update the ibs for AMD

Author: vickiegpt

README.md

@@ -106,6 +106,17 @@ The memory pool is managed by `SharedMemoryManager`. It can use POSIX shared mem
 
 The server CLI uses local parse-result structs in `src/main_server.cc`, `src/main.cc`, and `src/rob.cc`, so command-line parsing is handled in-tree.
 
+## CPU PMU Compatibility
+
+The application-level simulator uses Linux `perf_event_open()` for sampling. Intel systems use PEBS for load-miss samples and Intel CHA PMUs when the CPU model has a known uncore path. AMD systems use the AMD IBS op PMU when `/sys/bus/event_source/devices/ibs_op/type` is available, then fall back to generic hardware cache-miss sampling if IBS or physical-address sampling is unavailable. Intel CHA counters and LBR accounting are optional on non-Intel systems; unsupported PMU paths are logged and disabled instead of aborting the run.
+
+If monitor setup fails, check the CPU PMU support exposed by the kernel and the perf permission level:
+
+```bash
+cat /proc/sys/kernel/perf_event_paranoid
+ls /sys/bus/event_source/devices/
+```
+
 ## Coherency and Distributed Memory
 
 The distributed path is implemented in:

include/helper.h

@@ -106,10 +106,11 @@ struct LBRElem {
 };
 
 struct CPUInfo {
-    uint32_t max_cpuid;
-    uint32_t cpu_family;
-    uint32_t cpu_model;
-    uint32_t cpu_stepping;
+    char vendor_id[16]{};
+    uint32_t max_cpuid{};
+    uint32_t cpu_family{};
+    uint32_t cpu_model{};
+    uint32_t cpu_stepping{};
 };
 
 struct Elem {

include/lbr.h

@@ -49,7 +49,8 @@ struct lbr_sample {
     uint32_t tid;
     uint64_t timestamp;
     uint32_t cpu;
-    uint64_t nr2;
+    uint32_t cpu_reserved;
+    uint64_t nr;
     lbr lbrs[32];
     cntr counters[32];
 };
@@ -63,6 +64,7 @@ class LBR {
     size_t rdlen{};
     size_t mplen{};
     bool use_pe2 = false;
+    bool sample_has_branch_counters = false;
     perf_event_mmap_page *mp;
     explicit LBR(pid_t, uint64_t);
     ~LBR();

include/monitor.h

@@ -73,8 +73,7 @@ class Monitor {
           injected_delay(other.injected_delay), before(nullptr), // Will be set after copying elements
           after(nullptr), // Will be set after copying elements
           total_delay(other.total_delay), start_exec_ts(other.start_exec_ts), end_exec_ts(other.end_exec_ts),
-          is_process(other.is_process), pebs_ctx(other.pebs_ctx ? new PEBS(*other.pebs_ctx) : nullptr),
-          lbr_ctx(other.lbr_ctx ? new LBR(*other.lbr_ctx) : nullptr) {
+          is_process(other.is_process), pebs_ctx(nullptr), lbr_ctx(nullptr) {
         status.store(other.status.load());
         std::copy(std::begin(other.elem), std::end(other.elem), std::begin(elem));
         before = &elem[0];

include/pebs.h

@@ -30,6 +30,7 @@ class PEBS {
     uint32_t seq{};
     size_t rdlen{};
     size_t mplen{};
+    bool sample_has_phys_addr{};
     perf_event_mmap_page *mp;
     PEBS(pid_t, uint64_t);
     ~PEBS();

qemu_integration/restart_vms_shared.sh

@@ -1,74 +1,74 @@
 #!/bin/bash
-# 重启两个VM使用共享内存
+# Restart two VMs with shared memory
 
 echo "========================================="
-echo "重启VM以使用共享内存配置"
+echo "Restart VMs with shared-memory configuration"
 echo "========================================="
 echo ""
 
-# 找到QEMU进程并杀掉
-echo "[1/5] 停止现有VM..."
+# Find and stop QEMU processes
+echo "[1/5] Stopping existing VMs..."
 pkill -9 qemu-system-x86_64
 sleep 3
 
-# 确保cxlmemsim_server正在运行
-echo "[2/5] 检查cxlmemsim_server..."
+# Ensure cxlmemsim_server is running
+echo "[2/5] Checking cxlmemsim_server..."
 if ! pgrep -x cxlmemsim_server > /dev/null; then
-    echo "  错误: cxlmemsim_server未运行"
-    echo "  请先启动server"
+    echo "  ERROR: cxlmemsim_server is not running"
+    echo "  Start the server first"
     exit 1
 fi
-echo "  ✓ Server正在运行"
+echo "  OK: server is running"
 
