diff --git a/scripts/perfcct.py b/scripts/perfcct.py
index 70ae8b61740..b939f739ee6 100644
--- a/scripts/perfcct.py
+++ b/scripts/perfcct.py
@@ -1,92 +1,268 @@
-import sqlite3 as sql
+#!/usr/bin/env python3
+
 import argparse
+import sqlite3 as sql
+import subprocess
+import tempfile
+
 
+STAGES = {
+    "AtFetch": "f",
+    "AtDecode": "d",
+    "AtRename": "r",
+    "AtDispQue": "D",
+    "AtIssueQue": "i",
+    "AtIssueArb": "a",
+    "AtIssueReadReg": "g",
+    "AtFU": "e",
+    "AtBypassVal": "b",
+    "AtWriteVal": "w",
+    "AtCommit": "c",
+}
+
+LATENCIES = (
+    ("issue-to-readreg", "AtIssueQue", "AtIssueReadReg"),
+    ("readreg-to-fu", "AtIssueReadReg", "AtFU"),
+    ("issue-to-writeback", "AtIssueQue", "AtWriteVal"),
+    ("issue-to-commit", "AtIssueQue", "AtCommit"),
+)
+
+
+class DasmQuery:
+    def __init__(self, cmd="riscv64-linux-gnu-objdump", enable_cache=True):
+        self.cmd = cmd
+        self.cache = {} if enable_cache else None
+        if cmd.endswith("objdump"):
+            self.mode = "objdump"
+            self.spike = None
+        elif cmd == "spike-dasm":
+            self.mode = "spike-dasm"
+            self.spike = subprocess.Popen(
+                [self.cmd],
+                stdin=subprocess.PIPE,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                text=True,
+            )
+        else:
+            raise ValueError(f"unsupported disassembler: {cmd}")
 
+    def close(self):
+        if self.spike is not None:
+            self.spike.terminate()
+            self.spike.wait()
+            self.spike = None
 
-parser = argparse.ArgumentParser()
-parser.add_argument('sqldb')
-parser.add_argument('-v', '--visual', action='store_true', default=False)
-parser.add_argument('-z', '--zoom', action='store', type=float, default=1)
-parser.add_argument('-p', '--period', action='store', default=333)
+    def dasm(self, instr):
+        if self.cache is not None and instr in self.cache:
+            return self.cache[instr]
+        if self.mode == "spike-dasm":
+            self.spike.stdin.write(f"DASM(0x{instr:x})\n")
+            self.spike.stdin.flush()
+            res = self.spike.stdout.readline().strip()
+        else:
+            with tempfile.NamedTemporaryFile() as f:
+                f.write(instr.to_bytes(4, byteorder="little"))
+                f.flush()
+                proc = subprocess.run(
+                    [self.cmd, "-b", "binary", "-m", "riscv:rv64", "-M,max", "-D", f.name],
+                    capture_output=True,
+                    text=True,
+                    check=False,
+                )
+                res = ""
+                for line in proc.stdout.splitlines():
+                    if line.startswith("   0:\t"):
+                        res = "\t".join(line.split("\t")[2:])
+                        break
+        if self.cache is not None:
+            self.cache[instr] = res
+        return res
 
-args = parser.parse_args()
 
-sqldb = args.sqldb
+def parse_int(value):
+    return int(value, 0)
 
-tick_per_cycle = int(args.period)
-cycle_per_line = int(100 * args.zoom)
 
-stages = ['f','d','r','D','i','a','g','e','b','w','c']
+def unsigned64(value):
+    return value + (1 << 64) if value < 0 else value
 
