Merge branch 'main' into feat/merged-bitwise-bus

Author: jotabulacios

.github/actions/free-disk/action.yaml

@@ -5,5 +5,6 @@ runs:
   using: "composite"
   steps:
     - name: Remove .NET, Android SDK, Haskell
+      if: runner.environment == 'github-hosted'
       shell: bash
       run: sudo rm -rf /usr/share/dotnet/ /usr/local/lib/android /opt/ghc /usr/local/.ghcup

.github/workflows/bench-vs-nightly.yml

@@ -0,0 +1,58 @@
+name: Bench Vs Nightly
+
+on:
+  schedule:
+    # 03:00 America/Argentina/Buenos_Aires = 06:00 UTC
+    - cron: "0 6 * * *"
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: bench-vs-nightly-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  bench-vs:
+    runs-on: [self-hosted, bench]
+    timeout-minutes: 720
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Rust Environment
+        uses: ./.github/actions/setup-rust
+
+      - name: Add cargo to PATH
+        run: echo "$HOME/.cargo/bin" >> "$GITHUB_PATH"
+
+      - name: Add SP1 to PATH
+        run: echo "$HOME/.sp1/bin" >> "$GITHUB_PATH"
+
+      - name: Install SP1 toolchain
+        run: |
+          export PATH="$HOME/.cargo/bin:$HOME/.sp1/bin:$PATH"
+          if ! cargo prove --version 2>/dev/null | grep -qE '\b6\.0\.1\b'; then
+            curl -L https://sp1up.succinct.xyz | bash
+            export PATH="$HOME/.sp1/bin:$PATH"
+            sp1up --version v6.0.1
+          fi
+          cargo prove --version
+
+      - name: Run nightly benchmark
+        run: |
+          bash ./bench_vs/run.sh \
+            --steps 1000000 2000000 4000000 8000000 \
+            --report-dir bench_vs_artifacts \
+            --no-color
+
+      - name: Upload nightly benchmark artifact
+        uses: actions/upload-artifact@v4
+        with:
+          name: bench-vs-nightly-${{ github.run_number }}-${{ github.sha }}
+          path: bench_vs_artifacts
+          retention-days: 90
+
+      - name: Publish summary
+        run: cat bench_vs_artifacts/summary.md >> "$GITHUB_STEP_SUMMARY"

.github/workflows/pr_main.yaml

@@ -73,7 +73,7 @@ jobs:
           path: |
             executor/program_artifacts/rust
             executor/shared_target
-          key: rust-elf-artifacts-${{ hashFiles('executor/programs/rust/**', 'executor/programs/riscv64im-lambda-vm-elf.json', 'syscalls/**') }}
+          key: rust-elf-artifacts-${{ hashFiles('executor/programs/rust/**', 'executor/programs/riscv64im-lambda-vm-elf.json', 'syscalls/**', 'Makefile') }}
           restore-keys: |
             rust-elf-artifacts-
 
@@ -187,6 +187,26 @@ jobs:
         run: |
           make compile-programs-asm
 
+      - name: Cache compiled Rust ELF artifacts and build cache
+        id: cache-rust-elfs
+        uses: actions/cache@v4
+        with:
+          path: |
+            executor/program_artifacts/rust
+            executor/shared_target
+          key: rust-elf-artifacts-${{ hashFiles('executor/programs/rust/**', 'executor/programs/riscv64im-lambda-vm-elf.json', 'syscalls/**', 'Makefile') }}
+          restore-keys: |
+            rust-elf-artifacts-
+
+      - name: Setup Rust Environment
+        if: steps.cache-rust-elfs.outputs.cache-hit != 'true'
+        uses: ./.github/actions/setup-rust
+
+      - name: Compile Rust programs to ELF
+        if: steps.cache-rust-elfs.outputs.cache-hit != 'true'
+        run: |
+          make compile-programs-rust
+
       - name: Install nextest
         uses: taiki-e/install-action@v2
         with:
@@ -229,6 +249,26 @@ jobs:
         run: |
           make compile-programs-asm
 
