[rocprofiler-compute] Support new attach/detach API (#2642)
* Removed attach tool library path * Support new attach/detach API * New attach/detach API was introduced in https://github.com/ROCm/rocm-systems/pull/1653 * Provide backward compatibility with old api * Stabilize attach/detach tests by adding sleep to help workload get ready for attachment * Fix typo in test name --------- Co-authored-by: Vignesh Edithal <Vignesh.Edithal@amd.com> Co-authored-by: Fei Zheng <44449748+feizheng10@users.noreply.github.com>
Этот коммит содержится в:
abchoudh-amd
коммит произвёл
GitHub
GitHub
родитель
1c5aa2d4e7
Коммит
dd149d3957
@@ -92,7 +92,6 @@ class rocprofiler_sdk_profiler(RocProfCompute_Base):
|
||||
/ "librocprofiler-sdk-rocattach.so"
|
||||
)
|
||||
options.update({
|
||||
"ROCPROF_ATTACH_TOOL_LIBRARY": rocprofiler_sdk_tool_path,
|
||||
"ROCPROF_ATTACH_LIBRARY": rocprofiler_attach_library_path,
|
||||
"ROCPROF_ATTACH_PID": args.attach_pid,
|
||||
})
|
||||
|
||||
@@ -235,6 +235,101 @@ def detect_rocprof(args: argparse.Namespace) -> str:
|
||||
return rocprof_cmd
|
||||
|
||||
|
||||
def perform_attach_detach(new_env: dict[str, str], options: dict[str, Any]) -> None:
|
||||
@contextmanager
|
||||
def temporary_env(env_vars: dict[str, str]) -> Generator[None, None, None]:
|
||||
"""
|
||||
Temporarily change the environment variable of this application.
|
||||
"""
|
||||
original_env = os.environ.copy()
|
||||
os.environ.update({k: str(v) for k, v in env_vars.items()})
|
||||
try:
|
||||
yield
|
||||
finally:
|
||||
os.environ.clear()
|
||||
os.environ.update(original_env)
|
||||
|
||||
with temporary_env(new_env):
|
||||
libname = options["ROCPROF_ATTACH_LIBRARY"]
|
||||
|
||||
try:
|
||||
c_lib = ctypes.CDLL(libname)
|
||||
if c_lib is None:
|
||||
console_error(f"Error opening {libname}")
|
||||
except Exception as e:
|
||||
console_error(f"Error loading {libname}: {e}")
|
||||
|
||||
# Set argument and return types for attach/detach functions
|
||||
try:
|
||||
# old attach/detach API
|
||||
c_lib.attach.argtypes = [ctypes.c_uint]
|
||||
except Exception as e:
|
||||
console_debug(
|
||||
"Error setting old attach/detach API argument "
|
||||
f"types: {e}, trying new API"
|
||||
)
|
||||
try:
|
||||
# new attach/detach API
|
||||
c_lib.rocattach_attach.restype = ctypes.c_int
|
||||
c_lib.rocattach_attach.argtypes = [ctypes.c_int]
|
||||
c_lib.rocattach_detach.restype = ctypes.c_int
|
||||
c_lib.rocattach_detach.argtypes = [ctypes.c_int]
|
||||
except Exception as e:
|
||||
console_error(
|
||||
f"Error setting attach/detach function argument types: {e}"
|
||||
)
|
||||
|
||||
pid = options["ROCPROF_ATTACH_PID"]
|
||||
if pid is None:
|
||||
console_error("Mode of attach/detach must have setup for process ID")
|
||||
|
||||
try:
|
||||
# old attach/detach API
|
||||
c_lib.attach(int(pid))
|
||||
except Exception as e:
|
||||
console_debug(f"Error attaching with old API: {e}, trying new API")
|
||||
try:
|
||||
# new attach/detach API
|
||||
attach_status = c_lib.rocattach_attach(int(pid))
|
||||
if attach_status != 0:
|
||||
console_error(
|
||||
f"Error attaching to process {pid}, "
|
||||
f"rocattach_attach returned {attach_status}"
|
||||
)
|
||||
except Exception as e:
|
||||
console_error(f"Error attaching to process {pid}: {e}")
|
||||
|
||||
duration = os.environ.get("ROCPROF_ATTACH_DURATION", None)
|
||||
if duration is None:
|
||||
console_log(
|
||||
f"\033[93mAttach to process with ID {pid} is successful, "
|
||||
"Press Enter to detach...\033[0m"
|
||||
)
|
||||
input()
|
||||
else:
|
||||
console_log(
|
||||
f"\033[93mAttach to process with ID {pid} is successful, "
|
||||
f"detach will happen in {duration} milliseconds...\033[0m"
|
||||
)
|
||||
time.sleep(int(duration) / 1000)
|
||||
|
||||
try:
|
||||
# old attach/detach API
|
||||
c_lib.detach(int(pid))
|
||||
except Exception as e:
|
||||
console_debug(f"Error detaching with old API: {e}, trying new API")
|
||||
try:
|
||||
# new attach/detach API
|
||||
detach_status = c_lib.rocattach_detach(int(pid))
|
||||
if detach_status != 0:
|
||||
console_error(
|
||||
f"Error detaching from process {pid}, "
|
||||
f"rocattach_detach returned {detach_status}"
|
||||
)
|
||||
except Exception as e:
|
||||
console_error(f"Error detaching from process {pid}: {e}")
|
||||
|
||||
|
||||
def capture_subprocess_output(
|
||||
subprocess_args: list[str],
|
||||
new_env: Optional[dict[str, str]] = None,
|
||||
@@ -788,49 +883,7 @@ def run_prof(
|
||||
console_debug(f"rocprof sdk env vars: {new_env}")
|
||||
|
||||
if is_mode_live_attach:
|
||||
|
||||
@contextmanager
|
||||
def temporary_env(env_vars: dict[str, str]) -> Generator[None, None, None]:
|
||||
"""
|
||||
Temporarily change the environment variable of this application.
