-
Notifications
You must be signed in to change notification settings - Fork 7
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
feat: add verification to constant folding #1030
Conversation
83ce472
to
862419f
Compare
I recommend reviewing commit-by-commit. @cqc-alec I'm mostly asking for your view of the test fixes. |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #1030 +/- ##
==========================================
+ Coverage 86.26% 86.31% +0.05%
==========================================
Files 82 82
Lines 17264 17234 -30
Branches 17264 17234 -30
==========================================
- Hits 14893 14876 -17
+ Misses 1565 1538 -27
- Partials 806 820 +14
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
// pseudocode: | ||
// | ||
// x0 := int_s<5>(-3); | ||
// x1 := inarrow_s<5, 4>(x0); | ||
// output x1 == int_s<4>(-3); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This comment should be removed or rewritten in more general terms.
hugr/src/algorithm/const_fold.rs
Outdated
} | ||
|
||
#[derive(Debug, Clone, Copy, Default)] | ||
/// TODO |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What is TODO?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
docs, I missed this one.
hugr/src/algorithm/const_fold.rs
Outdated
for rem in removes { | ||
if let Ok(const_node) = h.apply_rewrite(rem) { | ||
// if the LoadConst was removed, try removing the Const too. | ||
let _ = h.apply_rewrite(RemoveConst(const_node)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is it safe to ignore all errors here?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes it is, but I'll add a comment to explain why.
@@ -395,4 +462,44 @@ mod test { | |||
let expected = Value::false_val(); | |||
assert_fully_folded(&h, &expected); | |||
} | |||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is it worth also including the test from #996 ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, will do.
Thanks for the review Alec. I will wait until @ss2165 has had a look before I merge, significant changes to his baby. |
b2698ea
to
1e7a6f5
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
PR looks good, thanks for the fix, main query about whether broader "pass level" design changes should be made in this PR
hugr/src/algorithm/const_fold.rs
Outdated
@@ -69,7 +156,10 @@ pub fn fold_leaf_op(op: &OpType, consts: &[(IncomingPort, Value)]) -> ConstFoldR | |||
ext_op.constant_fold(consts) | |||
} | |||
_ => None, | |||
} | |||
}; | |||
assert!(fold_result.as_ref().map_or(true, |x| x.len() |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
should this be a debug assert?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lol, I thought that's what assert was. will change.
hugr/src/algorithm/const_fold.rs
Outdated
} | ||
|
||
/// TODO | ||
pub fn run(&self, h: &mut impl HugrMut, reg: &ExtensionRegistry) -> Result<(), ConstFoldError> { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
not sure I like having a run
method on something with the suffix Config
. There's the beginnings of a general "algorithm" interface here (configuration, transformation, verfification). Is it worth deferring some of these changes in to a generic follow up PR?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, I can do that. I didn't want to through away verifying before and after the pass, but I can change that to be guarded behind #[cfg(test)]
.
hugr/src/algorithm/const_fold.rs
Outdated
// neighbouring nodes of those added. | ||
let rewrites = find_consts(h, h.nodes(), reg).collect_vec(); | ||
if rewrites.is_empty() { | ||
// We can only safely apply a single replacement. Applying a |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes, the original code was based on the assumption that find_consts
would only return rewrites for ops that only had const inputs (which would I think be safe to apply all of them?) but that's not what find_consts does
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes I think that would be safe, if that's what find_consts did.
@@ -16,6 +16,16 @@ use crate::{ | |||
|
|||
use super::IntOpDef; | |||
|
|||
use lazy_static::lazy_static; | |||
|
|||
lazy_static! { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm going to assume Alec has sufficiently reviewed the changes in this commit
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
thanks!
