mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-07-12 18:17:58 +00:00
Extract Python script and make PEP 723 compliant
- Extract analysis script from bash heredoc into standalone Python file - Add PEP 723 inline script metadata for dependency management - Make script compatible with pipx and uv for automatic dependency installation - Improve code organization with proper functions and docstrings - Update documentation with PEP 723 usage examples Changes: - New file: analyze_build_trace.py (PEP 723 compliant) - Modified: ck-build-analysis (now uses external Python script) - Modified: ck-build-analysis.md (added implementation details section) Benefits: - Script can be run standalone with pipx/uv - Better code organization and maintainability - Clear dependency declaration - Easier to test and develop independently Example standalone usage: pipx run .claude/skills/analyze_build_trace.py trace.json report.md target 100 22 templates/ Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Max Podkorytov
234
.claude/skills/analyze_build_trace.py
Executable file
234
.claude/skills/analyze_build_trace.py
Executable file
@@ -0,0 +1,234 @@
|
||||
#!/usr/bin/env python3
|
||||
# /// script
|
||||
# requires-python = ">=3.8"
|
||||
# dependencies = [
|
||||
# "jinja2>=3.0.0",
|
||||
# ]
|
||||
# ///
|
||||
"""
|
||||
Build Time Analysis Tool for Composable Kernel
|
||||
|
||||
Analyzes Clang -ftime-trace output to identify template instantiation
|
||||
bottlenecks and generate comprehensive build time reports.
|
||||
"""
|
||||
|
||||
import json
|
||||
import re
|
||||
import sys
|
||||
from collections import defaultdict
|
||||
from datetime import datetime
|
||||
from pathlib import Path
|
||||
|
||||
try:
|
||||
from jinja2 import Environment, FileSystemLoader
|
||||
except ImportError:
|
||||
print("Error: jinja2 is required but not installed.", file=sys.stderr)
|
||||
print("Install with: apt-get install python3-jinja2", file=sys.stderr)
|
||||
print("Or with pip: pip install jinja2", file=sys.stderr)
|
||||
sys.exit(1)
|
||||
|
||||
|
||||
def parse_arguments():
|
||||
"""Parse command-line arguments."""
|
||||
if len(sys.argv) < 7:
|
||||
print("Usage: analyze_build_trace.py <trace_file> <output_file> <target> <granularity> <build_time> <template_dir>")
|
||||
sys.exit(1)
|
||||
|
||||
return {
|
||||
'trace_file': sys.argv[1],
|
||||
'output_file': sys.argv[2],
|
||||
'target': sys.argv[3],
|
||||
'granularity': sys.argv[4],
|
||||
'build_time': sys.argv[5],
|
||||
'template_dir': sys.argv[6],
|
||||
}
|
||||
|
||||
|
||||
def load_trace_data(trace_file):
|
||||
"""Load and parse the trace JSON file."""
|
||||
print(f'Loading trace file: {trace_file}')
|
||||
with open(trace_file, 'r') as f:
|
||||
return json.load(f)
|
||||
|
||||
|
||||
def process_events(data):
|
||||
"""Process trace events and extract template instantiation statistics."""
|
||||
print('Processing events...')
|
||||
|
||||
template_stats = defaultdict(lambda: {'count': 0, 'total_dur': 0.0})
|
||||
phase_stats = defaultdict(float)
|
||||
top_individual = []
|
||||
|
||||
for event in data.get('traceEvents', []):
|
||||
name = event.get('name', '')
|
||||
dur = event.get('dur', 0) / 1000.0 # Convert to milliseconds
|
||||
|
||||
if name and dur > 0:
|
||||
phase_stats[name] += dur
|
||||
|
||||
if name in ['InstantiateFunction', 'InstantiateClass']:
|
||||
detail = event.get('args', {}).get('detail', '')
|
||||
top_individual.append({
|
||||
'detail': detail,
|
||||
'dur': dur,
|
||||
'type': name
|
||||
})
|
||||
|
||||
# Extract template name (everything before '<' or '(')
|
||||
match = re.match(r'^([^<(]+)', detail)
|
||||
if match:
|
||||
template_name = match.group(1).strip()
|
||||
# Normalize template names
|
||||
template_name = re.sub(r'^ck::', '', template_name)
|
||||
template_name = re.sub(r'^std::', 'std::', template_name)
|
||||
|
||||
template_stats[template_name]['count'] += 1
|
||||
template_stats[template_name]['total_dur'] += dur
|
||||
|
||||
return template_stats, phase_stats, top_individual
|
||||
|
||||
|
||||
def prepare_template_data(template_stats, phase_stats, top_individual):
|
||||
"""Prepare and calculate derived statistics for template rendering."""
