Advance native Windows RR integration, kernel tooling and local packaging
Implement native NGX feature-13 RR through a LUID-matched D3D12/Vulkan shared-memory and timeline-fence bridge. Add owned parameter snapshots, GPU resource lifetime handling, inherited/enhanced barriers, optional guides, sampled input formats and depth-plane support with stencil preservation. Separate RR initialization from the SR CUDA runtime and expose explicit backend/model selection through player configuration. Add synthetic bridge, snapshot and NGX lifecycle harnesses covering camera/history/reset, parameter mutation, delayed submission, early release, capability gating and unsupported-input rejection. Add exact gfx11 D/E model generation, verified caches, versioned local deployment with configuration backups, and isolated portable model tools with pinned Python, NumPy and glslang dependencies. Extend Windows K benchmarking and private native-cache selection. Fix LLVM 22 emitter/helper compatibility and preserve exact-output validation for local output-kernel optimization. Add native FSR RR provider diagnostics, local development package refresh tooling, installer/checker updates, FidelityFX headers and Windows parity documentation. Keep NVIDIA binaries, derived models and private optimized kernels out of distributable packages. Native RR evidence remains synthetic; game validation, FSR RR dispatch, installer integration and Linux performance parity remain incomplete.
This commit is contained in:
ayo committed
1 parent
3345fffcd3
commit
247fe1092d
60 files changed
+7168
-66
No files matched your search
@@ -0,0 +1,140 @@
|
||||
"""Compare K input batching using identical synthetic frames and alternating runs.
|
||||
|
||||
Uses an existing private runtime; never installs into a game. CPU interop timing
|
||||
includes synchronization and NGX execution, and is not a GPU network-only time.
|
||||
"""
|
||||
import argparse
|
||||
import hashlib
|
||||
import json
|
||||
import os
|
||||
from pathlib import Path
|
||||
import re
|
||||
import statistics
|
||||
import subprocess
|
||||
|
||||
|
||||
def main():
|
||||
parser = argparse.ArgumentParser(description=__doc__)
|
||||
for key in ('probe', 'shim', 'runtime', 'core', 'dlss', 'output'):
|
||||
parser.add_argument('--' + key, type=lambda s: Path(s).resolve(), required=True)
|
||||
parser.add_argument('--resolution', default='1920x1080')
|
||||
parser.add_argument('--input-resolution', help='defaults to nominal Quality (two-thirds per axis)')
|
||||
parser.add_argument('--frames', type=int, default=16)
|
||||
parser.add_argument('--profile-stages', action='store_true',
|
||||
help='diagnostic CPU stage logging; adds overhead to timings')
|
||||
parser.add_argument('--candidate-native', type=lambda s: Path(s).resolve(),
|
||||
help='compare an experimental native directory instead of input batching')
|
||||
parser.add_argument('--profile-kernels', action='store_true',
|
||||
help='diagnostic deferred GPU events, allowing legacy-stream dependencies; timings are instrumented')
|
||||
args = parser.parse_args()
|
||||
if args.frames < 8 or args.frames > 120:
|
||||
parser.error('frames must be 8..120')
|
||||
if not re.fullmatch(r'[1-9][0-9]*x[1-9][0-9]*', args.resolution):
|
||||
parser.error('resolution must be WIDTHxHEIGHT')
|
||||
ow, oh = map(int, args.resolution.split('x'))
|
||||
if not args.input_resolution:
|
||||
args.input_resolution = f'{ow * 2 // 3}x{oh * 2 // 3}'
|
||||
if not re.fullmatch(r'[1-9][0-9]*x[1-9][0-9]*', args.input_resolution):
|
||||
parser.error('input-resolution must be WIDTHxHEIGHT')
|
||||
build_root = Path(__file__).resolve().parents[2] / 'build'
|
||||
if not args.output.is_relative_to(build_root.resolve()):
|
||||
parser.error('output must stay under the repository ignored build directory')
|
||||
args.output.mkdir(parents=True, exist_ok=False)
|
||||
report = {'passed': False, 'resolution': args.resolution, 'warmupFrames': 4,
|
||||
'inputResolution': args.input_resolution,
|
||||
'stageProfiling': args.profile_stages,
|
||||
'kernelProfiling': args.profile_kernels,
|
||||
'comparison': 'native kernels' if args.candidate_native else 'input batching',
|
||||
'timing': 'CPU interop including NGX and synchronization', 'runs': []}
|
||||
runtime = args.runtime
|
||||
native = runtime / 'native'
|
||||
if (native / 'gfx1101').is_dir():
|
||||
native = native / 'gfx1101'
|
||||
nvapi = runtime / 'nvapi/nvapi64.dll'
|
||||
if not nvapi.is_file():
|
||||
nvapi = runtime / 'nvapi/gfx1101/nvapi64.dll'
|
||||
report['artifacts'] = {str(p): hashlib.sha256(p.read_bytes()).hexdigest()
|
||||
for p in (args.probe, args.shim, args.core, args.dlss)}
|
||||
report['nativeArtifacts'] = {p.name: hashlib.sha256(p.read_bytes()).hexdigest()
|
||||
for p in sorted(native.glob('*')) if p.is_file()}
|
||||
if args.candidate_native:
|
||||
report['candidateNativeArtifacts'] = {p.name: hashlib.sha256(p.read_bytes()).hexdigest()
|
||||
for p in sorted(args.candidate_native.glob('*')) if p.is_file()}
|
||||
env = os.environ.copy()
