Skip to content

Cache NPU device probe to avoid per-dispatch xrt-smi subprocess #86

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
erwei-xilinx merged 2 commits into from
Jul 31, 2026
+6 −1
Merged

Cache NPU device probe to avoid per-dispatch xrt-smi subprocess #86

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a67e1cd
Cache NPU device probe to avoid per-dispatch xrt-smi subprocess
erwei-xilinx Jul 31, 2026
ba6afbc
Remove duplicate tempfile import
erwei-xilinx Jul 31, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 6 additions & 1 deletion amd_triton_npu/backend/driver.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,8 +6,9 @@
import logging
import tempfile
import sys
import functools

import os, subprocess, tempfile, platform
import os, subprocess, platform
import importlib.util
import importlib.metadata
import shutil
Expand Down Expand Up @@ -406,7 +407,11 @@ def _dump_ir_if_needed(files):
shutil.copy(f, os.path.join(air_proj_path, os.path.basename(f)))


@functools.lru_cache(maxsize=1)
def get_npu_device_info():
# Cached: physical NPU devices do not change within a process, and this
# is called on the per-dispatch hot path (via detect_npu_version). Without
# caching, each kernel launch would spawn an `xrt-smi examine` subprocess.
try:
import re

Expand Down
Loading