#!/usr/bin/env python3
"""Summarize warm D4R_PROFILE samples without merging feature instances or sizes.

GPU pass fields and delayed vk_* fields are summarized independently. Their
per-line difference is deliberately not calculated; they describe different frames.
"""
import argparse
from collections import defaultdict
import json
import math
from pathlib import Path
import re
import statistics

GROUP_FIELDS = ('render', 'output', 'preset', 'quality', 'flags', 'input_wait',
                'feature', 'native_field', 'native_substitute', 'wire_requested', 'gpu_completion', 'vk_output_mode', 'vk_input_mode')

FIELDS = ('vk_total', 'dlss_gpu', 'vk_input', 'vk_dlss_wait', 'vk_output',
          'native_motion_gpu', 'wire_repair_gpu', 'marker_wait', 'pipeline',
          'throughput_interval')


def summarize(text, min_frame=300, since_clock=None, until_clock=None):
    groups = defaultdict(list)
    instance = 0
    for line in text.splitlines():
        if 'NVSDK_NGX_D3D12_CreateFeature(feature=' in line:
            instance += 1
        if 'D4R_PROFILE frame=' not in line:
            continue
        if since_clock is not None or until_clock is not None:
            clock = re.search(r'^\[\s*([0-9.]+)\]', line)
            if clock is None:
                continue
            seconds = float(clock[1])
            if since_clock is not None and seconds < since_clock:
                continue
            if until_clock is not None and seconds > until_clock:
                continue
        values = dict(re.findall(r'(\w+)=([^\s]+)', line))
        if int(values['frame']) < min_frame or values.get('reset') != '0':
            continue
        key = (instance, *(values.get(k, 'unknown') for k in
               GROUP_FIELDS))
        groups[key].append(values)
    result = []
    for key, samples in groups.items():
        row = dict(zip(('feature_instance', *GROUP_FIELDS), key))
        row.update(samples=len(samples), first_frame=int(samples[0]['frame']), last_frame=int(samples[-1]['frame']))
        row['timings_ms'] = {}
        for field in FIELDS:
            values = [float(s.get(field, '-1')) for s in samples]
            valid = [v for v in values if math.isfinite(v) and v >= 0]
            row['timings_ms'][field] = (dict(n=len(valid), median=statistics.median(valid),
                                           min=min(valid), max=max(valid)) if valid else None)
        result.append(row)
    return result


def main():
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument('log', type=Path)
    parser.add_argument('--min-frame', type=int, default=300)
    parser.add_argument('--since-clock', type=float, help='minimum bracketed log-clock seconds')
    parser.add_argument('--until-clock', type=float, help='maximum bracketed log-clock seconds')
    args = parser.parse_args()
    print(json.dumps(summarize(args.log.read_text(errors='replace'), args.min_frame,
                               args.since_clock, args.until_clock), indent=2))


if __name__ == '__main__':
    main()