 
 def non_stage():
-    return '.'
-
-def stage(x):
-    return stages[x]
-
-def dump_visual(pos, records):
-    pos_start = pos[0] % cycle_per_line
-    line = ''
-    line += '[' + non_stage() * pos_start
-    pos_next = pos_start
-    last_index = 0
-    for i in range(1, len(pos)):
-        if (pos[i] == pos[last_index]) or pos[i] == 0:
+    return "."
+
+
+def stage(name):
+    return STAGES[name]
+
+
+def record_header(records, dasm):
+    instr = records.get("DisAsm", "")
+    if dasm is not None and isinstance(instr, int):
+        instr = f"{instr:08x} {dasm.dasm(instr)}"
+    pc = records.get("PC", "")
+    sn = records.get("SN", "")
+    uop_idx = records.get("UopIdx", "")
+    return f"sn={sn} uop={uop_idx} pc={pc}: {instr}"
+
+
+def dump_visual(pos, records, args, dasm):
+    empty_line = list(f"[{non_stage() * args.cycle_per_line}]")
+    line_buffer = empty_line.copy()
+    line = ""
+    last_key = None
+    last_line = None
+
+    def fill_line_pos(offset, char):
+        line_buffer[1 + offset] = char
+
+    def commit_last_line(offset=None):
+        nonlocal line, line_buffer
+        if offset is None or (
+            last_line is not None
+            and offset // args.cycle_per_line != last_line
+            and line_buffer != empty_line
+        ):
+            prefix = f"{last_line * args.cycle_per_line:07d}:" if args.cycle else ""
+            line += prefix + "".join(line_buffer) + "\n"
+            line_buffer = empty_line.copy()
+
+    for key, value in sorted(pos.items(), key=lambda item: item[1]):
+        if value == 0:
             continue
-        if pos[i] - pos[last_index] >= cycle_per_line - pos_next:
-            diff = cycle_per_line - pos_next
-            line += f'{stage(last_index)}' * diff + ']\n'
-            diff_line = ((pos[i] - pos[last_index]) - diff - 1) // cycle_per_line
-            if diff_line > 0:
-                line += '[' + f'{stage(last_index)}' * cycle_per_line + ']\n'
-
-            pos_next = pos[i] % cycle_per_line
-            line += '[' + f'{stage(last_index)}' * pos_next
-        else:
-            diff = pos[i] - pos[last_index]
-            pos_next = pos[i] % cycle_per_line
-            line += f'{stage(last_index)}' * diff
-        last_index = i
-    if cycle_per_line - pos_next == 0:
-        line += ']\n'
-        line += f'[{stage(i)}{non_stage() * (cycle_per_line - 1)}]\n'
+        if last_key is not None:
+            if args.singleline:
+                for i in range(pos[last_key], value):
+                    fill_line_pos(i % args.cycle_per_line, stage(last_key))
+            else:
+                for i in range(pos[last_key], value):
+                    commit_last_line(i)
+                    fill_line_pos(i % args.cycle_per_line, stage(last_key))
+                    last_line = i // args.cycle_per_line
+        last_key = key
+
+    if last_key is None:
+        print(record_header(records, dasm))
+        return
+    if not args.singleline and last_line is not None:
+        commit_last_line(pos[last_key])
     else:
-        line += f'{stage(i)}' + non_stage() * (cycle_per_line - pos_next - 1) + ']'
-    line += str(records)
-    print(line)
+        last_line = pos[last_key] // args.cycle_per_line
+    fill_line_pos(pos[last_key] % args.cycle_per_line, stage(last_key))
+    commit_last_line()
+    line = line.strip()
+    print(f"{line} {record_header(records, dasm)}")
+
+
+def dump_txt(pos, records, args, dasm):
+    for key in STAGES:
+        if key in pos:
+            print(f"{stage(key)}{pos[key]}", end=" ")
+    print(record_header(records, dasm))
+
 
+def dump_latencies(pos):
+    parts = []
+    for name, start, end in LATENCIES:
+        if pos.get(start, 0) != 0 and pos.get(end, 0) != 0:
+            parts.append(f"{name}={pos[end] - pos[start]}")
+    if parts:
+        print("  " + " ".join(parts))
 
-def dump_txt(pos, records):
-    for i in range(len(pos)):
-        print(f'{stage(i)}{pos[i]}', end=' ')
-    print(records)
 
+def build_query(args):
+    where = []
+    params = []
+    if not args.spec or args.committed:
+        where.append("AtCommit != 0")
+    if args.pc is not None:
+        where.append("PC = ?")
+        params.append(args.pc)
+    if args.sn is not None:
+        where.append("SN = ?")
+        params.append(args.sn)
+    if args.uop_idx is not None:
+        where.append("UopIdx = ?")
+        params.append(args.uop_idx)
+    clause = " WHERE " + " AND ".join(where) if where else ""
+    return f"SELECT * FROM LifeTimeCommitTrace{clause} ORDER BY SN, UopIdx, ID", params
 
