Remove RAM profiling

Cargo.toml

@@ -46,7 +46,6 @@ smallvec = "1.13"
 structopt = "0.3.26"
 codegen = "0.2.0"
 regex = { version = "1.10.6", features = ["pattern"] }
-peak_alloc = "*"
 
 [dev-dependencies]
 rand = "0.4"

src/utils.rs

@@ -4,100 +4,6 @@ use crate::witness::tree::BinaryHasher;
 
 use circuit_definitions::encodings::BytesSerializable;
 
-use peak_alloc::PeakAlloc;
-use std::sync::{Mutex, OnceLock};
-use std::time::Duration;
-use std::time::Instant;
-
-#[global_allocator]
-static PEAK_ALLOC: PeakAlloc = PeakAlloc;
-
-fn mem_array() -> &'static Mutex<Vec<(String, f32)>> {
-    static MEM_ARRAY: OnceLock<Mutex<Vec<(String, f32)>>> = OnceLock::new();
-    MEM_ARRAY.get_or_init(|| Mutex::new(vec![]))
-}
-
-fn time_instant() -> &'static Mutex<Instant> {
-    static TIME_INSTANT: OnceLock<Mutex<Instant>> = OnceLock::new();
-    TIME_INSTANT.get_or_init(|| Mutex::new(Instant::now()))
-}
-
-fn time_array() -> &'static Mutex<Vec<(String, Duration)>> {
-    static TIME_ARRAY: OnceLock<Mutex<Vec<(String, Duration)>>> = OnceLock::new();
-    TIME_ARRAY.get_or_init(|| Mutex::new(vec![]))
-}
-
-fn peak_mem_array() -> &'static Mutex<Vec<(String, f32)>> {
-    static PEAK_MEM_ARRAY: OnceLock<Mutex<Vec<(String, f32)>>> = OnceLock::new();
-    PEAK_MEM_ARRAY.get_or_init(|| Mutex::new(vec![]))
-}
-
-fn peak_snapshot_mem(label: &str) {
-    let peak_mem = PEAK_ALLOC.peak_usage_as_mb();
-    println!("PEAK MEM: {}: {}", label.to_owned(), peak_mem);
-    peak_mem_array()
-        .lock()
-        .unwrap()
-        .push((label.to_owned(), peak_mem));
-}
-
-pub fn print_mem_snapshots() {
-    let mem = mem_array().lock().unwrap();
-    println!("MEMORY SNAPSHOTS");
-    for snapshot in mem.clone() {
-        println!("{}: {}", snapshot.0, snapshot.1);
-    }
-    println!();
-}
-
-pub fn print_peak_mem_snapshots() {
-    let mem = peak_mem_array().lock().unwrap();
-    println!("PEAK MEMORY SNAPSHOTS");
-    for snapshot in mem.clone() {
-        println!("{}: {}", snapshot.0, snapshot.1);
-    }
-    println!();
-}
-
-pub fn print_time_snapshots() {
-    let time = time_array().lock().unwrap();
-    println!("TIME SNAPSHOTS");
-    for snapshot in time.clone() {
-        println!("{}: {:.2?}", snapshot.0, snapshot.1);
-    }
-    println!();
-}
-
-pub fn snapshot_mem(label: &str) {
-    let current_mem = PEAK_ALLOC.current_usage_as_mb();
-    println!("MEM: {}: {}", label.to_owned(), current_mem);
-    mem_array()
-        .lock()
-        .unwrap()
-        .push((label.to_owned(), current_mem));
-}
-
-pub fn snapshot_prof(label: &str) {
-    snapshot_mem(label);
-    peak_snapshot_mem(label);
-    snapshot_time(label);
-    reset_peak_snapshot_mem();
-}
-
-pub fn snapshot_time(label: &str) {
-    let mut instant = time_instant().lock().unwrap();
-    println!("TIME: {}: {:.2?}", label.to_owned(), instant.elapsed());
-    time_array()
-        .lock()
-        .unwrap()
-        .push((label.to_owned(), instant.elapsed()));
-    *instant = Instant::now();
-}
-
-pub fn reset_peak_snapshot_mem() {
-    PEAK_ALLOC.reset_peak_usage();
-}
-
 pub fn u64_as_u32_le(value: u64) -> [u32; 2] {
     [value as u32, (value >> 32) as u32]
 }

src/witness/oracle.rs

@@ -14,7 +14,6 @@ use crate::boojum::field::SmallField;
 use crate::boojum::gadgets::queue::QueueState;
 use crate::boojum::gadgets::traits::allocatable::CSAllocatable;
 use crate::ethereum_types::U256;
-use crate::snapshot_prof;
 use crate::toolset::GeometryConfig;
 use crate::witness::artifacts::{DemuxedLogQueries, MemoryArtifacts, MemoryCircuitsArtifacts};
 use crate::witness::aux_data_structs::one_per_circuit_accumulator::{
@@ -1172,8 +1171,6 @@ fn process_memory_related_circuits<CB: FnMut(WitnessGenerationArtifact)>(
         - 1)
         / geometry.cycles_per_ram_permutation;
 
-    snapshot_prof("BEFORE QUEUES SIMULATION");
-
     // Memory queues simulation is a slowest part in basic witness generation.
     // Each queue simulation is sequential single-threaded computation of hashes.
     // We will simulate unsorted and sorted queues in separate threads.
@@ -1258,8 +1255,6 @@ fn process_memory_related_circuits<CB: FnMut(WitnessGenerationArtifact)>(
         implicit_memory_states.amount_of_states()
     );
 
-    snapshot_prof("AFTER QUEUES SIMULATION");
-
     use crate::witness::individual_circuits::memory_related::decommit_code::compute_decommitter_circuit_snapshots;
 
     tracing::debug!("Running code code decommitter simulation");
@@ -1362,8 +1357,6 @@ fn process_memory_related_circuits<CB: FnMut(WitnessGenerationArtifact)>(
 
     tracing::debug!("Running RAM permutation simulation");
 
-    snapshot_prof("BEFORE RAM");
-
     let (ram_permutation_circuits, ram_permutation_circuits_compact_forms_witnesses) =
         compute_ram_circuit_snapshots(
             amount_of_memory_queries,
@@ -1380,8 +1373,6 @@ fn process_memory_related_circuits<CB: FnMut(WitnessGenerationArtifact)>(
             &mut artifacts_callback,
         );
 
-    snapshot_prof("AFTER RAM");
-
     (
         circuits_data,
         memory_artifacts_for_main_vm,
@@ -1468,8 +1459,6 @@ pub(crate) fn create_artifacts_from_tracer<CB: FnMut(WitnessGenerationArtifact)>
         std::mem::take(&mut callstack_with_aux_data.flat_new_frames_history);
     drop(callstack_with_aux_data);
 
-    snapshot_prof("BEFORE MUX LOG QUEUE CIRCUITS");
-
     tracing::debug!("Running multiplexed log queue simulation");
 
     // We have all log queries in one multiplexed queue. We need to simulate this queue,
@@ -1487,8 +1476,6 @@ pub(crate) fn create_artifacts_from_tracer<CB: FnMut(WitnessGenerationArtifact)>
         *round_function,
     );
 
-    snapshot_prof("AFTER MUX LOG QUEUE CIRCUITS");
-
     use std::thread;
     let callstack_handle = {
         let log_rollback_tails_for_frames = log_rollback_tails_for_frames.clone();
@@ -1532,8 +1519,6 @@ pub(crate) fn create_artifacts_from_tracer<CB: FnMut(WitnessGenerationArtifact)>
 
     tracing::debug!("Processing log circuits");
 
-    snapshot_prof("AFTER LOG DEMUX CIRCUITS");
-
     // Process part of log circuits that do not use memory (I/O-like).
     // Precompiles will be processed in process_memory_related_circuits.
     // Also makes storage application circuits and compact form witnesses.
@@ -1547,8 +1532,6 @@ pub(crate) fn create_artifacts_from_tracer<CB: FnMut(WitnessGenerationArtifact)>
             &mut artifacts_callback,
         );
 
-    snapshot_prof("AFTER LOG CIRCUITS");
-
     tracing::debug!("Processing memory-related circuits");
 
     let precompiles_data = PrecompilesInputData {