Assign memo ingredients per salsa-struct-ingredient by ShoyuVanilla · Pull Request #614 · salsa-rs/salsa

ShoyuVanilla · 2024-11-13T15:01:46Z

Closes #600

netlify · 2024-11-13T15:02:07Z

✅ Deploy Preview for salsa-rs canceled.

Name	Link
🔨 Latest commit	`53c7eba`
🔍 Latest deploy log	https://app.netlify.com/sites/salsa-rs/deploys/6762f08f5314490008b8f2fd

codspeed-hq · 2024-11-13T15:04:21Z

CodSpeed Performance Report

Merging #614 will not alter performance

_{Comparing ShoyuVanilla:issue-600 (53c7eba) with master (3c7f169)}

Summary

✅ 9 untouched benchmarks

ShoyuVanilla · 2024-11-13T15:07:48Z

src/zalsa.rs


+    /// Map from the type-id of a salsa struct to the index of its first ingredient.
+    salsa_struct_map: Mutex<FxHashMap<TypeId, IngredientIndex>>,
+


I don't like this new table for SalsaStruct -> IngredientIndex but I couldn't find any other way to pass the salsa-struct-ingredient to its tracked function when creating the function's ingredient, as salsa-struct-ingredient is accessible only with its Configuration, that is private inside the macro-expanded code of that salsa struct.
Would there be some good alternatives?

I would be inclined to add a method ingredient_index to SalsaStructInDb but best to hear what @nikomatsakis thinks

We need to get the IngredientIndex for salsa struct here;

salsa/components/salsa-macro-rules/src/setup_tracked_fn.rs

Lines 196 to 205 in e4d36da

impl $zalsa::Jar for $Configuration {

fn create_ingredients(

&self,

aux: &dyn $zalsa::JarAux,

first_index: $zalsa::IngredientIndex,

) -> Vec<Box<dyn $zalsa::Ingredient>> {

let fn_ingredient = <$zalsa::function::IngredientImpl<$Configuration>>::new(

first_index,

aux,

);

and we can't get the actual instance of salsa struct here. All we have is JarAux - actually, this is Zalsa -, so that function should be a static method, I guess. (I'm afraid that this makes SalsaStructInDb dyn-incompatible but we are not using it as a trait object anywhere in salsa)

Thus, the function would be implemented inside our setup_* macros like;

fn ingredient_index(aux: &dyn JarAux) -> IngredientIndex { aux.lookup_jar_by_type(JarImpl::<$Configuration>::default()) }

and we need to expose some method like lookup_jar_by_type in JarAux, I think.

MichaReiser

I don't think we currently have a test for a tracked fn that has more than one argument (other than db) and thus uses the interner. Could you double check if that's indeed the case and, if so, add a test for it?

MichaReiser · 2024-11-13T17:02:14Z

src/zalsa.rs

+    /// [ingredient-indices](`IngredientIndex`)for tracked functions that have this salsa struct
+    /// as input.
+    memo_ingredients: RwLock<FxHashMap<IngredientIndex, Vec<IngredientIndex>>>,


Nit:

Suggested change

/// [ingredient-indices](`IngredientIndex`)for tracked functions that have this salsa struct

/// as input.

memo_ingredients: RwLock<FxHashMap<IngredientIndex, Vec<IngredientIndex>>>,

/// [ingredient-indices](`IngredientIndex`) for tracked functions that have this salsa struct

/// as input.

memo_ingredient_indices: RwLock<FxHashMap<IngredientIndex, Vec<IngredientIndex>>>,

I also have a slight preference to the name @nikomatsakis suggest in the mentoring instructions (memo_ingredient_indices)

MichaReiser · 2024-11-13T17:06:30Z

src/zalsa.rs

+    /// [ingredient-indices](`IngredientIndex`)for tracked functions that have this salsa struct
+    /// as input.
+    memo_ingredients: RwLock<FxHashMap<IngredientIndex, Vec<IngredientIndex>>>,


We could consider using a Vec<Vec<IngredientIndex>> here and directly index into the Vec using the IngredientIndex as an offset. This would result in wasting some memory because not every IngredientIndex is an index for a tracked struct but it would avoid one additional hashing step

ShoyuVanilla · 2024-11-14T03:09:58Z

I don't think we currently have a test for a tracked fn that has more than one argument (other than db) and thus uses the internet. Could you double check if that's indeed the case and, if so, add a test for it?