|
||||
print('Sorting data...')
|
||||
|
||||
# Sort data
|
||||
sorted_phases = sorted(phase_stats.items(), key=lambda x: x[1], reverse=True)
|
||||
top_individual.sort(key=lambda x: x['dur'], reverse=True)
|
||||
|
||||
# Calculate totals
|
||||
total_template_time = sum(s['total_dur'] for s in template_stats.values())
|
||||
total_trace_time = sum(phase_stats.values())
|
||||
total_inst = sum(s['count'] for s in template_stats.values())
|
||||
|
||||
# Prepare templates by time with calculated fields
|
||||
templates_by_time = []
|
||||
for name, stats in sorted(template_stats.items(), key=lambda x: x[1]['total_dur'], reverse=True):
|
||||
templates_by_time.append((name, {
|
||||
'count': stats['count'],
|
||||
'total_dur': stats['total_dur'],
|
||||
'avg': stats['total_dur'] / stats['count'] if stats['count'] > 0 else 0,
|
||||
'pct': 100 * stats['total_dur'] / total_template_time if total_template_time > 0 else 0
|
||||
}))
|
||||
|
||||
# Prepare templates by count
|
||||
templates_by_count = []
|
||||
for name, stats in sorted(template_stats.items(), key=lambda x: x[1]['count'], reverse=True):
|
||||
templates_by_count.append((name, {
|
||||
'count': stats['count'],
|
||||
'total_dur': stats['total_dur'],
|
||||
'avg': stats['total_dur'] / stats['count'] if stats['count'] > 0 else 0
|
||||
}))
|
||||
|
||||
# Add friendly type names to individual instantiations
|
||||
for inst in top_individual:
|
||||
inst['inst_type'] = 'Func' if inst['type'] == 'InstantiateFunction' else 'Class'
|
||||
|
||||
# Calculate additional metrics
|
||||
median_count = 0
|
||||
if len(template_stats) > 0:
|
||||
median_count = sorted([s["count"] for s in template_stats.values()])[len(template_stats) // 2]
|
||||
|
||||
top10_pct = 0
|
||||
if len(templates_by_time) >= 10:
|
||||
top10_pct = 100 * sum(s[1]["total_dur"] for s in templates_by_time[:10]) / total_template_time
|
||||
|
||||
return {
|
||||
'sorted_phases': sorted_phases,
|
||||
'top_individual': top_individual,
|
||||
'templates_by_time': templates_by_time,
|
||||
'templates_by_count': templates_by_count,
|
||||
'total_template_time': total_template_time,
|
||||
'total_trace_time': total_trace_time,
|
||||
'total_inst': total_inst,
|
||||
'median_count': median_count,
|
||||
'top10_pct': top10_pct,
|
||||
'unique_families': len(template_stats),
|
||||
}
|
||||
|
||||
|
||||
def setup_jinja_environment(template_dir):
|
||||
"""Set up Jinja2 environment with custom filters."""
|
||||
env = Environment(loader=FileSystemLoader(template_dir))
|
||||
|
||||
def format_number(value):
|
||||
"""Format number with thousand separators."""
|
||||
return f'{value:,}'
|
||||
|
||||
def truncate(value, length):
|
||||
"""Truncate string to length with ellipsis."""
|
||||
if len(value) > length:
|
||||
return value[:length - 3] + '...'
|
||||
return value
|
||||
|
||||
def pad(value, length):
|
||||
"""Pad string to specified length."""
|
||||
return f'{value:<{length}}'
|
||||
|
||||
env.filters['format_number'] = format_number
|
||||
env.filters['truncate'] = truncate
|
||||
env.filters['pad'] = pad
|
||||
|
||||
return env
|
||||
|
||||
|
||||
def generate_report(env, data, args, total_events):
|
||||
"""Generate the final report using Jinja2 template."""
|
||||
print('Rendering report with Jinja2...')
