[Merged by Bors] - Fix performance bottleneck in VM

[pdogr]

This Pull Request fixes/closes #1972 .

It changes the following:

• remove format! macro in execute_instruction and replace by &'static str

[codecov]

Codecov Report

Merging #1973 (dfaa29b) into main (e2630fa) will decrease coverage by 0.00%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #1973      +/-   ##
==========================================
- Coverage   45.90%   45.89%   -0.01%     
==========================================
  Files         206      206              
  Lines       17116    17116              
==========================================
- Hits         7857     7856       -1     
- Misses       9259     9260       +1     

Impacted Files	Coverage Δ
boa_engine/src/vm/mod.rs	70.95% <100.00%> (ø)
boa_engine/src/vm/opcode.rs	33.33% <100.00%> (-9.53%)	⬇️
boa_engine/src/string.rs	55.17% <0.00%> (-0.69%)	⬇️
boa_engine/src/builtins/regexp/mod.rs	67.28% <0.00%> (ø)
boa_engine/src/builtins/generator/mod.rs	7.77% <0.00%> (+0.08%)	⬆️
boa_engine/src/builtins/object/for_in_iterator.rs	94.11% <0.00%> (+2.68%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e2630fa...dfaa29b. Read the comment docs.

[jasonwilliams]

great work @pdogr how did you spot that? Will this not be duplication over https://github.com/boa-dev/boa/blob/main/boa_engine/src/vm/opcode.rs#L939 though now?

[pdogr]

great work @pdogr how did you spot that? Will this not be duplication over https://github.com/boa-dev/boa/blob/main/boa_engine/src/vm/opcode.rs#L939 though now?

I noticed that in execute_instruction a lot of time was spent in write_str which meant that string slices were being constantly written somewhere.

Yes, this is a duplication of as_str method, as the profiler had an "INST " prefix. We can instead use the as_str method in the profiler.

[jasonwilliams]

@pdogr did you use https://github.com/boa-dev/boa/blob/main/docs/profiling.md or you had your own setup for that flame graph? Would be curious to hear how you got setup

[pdogr]

No nothing specific just linux-perf and flamegraph. The profiling setup for boa https://github.com/boa-dev/boa/blob/main/docs/profiling.md is great to time bytecode execution though but this bottleneck was before that hence probably not caught.

[Razican]

Looks very good! Thank you! I would just document the function, but it's ready to merge from my side :)

[HalidOdat]

The benchmarks look awesome!

Benchmarks

Just execution tests.

name	baseDuration	changesDuration	difference
'Arithmetic operations (Execution)'	'1989.1±2.26ns'	'679.8±0.83ns'	'-66'
'Array access (Execution)'	'9.8±0.03µs'	'7.5±0.02µs'	'-23'
'Array creation (Execution)'	'3.2±0.01ms'	'2.4±0.00ms'	'-25'
'Array pop (Execution)'	'1403.2±3.52µs'	'1088.6±3.62µs'	'-22'
'Boolean Object Access (Execution)'	'5.7±0.02µs'	'3.9±0.03µs'	'-32'
'Clean js (Execution)'	'1087.9±7.34µs'	'687.4±4.20µs'	'-37'
'Dynamic Object Property Access (Execution)'	'6.7±0.04µs'	'5.0±0.01µs'	'-25'
'Fibonacci (Execution)'	'1782.5±13.85µs'	'1272.5±2.40µs'	'-29'
'For loop (Execution)'	'42.3±0.11µs'	'14.9±0.03µs'	'-65'
'Mini js (Execution)'	'1001.9±37.22µs'	'651.3±4.77µs'	'-35'
'Number Object Access (Execution)'	'4.5±0.01µs'	'3.0±0.01µs'	'-33'
'Object Creation (Execution)'	'6.0±0.04µs'	'4.7±0.02µs'	'-21'
'RegExp (Execution)'	'12.0±0.06µs'	'10.7±0.04µs'	'-11'
'RegExp Creation (Execution)'	'8.8±0.03µs'	'8.1±0.03µs'	'-8.3'
'RegExp Literal (Execution)'	'12.0±0.07µs'	'10.8±0.04µs'	'-9.9'
'RegExp Literal Creation (Execution)'	'8.8±0.04µs'	'8.0±0.02µs'	'-9.1'
'Static Object Property Access (Execution)'	'6.3±0.02µs'	'4.9±0.02µs'	'-22'
'String Object Access (Execution)'	'7.5±0.02µs'	'5.8±0.03µs'	'-23'
'String comparison (Execution)'	'5.8±0.01µs'	'4.2±0.01µs'	'-26'
'String concatenation (Execution)'	'5.3±0.02µs'	'4.0±0.01µs'	'-24'
'String copy (Execution)'	'4.7±0.03µs'	'3.8±0.02µs'	'-19'
'Symbols (Execution)'	'4.5±0.01µs'	'3.8±0.01µs'	'-14'

[jasonwilliams]

bors r+

[bors]

Pull request successfully merged into main.

Build succeeded:

• Clippy
• Documentation
• Examples
• Rustfmt
• Tests on Linux
• Tests on MacOS
• Tests on Windows