-# 检查共享内存文件
-echo "[3/5] 检查共享内存文件..."
+# Check the shared-memory file
+echo "[3/5] Checking shared-memory file..."
 if [ ! -f /dev/shm/cxlmemsim_shared ]; then
-    echo "  错误: /dev/shm/cxlmemsim_shared 不存在"
+    echo "  ERROR: /dev/shm/cxlmemsim_shared does not exist"
     exit 1
 fi
-echo "  ✓ 共享内存文件存在: $(ls -lh /dev/shm/cxlmemsim_shared | awk '{print $5}')"
+echo "  OK: shared-memory file exists: $(ls -lh /dev/shm/cxlmemsim_shared | awk '{print $5}')"
 
-# 创建LSA文件（如果不存在）
-echo "[4/5] 准备LSA文件..."
+# Create LSA files if they do not exist
+echo "[4/5] Preparing LSA files..."
 for i in 0 1; do
     LSA_FILE="/tmp/lsa${i}.raw"
     if [ ! -f "$LSA_FILE" ]; then
-        echo "  创建 $LSA_FILE..."
+        echo "  Creating $LSA_FILE..."
         truncate -s 256M "$LSA_FILE"
     fi
 done
-echo "  ✓ LSA文件准备完成"
+echo "  OK: LSA files are ready"
 
-# 启动VM
-echo "[5/5] 启动VM..."
+# Start VMs
+echo "[5/5] Starting VMs..."
 echo ""
-echo "启动Node 0..."
+echo "Starting Node 0..."
 cd /home/yhgan913/CXLMemSim/qemu_integration
 nohup ./launch_qemu_cxl.sh > /tmp/qemu0.log 2>&1 &
 QEMU0_PID=$!
 echo "  PID: $QEMU0_PID"
 
 sleep 5
 
-echo "启动Node 1..."
+echo "Starting Node 1..."
 nohup ./launch_qemu_cxl1.sh > /tmp/qemu1.log 2>&1 &
 QEMU1_PID=$!
 echo "  PID: $QEMU1_PID"
 
 echo ""
 echo "========================================="
-echo "VM已启动"
+echo "VMs started"
 echo "========================================="
 echo "Node 0 PID: $QEMU0_PID"
 echo "Node 1 PID: $QEMU1_PID"
 echo ""
-echo "查看日志:"
+echo "View logs:"
 echo "  tail -f /tmp/qemu0.log"
 echo "  tail -f /tmp/qemu1.log"
 echo ""
-echo "等待10秒让VM启动..."
+echo "Waiting 10 seconds for VM boot..."
 sleep 10
 
-echo "检查VM状态..."
-ping -c 1 -W 2 192.168.100.10 > /dev/null && echo "  ✓ Node 0 (192.168.100.10) 可达" || echo "  ✗ Node 0 不可达"
-ping -c 1 -W 2 192.168.100.11 > /dev/null && echo "  ✓ Node 1 (192.168.100.11) 可达" || echo "  ✗ Node 1 不可达"
+echo "Checking VM status..."
+ping -c 1 -W 2 192.168.100.10 > /dev/null && echo "  OK: Node 0 (192.168.100.10) reachable" || echo "  FAIL: Node 0 unreachable"
+ping -c 1 -W 2 192.168.100.11 > /dev/null && echo "  OK: Node 1 (192.168.100.11) reachable" || echo "  FAIL: Node 1 unreachable"

qemu_integration/test_shared_memory.sh

@@ -1,145 +1,145 @@
 #!/bin/bash
-# 测试共享内存是否工作
+# Test whether shared memory is working
 
 echo "========================================="
-echo "测试共享内存配置"
+echo "Testing shared-memory configuration"
 echo "========================================="
 echo ""
 
-# 1. 检查host共享内存
-echo "[1/6] 检查host共享内存文件..."
+# 1. Check host shared memory
+echo "[1/6] Checking host shared-memory file..."
 if [ ! -f /dev/shm/cxlmemsim_shared ]; then
-    echo "  ✗ 错误: /dev/shm/cxlmemsim_shared 不存在"
+    echo "  ERROR: /dev/shm/cxlmemsim_shared does not exist"
     exit 1
 fi
-echo "  ✓ 文件存在: $(ls -lh /dev/shm/cxlmemsim_shared | awk '{print $5}')"
+echo "  OK: file exists: $(ls -lh /dev/shm/cxlmemsim_shared | awk '{print $5}')"
 
-# 2. 检查哪些进程在使用
-echo "[2/6] 检查使用共享内存的进程..."
+# 2. Check which processes are using it
+echo "[2/6] Checking processes using shared memory..."
 USERS=$(lsof /dev/shm/cxlmemsim_shared 2>/dev/null | tail -n +2 | awk '{print $1}' | sort -u)
 if [ -z "$USERS" ]; then
-    echo "  ✗ 警告: 没有进程在使用共享内存"
+    echo "  WARN: no process is using shared memory"
 else
-    echo "  ✓ 使用中的进程:"
+    echo "  OK: processes using shared memory:"
     echo "$USERS" | while read proc; do
         COUNT=$(lsof /dev/shm/cxlmemsim_shared 2>/dev/null | grep -c "$proc")
-        echo "    - $proc (${COUNT}个实例)"
+        echo "    - $proc (${COUNT} instances)"
     done
 fi
 echo ""
 
-# 3. 检查VM可达性
-echo "[3/6] 检查VM网络连接..."
+# 3. Check VM reachability
+echo "[3/6] Checking VM network connectivity..."
 if ping -c 1 -W 2 192.168.100.10 > /dev/null 2>&1; then
-    echo "  ✓ Node 0 (192.168.100.10) 可达"
+    echo "  OK: Node 0 (192.168.100.10) reachable"
     NODE0_UP=1
 else
-    echo "  ✗ Node 0 (192.168.100.10) 不可达"
+    echo "  FAIL: Node 0 (192.168.100.10) unreachable"
     NODE0_UP=0
 fi
 
 if ping -c 1 -W 2 192.168.100.11 > /dev/null 2>&1; then
-    echo "  ✓ Node 1 (192.168.100.11) 可达"
+    echo "  OK: Node 1 (192.168.100.11) reachable"
     NODE1_UP=1
 else
-    echo "  ✗ Node 1 (192.168.100.11) 不可达"
+    echo "  FAIL: Node 1 (192.168.100.11) unreachable"
     NODE1_UP=0
 fi
 echo ""
 
 if [ $NODE0_UP -eq 0 ] || [ $NODE1_UP -eq 0 ]; then
-    echo "错误: 一个或多个VM不可达，无法继续测试"
+    echo "ERROR: one or more VMs are unreachable; cannot continue"
     exit 1
 fi
 
-# 4. 检查DAX设备
-echo "[4/6] 检查VM内DAX设备..."
+# 4. Check DAX devices
+echo "[4/6] Checking DAX devices inside VMs..."
 NODE0_DAX=$(ssh root@192.168.100.10 "ls -l /dev/dax0.0 2>&1" | grep -c "^c")
 NODE1_DAX=$(ssh root@192.168.100.11 "ls -l /dev/dax0.0 2>&1" | grep -c "^c")
 
 if [ $NODE0_DAX -eq 1 ]; then
-    echo "  ✓ Node 0: /dev/dax0.0 存在"
+    echo "  OK: Node 0: /dev/dax0.0 exists"
 else
-    echo "  ✗ Node 0: /dev/dax0.0 不存在"
+    echo "  FAIL: Node 0: /dev/dax0.0 does not exist"
 fi
 
 if [ $NODE1_DAX -eq 1 ]; then
-    echo "  ✓ Node 1: /dev/dax0.0 存在"
+    echo "  OK: Node 1: /dev/dax0.0 exists"
 else
-    echo "  ✗ Node 1: /dev/dax0.0 不存在"
+    echo "  FAIL: Node 1: /dev/dax0.0 does not exist"
 fi
 echo ""
 
-# 5. 测试共享内存写入/读取
-echo "[5/6] 测试共享内存读写..."
+# 5. Test shared-memory write/read
+echo "[5/6] Testing shared-memory write/read..."
 TEST_STRING="SHARED_MEMORY_TEST_$(date +%s)"
-echo "  写入测试字符串到Node 0: $TEST_STRING"
+echo "  Writing test string to Node 0: $TEST_STRING"
 
-# 在Node 0写入
+# Write on Node 0
 ssh root@192.168.100.10 "echo -n '$TEST_STRING' | dd of=/dev/dax0.0 bs=1 seek=1024 2>/dev/null"
 sleep 1
 
