add auxilery memory address verification (#418)

stwo_cairo_prover/crates/prover/src/cairo_air/mod.rs

@@ -21,6 +21,8 @@ use stwo_prover::core::prover::{prove, verify, ProvingError, VerificationError};
 use thiserror::Error;
 use tracing::{span, Level};
 
+use crate::components::memory::LOG_MEMORY_ADDRESS_BOUND;
+use crate::components::memory_address_to_id::component::MEMORY_ADDRESS_TO_ID_SPLIT;
 use crate::input::ProverInput;
 
 // TODO(Ohad): decide dynamically.
@@ -126,8 +128,15 @@ pub fn verify_cairo<MC: MerkleChannel>(
         stark_proof,
     }: CairoProof<MC::H>,
 ) -> Result<(), CairoVerificationError> {
-    // Verify.
-    // TODO(Ohad): Propogate config from CLI args.
+    // Auxiliary verifications.
+    // Assert that ADDRESS->ID component does not overflow.
+    assert!(
+        (1 << claim.memory_address_to_id.log_size) * MEMORY_ADDRESS_TO_ID_SPLIT
+            <= (1 << LOG_MEMORY_ADDRESS_BOUND)
+    );
+
+    // Setup STARK protocol.
+    // TODO(Ohad): Propagate config from CLI args.
     let config = PcsConfig {
         pow_bits: 0,
         fri_config: FriConfig {
@@ -147,6 +156,8 @@ pub fn verify_cairo<MC: MerkleChannel>(
     claim.mix_into(channel);
     commitment_scheme_verifier.commit(stark_proof.commitments[1], &log_sizes[1], channel);
     let interaction_elements = CairoInteractionElements::draw(channel);
+
+    // Verify lookup argument.
     if lookup_sum(&claim, &interaction_elements, &interaction_claim) != SecureField::zero() {
         return Err(CairoVerificationError::InvalidLogupSum);
     }
@@ -157,6 +168,7 @@ pub fn verify_cairo<MC: MerkleChannel>(
         CairoComponents::new(&claim, &interaction_elements, &interaction_claim);
     let components = component_generator.components();
 
+    // Verify stark.
     verify(
         &components,
         channel,

stwo_cairo_prover/crates/prover/src/components/memory/memory_address_to_id/component.rs

@@ -15,19 +15,20 @@ use crate::relations;
 /// Split the (ID , Multiplicity) columns to shorter chunks. This is done to improve the performance
 /// during The merkle commitment and FRI, as this component is usually the tallest in the Cairo AIR.
 ///
-/// 1. The ID and Multiplicity vectors are split to 'N_SPLIT_CHUNKS' chunks of size
-///    `ids.len()`/`N_SPLIT_CHUNKS`.
+/// 1. The ID and Multiplicity vectors are split to 'MEMORY_ADDRESS_TO_ID_SPLIT' chunks of size
+///    `ids.len()`/`MEMORY_ADDRESS_TO_ID_SPLIT`.
 /// 2. The chunks are padded with 0s to the next power of 2.
 ///
 /// #  Example
-/// ID = [id0..id10], N_SPLIT_CHUNKS = 4:
+/// ID = [id0..id10], MEMORY_ADDRESS_TO_ID_SPLIT = 4:
 /// ID0 = [id0, id1, id2, 0]
 /// ID1 = [id3, id4, id5, 0]
 /// ID2 = [id6, id7, id8, 0]
 /// ID3 = [id9, id10, 0, 0]
-pub(super) const N_SPLIT_CHUNKS: usize = 8;
+pub const MEMORY_ADDRESS_TO_ID_SPLIT: usize = 8;
 pub(super) const N_ID_AND_MULT_COLUMNS_PER_CHUNK: usize = 2;
-pub(super) const N_TRACE_COLUMNS: usize = N_SPLIT_CHUNKS * N_ID_AND_MULT_COLUMNS_PER_CHUNK;
+pub(super) const N_TRACE_COLUMNS: usize =
+    MEMORY_ADDRESS_TO_ID_SPLIT * N_ID_AND_MULT_COLUMNS_PER_CHUNK;
 
 pub type Component = FrameworkComponent<Eval>;
 
