feat(experimental): Always inline certain methods in sha

[vezenovm]

Description

Problem*

Optimizing bytecode sizes and execution traces in Brillig

Summary*

I know we are deprecating sha256 in the stdlib and moving to use noir-lang/sha256. I have replicated this PR here noir-lang/sha256#13. I am just pushing this draft for now as we have more complete benchmarking and visualization on this repo.

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 number of Brillig opcodes executed

Generated at commit: 60164846b3715aedfc09fc93e2d5a84d445b5621, compared to commit: 97afa52f5212be2d05af26b9e8dde9c3ea7a1d2e

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
sha256_var_size_regression_inliner_zero	-5,164 ✅	-24.23%
sha256_var_padding_regression_inliner_min	-57,130 ✅	-24.67%
sha256_var_padding_regression_inliner_zero	-56,770 ✅	-24.67%
sha256_regression_inliner_min	-42,603 ✅	-27.10%
sha256_regression_inliner_zero	-42,191 ✅	-27.19%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
sha2_byte_inliner_min	88,444 (-329)	-0.37%
sha2_byte_inliner_zero	71,391 (-280)	-0.39%
array_dynamic_nested_blackbox_input_inliner_zero	4,199 (-252)	-5.66%
array_dynamic_nested_blackbox_input_inliner_min	4,607 (-298)	-6.08%
sha256_inliner_zero	10,547 (-700)	-6.22%
sha256_var_witness_const_regression_inliner_zero	6,431 (-448)	-6.51%
conditional_1_inliner_zero	5,088 (-381)	-6.97%
conditional_1_inliner_min	5,509 (-429)	-7.22%
sha256_inliner_min	11,274 (-909)	-7.46%
sha256_var_witness_const_regression_inliner_min	7,092 (-580)	-7.56%
conditional_regression_short_circuit_inliner_zero	4,110 (-370)	-8.26%
6_inliner_zero	4,030 (-370)	-8.41%
conditional_regression_short_circuit_inliner_min	4,490 (-429)	-8.72%
6_inliner_min	4,384 (-429)	-8.91%
regression_4449_inliner_min	235,513 (-29,820)	-11.24%
regression_4449_inliner_zero	198,065 (-26,042)	-11.62%
ecdsa_secp256k1_inliner_min	6,908 (-1,241)	-15.23%
ecdsa_secp256k1_inliner_zero	6,378 (-1,192)	-15.75%
brillig_cow_regression_inliner_min	456,919 (-86,030)	-15.84%
brillig_cow_regression_inliner_zero	453,386 (-85,966)	-15.94%
ram_blowup_regression_inliner_min	685,985 (-131,322)	-16.07%
ram_blowup_regression_inliner_zero	681,794 (-131,255)	-16.14%
array_dynamic_blackbox_input_inliner_min	18,600 (-3,664)	-16.46%
array_dynamic_blackbox_input_inliner_zero	17,644 (-3,533)	-16.68%
sha256_brillig_performance_regression_inliner_zero	21,526 (-4,369)	-16.87%
sha256_brillig_performance_regression_inliner_min	21,842 (-4,434)	-16.87%
sha256_var_size_regression_inliner_min	16,784 (-5,282)	-23.94%
sha256_var_size_regression_inliner_zero	16,147 (-5,164)	-24.23%
sha256_var_padding_regression_inliner_min	174,492 (-57,130)	-24.67%
sha256_var_padding_regression_inliner_zero	173,353 (-56,770)	-24.67%
sha256_regression_inliner_min	114,591 (-42,603)	-27.10%
sha256_regression_inliner_zero	112,961 (-42,191)	-27.19%

[github-actions]

Changes to Brillig bytecode sizes

Generated at commit: 60164846b3715aedfc09fc93e2d5a84d445b5621, compared to commit: 97afa52f5212be2d05af26b9e8dde9c3ea7a1d2e

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
sha256_var_witness_const_regression_inliner_zero	-14 ✅	-1.76%
conditional_1_inliner_zero	-27 ✅	-2.44%
ecdsa_secp256k1_inliner_zero	-27 ✅	-3.00%
array_dynamic_nested_blackbox_input_inliner_zero	-27 ✅	-3.08%
regression_4449_inliner_zero	-37 ✅	-4.89%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
sha256_regression_inliner_zero	4,829 (-6)	-0.12%
sha256_var_padding_regression_inliner_zero	2,910 (-6)	-0.21%
sha256_regression_inliner_min	5,256 (-17)	-0.32%
sha256_inliner_zero	1,226 (-6)	-0.49%
sha256_var_padding_regression_inliner_min	3,175 (-17)	-0.53%
sha2_byte_inliner_min	2,946 (-17)	-0.57%
brillig_cow_regression_inliner_min	2,314 (-17)	-0.73%
brillig_cow_regression_inliner_zero	2,059 (-16)	-0.77%
sha256_brillig_performance_regression_inliner_zero	1,641 (-15)	-0.91%
sha256_brillig_performance_regression_inliner_min	1,791 (-17)	-0.94%
conditional_1_inliner_min	1,390 (-17)	-1.21%
sha256_inliner_min	1,368 (-17)	-1.23%
sha2_byte_inliner_zero	2,152 (-27)	-1.24%
sha256_var_size_regression_inliner_zero	1,071 (-14)	-1.29%
array_dynamic_blackbox_input_inliner_min	1,244 (-17)	-1.35%
sha256_var_size_regression_inliner_min	1,224 (-17)	-1.37%
array_dynamic_blackbox_input_inliner_zero	994 (-14)	-1.39%
ecdsa_secp256k1_inliner_min	1,189 (-17)	-1.41%
ram_blowup_regression_inliner_zero	908 (-13)	-1.41%
array_dynamic_nested_blackbox_input_inliner_min	1,176 (-17)	-1.42%
ram_blowup_regression_inliner_min	1,097 (-17)	-1.53%
conditional_regression_short_circuit_inliner_min	1,087 (-17)	-1.54%
conditional_regression_short_circuit_inliner_zero	894 (-14)	-1.54%
regression_4449_inliner_min	1,042 (-17)	-1.61%
6_inliner_min	986 (-17)	-1.69%
6_inliner_zero	812 (-14)	-1.69%
sha256_var_witness_const_regression_inliner_min	960 (-17)	-1.74%
sha256_var_witness_const_regression_inliner_zero	780 (-14)	-1.76%
conditional_1_inliner_zero	1,079 (-27)	-2.44%
ecdsa_secp256k1_inliner_zero	874 (-27)	-3.00%
array_dynamic_nested_blackbox_input_inliner_zero	851 (-27)	-3.08%
regression_4449_inliner_zero	720 (-37)	-4.89%

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Test Suite Duration'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 99beecc	Previous: 97afa52	Ratio
noir-lang_noir_bigcurve_	311 s	252 s	1.23

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench

[vezenovm]

Closing as noir-lang/sha256#13 is merged

[TomAFrench]

I've rolled these changes into the 0.1.2 release which is now being used in Aztec in AztecProtocol/aztec-packages#11696