Worked more on cleanup

Author: Can Gokmen

CMakeLists.txt

@@ -24,10 +24,9 @@ set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "-O2 -g")
 # add_executable(main main.cpp)
 # add_executable(insert_profiling insert_profiling.cpp)
 add_executable(run run.cpp)
-add_executable(test_compress test/test_compress.cpp)
 
 # TPC-C insert experiment (ART paper replication, warehouse=1)
-# Note: ArtNodeNewMethods.cpp is #include'd from the QuART headers,
+# Note: QuArtNodeMethods.cpp is #include'd from the QuART headers,
 # so it must NOT be listed as a separate translation unit here.
 if(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/tpcc/tpcc_art.cpp")
     add_executable(tpcc_art tpcc/tpcc_art.cpp)

ArtNodeBulkLoadMethods.cpp QuArtNodeBulkLoadMethods.cppArtNodeBulkLoadMethods.cpp renamed to QuArtNodeBulkLoadMethods.cpp

@@ -11,7 +11,8 @@
 # Environment variables (all optional):
 #   WORKLOAD_DIR  – directory containing BoDS workload .bin files
 #                   (default: /scratch/cgokmen/bods/workloads)
-#   REPEAT        – number of timed repetitions per configuration (default: 5)
+#   REPEAT        – number of timed repetitions per configuration (default: 1)
+#                   (set to 5 to reproduce the paper's numbers; substantially longer)
 #
 # Output:
 #   experiments/5.2-quart-vs-quit/results/results_<TIMESTAMP>.csv
@@ -42,7 +43,7 @@ if [[ ! -x "$QUIT_BUILD/run_quit_2k" ]]; then
 fi
 
 WORKLOAD_DIR="${WORKLOAD_DIR:-$WORKLOAD_DIR}"
-REPEAT="${REPEAT:-5}"
+REPEAT="${REPEAT:-1}"  # paper used 5; increase for publication-quality averages
 N=500000000
 
 SUFFIX=$(date +"%Y%m%d_%H%M%S")
@@ -64,7 +65,7 @@ echo ""
 # ---------------------------------------------------------------------------
 echo "workload,K,L,tree_type,avg_insert_ns,avg_query_ns" > "$RESULTS_FILE"
 
-TREES=(QuART_stail QuIT_2k)
+TREES=(QuART_stail QuIT)
 
 # ---------------------------------------------------------------------------
 # Helper: run one (workload, tree) configuration REPEAT times and average
@@ -73,7 +74,7 @@ run_config() {
     local FILE="$1" N_VAL="$2" K_VAL="$3" L_VAL="$4" TREE="$5"
     local WNAME LOG INSERT_SUM QUERY_SUM FAILED BINARY
     WNAME="$(basename "$FILE" .bin)"
-    LOG="$LOG_DIR/${WNAME}_${TREE}.log"
+    LOG="$LOG_DIR/${WNAME}.log"
     INSERT_SUM=0
     QUERY_SUM=0
     FAILED=0
@@ -85,7 +86,7 @@ run_config() {
         BINARY="$BUILD/run"
     fi
 
-    echo "=== workload=$WNAME  tree=$TREE ===" > "$LOG"
+    echo "=== workload=$WNAME  tree=$TREE ===" >> "$LOG"
 
     for ((i=1; i<=REPEAT; i++)); do
         echo "--- Run $i/$REPEAT ---" >> "$LOG"
@@ -98,7 +99,7 @@ run_config() {
         if [[ $STATUS -ne 0 || -z "$OUTPUT" ]]; then
             echo "  [WARN] run failed: K=$K_VAL L=$L_VAL tree=$TREE run=$i" >&2
             FAILED=1
-            break
+            continue
         fi
 
         CSV_LINE=$(echo "$OUTPUT" | tail -1)

ArtNodeNewMethods.cpp QuArtNodeMethods.cppArtNodeNewMethods.cpp renamed to QuArtNodeMethods.cpp

@@ -12,7 +12,8 @@
 # Environment variables (all optional):
 #   WORKLOAD_FILE  – path to the TPC-H .bin workload file
 #                    (default: /scratch/cgokmen/bods/workloads/workload_N6000000_K9667_L01.bin)
-#   REPEAT         – number of timed repetitions per configuration (default: 5)
+#   REPEAT         – number of timed repetitions per configuration (default: 10)
+#                    (set to 200 to reproduce the paper's numbers; substantially longer)
 #
 # Output:
 #   experiments/5.3-tpch/results/results_<TIMESTAMP>.csv
@@ -43,7 +44,7 @@ if [[ ! -x "$QUIT_BUILD/run_quit_2k" || ! -x "$QUIT_BUILD/run_quit_base_2k" ]];
 fi
 
 WORKLOAD_FILE="${WORKLOAD_FILE:-$WORKLOAD_DIR/workload_N6000000_K9667_L01.bin}"
-REPEAT="${REPEAT:-5}"
+REPEAT="${REPEAT:-10}"  # paper used 200; increase for publication-quality averages
 N=6000000
 
 SUFFIX=$(date +"%Y%m%d_%H%M%S")

compare_bulkload.sh

@@ -5,17 +5,13 @@
 # for increasing numbers of sorted keys (100M to 2B).  Output corresponds to
 # Figure 11 in the paper: insertion time vs. number of elements.
 #
-# The bulk-load path generates sequential keys 1..N synthetically (independent
-# of the workload file) to guarantee a fully sorted input.  The regular-insert
-# path reads the first N keys from the sorted workload file; sizes larger than
-# the file are skipped automatically.
+# Both regular and bulk-load paths generate sequential keys 1..N synthetically
+# (no workload file required), guaranteeing a fully sorted input.
 #
 # Usage:
 #   bash experiments/5.4-bulkload/run.sh
 #
 # Environment variables (all optional):
-#   WORKLOAD_FILE  – sorted (K=0) .bin workload file
-#                    (default: /scratch/cgokmen/bods/workloads/workload_N500000000_K0_L0.bin)
 #   REPEAT         – number of timed repetitions per configuration (default: 5)
 #
 # Output:
@@ -36,7 +32,6 @@ BUILD="$REPO_ROOT/build"
 # Bootstrap: clone deps, generate workloads, build binaries (idempotent)
 source "$SCRIPT_DIR/../setup.sh"
 
