diff --git a/Makefile b/Makefile
index 855c5ada75c9cda2194a3c2ba73b613a4b6d3b9c..7418a10d2292798eecd6b0c6b547500b0759cf81 100644
--- a/Makefile
+++ b/Makefile
@@ -63,7 +63,6 @@ EMU_CXXFLAGS  = -std=c++11 -static -Wall -I$(EMU_CSRC_DIR)
 EMU_CXXFLAGS += -DVERILATOR -Wno-maybe-uninitialized
 EMU_LDFLAGS   = -lpthread -lSDL2 -ldl
 
-# dump vcd: --debug --trace
 VERILATOR_FLAGS = --top-module $(SIM_TOP) \
   +define+VERILATOR=1 \
   +define+PRINTF_COND=1 \
@@ -72,6 +71,7 @@ VERILATOR_FLAGS = --top-module $(SIM_TOP) \
   --assert \
   --savable \
   --stats-vars \
+  --trace \
   --output-split 5000 \
   --output-split-cfuncs 5000 \
   -I$(abspath $(BUILD_DIR)) \
@@ -115,7 +115,7 @@ else
 SNAPSHOT_OPTION = --load-snapshot=$(REMOTE_PREFIX)/$(SNAPSHOT)
 endif
 
-EMU_FLAGS = -s $(SEED) -b $(B) -e $(E) $(SNAPSHOT_OPTION)
+EMU_FLAGS = -s $(SEED) -b $(B) -e $(E) $(SNAPSHOT_OPTION) --dump-wave
 
 emu: $(EMU)
 ifeq ($(REMOTE),localhost)
diff --git a/src/test/csrc/emu.cpp b/src/test/csrc/emu.cpp
index 98295f4d00945ef1f3563de4214e152b67893fd6..7e29b9a864387acd24a09f550cf9f2d9e59e1056 100644
--- a/src/test/csrc/emu.cpp
+++ b/src/test/csrc/emu.cpp
@@ -25,10 +25,20 @@ Emulator::Emulator(EmuArgs &args):
 
   init_difftest();
 
+  enable_waveform = args.enable_waveform;
+  if (enable_waveform) {
+    Verilated::traceEverOn(true);	// Verilator must compute traced signals
+    tfp = new VerilatedVcdC;
+    dut_ptr->trace(tfp, 99);	// Trace 99 levels of hierarchy
+    time_t now = time(NULL);
+    tfp->open(waveform_filename(now));	// Open the dump file
+  }
+
   // init core
   reset_ncycles(10);
 
   if (args.snapshot_path != NULL) {
+    printf("loading from snapshot `%s`...\n", args.snapshot_path);
     snapshot_load(args.snapshot_path);
     hascommit = 1;
   }
@@ -93,9 +103,13 @@ inline void Emulator::single_cycle() {
   dut_ptr->clock = 1;
   dut_ptr->eval();
 
-#if VM_TRACE
-  tfp->dump(cycles);
-#endif
+  if (enable_waveform) {
+    uint64_t cycle = dut_ptr->io_trap_cycleCnt;
+    uint64_t begin = dut_ptr->io_logCtrl_log_begin;
+    uint64_t end   = dut_ptr->io_logCtrl_log_end;
+    bool in_range = (begin <= cycle) && (cycle <= end);
+    if (in_range) { tfp->dump(cycle); }
+  }
 
   if (dut_ptr->io_uart_out_valid) {
     printf("%c", dut_ptr->io_uart_out_ch);
@@ -127,14 +141,6 @@ uint64_t Emulator::execute(uint64_t n) {
   diff.wdata = wdata;
   diff.wdst = wdst;
 
-#if VM_TRACE
-  Verilated::traceEverOn(true);	// Verilator must compute traced signals
-  VL_PRINTF("Enabling waves...\n");
-  tfp = new VerilatedVcdC;
-  dut_ptr->trace(tfp, 99);	// Trace 99 levels of hierarchy
-  tfp->open("vlt_dump.vcd");	// Open the dump file
-#endif
-
   while (trapCode == STATE_RUNNING && n > 0) {
     single_cycle();
     n --;
@@ -146,9 +152,6 @@ uint64_t Emulator::execute(uint64_t n) {
       eprintf("No instruction commits for %d cycles, maybe get stuck\n"
           "(please also check whether a fence.i instruction requires more than %d cycles to flush the icache)\n",
           stuck_limit, stuck_limit);
-#if VM_TRACE
-      tfp->close();
-#endif
       difftest_display(dut_ptr->io_difftest_priviledgeMode);
       trapCode = STATE_ABORT;
     }
@@ -176,9 +179,6 @@ uint64_t Emulator::execute(uint64_t n) {
       diff.priviledgeMode = dut_ptr->io_difftest_priviledgeMode;
 
       if (difftest_step(&diff)) {
-#if VM_TRACE
-        tfp->close();
-#endif
         trapCode = STATE_ABORT;
       }
       lastcommit = n;
@@ -197,17 +197,31 @@ uint64_t Emulator::execute(uint64_t n) {
     }
   }
 
