starknet_os: os resources test - support virtual builtins, add sha256

[dorimedini-starkware]

No description provided.

[dorimedini-starkware]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• starknet_os: os resources test - add send_message_to_l1 #14143 : 2 dependent PRs (#14144 , #14153 )
• starknet_os: os resources test - add replace_class #14142
• starknet_os: os resources test - support virtual builtins, add sha256 #14141 👈 (View in Graphite)
• starknet_os: define sha256 batch resources constants #14140
• starknet_os: os resources test - add keccak #14139
• starknet_os: os resources test - add 3 more syscalls #14138
• starknet_os: os resources test - remove all fee transfer syscalls #14137
• starknet_os: os resources test - add emit event #14136
• starknet_os: os resources test - add meta tx #14135
• starknet_os: os resources test - measure linear factor, add deploy #14134
• starknet_os: os resources test - library call #14129
• starknet_os: initial test, measure call contract resources #14128
• blockifier_test_utils: add option to deploy from zero in test accounts #14258
• starknet_os_flow_tests: helper function to declare explicit classes #14257
• starknet_os: os resources test - create fixed deployable account #14173
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[reviewable-StarkWare]

This change is 

[cursor]

PR Summary

Medium Risk
Touches versioned OS program hashes, syscall gas costs, and fee/resource accounting used by blockifier and proving; incorrect virtual-builtin math would misreport Sha256 OS overhead in regression tests and constants.

Overview
Enables OS syscall resource regression for Sha256ProcessBlock by treating it as a virtual builtin whose batch work runs outside OsLogger, and rebases OS / fee constants after the OS rebuild.

The flow test harness adds SYSCALLS_WITH_VIRTUAL_BUILTINS and update_resources_for_virtual_builtin_syscall, which adds a per-block share of SHA256_BATCH_RESOURCES_LINEAR_UNSCALED (using SHA256_BATCH_SIZE = 7) to measured syscall overhead. Sha256ProcessBlock is removed from UNMEASURABLE_SYSCALLS, and OsResourcesTestContract now invokes sha256_process_block_syscall during __execute__.

Constants and expectations move together: allowed virtual OS program hash slot 2, program_hash.json (os / virtual_os), SHA256_PROCESS_BLOCK_GAS_COST (839995), Sha256ProcessBlock n_steps 1854, blockifier sha256 syscall test gas (850445), and the 0.14.3→0.14.4 diff regression file. starknet_os test utils expose unscaled vs scaled SHA256 batch linear resources and make BUILTIN_INSTANCE_SIZES public for sharing with flow tests.

^{Reviewed by Cursor Bugbot for commit 07d79f0. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 035e95f. Configure here.}

[yoavGrs]

@yoavGrs reviewed 8 files and made 4 comments.
Reviewable status: 8 of 10 files reviewed, 4 unresolved discussions (waiting on dorimedini-starkware and Yoni-Starkware).

---

crates/starknet_os/src/test_utils.rs line 22 at r2 (raw file):

// Resources consumed by the SHA-256 batch phase, separated into linear and constant factors.
pub const SHA256_BLOCK_TO_ROUND: usize = 7;

Or how you called it in the previous PR.

Suggestion:

SHA256_BLOCKS_IN_ROUND

---

crates/starknet_os_flow_tests/src/os_resources_test.rs line 114 at r2 (raw file):

                n_steps: linear_steps,
                builtin_instance_counter: linear_builtin_instance_counter,
                n_memory_holes: linear_memory_holes,

Please

Suggestion:

                round_n_steps: linear_steps,
                round_builtin_instance_counter: linear_builtin_instance_counter,
                round_n_memory_holes: linear_memory_holes,

---

crates/starknet_os_flow_tests/src/os_resources_test.rs line 120 at r2 (raw file):

            for (builtin, count) in linear_builtin_instance_counter.iter() {
                *new_resources.builtin_instance_counter.entry(*builtin).or_insert(0) +=
                    BUILTIN_INSTANCE_SIZES.get(builtin).unwrap() * count / SHA256_BLOCK_TO_ROUND;

In SHA256_BATCH_RESOURCES_LINEAR the count was the number of instances, now you count memory cells.
There, the resources are divided by the builtin size, and here you multiply by the same number.

Code quote:

BUILTIN_INSTANCE_SIZES.get(builtin).unwrap() * count

---

crates/blockifier/resources/blockifier_versioned_constants_0_14_4.json line 468 at r2 (raw file):

                "builtin_instance_counter": {
                    "range_check_builtin": 65,
                    "bitwise_builtin": 1114

Number of builtins changes.

Code quote:

"bitwise_builtin": 1114

[yoavGrs]

@yoavGrs reviewed 2 files and all commit messages.
Reviewable status: all files reviewed, 4 unresolved discussions (waiting on dorimedini-starkware and Yoni-Starkware).

[dorimedini-starkware]

@dorimedini-starkware made 4 comments.
Reviewable status: all files reviewed, 4 unresolved discussions (waiting on yoavGrs and Yoni-Starkware).

---

crates/blockifier/resources/blockifier_versioned_constants_0_14_4.json line 468 at r2 (raw file):

Previously, yoavGrs wrote…

Number of builtins changes.

keep this blocking until this PR is close to main, I'll rerun the numbers...

---

crates/starknet_os/src/test_utils.rs line 22 at r2 (raw file):

Previously, yoavGrs wrote…

Or how you called it in the previous PR.

changed to batch size, to match the cairo-common terminology

---

crates/starknet_os_flow_tests/src/os_resources_test.rs line 114 at r2 (raw file):

Previously, yoavGrs wrote…

Please

Done.

---

crates/starknet_os_flow_tests/src/os_resources_test.rs line 120 at r2 (raw file):

Previously, yoavGrs wrote…

In SHA256_BATCH_RESOURCES_LINEAR the count was the number of instances, now you count memory cells.
There, the resources are divided by the builtin size, and here you multiply by the same number.

correct, this is for visibility - the numbers in cairo-common are given in builtin instances. I hard-code the numbers scaled thus so that they can be compared to the cairo-common impl

[yoavGrs]

@yoavGrs reviewed 8 files and all commit messages, and resolved 3 discussions.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on Yoni-Starkware).