Use scope tree depths to speed up nearest_common_ancestor.
#51394
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This patch adds depth markings to all entries in the `ScopeTree`'s `parent_map`. This change increases memory usage somewhat, but permits a much faster algorithm to be used: - If one scope has a greater depth than the other, the deeper scope is moved upward until they are at equal depths. - Then we move the two scopes upward in lockstep until they match. This avoids the need to keep track of which scopes have already been seen, which was the major part of the cost of the old algorithm. It also reduces the number of child-to-parent moves (which are hash table lookups) when the scopes start at different levels, because it never goes past the nearest common ancestor the way the old algorithm did. Finally, the case where one of the scopes is the root is now handled in advance, because that is moderately common and lets us skip everything. This change speeds up runs of several rust-perf benchmarks, the best by 6%.
rust-highfive
added
the
S-waiting-on-review
nikomatsakis
approved these changes
Jun 6, 2018
|
@bors r+ |
|
📌 Commit 5c36e01 has been approved by |
bors
added
S-waiting-on-bors
Mark-Simulacrum
added a commit
to Mark-Simulacrum/rust
that referenced
this pull request
Jun 6, 2018
Use scope tree depths to speed up `nearest_common_ancestor`.
This patch adds depth markings to all entries in the `ScopeTree`'s
`parent_map`. This change increases memory usage somewhat, but permits a
much faster algorithm to be used:
- If one scope has a greater depth than the other, the deeper scope is
moved upward until they are at equal depths.
- Then we move the two scopes upward in lockstep until they match.
This avoids the need to keep track of which scopes have already been
seen, which was the major part of the cost of the old algorithm. It also
reduces the number of child-to-parent moves (which are hash table
lookups) when the scopes start at different levels, because it never
goes past the nearest common ancestor the way the old algorithm did.
Finally, the case where one of the scopes is the root is now handled in
advance, because that is moderately common and lets us skip everything.
This change speeds up runs of several rust-perf benchmarks, the best by
6%.
A selection of the bigger improvements:
```
clap-rs-check
avg: -2.6% min: -6.6% max: 0.0%
syn-check
avg: -2.2% min: -5.0% max: 0.0%
style-servo-check
avg: -2.9%? min: -4.8%? max: 0.0%?
cargo-check
avg: -1.3% min: -2.8% max: 0.0%
sentry-cli-check
avg: -1.0% min: -2.1% max: 0.0%
webrender-check
avg: -0.9% min: -2.0% max: 0.0%
style-servo
avg: -0.9%? min: -1.8%? max: -0.0%?
ripgrep-check
avg: -0.7% min: -1.8% max: 0.1%
clap-rs
avg: -0.9% min: -1.6% max: -0.2%
regex-check
avg: -0.2% min: -1.3% max: 0.1%
syn
avg: -0.6% min: -1.3% max: 0.1%
hyper-check
avg: -0.5% min: -1.1% max: 0.0%
```
The idea came from multiple commenters on my blog and on Reddit. Thank you!
r? @nikomatsakis
Mark-Simulacrum
added a commit
to Mark-Simulacrum/rust
that referenced
this pull request
Jun 7, 2018
Use scope tree depths to speed up `nearest_common_ancestor`.
This patch adds depth markings to all entries in the `ScopeTree`'s
`parent_map`. This change increases memory usage somewhat, but permits a
much faster algorithm to be used:
- If one scope has a greater depth than the other, the deeper scope is
moved upward until they are at equal depths.
- Then we move the two scopes upward in lockstep until they match.
This avoids the need to keep track of which scopes have already been
seen, which was the major part of the cost of the old algorithm. It also
reduces the number of child-to-parent moves (which are hash table
lookups) when the scopes start at different levels, because it never
goes past the nearest common ancestor the way the old algorithm did.
Finally, the case where one of the scopes is the root is now handled in
advance, because that is moderately common and lets us skip everything.
This change speeds up runs of several rust-perf benchmarks, the best by
6%.
A selection of the bigger improvements:
```
clap-rs-check
avg: -2.6% min: -6.6% max: 0.0%
syn-check
avg: -2.2% min: -5.0% max: 0.0%
style-servo-check
avg: -2.9%? min: -4.8%? max: 0.0%?
cargo-check
avg: -1.3% min: -2.8% max: 0.0%
sentry-cli-check
avg: -1.0% min: -2.1% max: 0.0%
webrender-check
avg: -0.9% min: -2.0% max: 0.0%
style-servo
avg: -0.9%? min: -1.8%? max: -0.0%?
ripgrep-check
avg: -0.7% min: -1.8% max: 0.1%
clap-rs
avg: -0.9% min: -1.6% max: -0.2%
regex-check
avg: -0.2% min: -1.3% max: 0.1%
syn
avg: -0.6% min: -1.3% max: 0.1%
hyper-check
avg: -0.5% min: -1.1% max: 0.0%
```
The idea came from multiple commenters on my blog and on Reddit. Thank you!
r? @nikomatsakis
bors
added a commit
that referenced
this pull request
Jun 7, 2018
Use scope tree depths to speed up `nearest_common_ancestor`.
This patch adds depth markings to all entries in the `ScopeTree`'s
`parent_map`. This change increases memory usage somewhat, but permits a
much faster algorithm to be used:
- If one scope has a greater depth than the other, the deeper scope is
moved upward until they are at equal depths.
- Then we move the two scopes upward in lockstep until they match.