+  if (enable_waveform) tfp->close();
   display_trapinfo();
   return cycles;
 }
 
-inline char* Emulator::snapshot_filename(time_t t) {
-  static char buf[1024];
+inline char* Emulator::timestamp_filename(time_t t, char *buf) {
   char buf_time[64];
   strftime(buf_time, sizeof(buf_time), "%F@%T", localtime(&t));
   char *noop_home = getenv("NOOP_HOME");
   assert(noop_home != NULL);
-  snprintf(buf, 1024, "%s/build/%s.snapshot", noop_home, buf_time);
+  int len = snprintf(buf, 1024, "%s/build/%s", noop_home, buf_time);
+  return buf + len;
+}
+
+inline char* Emulator::snapshot_filename(time_t t) {
+  static char buf[1024];
+  char *p = timestamp_filename(t, buf);
+  strcpy(p, ".snapshot");
+  return buf;
+}
+
+inline char* Emulator::waveform_filename(time_t t) {
+  static char buf[1024];
+  char *p = timestamp_filename(t, buf);
+  strcpy(p, ".vcd");
   return buf;
 }
 
diff --git a/src/test/csrc/emu.h b/src/test/csrc/emu.h
index 1a2bfbe2f10eec681913325343fee2b99f44a648..90e2aba44d8cbdbb38e6cb00eac6c1c9cbcd4758 100644
--- a/src/test/csrc/emu.h
+++ b/src/test/csrc/emu.h
@@ -1,10 +1,7 @@
 #include "common.h"
 #include "snapshot.h"
 #include "VXSSimTop.h"
-
-#if VM_TRACE
 #include <verilated_vcd_c.h>	// Trace file format header
-#endif
 
 #define DIFFTEST_WIDTH 6
 #define SNAPSHOT_INTERVAL 10 // unit: second
@@ -15,6 +12,7 @@ struct EmuArgs {
   uint64_t log_begin, log_end;
   const char *image;
   const char *snapshot_path;
+  bool enable_waveform;
 
   EmuArgs() {
     seed = 0;
@@ -23,14 +21,14 @@ struct EmuArgs {
     log_end = -1;
     snapshot_path = NULL;
     image = NULL;
+    enable_waveform = false;
   }
 };
 
 class Emulator {
   VXSSimTop *dut_ptr;
-#if VM_TRACE
   VerilatedVcdC* tfp;
-#endif
+  bool enable_waveform;
   VerilatedSaveMem snapshot_slot[2];
 
   enum {
@@ -50,9 +48,11 @@ class Emulator {
   inline void reset_ncycles(size_t cycles);
   inline void single_cycle();
   void display_trapinfo();
+  inline char* timestamp_filename(time_t t, char *buf);
   inline char* snapshot_filename(time_t t);
   void snapshot_save(const char *filename);
   void snapshot_load(const char *filename);
+  inline char* waveform_filename(time_t t);
 
 public:
   Emulator(EmuArgs &args);
diff --git a/src/test/csrc/main.cpp b/src/test/csrc/main.cpp
index a7c5f99feafce22259b9bd6708fb8efd47ba67ee..1cec6474384ce91ac6190f3a2b92e0aa4d2ad224 100644
--- a/src/test/csrc/main.cpp
+++ b/src/test/csrc/main.cpp
@@ -16,6 +16,8 @@ static inline void print_help(const char *file) {
   printf("  -i, --image=FILE      run with this image file\n");
   printf("  -b, --log-begin=NUM   display log from NUM th cycle\n");
   printf("  -e, --log-end=NUM     stop display log at NUM th cycle\n");
+  printf("      --load-snapshot=PATH   load snapshot from PATH\n");
+  printf("      --dump-wave       dump waveform when log is enabled\n");
   printf("  -h, --help            print program help info\n");
   printf("\n");
 }
@@ -25,6 +27,7 @@ static inline EmuArgs parse_args(int argc, const char *argv[]) {
   int long_index = 0;
   const struct option long_options[] = {
     { "load-snapshot",  1, NULL,  0  },
+    { "dump-wave",      0, NULL,  0  },
     { "seed",           1, NULL, 's' },
     { "max-cycles",     1, NULL, 'C' },
     { "image",          1, NULL, 'i' },
@@ -39,9 +42,9 @@ static inline EmuArgs parse_args(int argc, const char *argv[]) {
           "-s:C:hi:m:b:e:", long_options, &long_index)) != -1) {
     switch (o) {
       case 0:
-        if (long_index == 0) {
-          args.snapshot_path = optarg;
-          break;
+        switch (long_index) {
+          case 0: args.snapshot_path = optarg; continue;
+          case 1: args.enable_waveform = true; continue;
         }
         // fall through
       default: