Refactor: redesign log system around V0..V9 + Python single source#703
Merged
jvjhfhg merged 1 commit intohw-native-sys:mainfrom Apr 30, 2026
Merged
Conversation
ChaoWao
force-pushed
the
refactor/redesign-log-system-v0-v9
branch
from
6f325f3 to
0310987
Compare
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces a unified logging system that synchronizes log levels across Python, host C++, and AICPU device code, replacing the environment-variable-based PTO_LOG_LEVEL with a single source of truth in the Python simpler logger. The new system uses two orthogonal axes (severity and INFO sub-verbosity) which are propagated through KernelArgs to the device. Feedback suggests optimizing the host-side unified logging implementation by passing va_list directly to HostLogger to avoid redundant formatting steps and intermediate buffer copies.
Replace the env-var-driven log config (PTO_LOG_LEVEL / SIMPLER_LOG_LEVEL /
PTO_LOG_FILE) with the Python "simpler" logger as single source of truth,
expose INFO sub-verbosity tiers V0..V9 mapped into Python's logging integer
space (15..24 with V5=20=INFO), and unify severity + verbosity into one
user-facing setLevel() call.
User API:
import simpler, logging
logging.getLogger("simpler").setLevel(simpler.V3) # V3..V9 visible
logging.getLogger("simpler").setLevel(simpler.NUL) # silent
Plumbing:
- python/simpler/_log.py: V0..V9 / NUL constants, _split_threshold, sets the
"simpler" logger to V5 at import (avoids root WARNING inheritance).
- Worker.init() reads the logger via _log.get_current_config() and forwards
(severity, info_v) to ChipWorker::init(..., log_level, log_info_v) once.
- New simpler_init(ctx, log_level, log_info_v) C entry in each platform SO:
pushes config to HostLogger + runner state; on onboard also calls
dlog_setlevel(-1, sev, 0) unless ASCEND_GLOBAL_LOG_LEVEL is externally set.
- AICPU receives the same values via KernelArgs.{log_level, log_info_v}
(cross-process).
- run_runtime no longer carries log args; CallConfig wire format reverts
to 5 int32 + 1024-byte output_prefix.
C++ macros:
- LOG_DEBUG / LOG_WARN / LOG_ERROR + 10 fixed LOG_INFO_V0..V9.
- LOG_ALWAYS / LOG_INFO sugar / unified_log_always / dev_log_always all
removed; ops table swaps log_info+log_always for log_info_v.
Behavior change:
- Old LOG_INFO sites mass-migrated to LOG_INFO_V0 (default-hidden noise).
- Old LOG_ALWAYS sites mass-migrated to LOG_INFO_V9 (must-see; shown by
default at V5 threshold).
- All log output goes to stderr (host + sim AICPU). Onboard AICPU still
routes through CANN dlog (severity managed by CANN, verbosity by simpler).
CI / scripts grepping legacy INFO output: pass --log-level v0 (or set the
"simpler" Python logger to simpler.V0) to see the verbose tier.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ChaoWao
force-pushed
the
refactor/redesign-log-system-v0-v9
branch
from
0310987 to
b2f65a3
Compare
jvjhfhg
approved these changes
Apr 30, 2026
4 tasks
ChaoWao
added a commit
to ChaoWao/simpler-fork
that referenced
this pull request
Apr 30, 2026
Three issues reported after hw-native-sys#703 merged: 1. `pytest --log-level v0` was rejected ("'V0' is not recognized as a logging level name"). pytest's CLI validator uses `int(getattr(logging, name.upper(), name))`, so the level names must be exposed as **attributes on the logging module** — `logging.addLevelName` alone is not enough. Set both: keep `addLevelName` for nice `%(levelname)s` formatting, and `setattr(logging, ...)` so pytest's getattr lookup resolves. Mirror the registration in conftest.py top-level too — pytest validates --log-level before conftest's first `import simpler`. Also expose `logging.NULL` for `pytest --log-level null` (pytest upcases the value before lookup). 2. L3/L4 chip subprocesses ignored the user's simpler logger level. The parent's `Worker.init()` snapshot was only forwarded along the L2 path; `_chip_process_loop` and `_chip_process_loop_with_bootstrap` called `ChipWorker.init()` with the binding defaults (V5/INFO). Added `log_level` / `log_info_v` parameters to both child loops and snapshot the parent's logger config once before the fork loop in `_start_hierarchical`, then pass it explicitly into each forked child. 3. `docs/testing.md` Configuration Sources paragraph still claimed `Worker.run()` snapshots into `CallConfig.log_level` / `log_info_v` per-run. Those CallConfig fields were dropped in C4; the actual snapshot is one-shot at `Worker.init()`. Rewrote the paragraph to match, documented the L3/L4 fork plumbing, and updated the `--log-level` row in the Option Reference. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ChaoWao
added a commit
that referenced
this pull request
Apr 30, 2026
Three issues reported after #703 merged: 1. `pytest --log-level v0` was rejected ("'V0' is not recognized as a logging level name"). pytest's CLI validator uses `int(getattr(logging, name.upper(), name))`, so the level names must be exposed as **attributes on the logging module** — `logging.addLevelName` alone is not enough. Set both: keep `addLevelName` for nice `%(levelname)s` formatting, and `setattr(logging, ...)` so pytest's getattr lookup resolves. Mirror the registration in conftest.py top-level too — pytest validates --log-level before conftest's first `import simpler`. Also expose `logging.NULL` for `pytest --log-level null` (pytest upcases the value before lookup). 2. L3/L4 chip subprocesses ignored the user's simpler logger level. The parent's `Worker.init()` snapshot was only forwarded along the L2 path; `_chip_process_loop` and `_chip_process_loop_with_bootstrap` called `ChipWorker.init()` with the binding defaults (V5/INFO). Added `log_level` / `log_info_v` parameters to both child loops and snapshot the parent's logger config once before the fork loop in `_start_hierarchical`, then pass it explicitly into each forked child. 3. `docs/testing.md` Configuration Sources paragraph still claimed `Worker.run()` snapshots into `CallConfig.log_level` / `log_info_v` per-run. Those CallConfig fields were dropped in C4; the actual snapshot is one-shot at `Worker.init()`. Rewrote the paragraph to match, documented the L3/L4 fork plumbing, and updated the `--log-level` row in the Option Reference.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Replace the env-var-driven log config (
PTO_LOG_LEVEL/SIMPLER_LOG_LEVEL/PTO_LOG_FILE) with the Pythonsimplerlogger as single source of truth, expose INFO sub-verbosity tiers V0..V9 mapped into Python'slogginginteger space (15..24 with V5=20=INFO), and unify severity + verbosity into one user-facingsetLevel()call.User API
Integer layout
DEBUGlogging.DEBUGsimpler.V0..V4simpler.V5=logging.INFOsimpler.V6..V9WARNING/ERRORsimpler.NULPlumbing
python/simpler/_log.py: V0..V9 / NUL constants,_split_threshold, sets thesimplerlogger to V5 at import (avoids root WARNING inheritance).Worker.init()reads the logger via_log.get_current_config()and forwards(severity, info_v)toChipWorker::init(..., log_level, log_info_v)once.simpler_init(ctx, log_level, log_info_v)C entry in each platform SO: pushes config toHostLogger+ runner state; on onboard also callsdlog_setlevel(-1, sev, 0)unlessASCEND_GLOBAL_LOG_LEVELis externally set.KernelArgs.{log_level, log_info_v}(cross-process).run_runtimeno longer carries log args;CallConfigwire format reverts to 5 int32 + 1024-byteoutput_prefix.C++ macros
LOG_DEBUG/LOG_WARN/LOG_ERROR+ 10 fixedLOG_INFO_V0..V9.LOG_ALWAYS/LOG_INFOsugar /unified_log_always/dev_log_alwaysall removed; ops table swapslog_info+log_alwaysforlog_info_v.Behavior change
LOG_INFOsites mass-migrated toLOG_INFO_V0(default-hidden noise).LOG_ALWAYSsites mass-migrated toLOG_INFO_V9(must-see; shown by default at V5 threshold).CI / scripts grepping legacy INFO output: pass
--log-level v0(or set thesimplerPython logger tosimpler.V0) to see the verbose tier.Test plan
build/ut_cppctest)simpler.V*constants,CallConfigno longer has log fields,get_current_config()splits correctly, default V5