feat(starknet_os): pack usize in felts

[yoavGrs]

No description provided.

[reviewable-StarkWare]

This change is 

[yoavGrs]

• feat(starknet_os): implement set decompressed dst hint #4666
• feat(starknet_os): implement compression hint #4660
• feat(starknet_os): stateless compression util #4585
• feat(starknet_os): bucket element of 252 bits is felt #4642
• feat(starknet_os): pack usize in felts #4573 👈 (View in Graphite)
• feat(starknet_os): compression set struct #4397 : 1 other dependent PR (#4398 )
• feat(starknet_os): buckets struct #4570
• feat(starknet_os): unique value bucket #4568
• feat(starknet_os): enum for sized bucket elements #4546
• feat(starknet_os): trait for unique bucket element #4545
• feat(starknet_os): sized bits vector #4391 : 1 other dependent PR (#4395 )
• feat(starknet_os): bit length enum for unique buckets #4517
• chore(starknet_os): stateless compression directory #4390
• main

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

[github-actions]

Artifacts upload workflows:

• Started at 2025-03-02T14:28:46Z
• Started at 2025-03-02T16:25:48Z
• Started at 2025-03-04T08:00:29Z
• Started at 2025-03-04T09:07:43Z
• Started at 2025-03-04T11:18:07Z
• Started at 2025-03-04T12:06:13Z
• Started at 2025-03-04T12:53:13Z

[github-actions]

Benchmark movements:
full_committer_flow performance regressed!
full_committer_flow time: [30.426 ms 30.502 ms 30.605 ms]
change: [+1.0206% +1.3006% +1.6629%] (p = 0.00 < 0.05)
Performance has regressed.
Found 4 outliers among 100 measurements (4.00%)
3 (3.00%) high mild
1 (1.00%) high severe

[dorimedini-starkware]

Reviewed 4 of 4 files at r1, all commit messages.
Reviewable status: all files reviewed, 4 unresolved discussions (waiting on @Yoni-Starkware)

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/tests.rs line 121 at r1 (raw file):

    let unpacked = unpack_felts_to_usize(packed.as_ref(), nums.len(), elm_bound);
    assert_eq!(nums, unpacked);
}

edge cases please:

1. Felt::MAX
2. powers of two
3. powers of two - 1
4. powers of two + 1

Code quote:

#[test]
fn test_usize_pack_and_unpack() {
    let nums = vec![34, 0, 11111, 1034, 3404];
    let elm_bound = 12345;
    let packed = pack_usize_in_felts(&nums, elm_bound);
    let unpacked = unpack_felts_to_usize(packed.as_ref(), nums.len(), elm_bound);
    assert_eq!(nums, unpacked);
}

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 415 at r1 (raw file):

}

/// Calculates the number of elements that can fit in a single felt value, given the element bound.

update docstring (I think there was a discussion on a previous iteration of this PR)

Code quote:

/// Calculates the number of elements that can fit in a single felt value, given the element bound.

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 419 at r1 (raw file):

    let n_bits_required =
        max(1, usize::try_from(usize::BITS - elm_bound.leading_zeros()).expect("usize overflow"));
    if n_bits_required > MAX_N_BITS {

can this ever happen?
isn't the correct answer 0 in this case?
consider making the input a Felt, not a usize, so that

1. we can give an elm_bound of more than usize::MAX (which is much less then 256 bits - or even 128 bits - unless running on a CPU from the future)
2. if you have Felt as input, converting to biguint and counting bits always results in at most MAX_N_BITS, right?

if (1) is never required (i.e. elm_bounds is always at most 128 bits), please change the type to u128 (or u64, smallest possible); usize depends on architecture and unclear from reading that the bits are always within allowed limit

Code quote:

if n_bits_required > MAX_N_BITS

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 434 at r1 (raw file):

/// Packs a chunk of elements into a single felt.
pub fn pack_usize_in_felt(elms: &[usize], elm_bound: usize) -> Felt {

this function can fail if you provide invalid input; make it private please

Suggestion:

fn pack_usize_in_felt

[yoavGrs]

Reviewable status: 1 of 4 files reviewed, 4 unresolved discussions (waiting on @dorimedini-starkware and @Yoni-Starkware)

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/tests.rs line 121 at r1 (raw file):

Previously, dorimedini-starkware wrote…

edge cases please:

1. Felt::MAX
2. powers of two
3. powers of two - 1
4. powers of two + 1

It for usize, Felt::MAX is too large.

