29, 2018

bors commented Jan 29, 2018

nikomatsakis Jan 27, 2018 •

edited

Loading