## 🤖 New release * `hugr`: 0.3.1 -> 0.4.0 (⚠️ API breaking changes) ###⚠️ `hugr` breaking changes ``` --- failure inherent_method_const_removed: pub method is no longer const --- Description: A publicly-visible method or associated fn is no longer `const` and can no longer be used in a `const` context. ref: https://doc.rust-lang.org/reference/const_eval.html impl: https://github.com/obi1kenobi/cargo-semver-checks/tree/v0.31.0/src/lints/inherent_method_const_removed.ron Failed in: ConstF64::new in /tmp/.tmpwo5blB/hugr/hugr/src/std_extensions/arithmetic/float_types.rs:43 --- failure struct_missing: pub struct removed or renamed --- Description: A publicly-visible struct cannot be imported by its prior path. A `pub use` may have been removed, or the struct itself may have been renamed or removed entirely. ref: https://doc.rust-lang.org/cargo/reference/semver.html#item-remove impl: https://github.com/obi1kenobi/cargo-semver-checks/tree/v0.31.0/src/lints/struct_missing.ron Failed in: struct hugr::ops::constant::ExtensionValue, previously in file /tmp/.tmpq1W6bC/hugr/src/ops/constant.rs:184 ``` <details><summary><i><b>Changelog</b></i></summary><p> <blockquote> ## 0.4.0 (2024-05-20) ### Bug Fixes - Disallow non-finite values for `ConstF64` ([#1075](#1075)) - Serialization round-trips ([#948](#948)) - [**breaking**] Combine `ConstIntU` and `ConstIntS` ([#974](#974)) - Disable serialisation tests when miri is active ([#977](#977)) - [**breaking**] Serialisation schema ([#968](#968)) - Correct constant fold for `fne`. ([#995](#995)) - [**breaking**] Serialisation fixes ([#997](#997)) - [**breaking**] OpDef serialisation ([#1013](#1013)) - NaryLogicOp constant folding ([#1026](#1026)) ### Features - Add verification to constant folding ([#1030](#1030)) - Add `Const::get_custom_value` ([#1037](#1037)) - Add serialization schema for metadata ([#1038](#1038)) - 'Replace' rewrite returns node map ([#929](#929)) - `new` methods for leaf ops ([#940](#940)) - Add `string` type and `print` function to `prelude` ([#942](#942)) - `CustomOp::extension` utility function ([#951](#951)) - [**breaking**] Add `non_exhaustive` to various enums ([#952](#952)) - Encoder metadata in serialized hugr ([#955](#955)) - [**breaking**] Bring back Value ([#967](#967)) - Add LoadFunction node ([#947](#947)) - Add From impls for TypeArg ([#1002](#1002)) - Constant-folding of integer and logic operations ([#1009](#1009)) - [**breaking**] Update serialisation schema, implement `CustomConst` serialisation ([#1005](#1005)) - Merge basic blocks algorithm ([#956](#956)) - [**breaking**] Allow panic operation to have any input and output wires ([#1024](#1024)) ### Refactor - [**breaking**] Rename `crate::ops::constant::ExtensionValue` => `OpaqueValue` ([#1036](#1036)) - Outline hugr::serialize::test ([#976](#976)) - [**breaking**] Replace SmolStr identifiers with wrapper types. ([#959](#959)) - Separate extension validation from the rest ([#1011](#1011)) - Remove "trait TypeParametrised" ([#1019](#1019)) ### Testing - Reorg OutlineCfg/nest_cfgs tests so hugr doesn't depend on algorithm ([#1007](#1007)) - Ignore tests which depend on typetag when cfg(miri) ([#1051](#1051)) - Really ignore tests which depend on typetag when cfg(miri) ([#1058](#1058)) - Proptests for round trip serialisation of `Type`s and `Op`s. ([#981](#981)) - Add a test of instantiating an extension set ([#939](#939)) - Ignore serialisation tests when using miri ([#975](#975)) - [**breaking**] Test roundtrip serialisation against strict + lax schema ([#982](#982)) - Fix some bad assert_matches ([#1006](#1006)) - Expand test of instantiating extension sets ([#1003](#1003)) - Fix unconnected ports in extension test ([#1010](#1010)) </blockquote> </p></details> --- This PR was generated with [release-plz](https://github.com/MarcoIeni/release-plz/). --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Douglas Wilson <douglas.wilson@quantinuum.com>
Fixes #996. Fixes
test_fold_inarrow
tests which were shown to be wrong by verification, andtest_fold_int_ops
test which was causing constant folding to panic.