-dump = dump_txt
-if args.visual:
-    dump = dump_visual
 
-with sql.connect(sqldb) as con:
-    cur = con.cursor()
-    cur.execute("SELECT * FROM LifeTimeCommitTrace")
-    col_name = [i[0] for i in cur.description]
-    col_name = col_name[1:]
-    col_name = [i.lower() for i in col_name]
-    rows = cur.fetchall()
+def select_index(rows, col_names, index):
+    if index is None:
+        return rows
+    sn_idx = col_names.index("SN")
+    selected = None
+    seen = []
     for row in rows:
-        row = row[1:]
-        pos = []
-        records = []
-        i = 0
-        for val in row:
-            if col_name[i].startswith('at'):
-                pos.append(val//tick_per_cycle)
-            elif col_name[i].startswith('pc'):
-                if val < 0:
-                    val = val + 1 << 64
-                records.append(hex(val))
-            else:
-                records.append(val)
-            i += 1
-        dump(pos, records)
\ No newline at end of file
+        sn = row[sn_idx]
+        if sn not in seen:
+            seen.append(sn)
+        if len(seen) == index:
+            selected = sn
+            break
+    if selected is None:
+        return []
+    return [row for row in rows if row[sn_idx] == selected]
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("sqldb")
+    parser.add_argument("-v", "--visual", action="store_true", default=False)
+    parser.add_argument("-z", "--zoom", action="store", type=float, default=1)
+    parser.add_argument("-p", "--period", action="store", type=int, default=1)
+    parser.add_argument(
+        "-d",
+        "--dasm",
+        action="store",
+        help='disassembler command, "spike-dasm" or an objdump path',
+        default=None,
+    )
+    parser.add_argument("-s", "--spec", action="store_true", default=False, help="show speculative execution")
+    parser.add_argument("-l", "--singleline", action="store_true", default=False, help="single line per instruction")
+    parser.add_argument("-c", "--cycle", action="store_true", default=False, help="show cycle number in visual mode")
+    parser.add_argument("--pc", type=parse_int, help="filter by PC, for example 0x80000e48")
+    parser.add_argument("--sn", type=parse_int, help="filter by sequence number")
+    parser.add_argument("--uop-idx", type=parse_int, help="filter by uop index")
+    parser.add_argument("--committed", action="store_true", default=False, help="show committed records only")
+    parser.add_argument("--index", type=int, help="select the Nth dynamic instruction after filters, 1-based")
+    parser.add_argument("--limit", type=int, help="limit number of output rows")
+    parser.add_argument("--latency", action="store_true", default=False, help="print common per-uop latency metrics")
+    args = parser.parse_args()
+    args.cycle_per_line = int(100 * args.zoom)
+
+    dasm = DasmQuery(args.dasm) if args.dasm is not None else None
+    try:
+        with sql.connect(args.sqldb) as con:
+            cur = con.cursor()
+            query, params = build_query(args)
+            cur.execute(query, params)
+            col_names = [col[0] for col in cur.description]
+            rows = select_index(cur.fetchall(), col_names, args.index)
+            if args.limit is not None:
+                rows = rows[: args.limit]
+
+            dump = dump_visual if args.visual else dump_txt
+            for row in rows:
+                pos = {}
+                records = {}
+                for name, val in zip(col_names, row):
+                    if name == "ID":
+                        continue
+                    if name in STAGES:
+                        pos[name] = val // args.period
+                    elif name == "PC":
+                        records[name] = f"{unsigned64(val):016x}"
+                    else:
+                        records[name] = val
+                dump(pos, records, args, dasm)
+                if args.latency:
+                    dump_latencies(pos)
+    finally:
+        if dasm is not None:
+            dasm.close()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/src/main/scala/xiangshan/backend/decode/DecodeStage.scala b/src/main/scala/xiangshan/backend/decode/DecodeStage.scala
index 17de49133c6..6bb3fdb4ea3 100644
--- a/src/main/scala/xiangshan/backend/decode/DecodeStage.scala
+++ b/src/main/scala/xiangshan/backend/decode/DecodeStage.scala
@@ -260,9 +260,7 @@ class DecodeStage(implicit p: Parameters) extends XSModule
       SrcType.isVp(s) && (l === 0.U)
     }.reduce(_ || _)
     inst.bits.srcType(3) := Mux(srcType0123HasV0, SrcType.v0, finalDecodedInst(i).srcType(3))
-    when (inst.bits.uopIdx =/= 0.U) {
-      inst.bits.debug_seqNum := 0.U.asTypeOf(new InstSeqNum)
-    }
+    inst.bits.debug_seqNum.uopIdx := inst.bits.uopIdx
   }
 
   io.out.map(x =>