|
||||
"""
|
||||
original_env = os.environ.copy()
|
||||
os.environ.update({k: str(v) for k, v in env_vars.items()})
|
||||
try:
|
||||
yield
|
||||
finally:
|
||||
os.environ.clear()
|
||||
os.environ.update(original_env)
|
||||
|
||||
with temporary_env(new_env):
|
||||
libname = options["ROCPROF_ATTACH_LIBRARY"]
|
||||
c_lib = ctypes.CDLL(libname)
|
||||
if c_lib is None:
|
||||
console_error(f"Error opening {libname}")
|
||||
c_lib.attach.argtypes = [ctypes.c_uint]
|
||||
|
||||
pid = options["ROCPROF_ATTACH_PID"]
|
||||
if pid is None:
|
||||
console_error(
|
||||
"Mode of attach/detach must have setup for process ID"
|
||||
)
|
||||
|
||||
c_lib.attach(int(pid))
|
||||
duration = os.environ.get("ROCPROF_ATTACH_DURATION", None)
|
||||
if duration is None:
|
||||
console_log(
|
||||
f"\033[93mAttach to process with ID {pid} is successful, "
|
||||
"Press Enter to detach...\033[0m"
|
||||
)
|
||||
input()
|
||||
else:
|
||||
console_log(
|
||||
f"\033[93mAttach to process with ID {pid} is successful, "
|
||||
f"detach will happen in {duration} milliseconds...\033[0m"
|
||||
)
|
||||
time.sleep(int(duration) / 1000)
|
||||
c_lib.detach(int(pid))
|
||||
|
||||
perform_attach_detach(new_env, options)
|
||||
else:
|
||||
if app_cmd is None:
|
||||
console_error(
|
||||
|
||||
@@ -29,6 +29,7 @@ import re
|
||||
import sqlite3
|
||||
import subprocess
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
|
||||
import numpy as np
|
||||
@@ -2268,6 +2269,7 @@ def test_live_attach_detach_block(binary_handler_profile_rocprof_compute):
|
||||
try:
|
||||
# Start workload
|
||||
process_workload = subprocess.Popen(config["app_hip_dynamic_shared"], env=env)
|
||||
time.sleep(5) # Give workload time to start
|
||||
|
||||
attach_detach = {
|
||||
"attach_pid": process_workload.pid,
|
||||
@@ -2316,8 +2318,9 @@ def test_live_attach_detach_block_thread_sleep(binary_handler_profile_rocprof_co
|
||||
try:
|
||||
# Start workload with sleep mode enabled
|
||||
process_workload = subprocess.Popen(
|
||||
[config["app_hip_dynamic_shared"], "--enable-sleep"], env=env
|
||||
[*config["app_hip_dynamic_shared"], "--enable-sleep"], env=env
|
||||
)
|
||||
time.sleep(5) # Give workload time to start
|
||||
|
||||
attach_detach = {
|
||||
"attach_pid": process_workload.pid,
|
||||
@@ -2358,7 +2361,7 @@ def test_live_attach_detach_block_thread_sleep(binary_handler_profile_rocprof_co
|
||||
|
||||
|
||||
@pytest.mark.live_attach_detach
|
||||
def test_live_attach_detach_singlepath_launch_stats(
|
||||
def test_live_attach_detach_singlepass_launch_stats(
|
||||
binary_handler_profile_rocprof_compute,
|
||||
):
|
||||
options = ["--set", "launch_stats"]
|
||||
@@ -2374,6 +2377,7 @@ def test_live_attach_detach_singlepath_launch_stats(
|
||||
try:
|
||||
# Start workload
|
||||
process_workload = subprocess.Popen(config["app_hip_dynamic_shared"], env=env)
|
||||
time.sleep(5) # Give workload time to start
|
||||
|
||||
attach_detach = {
|
||||
"attach_pid": process_workload.pid,
|
||||
|
||||
Ссылка в новой задаче
Block a user