Sure! I'll add tests and edit things you commented when I get home. Thanks for the review

ShoyuVanilla · 2024-11-14T16:43:44Z

src/zalsa.rs

-impl JarAux for Zalsa {
-    fn next_memo_ingredient_index(&self, ingredient_index: IngredientIndex) -> MemoIngredientIndex {
-        let mut memo_ingredients = self.memo_ingredients.lock();
+struct JarAuxImpl<'a>(&'a Zalsa, &'a FxHashMap<TypeId, IngredientIndex>);


I changed salsa struct ingredient lookup as I said here but I had to change this to avoid deadlocks 😢

MichaReiser · 2024-12-06T17:25:45Z

src/zalsa.rs

+            });
+            if should_create {
+                let aux = JarAuxImpl(self, &jar_map);
+                let ingredients = jar.create_ingredients(&aux, index);


There's a comment on jar_map that suggests that it's important to only access ingredients_vec while the lock on jar_map is held

https://github.com/MichaReiser/salsa/blob/38ad6555ebe4374bbc69022e005f4136a1839ad9/src/zalsa.rs#L127-L133

So moving it out to avoid the deadlock might not be entirely safe

For example. What could happen now is that two calls end up with the same index because the new ingredient isn't pushed on the ingredients_vec until line 206, so that self.ingredients_vec.len() returns the same length.

Okay, I'll fix this

nikomatsakis · 2024-12-11T10:18:02Z

@ShoyuVanilla hey, cool! My apologies for radio silence. I've been overwhelmed. I'm going to carve out more time for salsa though so let me take a look...

nikomatsakis

This looks good! Took me a bit to get my head in the right space. I left comments, can you check that they are accurate and merge them in? After that, I'd say we can land this.

nikomatsakis · 2024-12-11T10:24:33Z

src/zalsa.rs

-            .or_insert_with(|| {
-                let index = IngredientIndex::from(self.ingredients_vec.len());
-                let ingredients = jar.create_ingredients(self, index);
+            let mut should_create = false;


I don't quite follow this -- why move this logic out from the or_insert_with call?

I guess it is so that ingredient construction can call lookup_jar_by_type?

It'd be nice to have a comment here explaining that interaction

I don't remember for sure - as it was weeks before 😅 -, but I think that it was because it had some issues due to creation order like using ingredients index before they are created.
I'll look into such order related things along with the one you commented here

src/ingredient.rs

nikomatsakis · 2024-12-11T10:31:03Z

tests/tracked_fn_multiple_args.rs

+}
+
+#[salsa::tracked]
+fn tracked_fn<'db>(db: &'db dyn salsa::Database, input: MyInput, interned: MyInterned<'db>) -> u32 {


Why this particular test? Tracked functions with multiple arguments today compile to a tracked function on a hidden interned value -- is that what you were intending to test?

Yes. Though this PR is not directly related to that, but as it changes some codes in such hidden interned value, I was intending to add a regression test for it. Other case are pre-existing

nikomatsakis · 2024-12-11T10:38:25Z

components/salsa-macro-rules/src/setup_tracked_fn.rs

                ) -> Vec<Box<dyn $zalsa::Ingredient>> {
+                    let struct_index = $zalsa::macro_if! {
+                        if $needs_interner {
+                            first_index.successor(0)


I...guess this is fine. It's a bit surprising to use the index of something that hasn't been made yet. An alternative would be to flip the order so that we create the interned struct first and then pass its index to the function. There'd probably be some minor adjustments made elsewhere in the file as a result but shouldn't be too bad.

nikomatsakis · 2024-12-11T10:39:00Z

components/salsa-macro-rules/src/setup_tracked_fn.rs

+                        if $needs_interner {
+                            first_index.successor(0)
+                        } else {
+                            <$InternedData as $zalsa::SalsaStructInDb>::lookup_ingredient_index(aux)


OK, I see now the role of lookup_ingredient_index.

nikomatsakis · 2024-12-11T10:40:52Z

components/salsa-macro-rules/src/setup_tracked_fn.rs

+                            first_index.successor(0)
+                        } else {
+                            <$InternedData as $zalsa::SalsaStructInDb>::lookup_ingredient_index(aux)
+                                .expect(


I'm not 1000% percent sure that the struct ingredient will always have been created first. If this code doesn't run until the tracked functon is first called, I suppose that's true, but if it could run at some other time maybe not? Still, this seems like a good assertion for now to simplify our lives. If it's not true we can probably add some kind of dependency mechanism to force it to be true.

I thought - and experimented with some cases - that the structs are created first to be passed to a tracked function but my understanding on salsa is very limited and I might be wrong. I'll investigate this more

src/ingredient.rs

src/zalsa.rs

nikomatsakis · 2024-12-11T10:49:19Z

src/zalsa.rs

@@ -186,21 +188,22 @@ impl Zalsa {
        {
            let jar_type_id = jar.type_id();
            let mut jar_map = self.jar_map.lock();


Suggested change

let mut jar_map = self.jar_map.lock();

// Important: we hold the lock on `jar_map` during ingredient construction to

// ensure all ingredients are created atomically.

let mut jar_map = self.jar_map.lock();

ShoyuVanilla · 2024-12-11T18:52:28Z

This looks good! Took me a bit to get my head in the right space. I left comments, can you check that they are accurate and merge them in? After that, I'd say we can land this.