+      - name: Cache compiled Rust ELF artifacts and build cache
+        id: cache-rust-elfs
+        uses: actions/cache@v4
+        with:
+          path: |
+            executor/program_artifacts/rust
+            executor/shared_target
+          key: rust-elf-artifacts-${{ hashFiles('executor/programs/rust/**', 'executor/programs/riscv64im-lambda-vm-elf.json', 'syscalls/**', 'Makefile') }}
+          restore-keys: |
+            rust-elf-artifacts-
+
+      - name: Setup Rust Environment
+        if: steps.cache-rust-elfs.outputs.cache-hit != 'true'
+        uses: ./.github/actions/setup-rust
+
+      - name: Compile Rust programs to ELF
+        if: steps.cache-rust-elfs.outputs.cache-hit != 'true'
+        run: |
+          make compile-programs-rust
+
       - name: Install nextest
         uses: taiki-e/install-action@v2
         with:

bench_vs/README.md

@@ -0,0 +1,98 @@
+# Lambda VM vs SP1 v6 Benchmark
+
+Compares proving time for an identical u64 wrapping Fibonacci computation.
+
+## Prerequisites
+
+1. **Lambda VM CLI** (built from this repo):
+   ```bash
+   cargo build --release -p cli
+   ```
+
+2. **SP1 toolchain** (Succinct's prover):
+   ```bash
+   curl -L https://sp1up.succinct.xyz | bash
+   sp1up
+   ```
+
+3. **Rust nightly** (for cross-compiling Lambda VM guest):
+   ```bash
+   rustup toolchain install nightly
+   ```
+
+## Usage
+
+```bash
+# Default series: 1k, 10k, 100k, 300k iterations
+./bench_vs/run.sh
+
+# Custom series
+./bench_vs/run.sh -n 1000 50000
+
+# Approximate workload steps (converted with 5 steps/iteration)
+./bench_vs/run.sh --steps 1000000 2000000 4000000 8000000
+
+# Project to a target cycle count
+./bench_vs/run.sh --target-cycles 500000000
+
+# Run only one prover
+./bench_vs/run.sh --lambda-only
+./bench_vs/run.sh --sp1-only
+```
+
+## What is measured
+
+Both provers execute the same program: iterative Fibonacci with `u64::wrapping_add`.
+
+The timing window on both sides is **end-to-end single-shot proving, with no
+verification and no recursion/compression**. Concretely:
+
+| Phase                                      | Lambda VM timer | SP1 v6 timer |
+|--------------------------------------------|:---------------:|:------------:|
+| Read ELF + input from disk                 |        ❌       |       ❌     |
+| Pre-pass execution to count cycles         |        ❌       |       ❌     |
+| `setup` / verifying-key derivation         |  N/A (none)     |       ✅     |
+| ELF parse + guest execution (inside prove) |        ✅       |       ✅     |
+| Trace build                                |        ✅       |       ✅     |
+| AIR construction                           |        ✅       |       ✅     |
+| STARK prove (`core` mode)                  |        ✅       |       ✅     |
+| Proof serialization / write                |        ❌       |       ❌     |
+| Verify                                     |        ❌       |       ❌     |
+
+Both sides run one extra execution pass **outside** the timer to report dynamic
+instruction counts (SP1's `execute(...)` / Lambda's executor pre-pass). This
+costs wall-clock time in the CI job but does not inflate the measured proving
+time, and the cost is symmetric between the two provers.
+
+Lambda VM uses the default proof options from `prover::prove_with_inputs`
+(`GoldilocksCubicProofOptions::with_blowup(2)`, 50 FRI queries). SP1 v6 uses
+the `core` proof mode exposed by `sp1-sdk::ProverClient::from_env()`.
+
+## Projection axis
+
+The linear projection uses **measured cycles** per prover — Lambda's executor
+log count and SP1's `report.total_instruction_count()`. For Fibonacci the two
+values agree to within ~1% (both compile to the same inner loop shape on
+RISC-V). When cycle data is missing, the script falls back to the approximate
+`target_workload_steps ~= 5 * n` label that was passed on the command line.
+
+## Output
+
+```
+=== Summary ===
+Program: Fibonacci (u64 wrapping)
+
+  Target steps  Iterations      Lambda (s)   Lambda cycles         SP1 (s)      SP1 cycles     Ratio
+  ------------  ----------      ----------   -------------         -------      ----------     -----
+  1000000       200000              ...s         1004794             ...s         1004794       ...
+  2000000       400000              ...s         2004794             ...s         2004794       ...
+
+Timing window covers single-shot end-to-end proving; SP1 includes setup; both exclude verification.
+Green ratio = Lambda VM faster, Red = SP1 faster
+```
+
+With `--report-dir DIR` the script writes:
+- `results.tsv` — raw per-run data (`target_steps`, `iterations`, `lambda_time_s`, `lambda_axis_value`, `lambda_cycles`, `sp1_time_s`, `sp1_axis_value`, `sp1_cycles`, `ratio`).
+- `metrics.txt` — key=value pairs including `timing_window=setup_plus_end_to_end_prove_no_verify`.
+- `summary.md` — the same table plus linear projection to `TARGET_CYCLES` cycles.
+- `raw/` — stdout/stderr of every individual run.

