Add support for subtractions contenting references as right hand side operands

[FrancoGiachetta]

Add support for subtractions contenting references as right hand side operands

Description

The VM reads let expressions like c - tmp as an addition instead of a subtraction, where c is a constant and tmp is a local variable. This is because currently only additions are supported. This is not an issue with variables declared with the let keyword because those are resolved at compile time.

This PR introduces a little refactor to allow subtractions with references as the rhs operand. It does it by saving the sign with which the reference is defined.

Checklist

• Linked to Github Issue
• Unit tests added
• Integration tests added.
• This change requires new documentation.
  • Documentation has been added/updated.
  • CHANGELOG has been updated.

[github-actions]

**Hyper Thereading Benchmark results**




hyperfine -r 2 -n "hyper_threading_main threads: 1" 'RAYON_NUM_THREADS=1 ./hyper_threading_main' -n "hyper_threading_pr threads: 1" 'RAYON_NUM_THREADS=1 ./hyper_threading_pr'
Benchmark 1: hyper_threading_main threads: 1
  Time (mean ± σ):     31.650 s ±  0.134 s    [User: 30.869 s, System: 0.780 s]
  Range (min … max):   31.555 s … 31.744 s    2 runs
 
Benchmark 2: hyper_threading_pr threads: 1
  Time (mean ± σ):     31.825 s ±  0.008 s    [User: 31.094 s, System: 0.729 s]
  Range (min … max):   31.819 s … 31.830 s    2 runs
 
Summary
  hyper_threading_main threads: 1 ran
    1.01 ± 0.00 times faster than hyper_threading_pr threads: 1




hyperfine -r 2 -n "hyper_threading_main threads: 2" 'RAYON_NUM_THREADS=2 ./hyper_threading_main' -n "hyper_threading_pr threads: 2" 'RAYON_NUM_THREADS=2 ./hyper_threading_pr'
Benchmark 1: hyper_threading_main threads: 2
  Time (mean ± σ):     17.634 s ±  0.039 s    [User: 30.800 s, System: 0.818 s]
  Range (min … max):   17.607 s … 17.661 s    2 runs
 
Benchmark 2: hyper_threading_pr threads: 2
  Time (mean ± σ):     17.715 s ±  0.008 s    [User: 30.963 s, System: 0.790 s]
  Range (min … max):   17.709 s … 17.720 s    2 runs
 
Summary
  hyper_threading_main threads: 2 ran
    1.00 ± 0.00 times faster than hyper_threading_pr threads: 2




hyperfine -r 2 -n "hyper_threading_main threads: 4" 'RAYON_NUM_THREADS=4 ./hyper_threading_main' -n "hyper_threading_pr threads: 4" 'RAYON_NUM_THREADS=4 ./hyper_threading_pr'
Benchmark 1: hyper_threading_main threads: 4
  Time (mean ± σ):     12.770 s ±  0.096 s    [User: 43.259 s, System: 0.922 s]
  Range (min … max):   12.702 s … 12.839 s    2 runs
 
Benchmark 2: hyper_threading_pr threads: 4
  Time (mean ± σ):     12.282 s ±  0.664 s    [User: 43.648 s, System: 0.932 s]
  Range (min … max):   11.812 s … 12.751 s    2 runs
 
Summary
  hyper_threading_pr threads: 4 ran
    1.04 ± 0.06 times faster than hyper_threading_main threads: 4




hyperfine -r 2 -n "hyper_threading_main threads: 6" 'RAYON_NUM_THREADS=6 ./hyper_threading_main' -n "hyper_threading_pr threads: 6" 'RAYON_NUM_THREADS=6 ./hyper_threading_pr'
Benchmark 1: hyper_threading_main threads: 6
  Time (mean ± σ):     12.274 s ±  0.163 s    [User: 43.174 s, System: 0.995 s]
  Range (min … max):   12.158 s … 12.389 s    2 runs
 
Benchmark 2: hyper_threading_pr threads: 6
  Time (mean ± σ):     11.801 s ±  0.122 s    [User: 44.241 s, System: 1.055 s]
  Range (min … max):   11.715 s … 11.888 s    2 runs
 
Summary
  hyper_threading_pr threads: 6 ran
    1.04 ± 0.02 times faster than hyper_threading_main threads: 6




hyperfine -r 2 -n "hyper_threading_main threads: 8" 'RAYON_NUM_THREADS=8 ./hyper_threading_main' -n "hyper_threading_pr threads: 8" 'RAYON_NUM_THREADS=8 ./hyper_threading_pr'
Benchmark 1: hyper_threading_main threads: 8
  Time (mean ± σ):     11.952 s ±  0.356 s    [User: 44.304 s, System: 1.149 s]
  Range (min … max):   11.700 s … 12.204 s    2 runs
 
Benchmark 2: hyper_threading_pr threads: 8
  Time (mean ± σ):     12.116 s ±  0.101 s    [User: 44.296 s, System: 1.060 s]
  Range (min … max):   12.045 s … 12.188 s    2 runs
 
Summary
  hyper_threading_main threads: 8 ran
    1.01 ± 0.03 times faster than hyper_threading_pr threads: 8




hyperfine -r 2 -n "hyper_threading_main threads: 16" 'RAYON_NUM_THREADS=16 ./hyper_threading_main' -n "hyper_threading_pr threads: 16" 'RAYON_NUM_THREADS=16 ./hyper_threading_pr'
Benchmark 1: hyper_threading_main threads: 16
  Time (mean ± σ):     12.069 s ±  0.214 s    [User: 44.257 s, System: 1.113 s]
  Range (min … max):   11.918 s … 12.220 s    2 runs
 