Thanks for the review! I'll fix those things by this weekend

ShoyuVanilla · 2024-12-18T15:31:04Z

Well, I was going to continue working on this but it seems that I'm too late, sorry. I had family affairs last week 😢
I'll issue a new PR with fixes for quirks on this PR

Veykril · 2024-12-18T15:49:01Z

No need to apologize you have no obligations so do not feel pressured to work on things asap

nikomatsakis · 2024-12-18T15:52:04Z

@ShoyuVanilla yeah, no worries, I'm grateful you did the original PR! Very big impact.

The main thing this PR needs is cargo fmt

davidbarsky · 2024-12-18T16:13:37Z

@ShoyuVanilla if it's okay, i'll merge this PR: it'll make managing the stack of changes that Lukas and I have for master...davidbarsky:salsa:david-and-lukas/required-salsa-changes-for-ra a lot easier. Can you do the followups on a separate PR?

ShoyuVanilla · 2024-12-18T16:15:18Z

@ShoyuVanilla if it's okay, i'll merge this PR: it'll make managing the stack of changes that Lukas and I have for master...davidbarsky:salsa:david-and-lukas/required-salsa-changes-for-ra a lot easier. Can you do the followups on a separate PR?

Yes, I'll issue a new follow-up PR (along with #599, maybe?)

davidbarsky · 2024-12-18T16:18:49Z

Yes, I'll issue a new follow-up PR (along with #599, maybe?)

Up to you!

Assign memo ingredients per salsa-struct-ingredient

fac6053

ShoyuVanilla commented Nov 13, 2024

View reviewed changes

MichaReiser reviewed Nov 13, 2024

View reviewed changes

ShoyuVanilla added 2 commits November 15, 2024 01:40

Modify struct ingredient lookups

2cae1ef

Add a test for tracked function with multiple salsa struct args

d34ed14

ShoyuVanilla commented Nov 14, 2024

View reviewed changes

ShoyuVanilla added 2 commits November 15, 2024 01:46

Fix clippy

738d5f9

Clean up code a bit

38ad655

MichaReiser reviewed Dec 6, 2024

View reviewed changes

nikomatsakis requested changes Dec 11, 2024

View reviewed changes

Apply suggestions from code review

7aa1518

nikomatsakis approved these changes Dec 18, 2024

View reviewed changes

Run rustfmt

53c7eba

davidbarsky added this pull request to the merge queue Dec 18, 2024

Merged via the queue into salsa-rs:master with commit 0ac5c1c Dec 18, 2024

ShoyuVanilla deleted the issue-600 branch December 18, 2024 16:25

MichaReiser mentioned this pull request Jan 3, 2025

Update salsa astral-sh/ruff#15243

Merged


		/// Map from the type-id of a salsa struct to the index of its first ingredient.
		salsa_struct_map: Mutex<FxHashMap<TypeId, IngredientIndex>>,

	impl $zalsa::Jar for $Configuration {
	fn create_ingredients(
	&self,
	aux: &dyn $zalsa::JarAux,
	first_index: $zalsa::IngredientIndex,
	) -> Vec<Box<dyn $zalsa::Ingredient>> {
	let fn_ingredient = <$zalsa::function::IngredientImpl<$Configuration>>::new(
	first_index,
	aux,
	);

Conversation

ShoyuVanilla commented Nov 13, 2024

Uh oh!

netlify bot commented Nov 13, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Deploy Preview for salsa-rs canceled.

Uh oh!

codspeed-hq bot commented Nov 13, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging #614 will not alter performance

Summary

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MichaReiser left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ShoyuVanilla commented Nov 14, 2024

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nikomatsakis commented Dec 11, 2024

Uh oh!

nikomatsakis left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ShoyuVanilla commented Dec 11, 2024

Uh oh!

ShoyuVanilla commented Dec 18, 2024

Uh oh!

Veykril commented Dec 18, 2024

Uh oh!

nikomatsakis commented Dec 18, 2024

Uh oh!

davidbarsky commented Dec 18, 2024

Uh oh!

ShoyuVanilla commented Dec 18, 2024

Uh oh!

davidbarsky commented Dec 18, 2024

netlify bot commented Nov 13, 2024 •

edited

Loading

codspeed-hq bot commented Nov 13, 2024 •

edited

Loading

MichaReiser left a comment •

edited

Loading