-WORKLOAD_FILE="${WORKLOAD_FILE:-$WORKLOAD_DIR/workload_N500000000_K0_L0.bin}"
 REPEAT="${REPEAT:-5}"
 
 # Test sizes matching Figure 11 (100M to 2B)
@@ -49,10 +44,7 @@ LOG_DIR="$RESULTS_DIR/logs_${SUFFIX}"
 
 mkdir -p "$RESULTS_DIR" "$LOG_DIR"
 
-# Determine number of keys available in the workload file
-FILE_KEYS=$(( $(stat -c%s "$WORKLOAD_FILE") / 4 ))
 echo "Experiment 5.4 — Bulk Loading Performance"
-echo "  WORKLOAD_FILE : $WORKLOAD_FILE  (${FILE_KEYS} keys)"
 echo "  REPEAT        : $REPEAT"
 echo "  Results       : $RESULTS_FILE"
 echo "  Logs          : $LOG_DIR"
@@ -68,8 +60,8 @@ echo "N,insertion_method,avg_insert_ns,avg_query_ns" > "$RESULTS_FILE"
 # ---------------------------------------------------------------------------
 run_config() {
     local N_VAL="$1" METHOD="$2"
-    local EXTRA_ARGS=""
-    [[ "$METHOD" == "bulkload" ]] && EXTRA_ARGS="--bulkload"
+    local EXTRA_ARGS="--synthetic"
+    [[ "$METHOD" == "bulkload" ]] && EXTRA_ARGS="--synthetic --bulkload"
 
     local LOG="$LOG_DIR/N${N_VAL}_${METHOD}.log"
     local INSERT_SUM=0 QUERY_SUM=0 FAILED=0
@@ -79,7 +71,7 @@ run_config() {
     for ((i=1; i<=REPEAT; i++)); do
         echo "--- Run $i/$REPEAT ---" >> "$LOG"
         # shellcheck disable=SC2086
-        OUTPUT=$("$BUILD/run" -f "$WORKLOAD_FILE" -N "$N_VAL" -t ART $EXTRA_ARGS 2>>"$LOG") || true
+        OUTPUT=$("$BUILD/run" -N "$N_VAL" -t ART $EXTRA_ARGS 2>>"$LOG") || true
         STATUS=$?
         echo "$OUTPUT" >> "$LOG"
 
@@ -110,16 +102,10 @@ run_config() {
 for N in "${TEST_SIZES[@]}"; do
     echo ">>> N=$N"
 
-    # Regular insertion requires the file to have enough keys
-    if (( N <= FILE_KEYS )); then
-        echo "  [regular]"
-        run_config "$N" "regular"
-    else
-        echo "  [regular] SKIP — workload file has only ${FILE_KEYS} keys (need $N)" >&2
-        echo "$N,regular,SKIPPED,SKIPPED" >> "$RESULTS_FILE"
-    fi
+    echo "  [regular]"
+    run_config "$N" "regular"
 
-    # Bulk load always works: it generates sequential keys 1..N synthetically
+    # Bulk load uses grouped insertion on the same synthetic key sequence
     echo "  [bulkload]"
     run_config "$N" "bulkload"
 done

experiments/5.2-quart-vs-quit/run.sh

@@ -11,7 +11,7 @@
 #include "trees/QuART_lil.h"
 #include "trees/QuART_tail.h"
 #include "trees/QuART_stail.h"
-#include "ArtNodeBulkLoadMethods.cpp"
+#include "QuArtNodeBulkLoadMethods.cpp"
 
 using namespace std;
 
@@ -47,9 +47,10 @@ std::vector<key_type> read_file(const std::string& filename) {
 int main(int argc, char** argv) {
     bool verbose = false;      // optional argument
     int N = 500000000;         // optional argument
-    string input_file = "/scratch/cgokmen/Qu-ART/tpch/benchmarksql_workdir/workload.txt";         // required argument
+    string input_file = "";                // optional argument (-f)
     string tree_type = "ART";  // default tree type
     bool use_bulkload = false; // optional argument
+    bool use_synthetic = false; // optional argument: generate keys 1..N instead of reading a file
 
     // Query 1% of entries
     uint64_t minval = 0;
@@ -71,6 +72,9 @@ int main(int argc, char** argv) {
         } else if (string(argv[i]) == "--bulkload") {
             use_bulkload = true;
             i++;
+        } else if (string(argv[i]) == "--synthetic") {
+            use_synthetic = true;
+            i++;
         } else {
             i++;
         }
@@ -79,7 +83,13 @@ int main(int argc, char** argv) {
     uint64_t maxval = N-1;
 
     // read data
-    auto keys = read_file<uint32_t>(input_file);
+    std::vector<uint32_t> keys;
+    if (use_synthetic) {
+        keys.resize(N);
+        for (int i = 0; i < N; i++) keys[i] = static_cast<uint32_t>(i + 1);
+    } else {
+        keys = read_file<uint32_t>(input_file);
+    }
 
     if (tree_type == "ART") {
         ART::ART* tree = new ART::ART();