Benchmark 2: hyper_threading_pr threads: 16
  Time (mean ± σ):     12.133 s ±  0.076 s    [User: 44.219 s, System: 1.101 s]
  Range (min … max):   12.080 s … 12.187 s    2 runs
 
Summary
  hyper_threading_main threads: 16 ran
    1.01 ± 0.02 times faster than hyper_threading_pr threads: 16

[github-actions]

Benchmark Results for unmodified programs 🚀

Command	Mean [s]	Min [s]	Max [s]	Relative
base big_factorial	2.496 ± 0.042	2.453	2.575	1.00 ± 0.02
head big_factorial	2.486 ± 0.035	2.444	2.553	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base big_fibonacci	2.457 ± 0.027	2.402	2.503	1.01 ± 0.02
head big_fibonacci	2.445 ± 0.044	2.382	2.537	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base blake2s_integration_benchmark	9.240 ± 0.142	9.102	9.597	1.00
head blake2s_integration_benchmark	9.308 ± 0.318	9.004	9.755	1.01 ± 0.04

Command	Mean [s]	Min [s]	Max [s]	Relative
base compare_arrays_200000	2.583 ± 0.025	2.555	2.647	1.01 ± 0.01
head compare_arrays_200000	2.554 ± 0.027	2.522	2.599	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base dict_integration_benchmark	1.725 ± 0.012	1.704	1.745	1.01 ± 0.03
head dict_integration_benchmark	1.706 ± 0.049	1.669	1.816	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base field_arithmetic_get_square_benchmark	1.444 ± 0.018	1.428	1.493	1.01 ± 0.02
head field_arithmetic_get_square_benchmark	1.428 ± 0.025	1.401	1.479	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base integration_builtins	9.285 ± 0.059	9.200	9.358	1.01 ± 0.01
head integration_builtins	9.193 ± 0.077	9.104	9.309	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base keccak_integration_benchmark	9.536 ± 0.132	9.417	9.870	1.00 ± 0.02
head keccak_integration_benchmark	9.492 ± 0.070	9.348	9.562	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base linear_search	2.611 ± 0.059	2.548	2.751	1.01 ± 0.04
head linear_search	2.595 ± 0.082	2.518	2.756	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base math_cmp_and_pow_integration_benchmark	1.766 ± 0.020	1.745	1.804	1.00 ± 0.03
head math_cmp_and_pow_integration_benchmark	1.759 ± 0.044	1.720	1.861	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base math_integration_benchmark	1.715 ± 0.027	1.689	1.764	1.02 ± 0.03
head math_integration_benchmark	1.686 ± 0.043	1.655	1.801	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base memory_integration_benchmark	1.454 ± 0.034	1.419	1.534	1.01 ± 0.04
head memory_integration_benchmark	1.435 ± 0.038	1.406	1.520	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base operations_with_data_structures_benchmarks	1.836 ± 0.026	1.815	1.905	1.01 ± 0.02
head operations_with_data_structures_benchmarks	1.810 ± 0.027	1.781	1.866	1.00

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
base pedersen	593.1 ± 5.8	585.7	605.4	1.01 ± 0.02
head pedersen	587.8 ± 7.2	581.1	603.9	1.00

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
base poseidon_integration_benchmark	708.4 ± 7.0	695.6	721.6	1.00 ± 0.02
head poseidon_integration_benchmark	707.5 ± 13.2	691.6	731.3	1.00

Command	Mean [s]	Min [s]	Max [s]	Relative
base secp_integration_benchmark	2.089 ± 0.024	2.063	2.124	1.01 ± 0.01
head secp_integration_benchmark	2.060 ± 0.018	2.038	2.099	1.00

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
base set_integration_benchmark	685.5 ± 7.4	677.8	703.8	1.00
head set_integration_benchmark	696.9 ± 3.2	692.7	702.1	1.02 ± 0.01

Command	Mean [s]	Min [s]	Max [s]	Relative
base uint256_integration_benchmark	5.067 ± 0.020	5.040	5.098	1.01 ± 0.02
head uint256_integration_benchmark	5.010 ± 0.111	4.923	5.241	1.00

[codecov]

Codecov Report

Attention: Patch coverage is 98.57143% with 2 lines in your changes missing coverage. Please review.

Project coverage is 96.34%. Comparing base (052e7ce) to head (921e633).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
vm/src/hint_processor/hint_processor_utils.rs	94.11%	1 Missing ⚠️
vm/src/serde/deserialize_utils.rs	98.11%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1898   +/-   ##
=======================================
  Coverage   96.33%   96.34%           
=======================================
  Files         102      102           
  Lines       40751    40830   +79     
=======================================
+ Hits        39259    39336   +77     
- Misses       1492     1494    +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[JulianGCalderon]

I think this is OK as it's a simple and fast fix, but we should create an issue to tackle the underlying problem later on: supporting arbitrary expressions like cairo-lang does. I think we would eventually need to define an AST for the expression.

[gabrielbosio]

Since the compiler simplifies the expression whenever it can, there would be no problem in let expressions which only use constants or even variables defined with the let keyword. However, when we use variables defined with the local keyword this is not the case anymore because in doing so we are introducing the use of references.

PR looks good but the description could be rephrased, I think it's not clear what's the issue. It can be simplified to something like:

The VM reads let expressions as c - tmp as an addition instead of a subtraction, where c is a constant and tmp is a local variable.

Then it can describe why this happens.
Also:

Thus, only supporting additions would make expressiones like this one: 17 - [fp] to equivalent to 17 + [fp] because the sign of the reference is never taken into account.

Is [fp] interpreted as a local variable?

[FrancoGiachetta]

The fp points to the first local declared, fp + 1 to the second and so on. I'll change the description to make it clearer.