-# 在Node 1读取
+# Read on Node 1
 RESULT=$(ssh root@192.168.100.11 "dd if=/dev/dax0.0 bs=1 skip=1024 count=${#TEST_STRING} 2>/dev/null")
 
 if [ "$RESULT" = "$TEST_STRING" ]; then
-    echo "  ✓ 成功！Node 1 读取到Node 0写入的数据"
-    echo "    写入: $TEST_STRING"
-    echo "    读取: $RESULT"
+    echo "  OK: Node 1 read data written by Node 0"
+    echo "    wrote: $TEST_STRING"
+    echo "    read: $RESULT"
     SHARED_WORKS=1
 else
-    echo "  ✗ 失败！共享内存不工作"
-    echo "    写入: $TEST_STRING"
-    echo "    读取: $RESULT"
+    echo "  FAIL: shared memory is not working"
+    echo "    wrote: $TEST_STRING"
+    echo "    read: $RESULT"
     SHARED_WORKS=0
 fi
 echo ""
 
-# 6. 测试反向（Node 1写，Node 0读）
-echo "[6/6] 测试反向读写..."
+# 6. Test reverse direction (Node 1 writes, Node 0 reads)
+echo "[6/6] Testing reverse write/read..."
 TEST_STRING2="REVERSE_TEST_$(date +%s)"
-echo "  写入测试字符串到Node 1: $TEST_STRING2"
+echo "  Writing test string to Node 1: $TEST_STRING2"
 
-# 在Node 1写入
+# Write on Node 1
 ssh root@192.168.100.11 "echo -n '$TEST_STRING2' | dd of=/dev/dax0.0 bs=1 seek=2048 2>/dev/null"
 sleep 1
 
-# 在Node 0读取
+# Read on Node 0
 RESULT2=$(ssh root@192.168.100.10 "dd if=/dev/dax0.0 bs=1 skip=2048 count=${#TEST_STRING2} 2>/dev/null")
 
 if [ "$RESULT2" = "$TEST_STRING2" ]; then
-    echo "  ✓ 成功！Node 0 读取到Node 1写入的数据"
-    echo "    写入: $TEST_STRING2"
-    echo "    读取: $RESULT2"
+    echo "  OK: Node 0 read data written by Node 1"
+    echo "    wrote: $TEST_STRING2"
+    echo "    read: $RESULT2"
     REVERSE_WORKS=1
 else
-    echo "  ✗ 失败！反向共享不工作"
-    echo "    写入: $TEST_STRING2"
-    echo "    读取: $RESULT2"
+    echo "  FAIL: reverse shared-memory path is not working"
+    echo "    wrote: $TEST_STRING2"
+    echo "    read: $RESULT2"
     REVERSE_WORKS=0
 fi
 echo ""
 
-# 总结
+# Summary
 echo "========================================="
-echo "测试总结"
+echo "Test summary"
 echo "========================================="
 if [ $SHARED_WORKS -eq 1 ] && [ $REVERSE_WORKS -eq 1 ]; then
-    echo "✓ 共享内存配置正确！"
+    echo "OK: shared-memory configuration is correct"
     echo ""
-    echo "可以运行Tigon多节点测试了："
+    echo "You can now run the Tigon multi-node test:"
     echo "  cd /home/yhgan913/CXLMemSim/workloads/tigon"
     echo "  export CXL_BACKEND=dax"
     echo "  export CXL_MEMORY_RESOURCE=/dev/dax0.0"
     echo "  ./scripts/run.sh TPCC TwoPLPasha 2 3 mixed 10 15 1 0 1 Clock OnDemand 200000000 1 WriteThrough None 30 10 BLACKHOLE 20000 0 0"
     exit 0
 else
-    echo "✗ 共享内存配置有问题"
+    echo "FAIL: shared-memory configuration has a problem"
     echo ""
-    echo "故障排查："
-    echo "  1. 确认两个VM都使用 /dev/shm/cxlmemsim_shared"
-    echo "  2. 检查 lsof /dev/shm/cxlmemsim_shared"
-    echo "  3. 查看 /tmp/qemu0.log 和 /tmp/qemu1.log"
-    echo "  4. 重启VM: ./restart_vms_shared.sh"
+    echo "Troubleshooting:"
+    echo "  1. Confirm both VMs use /dev/shm/cxlmemsim_shared"
+    echo "  2. Check lsof /dev/shm/cxlmemsim_shared"
+    echo "  3. Review /tmp/qemu0.log and /tmp/qemu1.log"
+    echo "  4. Restart VMs: ./restart_vms_shared.sh"
     exit 1
 fi