feat: avoid generating a new witness when checking if linear expression is zero

[TomAFrench]

…sion is zero

Description

Problem*

Resolves

Summary*

We're currently creating a new witness when checking whether linear expressions are equal to zero. This is unnecessary as these will neatly fit into the following constraints (at least for direct scaling of a single witness)

Additional Context

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[github-actions]

Changes to circuit sizes

Generated at commit: 1058cd1d47bb4bee7acb578206c17b4c02597274, compared to commit: db28cb9ffb710c286b54dbfcf57292ae3dffb03d

🧾 Summary (10% most significant diffs)

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
conditional_1	-141 ✅	-4.70%	-141 ✅	-1.35%
references	-1 ✅	-12.50%	-1 ✅	-4.35%
type_aliases	-1 ✅	-16.67%	-1 ✅	-4.76%
brillig_conditional	-1 ✅	-16.67%	-1 ✅	-5.00%
negated_jmpif_condition	-1 ✅	-20.00%	-1 ✅	-5.00%
debug_logs	-1 ✅	-2.50%	-1 ✅	-5.26%

---

Full diff report 👇

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
bench_eddsa_poseidon	16,438 (-1)	-0.01%	19,585 (-1)	-0.01%
keccak256	123 (-2)	-1.60%	20,145 (-2)	-0.01%
nested_array_dynamic	3,351 (-2)	-0.06%	12,994 (-2)	-0.02%
embedded_curve_ops	22 (-1)	-4.35%	5,849 (-1)	-0.02%
nested_array_in_slice	958 (-1)	-0.10%	5,554 (-1)	-0.02%
hashmap	36,900 (-22)	-0.06%	98,580 (-22)	-0.02%
binary_operator_overloading	122 (-1)	-0.81%	4,262 (-1)	-0.02%
sha256_regression	22,342 (-44)	-0.20%	180,386 (-44)	-0.02%
regression_capacity_tracker	115 (-1)	-0.86%	3,970 (-1)	-0.03%
regression	292 (-1)	-0.34%	3,827 (-1)	-0.03%
conditional_regression_short_circuit	367 (-3)	-0.81%	11,139 (-3)	-0.03%
regression_mem_op_predicate	44 (-1)	-2.22%	3,567 (-1)	-0.03%
5_over	20 (-1)	-4.76%	3,488 (-1)	-0.03%
4_sub	13 (-1)	-7.14%	3,477 (-1)	-0.03%
xor	7 (-1)	-12.50%	3,146 (-1)	-0.03%
global_var_regression_simple	130 (-1)	-0.76%	3,068 (-1)	-0.03%
regression_4709	69 (-1)	-1.43%	2,916 (-1)	-0.03%
u16_support	69 (-1)	-1.43%	2,892 (-1)	-0.03%
arithmetic_binary_operations	15 (-1)	-6.25%	2,787 (-1)	-0.04%
array_dynamic	100 (-2)	-1.96%	3,718 (-2)	-0.05%
tuple_inputs	41 (-2)	-4.65%	3,647 (-2)	-0.05%
simple_comparison	22 (-2)	-8.33%	3,494 (-2)	-0.06%
brillig_not	22 (-2)	-8.33%	2,800 (-2)	-0.07%
conditional_2	17 (-2)	-10.53%	2,776 (-2)	-0.07%
bit_shifts_runtime	284 (-3)	-1.05%	3,945 (-3)	-0.08%
signed_division	200 (-3)	-1.48%	3,675 (-3)	-0.08%
sha256	1,436 (-24)	-1.64%	24,583 (-24)	-0.10%
wrapping_operations	31 (-3)	-8.82%	2,942 (-3)	-0.10%
regression_5252	32,862 (-54)	-0.16%	44,465 (-54)	-0.12%
bit_and	42 (-4)	-8.70%	3,058 (-4)	-0.13%
sha256_var_witness_const_regression	1,075 (-23)	-2.09%	16,477 (-23)	-0.14%
slice_dynamic_index	954 (-9)	-0.93%	6,345 (-9)	-0.14%
sha256_var_size_regression	10,192 (-93)	-0.90%	62,963 (-93)	-0.15%
bigint	1,031 (-31)	-2.92%	8,413 (-31)	-0.37%
u128	466 (-20)	-4.12%	4,425 (-20)	-0.45%
6_array	392 (-21)	-5.08%	4,031 (-21)	-0.52%
hash_to_field	454 (-30)	-6.20%	3,489 (-30)	-0.85%
sha2_byte	15,229 (-711)	-4.46%	81,964 (-711)	-0.86%
slices	678 (-37)	-5.17%	3,779 (-37)	-0.97%
ram_blowup_regression	140,756 (-7,680)	-5.17%	643,259 (-7,680)	-1.18%
conditional_1	2,860 (-141)	-4.70%	10,289 (-141)	-1.35%
references	7 (-1)	-12.50%	22 (-1)	-4.35%
type_aliases	5 (-1)	-16.67%	20 (-1)	-4.76%
brillig_conditional	5 (-1)	-16.67%	19 (-1)	-5.00%
negated_jmpif_condition	4 (-1)	-20.00%	19 (-1)	-5.00%
debug_logs	39 (-1)	-2.50%	18 (-1)	-5.26%

[github-actions]

Execution Memory Report

Program	Peak Memory	%
keccak256	74.67M	0%
workspace	123.79M	0%
regression_4709	316.00M	0%
ram_blowup_regression	512.61M	0%
rollup-base-public	478.82M	-1%
rollup-base-private	325.41M	0%
private-kernel-tail	180.42M	0%
private-kernel-reset	245.11M	0%
private-kernel-inner	208.50M	0%

[github-actions]

Compilation Memory Report

Program	Peak Memory
keccak256	78.13M
workspace	123.77M
regression_4709	424.09M
ram_blowup_regression	1.58G
rollup-base-public	4.85G
rollup-base-private	1.26G
private-kernel-tail	207.13M
private-kernel-reset	669.21M
private-kernel-inner	294.35M

[github-actions]

Execution Report

Program	Execution Time	%
sha256_regression	0.052s	-2%
regression_4709	0.001s	0%
ram_blowup_regression	0.602s	-6%
rollup-root	0.105s	0%
rollup-block-merge	0.105s	-1%
rollup-base-public	1.224s	-2%
rollup-base-private	0.450s	-2%
private-kernel-tail	0.019s	0%
private-kernel-reset	0.312s	-1%
private-kernel-inner	0.069s	-5%

[github-actions]

Compilation Report

Program	Compilation Time	%
sha256_regression	1.268s	0%
regression_4709	0.800s	-2%
ram_blowup_regression	17.900s	-1%
rollup-root	3.528s	5%
rollup-block-merge	3.392s	0%
rollup-base-public	41.200s	-3%
rollup-base-private	13.180s	1%
private-kernel-tail	1.020s	-5%
private-kernel-reset	6.896s	-5%
private-kernel-inner	2.202s	0%

[aakoshh]

Nice one, great reduction in bytecode size 👏

As I mentioned it would be nice to make the y_is_boolean_constraint expression more fluid, but otherwise looks correct 👍