bench_vs/lambda/fibonacci/.cargo/config.toml

@@ -0,0 +1,6 @@
+[target.riscv64im-lambda-vm-elf]
+rustflags = [
+  "-C", "link-arg=-e",
+  "-C", "link-arg=main",
+  "-C", "passes=lower-atomic"
+]

bench_vs/lambda/fibonacci/Cargo.lock

@@ -0,0 +1,8 @@
+[workspace]
+
+[package]
+name = "fibonacci-bench"
+version = "0.1.0"
+edition = "2024"
+
+[dependencies]

bench_vs/lambda/fibonacci/Cargo.toml

@@ -0,0 +1,68 @@
+#![no_std]
+#![no_main]
+
+use core::arch::asm;
+use core::panic::PanicInfo;
+
+const SYSCALL_GET_PRIVATE_INPUTS: u64 = 4;
+const SYSCALL_COMMIT: u64 = 64;
+const SYSCALL_HALT: u64 = 93;
+
+#[panic_handler]
+fn panic(_info: &PanicInfo) -> ! {
+    loop {}
+}
+
+fn read_n() -> u64 {
+    let mut input = [0u8; 12];
+
+    unsafe {
+        asm!(
+            "ecall",
+            in("a0") input.as_mut_ptr(),
+            in("a7") SYSCALL_GET_PRIVATE_INPUTS,
+        );
+    }
+
+    let mut n_bytes = [0u8; 8];
+    n_bytes.copy_from_slice(&input[4..12]);
+    u64::from_le_bytes(n_bytes)
+}
+
+fn commit(bytes: &[u8]) {
+    unsafe {
+        asm!(
+            "ecall",
+            in("a0") 1u64,
+            in("a1") bytes.as_ptr(),
+            in("a2") bytes.len(),
+            in("a7") SYSCALL_COMMIT,
+        );
+    }
+}
+
+fn halt() -> ! {
+    unsafe {
+        asm!(
+            "ecall",
+            in("a0") 0u64,
+            in("a7") SYSCALL_HALT,
+            options(noreturn),
+        );
+    }
+}
+
+#[unsafe(no_mangle)]
+pub fn main() -> ! {
+    let n = read_n();
+    let mut a: u64 = 0;
+    let mut b: u64 = 1;
+    for _ in 0..n {
+        let c = a.wrapping_add(b);
+        a = b;
+        b = c;
+    }
+
+    commit(&b.to_le_bytes());
+    halt()
+}