free input data (#2239)

This PR allows for free runtime data besides free compile time data,
based on `initial_memory`.

- Replaces `cbor` by `bincode` in prover queries (bincode is more
efficient and cbor crashed with `u128`)
- Allows the prover to pass a runtime initial memory, only possible with
continuations (from #2251, was
already merged into here, see commits list)
- Changes the `powdr` lib, which already always uses continuations, to
always use this mechanism for prover data
- Provides a new stdin-stream-like function to read inputs in sequence,
like other zkVMs.
- The function above is called `read_stdin` which I'm not super happy
with, ideally it'd just be called `read` but the QueryCalldata function
is already called `read`. I think we could just keep this as is and
change later.

---------

Co-authored-by: Lucas Clemente Vella <[email protected]>

.github/workflows/pr-tests.yml

@@ -146,8 +146,14 @@ jobs:
         run: rustup target add riscv32imac-unknown-none-elf --toolchain nightly-2024-08-01-x86_64-unknown-linux-gnu
       - name: Install test dependencies
         run: sudo apt-get update && sudo apt-get install -y binutils-riscv64-unknown-elf lld
-      - name: Run examples
-        run: cargo run --profile pr-tests --example hello_world && cargo run --profile pr-tests --example sqrt_with_publics && cargo run --profile pr-tests --example fibonacci
+      - name: Run examples that cargo accepts as examples
+        run: cargo run --profile pr-tests --example hello_world && cargo run --profile pr-tests --example sqrt_with_publics
+      - name: Run crate example serialized_inputs with the given branch
+        run: cd powdr-test/examples/serialized-inputs && cargo run -r
+      - name: Run crate example fibonacci with the given branch
+        run: cd powdr-test/examples/fibonacci && cargo run -r
+      - name: Run crate example fibonacci with the latest powdr release
+        run: cd examples/fibonacci && cargo run -r
 
   test_estark_polygon:
     needs: build

pipeline/src/pipeline.rs

@@ -148,6 +148,8 @@ pub struct Pipeline<T: FieldElement> {
     arguments: Arguments<T>,
     /// The context for the host.
     host_context: HostContext,
+    /// Initial memory given by the prover.
+    initial_memory: Vec<Vec<u8>>,
 }
 
 impl<T: FieldElement> Clone for Artifacts<T> {
@@ -193,6 +195,7 @@ where
             pilo: false,
             arguments: Arguments::default(),
             host_context: ctx,
+            initial_memory: vec![],
         }
         // We add the basic callback functionalities to support PrintChar and Hint.
         .add_query_callback(Arc::new(handle_simple_queries_callback()))
@@ -322,6 +325,18 @@ impl<T: FieldElement> Pipeline<T> {
         self
     }
 
+    /// Adds data to the initial memory given by the prover.
+    /// This is a more efficient method of passing bytes from the host
+    /// to the guest.
+    pub fn add_to_initial_memory(mut self, data: Vec<u8>) -> Self {
+        self.initial_memory.push(data);
+        self
+    }
+
+    pub fn initial_memory(&self) -> &[Vec<u8>] {
+        &self.initial_memory
+    }
+
     pub fn add_data<S: serde::Serialize>(self, channel: u32, data: &S) -> Self {
         let bytes = serde_cbor::to_vec(&data).unwrap();
         self.add_query_callback(Arc::new(serde_data_to_query_callback(channel, bytes)))

powdr-test/examples/fibonacci/Cargo.toml

@@ -0,0 +1,25 @@
+[package]
+name = "fibonacci"
+version = "0.1.0"
+edition = "2021"
+
+[features]
+default = []
+simd = ["powdr/plonky3-simd"]
+
+[dependencies]
+powdr = { path = "../../../powdr", features = ["plonky3"] }
+
+serde = { version = "1.0", default-features = false, features = [
+  "alloc",
+  "derive",
+  "rc",
+] }
+serde_cbor = { version = "0.11.2", default-features = false, features = [
+  "alloc",
+] }
+
+env_logger = "0.10.2"
+log = "0.4.17"
+
+[workspace]

powdr-test/examples/fibonacci/guest/Cargo.toml

@@ -0,0 +1,9 @@
+[package]
+name = "powdr-guest"
+version = "0.1.0"
+edition = "2021"
+
+[dependencies]
+powdr-riscv-runtime = { path = "../../../../riscv-runtime", features = ["std"]}
+
+[workspace]

