Release main-d9fd108

Fix: Restore decimal precision in summary MHz values (#32)

## Summary

The benchmark summary table and JSON output were truncating MHz values
to whole numbers (e.g., `20` instead of `20.596`). This was caused by
the `BenchSummary` struct fields being `u32` with explicit `as u32`
casts in `get_bench_summary()`.

## Changes

- Changed all `BenchSummary` MHz fields from `u32` → `f64` (and
`Option<u32>` → `Option<f64>`)
- Removed the `as u32` truncating casts in both branches of
`get_bench_summary()`

## Before

```
│ Exec Min MHz     │ 18 │
│ Exec Avg MHz     │ 20 │
│ Prove Min MHz    │  1 │
```

## After

```
│ Exec Min MHz     │ 18.807 │
│ Exec Avg MHz     │ 20.487 │
│ Prove Min MHz    │ 1.874  │
```

Closes #31

Release main-0d4ec71

Fix: Restore decimal precision in summary MHz values (#32)

Release main-b12573b

Change units from KHz to MHz (#29)

Closes #28

## Summary

Switch the benchmark output unit from KHz to MHz for all cycle-based
throughput measurements. This aligns with the magnitude typically used
when discussing prover performance.

## Changes

- **`prove_stark`**: Divide by `1_000_000.0` instead of `1000.0` to
produce MHz values. The variable name is updated from `khz` to `mhz`.
- **`BenchResult`**: Rename `exec_khz` → `exec_mhz`, `prove_khz` →
`prove_mhz`. Update `tabled` column headers to `"Exec MHz"` and `"Prove
MHz"`.
- **`BenchSummary`**: Rename all `*_khz` fields to `*_mhz`. Update
column headers accordingly.
- **`prove_snark` comment**: Updated `KHz` → `MHz` in the PostgreSQL
comment.

## Testing

- `cargo build --release -p bento-bench` — passes
- `RISC0_SKIP_BUILD=1 cargo clippy -p bento-bench` — no new warnings
- `RISC0_SKIP_BUILD=1 cargo test -p bento-bench` — 3/3 tests pass

Release main-1401c2e

feat: add --fetch flag for direct benchmark suite download (#26)

## Summary        
- Add `--fetch <url>` flag to `run` subcommand that downloads a
`.tar.zst` benchmark suite, extracts it to a
- Streaming decompression (no disk buffering of the archive — memory
efficient)
- Auto-detect single-subdirectory wrapping (handles archives that wrap
content in a subdirectory)
- Mutually exclusive with `--data-dir` (conflicts_with enforced by clap)
- 3 unit tests covering flat extraction, subdirectory extraction, and
bad URL rejection
 - Clippy clean, all tests pass             
                                                                       
Closes #14 

## Test Plan
- [x] `cargo test -p bento-bench` — 3/3 pass
- [x] `cargo clippy -p bento-bench -- -D warnings` — zero warnings
- [ ] Manual end-to-end: `bento-bench run --fetch <real-url>
--iterations 1`

---------

Co-authored-by: zeroklaw <267418690+zeroklaw@users.noreply.github.com>

Release main-43a2809

Fail on run if no manifest exists (#25)

Release main-69df2f1

Add license (#22)

Closes #21

Release main-c63bd17

Spring cleaning (#20)

Closes #18, Closes #19. 

Additionally:
- Add details to README, cleanup a bit
- Introduce AGENTS.md
- Enable docker builds on PR (no push); add cargo fmt check
- Fix minor style issues

Release main-694fc96

Minor fixes (#17)

- Clippy suggestions
- Update repo name
- Add tagline to readme

Release main-5ff2b34

Add snark support (#16)

Release main-dca84ee

Manifest fixes (#12)

- Closes #11 
- Closes #10 
- Closes #9 
- All suites have been migrated and re-uploaded