perf: reference management optimizations

[Oppen]

• Store the reference list rather than the manager structure
• Process the reference list only on Program creation rather than in each call to run_for_steps
• Use a Vec<_> rather than a HashMap<_, _> since the instance is dense
• Keep this data in the SharedProgramData structure instead of copying it each time
• Improve the accuracy of the benchmarks

Benchmarks

build runner            time:   [773.76 ns 774.50 ns 775.47 ns]                          
                        change: [-93.906% -93.895% -93.885%] (p = 0.00 < 0.05)
initialize              time:   [20.632 µs 20.649 µs 20.672 µs]                        
                        change: [-16.707% -15.351% -14.383%] (p = 0.00 < 0.05)
parse program           time:   [9.2019 ms 9.2176 ms 9.2341 ms]                          
                        change: [+0.1218% +0.2751% +0.4491%] (p = 0.00 < 0.05)

[github-actions]

Benchmark Results for unmodified programs 🚀

Command	Mean [s]	Min [s]	Max [s]	Relative
base factorial_multirun	3.622 ± 0.058	3.575	3.730	1.00
head factorial_multirun	3.689 ± 0.031	3.649	3.734	1.02 ± 0.02

Command	Mean [s]	Min [s]	Max [s]	Relative
base fibonacci_1000_multirun	3.094 ± 0.046	3.051	3.211	1.00
head fibonacci_1000_multirun	3.138 ± 0.030	3.112	3.199	1.01 ± 0.02

[unbalancedparentheses]

[github-actions]

Benchmark Results for modified programs 🚀

Command	Mean [s]	Min [s]	Max [s]	Relative
head blake2s_integration_benchmark	13.242 ± 0.099	13.141	13.417	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head compare_arrays_200000	4.600 ± 0.042	4.552	4.656	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head dict_integration_benchmark	3.047 ± 0.020	3.025	3.090	1.00

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
head field_arithmetic_get_square_benchmark	197.0 ± 14.6	186.9	234.0	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head integration_builtins	11.686 ± 0.149	11.512	11.973	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head keccak_integration_benchmark	13.550 ± 0.635	13.227	15.326	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head linear_search	4.598 ± 0.131	4.530	4.963	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head math_cmp_and_pow_integration_benchmark	3.262 ± 0.034	3.241	3.356	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head math_integration_benchmark	3.009 ± 0.026	2.983	3.076	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head memory_integration_benchmark	2.631 ± 0.022	2.612	2.692	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head operations_with_data_structures_benchmarks	2.986 ± 0.008	2.974	3.002	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head pedersen	1.042 ± 0.005	1.035	1.053	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head poseidon_integration_benchmark	1.868 ± 0.012	1.853	1.889	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head secp_integration_benchmark	3.562 ± 0.023	3.540	3.621	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head set_integration_benchmark	2.005 ± 0.008	1.990	2.016	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
head uint256_integration_benchmark	8.404 ± 0.032	8.374	8.482	1.00

[codecov]

Codecov Report

Merging #1214 (51a72b4) into main (4221723) will decrease coverage by 0.04%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #1214      +/-   ##
==========================================
- Coverage   97.61%   97.57%   -0.04%     
==========================================
  Files          88       88              
  Lines       36112    36110       -2     
==========================================
- Hits        35251    35235      -16     
- Misses        861      875      +14     

Impacted Files	Coverage Δ
...processor/cairo_1_hint_processor/hint_processor.rs	85.29% <100.00%> (ø)
vm/src/hint_processor/hint_processor_definition.rs	79.10% <100.00%> (-20.90%)	⬇️
vm/src/serde/deserialize_program.rs	96.75% <100.00%> (ø)
vm/src/types/program.rs	99.47% <100.00%> (+0.01%)	⬆️
vm/src/utils.rs	99.49% <100.00%> (ø)
vm/src/vm/runners/cairo_runner.rs	98.01% <100.00%> (-0.01%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more