This avoids the need to keep track of which scopes have already been
seen, which was the major part of the cost of the old algorithm. It also
reduces the number of child-to-parent moves (which are hash table
lookups) when the scopes start at different levels, because it never
goes past the nearest common ancestor the way the old algorithm did.
Finally, the case where one of the scopes is the root is now handled in
advance, because that is moderately common and lets us skip everything.
This change speeds up runs of several rust-perf benchmarks, the best by
6%.
A selection of the bigger improvements:
```
clap-rs-check
avg: -2.6% min: -6.6% max: 0.0%
syn-check
avg: -2.2% min: -5.0% max: 0.0%
style-servo-check
avg: -2.9%? min: -4.8%? max: 0.0%?
cargo-check
avg: -1.3% min: -2.8% max: 0.0%
sentry-cli-check
avg: -1.0% min: -2.1% max: 0.0%
webrender-check
avg: -0.9% min: -2.0% max: 0.0%
style-servo
avg: -0.9%? min: -1.8%? max: -0.0%?
ripgrep-check
avg: -0.7% min: -1.8% max: 0.1%
clap-rs
avg: -0.9% min: -1.6% max: -0.2%
regex-check
avg: -0.2% min: -1.3% max: 0.1%
syn
avg: -0.6% min: -1.3% max: 0.1%
hyper-check
avg: -0.5% min: -1.1% max: 0.0%
```
The idea came from multiple commenters on my blog and on Reddit. Thank you!
r? @nikomatsakis
|
💔 Test failed - status-travis |
bors
added
S-waiting-on-review
|
The job Click to expand the log.I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
|
The job Click to expand the log.I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
|
Looks like an infra failure. |
|
@bors retry - network failure |
bors
added
S-waiting-on-bors
Mark-Simulacrum
added a commit
to Mark-Simulacrum/rust
that referenced
this pull request
Jun 8, 2018
Use scope tree depths to speed up `nearest_common_ancestor`.
This patch adds depth markings to all entries in the `ScopeTree`'s
`parent_map`. This change increases memory usage somewhat, but permits a
much faster algorithm to be used:
- If one scope has a greater depth than the other, the deeper scope is
moved upward until they are at equal depths.
- Then we move the two scopes upward in lockstep until they match.
This avoids the need to keep track of which scopes have already been
seen, which was the major part of the cost of the old algorithm. It also
reduces the number of child-to-parent moves (which are hash table
lookups) when the scopes start at different levels, because it never
goes past the nearest common ancestor the way the old algorithm did.
Finally, the case where one of the scopes is the root is now handled in
advance, because that is moderately common and lets us skip everything.
This change speeds up runs of several rust-perf benchmarks, the best by
6%.
A selection of the bigger improvements:
```
clap-rs-check
avg: -2.6% min: -6.6% max: 0.0%
syn-check
avg: -2.2% min: -5.0% max: 0.0%
style-servo-check
avg: -2.9%? min: -4.8%? max: 0.0%?
cargo-check
avg: -1.3% min: -2.8% max: 0.0%
sentry-cli-check
avg: -1.0% min: -2.1% max: 0.0%
webrender-check
avg: -0.9% min: -2.0% max: 0.0%
style-servo
avg: -0.9%? min: -1.8%? max: -0.0%?
ripgrep-check
avg: -0.7% min: -1.8% max: 0.1%
clap-rs
avg: -0.9% min: -1.6% max: -0.2%
regex-check
avg: -0.2% min: -1.3% max: 0.1%
syn
avg: -0.6% min: -1.3% max: 0.1%
hyper-check
avg: -0.5% min: -1.1% max: 0.0%
```
The idea came from multiple commenters on my blog and on Reddit. Thank you!
r? @nikomatsakis
bors
added a commit
that referenced
this pull request
Jun 8, 2018
Rollup of 13 pull requests Successful merges: - #50143 (Add deprecation lint for duplicated `macro_export`s) - #51099 (Fix Issue 38777) - #51276 (Dedup auto traits in trait objects.) - #51298 (Stabilize unit tests with non-`()` return type) - #51360 (Suggest parentheses when a struct literal needs them) - #51391 (Use spans pointing at the inside of a rustdoc attribute) - #51394 (Use scope tree depths to speed up `nearest_common_ancestor`.) - #51396 (Make the size of Option<NonZero*> a documented guarantee.) - #51401 (Warn on `repr` without hints) - #51412 (Avoid useless Vec clones in pending_obligations().) - #51427 (compiletest: autoremove duplicate .nll.* files (#51204)) - #51436 (Do not require stage 2 compiler for rustdoc) - #51437 (rustbuild: generate full list of dependencies for metadata) Failed merges:
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch adds depth markings to all entries in the
ScopeTree'sparent_map. This change increases memory usage somewhat, but permits amuch faster algorithm to be used:
If one scope has a greater depth than the other, the deeper scope is
moved upward until they are at equal depths.
Then we move the two scopes upward in lockstep until they match.
This avoids the need to keep track of which scopes have already been
seen, which was the major part of the cost of the old algorithm. It also
reduces the number of child-to-parent moves (which are hash table
lookups) when the scopes start at different levels, because it never
goes past the nearest common ancestor the way the old algorithm did.
Finally, the case where one of the scopes is the root is now handled in
advance, because that is moderately common and lets us skip everything.
This change speeds up runs of several rust-perf benchmarks, the best by
6%.
A selection of the bigger improvements:
The idea came from multiple commenters on my blog and on Reddit. Thank you!
r? @nikomatsakis