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 415 at r1 (raw file):

Previously, dorimedini-starkware wrote…

update docstring (I think there was a discussion on a previous iteration of this PR)

What's wrong with the current doc?

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 419 at r1 (raw file):

Previously, dorimedini-starkware wrote…

can this ever happen?
isn't the correct answer 0 in this case?
consider making the input a Felt, not a usize, so that

1. we can give an elm_bound of more than usize::MAX (which is much less then 256 bits - or even 128 bits - unless running on a CPU from the future)
2. if you have Felt as input, converting to biguint and counting bits always results in at most MAX_N_BITS, right?

if (1) is never required (i.e. elm_bounds is always at most 128 bits), please change the type to u128 (or u64, smallest possible); usize depends on architecture and unclear from reading that the bits are always within allowed limit

The upper bound is 20 bits.

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 434 at r1 (raw file):

Previously, dorimedini-starkware wrote…

this function can fail if you provide invalid input; make it private please

Done.

[dorimedini-starkware]

Reviewed 3 of 3 files at r2, all commit messages.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @yoavGrs and @Yoni-Starkware)

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 415 at r1 (raw file):

Previously, yoavGrs wrote…

What's wrong with the current doc?

as yoni mentioned, this function doesn't exactly return the number of elements up to the bound that can be packed - the packing is assuming an actual bound of ceil(log2(elm_bound)) bits

[yoavGrs]

Reviewable status: 1 of 4 files reviewed, 1 unresolved discussion (waiting on @dorimedini-starkware and @Yoni-Starkware)

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 415 at r1 (raw file):

Previously, dorimedini-starkware wrote…

as yoni mentioned, this function doesn't exactly return the number of elements up to the bound that can be packed - the packing is assuming an actual bound of ceil(log2(elm_bound)) bits

Changed.

[dorimedini-starkware]

Reviewed 3 of 3 files at r3, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @Yoni-Starkware)

[Yoni-Starkware]

Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @yoavGrs)

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 426 at r3 (raw file):

        return MAX_N_BITS;
    }
    let n_bits_required = 32 - elm_bound.leading_zeros();

Why not elm_bound.ilog2() + 1?
Seems safer. You are using here implicitly that elm_bound is a u32.

Code quote:

let n_bits_required = 32 - elm_bound.leading_zeros();

[yoavGrs]

Reviewable status: 3 of 4 files reviewed, 1 unresolved discussion (waiting on @dorimedini-starkware and @Yoni-Starkware)

---

crates/starknet_os/src/hints/hint_implementation/stateless_compression/utils.rs line 426 at r3 (raw file):

Previously, Yoni-Starkware (Yoni) wrote…

Why not elm_bound.ilog2() + 1?
Seems safer. You are using here implicitly that elm_bound is a u32.

Done.

[dorimedini-starkware]

Reviewed 1 of 1 files at r4, all commit messages.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @Yoni-Starkware)

[dorimedini-starkware]

Reviewed 1 of 1 files at r5, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @yoavGrs)

[dorimedini-starkware]

Reviewed 1 of 1 files at r6, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @yoavGrs)

[Yoni-Starkware]

Reviewed 1 of 4 files at r1, 1 of 1 files at r4, 1 of 1 files at r5, 1 of 1 files at r6, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @yoavGrs)

[Yoni-Starkware]

Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @yoavGrs)