|
||||
|
||||
template = env.get_template('build_analysis_report.md.jinja')
|
||||
|
||||
report_content = template.render(
|
||||
timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
|
||||
target=args['target'],
|
||||
granularity=args['granularity'],
|
||||
build_time=args['build_time'],
|
||||
trace_time_sec=f'{data["total_trace_time"] / 1000:.1f}',
|
||||
template_time_sec=f'{data["total_template_time"] / 1000:.1f}',
|
||||
template_pct=f'{100 * data["total_template_time"] / data["total_trace_time"]:.1f}',
|
||||
total_events=total_events,
|
||||
total_instantiations=data['total_inst'],
|
||||
unique_families=data['unique_families'],
|
||||
total_trace_time=data['total_trace_time'],
|
||||
total_template_time=data['total_template_time'],
|
||||
phases=data['sorted_phases'],
|
||||
top_individual=data['top_individual'],
|
||||
templates_by_time=data['templates_by_time'],
|
||||
templates_by_count=data['templates_by_count'],
|
||||
median_count=data['median_count'],
|
||||
top10_pct=f'{data["top10_pct"]:.1f}'
|
||||
)
|
||||
|
||||
return report_content
|
||||
|
||||
|
||||
def main():
|
||||
"""Main entry point for the analysis tool."""
|
||||
args = parse_arguments()
|
||||
|
||||
# Load trace data
|
||||
trace_data = load_trace_data(args['trace_file'])
|
||||
total_events = len(trace_data.get('traceEvents', []))
|
||||
|
||||
# Process events
|
||||
template_stats, phase_stats, top_individual = process_events(trace_data)
|
||||
|
||||
# Prepare template data
|
||||
data = prepare_template_data(template_stats, phase_stats, top_individual)
|
||||
|
||||
# Setup Jinja2 environment
|
||||
env = setup_jinja_environment(args['template_dir'])
|
||||
|
||||
# Generate report
|
||||
report_content = generate_report(env, data, args, total_events)
|
||||
|
||||
# Write output
|
||||
with open(args['output_file'], 'w') as f:
|
||||
f.write(report_content)
|
||||
|
||||
print(f'Report generated: {args["output_file"]}')
|
||||
print(f'Report size: {len(report_content)} bytes')
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -171,158 +171,16 @@ fi
|
||||
TRACE_SIZE=$(docker exec "${CONTAINER_NAME}" bash -c "ls -lh ${TRACE_FILE} | awk '{print \$5}'")
|
||||
echo "Found trace file: ${TRACE_FILE} (${TRACE_SIZE})"
|
||||
|
||||
# Generate analysis script
|
||||
# Generate analysis report
|
||||
echo ""
|
||||
echo "Generating analysis report..."
|
||||
|
||||
ANALYSIS_SCRIPT="/tmp/analyze_${TARGET}_$$.py"
|
||||
cat > "${ANALYSIS_SCRIPT}" << 'PYSCRIPT'
|
||||
#!/usr/bin/env python3
|
||||
import json
|
||||
import re
|
||||
import sys
|
||||
from collections import defaultdict
|
||||
from datetime import datetime
|
||||
from jinja2 import Environment, FileSystemLoader
|
||||
|
||||
if len(sys.argv) < 4:
|
||||
print("Usage: analyze.py <trace_file> <output_file> <target> <granularity> <build_time> <template_dir>")
|
||||
sys.exit(1)
|
||||
|
||||
trace_file = sys.argv[1]
|
||||
output_file = sys.argv[2]
|
||||
target = sys.argv[3]
|
||||
granularity = sys.argv[4]
|
||||
build_time = sys.argv[5]
|
||||
template_dir = sys.argv[6]
|
||||
|
||||
print(f'Loading trace file: {trace_file}')
|
||||
with open(trace_file, 'r') as f:
|
||||
data = json.load(f)
|
||||
|
||||
print('Processing events...')
