fix: remove unused fields in rvstate

[mansur20478]

removes unused reservation_addr, reservation_valid, brk, start_brk, csrs fields from RvState.

[shuklaayush]

@claude @codex there might've been a performance reason why we added this. do we not need it anymore?

    /// Alignment padding.
    _pad0: u8,

[claude]

Claude finished @shuklaayush's task in 1m 48s —— View job

---

Investigating _pad0 alignment padding field

• Read the current state of RvState and the PR diff
• Investigate git history for why _pad0 was added
• Analyze whether the padding is still needed for performance/alignment

Answer

The _pad0: u8 field is not needed anymore and is safe to remove. It was explicit alignment padding, not a performance optimization.

Why it existed (old layout):

In the old struct, the field sequence around _pad0 was:

reservation_addr: u32    // offset 280, size 4
reservation_valid: u8    // offset 284, size 1
has_exited: u8           // offset 285, size 1
exit_code: u8            // offset 286, size 1
_pad0: u8                // offset 287, size 1  ← explicit pad
brk: u32                 // offset 288, size 4  ← needs 4-byte alignment
start_brk: u32           // offset 292, size 4
memory: *mut u8          // offset 296, size 8

After three consecutive u8 fields (reservation_valid, has_exited, exit_code), we're at offset 287. The next field brk: u32 needs 4-byte alignment (offset 288). _pad0 was an explicit padding byte to keep the Rust repr(C) layout in exact sync with the handwritten C struct, ensuring no implicit padding surprises in the FFI boundary.

Why it's safe to remove (new layout):

With reservation_addr, reservation_valid, brk, start_brk, and csrs all gone, the layout becomes:

target_instret: u64      // offset 272, size 8
has_exited: u8           // offset 280, size 1
exit_code: u8            // offset 281, size 1
// (6 bytes implicit padding — repr(C) aligns to next field's requirement)
memory: *mut u8          // offset 288, size 8  ← needs 8-byte alignment

The compiler handles the padding between exit_code and memory automatically in both repr(C) Rust and C. Both the Rust struct and C struct agree on the layout without any explicit padding field.

tl;dr: _pad0 existed to align the now-removed brk: u32 field. No performance concern — just FFI layout matching.

---

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. What shall we delve into next?

Reviewed commit: 4813b4b58b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[shuklaayush]

lgtm. let's merge this after the parent is merged

[github-actions]

group	app.proof_time_ms	app.cycles	leaf.proof_time_ms
fibonacci	1,661	4,000,051	527
keccak	16,369	14,365,133	3,039
sha2_bench	10,404	11,167,961	1,946
regex	1,543	4,090,656	432
ecrecover	482	112,210	312
pairing	627	592,827	304
kitchen_sink	3,934	1,979,971	867

Note: cells_used metrics omitted because CUDA tracegen does not expose unpadded trace heights.

Commit: 0924674

Benchmark Workflow