@@ -63,7 +64,7 @@ impl FrameworkEval for Eval {
         // Addresses are offseted by 1, as 0 address is reserved.
         let seq_plus_one =
             eval.get_preprocessed_column(Seq::new(self.log_size()).id()) + E::F::from(M31(1));
-        for i in 0..N_SPLIT_CHUNKS {
+        for i in 0..MEMORY_ADDRESS_TO_ID_SPLIT {
             let id = eval.next_trace_mask();
             let multiplicity = eval.next_trace_mask();
             let address =
@@ -89,7 +90,7 @@ impl Claim {
         let preprocessed_log_sizes = vec![self.log_size];
         let trace_log_sizes = vec![self.log_size; N_TRACE_COLUMNS];
         let interaction_log_sizes =
-            vec![self.log_size; SECURE_EXTENSION_DEGREE * N_SPLIT_CHUNKS.div_ceil(2)];
+            vec![self.log_size; SECURE_EXTENSION_DEGREE * MEMORY_ADDRESS_TO_ID_SPLIT.div_ceil(2)];
         TreeVec::new(vec![
             preprocessed_log_sizes,
             trace_log_sizes,

stwo_cairo_prover/crates/prover/src/components/memory/memory_address_to_id/prover.rs

@@ -15,7 +15,7 @@ use stwo_prover::core::pcs::TreeBuilder;
 use stwo_prover::core::poly::circle::{CanonicCoset, CircleEvaluation};
 use stwo_prover::core::poly::BitReversedOrder;
 
-use super::component::{Claim, InteractionClaim, N_SPLIT_CHUNKS};
+use super::component::{Claim, InteractionClaim, MEMORY_ADDRESS_TO_ID_SPLIT};
 use crate::cairo_air::preprocessed::Seq;
 use crate::components::memory_address_to_id::component::{
     N_ID_AND_MULT_COLUMNS_PER_CHUNK, N_TRACE_COLUMNS,
@@ -121,7 +121,7 @@ impl ClaimGenerator {
         SimdBackend: BackendForChannel<MC>,
     {
         let size = std::cmp::max(
-            (self.address_to_raw_id.len() / N_SPLIT_CHUNKS).next_power_of_two(),
+            (self.address_to_raw_id.len() / MEMORY_ADDRESS_TO_ID_SPLIT).next_power_of_two(),
             N_LANES,
         );
         let n_packed_rows = size.div_ceil(N_LANES);
@@ -146,9 +146,9 @@ impl ClaimGenerator {
         }
 
         // Lookup data.
-        let ids: [_; N_SPLIT_CHUNKS] =
+        let ids: [_; MEMORY_ADDRESS_TO_ID_SPLIT] =
             std::array::from_fn(|i| trace[i * N_ID_AND_MULT_COLUMNS_PER_CHUNK].data.clone());
-        let multiplicities: [_; N_SPLIT_CHUNKS] =
+        let multiplicities: [_; MEMORY_ADDRESS_TO_ID_SPLIT] =
             std::array::from_fn(|i| trace[1 + i * N_ID_AND_MULT_COLUMNS_PER_CHUNK].data.clone());
 
         // Commit on trace.
@@ -173,8 +173,8 @@ impl ClaimGenerator {
 }
 
 pub struct InteractionClaimGenerator {
-    pub ids: [Vec<PackedM31>; N_SPLIT_CHUNKS],
-    pub multiplicities: [Vec<PackedM31>; N_SPLIT_CHUNKS],
+    pub ids: [Vec<PackedM31>; MEMORY_ADDRESS_TO_ID_SPLIT],
+    pub multiplicities: [Vec<PackedM31>; MEMORY_ADDRESS_TO_ID_SPLIT],
 }
 impl InteractionClaimGenerator {
     pub fn write_interaction_trace<MC: MerkleChannel>(