|
||||
template_stats = defaultdict(lambda: {'count': 0, 'total_dur': 0.0})
|
||||
phase_stats = defaultdict(float)
|
||||
top_individual = []
|
||||
|
||||
for event in data.get('traceEvents', []):
|
||||
name = event.get('name', '')
|
||||
dur = event.get('dur', 0) / 1000.0
|
||||
|
||||
if name and dur > 0:
|
||||
phase_stats[name] += dur
|
||||
|
||||
if name in ['InstantiateFunction', 'InstantiateClass']:
|
||||
detail = event.get('args', {}).get('detail', '')
|
||||
top_individual.append({'detail': detail, 'dur': dur, 'type': name})
|
||||
|
||||
match = re.match(r'^([^<(]+)', detail)
|
||||
if match:
|
||||
template_name = match.group(1).strip()
|
||||
template_name = re.sub(r'^ck::', '', template_name)
|
||||
template_name = re.sub(r'^std::', 'std::', template_name)
|
||||
|
||||
template_stats[template_name]['count'] += 1
|
||||
template_stats[template_name]['total_dur'] += dur
|
||||
|
||||
print('Sorting data...')
|
||||
sorted_phases = sorted(phase_stats.items(), key=lambda x: x[1], reverse=True)
|
||||
top_individual.sort(key=lambda x: x['dur'], reverse=True)
|
||||
|
||||
total_template_time = sum(s['total_dur'] for s in template_stats.values())
|
||||
total_trace_time = sum(phase_stats.values())
|
||||
total_events = len(data.get('traceEvents', []))
|
||||
total_inst = sum(s['count'] for s in template_stats.values())
|
||||
|
||||
# Prepare templates by time with calculated fields
|
||||
templates_by_time = []
|
||||
for name, stats in sorted(template_stats.items(), key=lambda x: x[1]['total_dur'], reverse=True):
|
||||
templates_by_time.append((name, {
|
||||
'count': stats['count'],
|
||||
'total_dur': stats['total_dur'],
|
||||
'avg': stats['total_dur'] / stats['count'] if stats['count'] > 0 else 0,
|
||||
'pct': 100 * stats['total_dur'] / total_template_time if total_template_time > 0 else 0
|
||||
}))
|
||||
|
||||
# Prepare templates by count
|
||||
templates_by_count = []
|
||||
for name, stats in sorted(template_stats.items(), key=lambda x: x[1]['count'], reverse=True):
|
||||
templates_by_count.append((name, {
|
||||
'count': stats['count'],
|
||||
'total_dur': stats['total_dur'],
|
||||
'avg': stats['total_dur'] / stats['count'] if stats['count'] > 0 else 0
|
||||
}))
|
||||
|
||||
# Prepare top individual instantiations with friendly type names
|
||||
for inst in top_individual:
|
||||
inst['inst_type'] = 'Func' if inst['type'] == 'InstantiateFunction' else 'Class'
|
||||
|
||||
# Calculate additional metrics
|
||||
median_count = 0
|
||||
if len(template_stats) > 0:
|
||||
median_count = sorted([s["count"] for s in template_stats.values()])[len(template_stats)//2]
|
||||
|
||||
top10_pct = 0
|
||||
if len(templates_by_time) >= 10:
|
||||
top10_pct = 100 * sum(s[1]["total_dur"] for s in templates_by_time[:10]) / total_template_time
|
||||
|
||||
print('Rendering report with Jinja2...')
|
||||
# Set up Jinja2 environment with custom filters
|
||||
env = Environment(loader=FileSystemLoader(template_dir))
|
||||
|
||||
def format_number(value):
|
||||
"""Format number with thousand separators"""
|
||||
return f'{value:,}'
|
||||
|
||||
def truncate(value, length):
|
||||
"""Truncate string to length with ellipsis"""
|
||||
if len(value) > length:
|
||||
return value[:length-3] + '...'