powdr-test/examples/fibonacci/guest/src/main.rs

@@ -0,0 +1,20 @@
+use powdr_riscv_runtime;
+use powdr_riscv_runtime::commit;
+use powdr_riscv_runtime::io::{read, write};
+
+fn fib(n: u32) -> u32 {
+    if n <= 1 {
+        return n;
+    }
+    fib(n - 1) + fib(n - 2)
+}
+
+fn main() {
+    // Read input from stdin.
+    let n: u32 = read();
+    let r = fib(n);
+    // Write result to stdout.
+    write(1, r);
+    // Commit the result as a public.
+    commit::commit(r);
+}

powdr-test/examples/fibonacci/rust-toolchain.toml

@@ -0,0 +1,2 @@
+[toolchain]
+channel = "nightly-2024-09-21"

powdr-test/examples/fibonacci.rs → powdr-test/examples/fibonacci/src/main.rs

@@ -5,10 +5,10 @@ fn main() {
 
     let n = 11;
     let mut session = Session::builder()
-        .guest_path("./examples/fibonacci/guest")
+        .guest_path("./guest")
         .out_path("powdr-target")
         .build()
-        .write(0, &n);
+        .write(&n);
 
     // Fast dry run to test execution.
     session.run();
@@ -24,9 +24,6 @@ fn main() {
     let publics = session.publics();
     assert_eq!(
         publics,
-        [
-            555233681, 1854640251, 3298928347, 2857173302, 2660189392, 1608424695, 543896544,
-            3870154745
-        ]
+        [555233681, 1854640251, 3298928347, 2857173302, 2660189392, 1608424695, 543896544, 3870154745]
     );
 }

powdr-test/examples/serialized-inputs/Cargo.toml

@@ -0,0 +1,22 @@
+[package]
+name = "serialized-inputs"
+version = "0.1.0"
+edition = "2021"
+
+[features]
+default = []
+simd = ["powdr/plonky3-simd"]
+
+[dependencies]
+powdr = { path = "../../../powdr", features = ["plonky3"] }
+
+serde = { version = "1.0", default-features = false, features = [
+  "alloc",
+  "derive",
+  "rc",
+] }
+
+env_logger = "0.10.2"
+log = "0.4.17"
+
+[workspace]

powdr-test/examples/serialized-inputs/guest/Cargo.toml

@@ -0,0 +1,15 @@
+[package]
+name = "powdr-guest"
+version = "0.1.0"
+edition = "2021"
+
+[dependencies]
+powdr-riscv-runtime = { path = "../../../../riscv-runtime", features = ["std"]}
+
+serde = { version = "1.0", default-features = false, features = [
+  "alloc",
+  "derive",
+  "rc",
+] }
+
+[workspace]

powdr-test/examples/serialized-inputs/guest/src/main.rs

@@ -0,0 +1,16 @@
+use powdr_riscv_runtime;
+use powdr_riscv_runtime::io::read;
+
+#[derive(serde::Serialize, serde::Deserialize)]
+struct Data {
+    numbers: Vec<u32>,
+    sum: u32,
+}
+
+fn main() {
+    let data: Data  = read();
+    let s: String = read();
+
+    assert_eq!(data.numbers.iter().sum::<u32>(), data.sum);
+    assert_eq!(s, "test");
+}

powdr-test/examples/serialized-inputs/rust-toolchain.toml

@@ -0,0 +1,2 @@
+[toolchain]
+channel = "nightly-2024-09-21"

powdr-test/examples/serialized-inputs/src/main.rs

@@ -0,0 +1,32 @@
+use powdr::Session;
+
+#[derive(serde::Serialize, serde::Deserialize)]
+struct Data {
+    numbers: Vec<u32>,
+    sum: u32,
+}
+
+fn main() {
+    env_logger::init();
+
+    let some_data = Data {
+        numbers: vec![1, 2, 3, 4, 5],
+        sum: 15,
+    };
+
+    let s: String = "test".to_string();
+
+    let mut session = Session::builder()
+        .guest_path("./guest")
+        .out_path("powdr-target")
+        .chunk_size_log2(18)
+        .build()
+        .write(&some_data)
+        .write(&s);
+
+    // Fast dry run to test execution.
+    session.run();
+
+    // Uncomment to compute the proof.
+    //session.prove();
+}

powdr/Cargo.toml

@@ -25,6 +25,7 @@ serde = { version = "1.0", default-features = false, features = [
   "derive",
   "alloc",
 ] }
+serde_cbor = "0.11.2"
 
