Skip to content

Optimize the size of Instruction#2365

Merged
coolreader18 merged 10 commits into
masterfrom
coolreader18/optimize-bytecode-format
Dec 16, 2020
Merged

Optimize the size of Instruction#2365
coolreader18 merged 10 commits into
masterfrom
coolreader18/optimize-bytecode-format

Conversation

@coolreader18

@coolreader18 coolreader18 commented Dec 10, 2020 •
edited
Loading

Copy link
Copy Markdown
Member

This reduces the size of the Instruction enum from 56 bytes to 16 bytes, which should help with memory usage as well as a (very) slight performance boost:

Benchmark #1: ./rustpython-norm benchmarks/pystone.py 50000
  Time (mean ± σ):      3.655 s ±  0.119 s    [User: 3.637 s, System: 0.007 s]
  Range (min … max):    3.522 s …  3.853 s    10 runs
 
Benchmark #2: ./rustpython-opt benchmarks/pystone.py 50000
  Time (mean ± σ):      3.624 s ±  0.074 s    [User: 3.601 s, System: 0.008 s]
  Range (min … max):    3.557 s …  3.783 s    10 runs
 
Summary
  './rustpython-opt benchmarks/pystone.py 50000' ran
    1.01 ± 0.04 times faster than './rustpython-norm benchmarks/pystone.py 50000'

The instructions I changed were either a) flattening an inner enum, like CallFunction { CallType } -> CallFunction{Positional,Keyword,Ex} b) not in many hot paths, like Import, or c) provided redundant information, like SetupLoop, whose start field always pointed to lasti+1

Edit: well, SetupLoop isn't actually like that, but there was a way around it

@coolreader18 coolreader18 force-pushed the coolreader18/optimize-bytecode-format branch from d1a5081 to b90bb34 Compare December 10, 2020 19:37
@coolreader18

coolreader18 commented Dec 10, 2020

Copy link
Copy Markdown
Member Author

Here's the benchmark for the lz4_flex change:

Benchmark #1: ./rustpython-norm -c ""
  Time (mean ± σ):      53.1 ms ±  13.2 ms    [User: 45.4 ms, System: 7.3 ms]
  Range (min … max):    36.3 ms …  77.2 ms    66 runs
 
Benchmark #2: ./rustpython-opt -c ""
  Time (mean ± σ):      15.2 ms ±   7.4 ms    [User: 11.0 ms, System: 4.2 ms]
  Range (min … max):     7.7 ms …  33.8 ms    206 runs
 
Summary
  './rustpython-opt -c ""' ran
    3.50 ± 1.92 times faster than './rustpython-norm -c ""'

@coolreader18 coolreader18 force-pushed the coolreader18/optimize-bytecode-format branch 2 times, most recently from 8bbe430 to 9b65cd3 Compare December 11, 2020 18:32
@coolreader18 coolreader18 force-pushed the coolreader18/optimize-bytecode-format branch from 9b65cd3 to 1be79c9 Compare December 14, 2020 21:56
youknowone
youknowone approved these changes Dec 16, 2020
View reviewed changes
Comment thread bytecode/src/bytecode.rs
pub fn from_bytes(data: &[u8]) -> Result<Self, Box<dyn std::error::Error>> {
let reader = lz_fear::framed::LZ4FrameReader::new(data)?;
Ok(bincode::deserialize_from(reader.into_read())?)
// TODO: PR to lz4_flex to make it not panic

@youknowone youknowone Dec 16, 2020

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is this solved comment?

@coolreader18 coolreader18 Dec 16, 2020

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No, not yet, I haven't made a PR

@coolreader18 coolreader18 merged commit 2d246f0 into master Dec 16, 2020
@youknowone youknowone deleted the coolreader18/optimize-bytecode-format branch December 16, 2020 15:02
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@youknowone youknowone youknowone approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@coolreader18 @youknowone