|
||||
return value
|
||||
|
||||
def pad(value, length):
|
||||
"""Pad string to specified length"""
|
||||
return f'{value:<{length}}'
|
||||
|
||||
env.filters['format_number'] = format_number
|
||||
env.filters['truncate'] = truncate
|
||||
env.filters['pad'] = pad
|
||||
|
||||
# Load and render template
|
||||
template = env.get_template('build_analysis_report.md.jinja')
|
||||
report_content = template.render(
|
||||
timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
|
||||
target=target,
|
||||
granularity=granularity,
|
||||
build_time=build_time,
|
||||
trace_time_sec=f'{total_trace_time/1000:.1f}',
|
||||
template_time_sec=f'{total_template_time/1000:.1f}',
|
||||
template_pct=f'{100*total_template_time/total_trace_time:.1f}',
|
||||
total_events=total_events,
|
||||
total_instantiations=total_inst,
|
||||
unique_families=len(template_stats),
|
||||
total_trace_time=total_trace_time,
|
||||
total_template_time=total_template_time,
|
||||
phases=sorted_phases,
|
||||
top_individual=top_individual,
|
||||
templates_by_time=templates_by_time,
|
||||
templates_by_count=templates_by_count,
|
||||
median_count=median_count,
|
||||
top10_pct=f'{top10_pct:.1f}'
|
||||
)
|
||||
|
||||
with open(output_file, 'w') as f:
|
||||
f.write(report_content)
|
||||
|
||||
print(f'Report generated: {output_file}')
|
||||
print(f'Report size: {len(report_content)} bytes')
|
||||
PYSCRIPT
|
||||
|
||||
# Copy analysis script and templates to container
|
||||
docker cp "${ANALYSIS_SCRIPT}" "${CONTAINER_NAME}:/tmp/analyze.py"
|
||||
docker cp "${SCRIPT_DIR}/analyze_build_trace.py" "${CONTAINER_NAME}:/tmp/analyze_build_trace.py"
|
||||
docker cp "${SCRIPT_DIR}/templates" "${CONTAINER_NAME}:/tmp/ck_build_analysis_templates"
|
||||
|
||||
docker exec "${CONTAINER_NAME}" python3 /tmp/analyze.py \
|
||||
# Run analysis
|
||||
docker exec "${CONTAINER_NAME}" python3 /tmp/analyze_build_trace.py \
|
||||
"${TRACE_FILE}" \
|
||||
"/workspace/${OUTPUT_FILE}" \
|
||||
"${TARGET}" \
|
||||
@@ -334,8 +192,7 @@ docker exec "${CONTAINER_NAME}" python3 /tmp/analyze.py \
|
||||
docker cp "${CONTAINER_NAME}:/workspace/${OUTPUT_FILE}" "${PROJECT_ROOT}/${OUTPUT_FILE}"
|
||||
|
||||
# Cleanup
|
||||
rm -f "${ANALYSIS_SCRIPT}"
|
||||
docker exec "${CONTAINER_NAME}" rm -f /tmp/analyze.py
|
||||
docker exec "${CONTAINER_NAME}" rm -f /tmp/analyze_build_trace.py
|
||||
docker exec "${CONTAINER_NAME}" rm -rf /tmp/ck_build_analysis_templates
|
||||
|
||||
echo ""
|
||||
|
||||
@@ -110,3 +110,42 @@ The report includes:
|
||||
- Default 500µs captures major bottlenecks (filters out 86% of instantiations)
|
||||
- 1µs granularity reveals all 36,000+ instantiations but takes longer to analyze
|
||||
- 100µs is a good middle ground for most use cases
|
||||
|
||||
## Implementation Details
|
||||
|
||||
### PEP 723 Compliance
|
||||
|
||||
The analysis script (`analyze_build_trace.py`) is PEP 723 compliant with inline dependency metadata:
|
||||
|
||||
```python
|
||||
# /// script
|
||||
# requires-python = ">=3.8"
|
||||
# dependencies = [
|
||||
# "jinja2>=3.0.0",
|
||||
# ]
|
||||
# ///
|
||||
```
|
||||
|
||||
This allows tools like `pipx run` or `uv run` to automatically manage dependencies:
|
||||
|
||||
```bash
|
||||
# Run standalone with pipx (auto-installs dependencies)
|
||||
pipx run .claude/skills/analyze_build_trace.py trace.json report.md target 100 22 templates/
|
||||
|
||||
# Or with uv
|
||||
uv run .claude/skills/analyze_build_trace.py trace.json report.md target 100 22 templates/
|
||||
```
|
||||
|
||||
### Components
|
||||
|
||||
- **ck-build-analysis** - Main bash script that orchestrates Docker, CMake, and analysis
|
||||
- **analyze_build_trace.py** - PEP 723 compliant Python script for trace analysis
|
||||
- **templates/build_analysis_report.md.jinja** - Jinja2 template for report generation
|
||||
|
||||
### Requirements
|
||||
|
||||
In Docker container:
|
||||
- `python3-jinja2` (installed via `apt-get install python3-jinja2`)
|
||||
|
||||
For standalone use:
|
||||
- Python 3.8+ with `jinja2>=3.0.0` (auto-managed if using `pipx` or `uv`)
|
||||
|
||||
Reference in New Issue
Block a user