 [features]
 default = ["halo2", "plonky3"]
@@ -39,7 +40,10 @@ estark-polygon = [
 stwo = ["powdr-backend/stwo", "powdr-pipeline/stwo"]
 
 plonky3-simd = ["powdr-backend/plonky3-simd", "powdr-pipeline/plonky3-simd"]
-estark-starky-simd = ["powdr-backend/estark-starky-simd", "powdr-pipeline/estark-starky-simd"]
+estark-starky-simd = [
+  "powdr-backend/estark-starky-simd",
+  "powdr-pipeline/estark-starky-simd",
+]
 
 [lints.clippy]
 uninlined_format_args = "deny"

powdr/src/lib.rs

@@ -125,9 +125,17 @@ impl Session {
         }
     }
 
-    pub fn write<S: serde::Serialize>(self, channel: u32, data: &S) -> Self {
-        Session {
-            pipeline: self.pipeline.add_data(channel, data),
+    pub fn write<S: serde::Serialize>(self, data: &S) -> Self {
+        let bytes = serde_cbor::to_vec(&data).unwrap();
+        Self {
+            pipeline: self.pipeline.add_to_initial_memory(bytes),
+            ..self
+        }
+    }
+
+    pub fn write_bytes(self, bytes: Vec<u8>) -> Self {
+        Self {
+            pipeline: self.pipeline.add_to_initial_memory(bytes),
             ..self
         }
     }
@@ -278,7 +286,8 @@ pub fn run(pipeline: &mut Pipeline<GoldilocksField>) {
     let start = Instant::now();
 
     let asm = pipeline.compute_analyzed_asm().unwrap().clone();
-    let initial_memory = riscv::continuations::load_initial_memory(&asm);
+    let initial_memory = riscv::continuations::load_initial_memory(&asm, pipeline.initial_memory());
+
     let trace_len = riscv_executor::execute_fast(
         &asm,
         initial_memory,

riscv-runtime/Cargo.toml

@@ -12,6 +12,7 @@ serde = { version = "1.0", default-features = false, features = ["alloc", "deriv
 serde_cbor = { version = "0.11.2", default-features = false, features = ["alloc"] }
 powdr-riscv-syscalls = { path = "../riscv-syscalls", version = "0.1.4" }
 getrandom = { version = "0.2", features = ["custom"], optional = true }
+spin = "0.9"
 
 [features]
 std = ["serde/std", "serde_cbor/std"]

riscv-runtime/powdr.x

@@ -1,17 +1,32 @@
 # Powdr linker script.
 #
-# If you are using powdr-riscv-runtime, it expects the symbols
-# "__global_pointer$" and "__powdr_stack_start" to be defined.
+# If you are using powdr-riscv-runtime, it expects the following
+# symbols to be defined:
+#   __global_pointer$
+#   __powdr_stack_start
+#   __powdr_prover_data_init
+#   __powdr_prover_data_end
+#
+# Where "__powdr_prover_data_init" and "__powdr_prover_data_end"
+# defines a region bigger than 1 page (2 KB), whose size is a
+# power of 2, and aligned to its size.
 #
 # This linker script provides usable definitions to these
-# symbols, with a 256 MB stack. If you are not building via
-# powdr-rs, you must manually specify "-C link-arg=-Tpowdr.x"
+# symbols, with a stack of almost 256 MB. If you are not building
+# via powdr-rs, you must manually specify "-C link-arg=-Tpowdr.x"
 # in rustc to use this linker script (e.g. via RUSTFLAGS).
 
 SECTIONS
 {
-  # Data starts here, before is the stack.
-  . = 0x10000100;
+  # Stack starts backwards from here (one 2 KB page short of 256 MB)
+  __powdr_stack_start = 0x10000000 - 0x800;
+
+  # The prover data (the second 256 MB chunk of the address space)
+  __powdr_prover_data_start = 0x10000000;
+  __powdr_prover_data_end = 0x20000000;
+
+  # Data starts here, one page after the prover data.
+  . = 0x20000000 + 0x800;
   .data : {
     *(.data)
     PROVIDE( __global_pointer$ = . + 0x800 );
@@ -20,8 +35,6 @@ SECTIONS
 
   # Text addresses are fake in powdr, we use a different address space.
   .text : { *(.text) }
-
-  __powdr_stack_start = 0x10000000;
 }
 
 # Specify the entry point function provided by powdr-riscv-runtime: