Merge branch 'master' into spy/sandbox-for-all-e2e

barretenberg/.gitrepo

@@ -6,7 +6,7 @@
 [subrepo]
 	remote = https://github.com/AztecProtocol/barretenberg
 	branch = master
-	commit = 491aa9d0f86d0b19b05d0805bfe3dbcccb249472
-	parent = a60c70dca1d920ad88511f77be3ad186afab7bdb
+	commit = 8b5ad73c5eff8daf734bda54acfbc28f97e73c0f
+	parent = 73106008f464328935997028ca18698965b579a5
 	method = merge
 	cmdver = 0.4.6

barretenberg/cpp/CMakePresets.json

@@ -183,34 +183,40 @@
         "MULTITHREADING": "ON"
       }
     },
-    {
-      "name": "xray-1thread",
-      "displayName": "Build with single-threaded XRay Profiling",
-      "description": "Build with Clang and enable single-threaded LLVM XRay for profiling",
-      "generator": "Unix Makefiles",
-      "inherits": "clang16",
-      "environment": {
-        "CFLAGS": "-fxray-instrument -fxray-instruction-threshold=10",
-        "CXXFLAGS": "-fxray-instrument -fxray-instruction-threshold=10",
-        "LDFLAGS": "-fxray-instrument -fxray-instruction-threshold=10"
-      },
-      "cacheVariables": {
-        "MULTITHREADING": "OFF"
-      },
-      "binaryDir": "build-xray-1thread"
-    },
     {
       "name": "xray",
       "displayName": "Build with multi-threaded XRay Profiling",
       "description": "Build with Clang and enable multi-threaded LLVM XRay for profiling",
       "generator": "Unix Makefiles",
       "inherits": "clang16",
       "environment": {
-        "CFLAGS": "-fxray-instrument -fxray-instruction-threshold=10",
-        "CXXFLAGS": "-fxray-instrument -fxray-instruction-threshold=10",
-        "LDFLAGS": "-fxray-instrument -fxray-instruction-threshold=10"
+        "CFLAGS": "-fxray-instrument -fxray-instruction-threshold=100",
+        "CXXFLAGS": "-fxray-instrument -fxray-instruction-threshold=100",
+        "LDFLAGS": "-fxray-instrument -fxray-instruction-threshold=100"
       },
       "binaryDir": "build-xray"
+    },
+    {
+      "name": "xray-verbose",
+      "displayName": "Build with detailed XRay Profiling",
+      "description": "Build with Clang and enable detailed LLVM XRay for profiling",
+      "inherits": "xray",
+      "environment": {
+        "CFLAGS": "-fxray-instrument -fxray-instruction-threshold=100 -finline-max-stacksize=150",
+        "CXXFLAGS": "-fxray-instrument -fxray-instruction-threshold=100 -finline-max-stacksize=150",
+        "LDFLAGS": "-fxray-instrument -fxray-instruction-threshold=100 -finline-max-stacksize=150"
+      },
+      "binaryDir": "build-xray-verbose"
+    },
+    {
+      "name": "xray-1thread",
+      "displayName": "Build with single-threaded XRay Profiling",
+      "description": "Build with Clang and enable single-threaded LLVM XRay for profiling",
+      "inherits": "xray",
+      "cacheVariables": {
+        "MULTITHREADING": "OFF"
+      },
+      "binaryDir": "build-xray-1thread"
     }
   ],
   "buildPresets": [
@@ -303,6 +309,11 @@
       "jobs": 0,
       "targets": ["barretenberg.wasm"]
     },
+    {
+      "name": "xray-verbose",
+      "configurePreset": "xray-verbose",
+      "inherits": "default"
+    },
     {
       "name": "xray-1thread",
       "configurePreset": "xray-1thread",

barretenberg/cpp/scripts/collect_profile_information.sh

@@ -41,5 +41,6 @@ llvm-xray-16 stack xray-log.honk_bench_main_simple.* \
   --instr_map=./bin/honk_bench_main_simple --stack-format=flame --aggregate-threads --aggregation-type=time --all-stacks \
   | node ../scripts/llvm_xray_stack_flame_corrector.js \
   | shorten_cpp_names \
-  | ../scripts/flamegraph.pl > xray.svg
+  | ../scripts/flamegraph.pl --width 1200 --fontsize 10 \
+  > xray.svg
 echo "Profiling complete, now you can do e.g. 'scp mainframe:`readlink -f xray.svg` .' on a local terminal and open the SVG in a browser."

barretenberg/cpp/src/barretenberg/benchmark/CMakeLists.txt

@@ -2,4 +2,4 @@ add_subdirectory(decrypt_bench)
 add_subdirectory(pippenger_bench)
 add_subdirectory(plonk_bench)
 add_subdirectory(honk_bench)
-add_subdirectory(relations_bench)
+add_subdirectory(relations_bench)

barretenberg/cpp/src/barretenberg/benchmark/compare_branch_vs_baseline.sh

@@ -10,7 +10,7 @@ BASELINE_BRANCH="master"
 
 echo -e "\nComparing $BENCH_TARGET between $BASELINE_BRANCH and current branch:"
 # Set some directories
-BASE_DIR="$HOME/barretenberg/cpp"
+BASE_DIR="$HOME/aztec-packages/barretenberg/cpp"
 BUILD_DIR="$BASE_DIR/build-bench" # matches build dir specified in bench preset
 BENCH_RESULTS_DIR="$BASE_DIR/tmp_bench_results"
 BENCH_TOOLS_DIR="$BUILD_DIR/_deps/benchmark-src/tools"

barretenberg/cpp/src/barretenberg/benchmark/honk_bench/CMakeLists.txt

@@ -1,8 +1,8 @@
 # Each source represents a separate benchmark suite 
 set(BENCHMARK_SOURCES
-standard_plonk.bench.cpp
-ultra_honk.bench.cpp
-ultra_plonk.bench.cpp
+  standard_plonk.bench.cpp
+  ultra_honk.bench.cpp
+  ultra_plonk.bench.cpp
 )
 
 # Required libraries for benchmark suites

barretenberg/cpp/src/barretenberg/benchmark/honk_bench/benchmark_utilities.hpp

@@ -45,6 +45,9 @@ template <typename Builder> void generate_basic_arithmetic_circuit(Builder& buil
     proof_system::plonk::stdlib::field_t b(
         proof_system::plonk::stdlib::witness_t(&builder, barretenberg::fr::random_element()));
     proof_system::plonk::stdlib::field_t c(&builder);
+    if (num_gates < 4) {
+        throw std::runtime_error("too few gates");
+    }
     for (size_t i = 0; i < (num_gates / 4) - 4; ++i) {
         c = a + b;
         c = a * c;

barretenberg/cpp/src/barretenberg/benchmark/honk_bench/compare_honk_to_plonk_ultra.sh

@@ -6,7 +6,7 @@
 
 echo -e '\nComparing Ultra Plonk/Honk benchmarks.'
 # Set some directories
-BASE_DIR="$HOME/barretenberg/cpp"
+BASE_DIR="$HOME/aztec-packages/barretenberg/cpp"
 BUILD_DIR="$BASE_DIR/build-bench"
 BENCH_RESULTS_DIR="$BASE_DIR/tmp_bench_results"
 BENCH_TOOLS_DIR="$BUILD_DIR/_deps/benchmark-src/tools"

barretenberg/cpp/src/barretenberg/benchmark/honk_bench/main.simple.cpp

@@ -22,18 +22,24 @@
 #include "barretenberg/stdlib/primitives/packed_byte_array/packed_byte_array.hpp"
 #include "barretenberg/stdlib/primitives/witness/witness.hpp"
 
-using namespace proof_system::plonk;
-
-using UltraBuilder = proof_system::UltraCircuitBuilder;
-using UltraHonk = proof_system::honk::UltraComposer;
+using namespace proof_system;
 
 template <typename Builder> void generate_sha256_test_circuit(Builder& builder, size_t num_iterations)
 {
     std::string in;
     in.resize(32);
-    proof_system::plonk::stdlib::packed_byte_array<Builder> input(&builder, in);
+    plonk::stdlib::packed_byte_array<Builder> input(&builder, in);
     for (size_t i = 0; i < num_iterations; i++) {
-        input = proof_system::plonk::stdlib::sha256<Builder>(input);
+        input = plonk::stdlib::sha256<Builder>(input);
+    }
+}
+
+BBERG_INSTRUMENT BBERG_NOINLINE void sumcheck_profiling(honk::UltraProver& ext_prover)
+{
+    ext_prover.construct_proof();
+    for (size_t i = 0; i < 200; i++) {
+        // Bench sumcheck
+        ext_prover.execute_relation_check_rounds();
     }
 }
 
@@ -44,15 +50,14 @@ void construct_proof_ultra() noexcept
 {
     barretenberg::srs::init_crs_factory("../srs_db/ignition");
     // Constuct circuit and prover; don't include this part in measurement
-    auto builder = typename UltraHonk::CircuitBuilder();
-    generate_sha256_test_circuit<UltraBuilder>(builder, 1);
+    honk::UltraComposer::CircuitBuilder builder;
+    generate_sha256_test_circuit(builder, 1);
+    std::cout << "gates: " << builder.get_total_circuit_size() << std::endl;
 
-    auto composer = UltraHonk();
-    auto instance = composer.create_instance(builder);
-    auto ext_prover = composer.create_prover(instance);
-    for (size_t i = 0; i < 10; i++) {
-        auto proof = ext_prover.construct_proof();
-    }
+    honk::UltraComposer composer;
+    std::shared_ptr<honk::UltraComposer::Instance> instance = composer.create_instance(builder);
+    honk::UltraProver ext_prover = composer.create_prover(instance);
+    sumcheck_profiling(ext_prover);
 }
 
 int main()

barretenberg/cpp/src/barretenberg/benchmark/honk_bench/ultra_honk.bench.cpp

@@ -30,22 +30,22 @@ void construct_proof_ultra(State& state, void (*test_circuit_function)(UltraBuil
 BENCHMARK_CAPTURE(construct_proof_ultra, sha256, &bench_utils::generate_sha256_test_circuit<UltraBuilder>)
     ->DenseRange(MIN_NUM_ITERATIONS, MAX_NUM_ITERATIONS)
     ->Repetitions(NUM_REPETITIONS)
-    ->Unit(::benchmark::kSecond);
+    ->Unit(::benchmark::kMillisecond);
 BENCHMARK_CAPTURE(construct_proof_ultra, keccak, &bench_utils::generate_keccak_test_circuit<UltraBuilder>)
     ->DenseRange(MIN_NUM_ITERATIONS, MAX_NUM_ITERATIONS)
     ->Repetitions(NUM_REPETITIONS)
-    ->Unit(::benchmark::kSecond);
+    ->Unit(::benchmark::kMillisecond);
 BENCHMARK_CAPTURE(construct_proof_ultra,
                   ecdsa_verification,
                   &bench_utils::generate_ecdsa_verification_test_circuit<UltraBuilder>)
     ->DenseRange(MIN_NUM_ITERATIONS, MAX_NUM_ITERATIONS)
     ->Repetitions(NUM_REPETITIONS)
-    ->Unit(::benchmark::kSecond);
+    ->Unit(::benchmark::kMillisecond);
 BENCHMARK_CAPTURE(construct_proof_ultra,
                   merkle_membership,
                   &bench_utils::generate_merkle_membership_test_circuit<UltraBuilder>)
     ->DenseRange(MIN_NUM_ITERATIONS, MAX_NUM_ITERATIONS)
     ->Repetitions(NUM_REPETITIONS)
-    ->Unit(::benchmark::kSecond);
+    ->Unit(::benchmark::kMillisecond);
 
 } // namespace ultra_honk_bench

barretenberg/cpp/src/barretenberg/benchmark/honk_bench/ultra_plonk.bench.cpp

@@ -26,22 +26,22 @@ void construct_proof_ultra(State& state, void (*test_circuit_function)(UltraBuil
 BENCHMARK_CAPTURE(construct_proof_ultra, sha256, &bench_utils::generate_sha256_test_circuit<UltraBuilder>)
     ->DenseRange(MIN_NUM_ITERATIONS, MAX_NUM_ITERATIONS)
     ->Repetitions(NUM_REPETITIONS)
-    ->Unit(::benchmark::kSecond);
+    ->Unit(::benchmark::kMillisecond);
 BENCHMARK_CAPTURE(construct_proof_ultra, keccak, &bench_utils::generate_keccak_test_circuit<UltraBuilder>)
     ->DenseRange(MIN_NUM_ITERATIONS, MAX_NUM_ITERATIONS)
     ->Repetitions(NUM_REPETITIONS)
-    ->Unit(::benchmark::kSecond);
+    ->Unit(::benchmark::kMillisecond);
 BENCHMARK_CAPTURE(construct_proof_ultra,
                   ecdsa_verification,
                   &bench_utils::generate_ecdsa_verification_test_circuit<UltraBuilder>)
     ->DenseRange(MIN_NUM_ITERATIONS, MAX_NUM_ITERATIONS)
     ->Repetitions(NUM_REPETITIONS)
-    ->Unit(::benchmark::kSecond);
+    ->Unit(::benchmark::kMillisecond);
 BENCHMARK_CAPTURE(construct_proof_ultra,
                   merkle_membership,
                   &bench_utils::generate_merkle_membership_test_circuit<UltraBuilder>)
     ->DenseRange(MIN_NUM_ITERATIONS, MAX_NUM_ITERATIONS)
     ->Repetitions(NUM_REPETITIONS)
-    ->Unit(::benchmark::kSecond);
+    ->Unit(::benchmark::kMillisecond);
 
 } // namespace ultra_plonk_bench

barretenberg/cpp/src/barretenberg/benchmark/relations_bench/CMakeLists.txt

@@ -1,14 +1,17 @@
 # Each source represents a separate benchmark suite 
 set(BENCHMARK_SOURCES
-barycentric.bench.cpp
-relations.bench.cpp
+  barycentric.bench.cpp
+  relations.bench.cpp
+  widget.bench.cpp
 )
 
 # Required libraries for benchmark suites
 set(LINKED_LIBRARIES
   polynomials
   proof_system
   benchmark::benchmark
+  transcript
+  stdlib_primitives
 )
 
 # Add executable and custom target for each suite, e.g. ultra_honk_bench

barretenberg/cpp/src/barretenberg/benchmark/relations_bench/widget.bench.cpp

@@ -0,0 +1,35 @@
+#include "barretenberg/benchmark/honk_bench/benchmark_utilities.hpp"
+#include "barretenberg/honk/flavor/goblin_ultra.hpp"
+#include "barretenberg/honk/flavor/ultra.hpp"
+#include "barretenberg/plonk/composer/standard_composer.hpp"
+#include "barretenberg/plonk/composer/ultra_composer.hpp"
+#include "barretenberg/plonk/proof_system/widgets/transition_widgets/plookup_auxiliary_widget.hpp"
+#include <benchmark/benchmark.h>
+
+namespace {
+auto& engine = numeric::random::get_debug_engine();
+}
+
+namespace proof_system::plonk {
+
+template <typename Flavor, typename Widget> void execute_widget(::benchmark::State& state)
+{
+    barretenberg::srs::init_crs_factory("../srs_db/ignition");
+    auto inner_composer = plonk::UltraComposer();
+    auto builder = typename plonk::UltraComposer::CircuitBuilder();
+    bench_utils::generate_basic_arithmetic_circuit(builder, 80);
+    auto inner_prover = inner_composer.create_prover(builder);
+    auto inner_proof = inner_prover.construct_proof();
+    for (auto _ : state) {
+        Widget widget(inner_composer.circuit_proving_key.get());
+        widget.compute_quotient_contribution(barretenberg::fr::random_element(), inner_prover.transcript);
+    }
+}
+
+void plookup_auxiliary_widget(::benchmark::State& state) noexcept
+{
+    execute_widget<honk::flavor::Ultra, ProverPlookupAuxiliaryWidget<ultra_settings>>(state);
+}
+BENCHMARK(plookup_auxiliary_widget);
+
+} // namespace proof_system::plonk

barretenberg/cpp/src/barretenberg/common/compiler_hints.hpp

@@ -0,0 +1,18 @@
+#pragma once
+
+#ifdef _WIN32
+#define BBERG_INLINE __forceinline inline
+#else
+#define BBERG_INLINE __attribute__((always_inline)) inline
+#endif
+
+// TODO(AD): Other compilers
+#if defined(__clang__)
+#define BBERG_INSTRUMENT [[clang::xray_always_instrument]]
+#define BBERG_NO_INSTRUMENT [[clang::xray_never_instrument]]
+#define BBERG_NOINLINE [[clang::noinline]]
+#else
+#define BBERG_INSTRUMENT
+#define BBERG_NO_INSTRUMENT
+#define BBERG_NOINLINE
+#endif