|
||||
# Keep profiling and arithmetic settings independent of the caller's shell.
|
||||
for key in list(env):
|
||||
if key.startswith('D4R_') or key in ('ZLUDA_LOG_DIR', 'ZLUDA_CACHE_DIR', 'ZLUDA_CUDA_LIB'):
|
||||
del env[key]
|
||||
env.update(HIP_PATH=str(runtime / 'hip'),
|
||||
ZLUDA_CUDA_LIB=str(runtime / 'zluda/nvcuda.dll'),
|
||||
D4R_NVAPI_BACKEND=str(runtime / 'nvapi/backend/nvapi64.dll'),
|
||||
D4R_FORMAT_MODULE=str(runtime / 'pixel_convert_gfx1101.hsaco'),
|
||||
D4R_ZLUDA_NATIVE_DIR=str(native),
|
||||
D4R_ZLUDA_WMMA='1', D4R_ZLUDA_WMMA_FP8='1',
|
||||
D4R_ZLUDA_WMMA_F16_REFERENCE='1', D4R_QUIET_API='1',
|
||||
ZLUDA_CACHE_DIR=str(Path(os.environ['LOCALAPPDATA']) / 'd4r/cache'),
|
||||
D4R_NGX_DATA_DIR=str(args.output / 'ngx-data'))
|
||||
env['PATH'] = str(runtime / 'hip/bin') + os.pathsep + str(runtime / 'zluda') + os.pathsep + env['PATH']
|
||||
if args.profile_stages:
|
||||
env['D4R_PROFILE_STAGES'] = '1'
|
||||
if args.profile_kernels:
|
||||
env.update(D4R_ZLUDA_PROFILE_DEFERRED='1', D4R_ZLUDA_PROFILE_ALLOW_LEGACY='1',
|
||||
D4R_ZLUDA_PROFILE_EVERY='1')
|
||||
try:
|
||||
reference = None
|
||||
for index, batched in enumerate((False, True, True, False)):
|
||||
label = ('native-candidate' if args.candidate_native else 'batch') if batched else 'control'
|
||||
output = args.output / f'{index}-{label}'
|
||||
output.mkdir()
|
||||
env['D4R_DIAG_DIR'] = str(output)
|
||||
if batched and not args.candidate_native:
|
||||
env['D4R_BATCH_INPUT_COPIES'] = '1'
|
||||
else:
|
||||
env.pop('D4R_BATCH_INPUT_COPIES', None)
|
||||
env['D4R_ZLUDA_NATIVE_DIR'] = str(args.candidate_native if batched and args.candidate_native else native)
|
||||
command = [str(args.probe), '--module', str(args.shim), '--hip-root', str(runtime / 'hip'),
|
||||
'--cuda-dll', str(runtime / 'zluda/nvcuda.dll'), '--ngx-core', str(args.core),
|
||||
'--dlss-dll', str(args.dlss), '--nvapi-dll', str(nvapi),
|
||||
'--ngx-abi', 'project', '--interop-mode', 'command-list', '--preset', '11',
|
||||
'--ngx-output-resolution', args.resolution,
|
||||
'--ngx-input-resolution', args.input_resolution, '--iterations', str(args.frames)]
|
||||
with (output / 'stdout.log').open('wb') as out, (output / 'stderr.log').open('wb') as err:
|
||||
result = subprocess.run(command, env=env, stdout=out, stderr=err, timeout=180)
|
||||
text = (output / 'stdout.log').read_text(errors='replace')
|
||||
if result.returncode or 'PASS D3D12_DLSS' not in text:
|
||||
raise RuntimeError(f'run {index} failed ({result.returncode}); inspect {output}')
|
||||
iw, ih = map(int, args.input_resolution.split('x'))
|
||||
expected_size = f'D3D12_NGX_DIMENSIONS input={iw}x{ih} output={ow}x{oh}'
|
||||
if expected_size not in text or f'D4R_INTEROP async=0 batch_inputs={int(batched and not args.candidate_native)}' not in text:
|
||||
raise RuntimeError(f'run {index}: requested dimensions/interop mode not applied')
|
||||
if len(re.findall(r'D3D12_OUTPUT preset=11 frame=\d+ finite=1 ', text)) != args.frames:
|
||||
raise RuntimeError(f'run {index}: missing finite-output checks')
|
||||
times = [float(x) for x in re.findall(r'D4R_FRAME .*?interop_ngx_ms=([0-9.]+)', text)]
|
||||
if len(times) != args.frames:
|
||||
raise RuntimeError(f'run {index}: incomplete frame timing')
|
||||
hashes = {p.name: hashlib.sha256(p.read_bytes()).hexdigest() for p in output.glob('output-*.rgba16f')}
|
||||
if len(hashes) != args.frames:
|
||||
raise RuntimeError(f'run {index}: incomplete output readback')
|
||||
if reference is None:
|
||||
reference = hashes
|
||||
elif hashes != reference:
|
||||
raise RuntimeError(f'run {index}: output differs from control')
|
||||
report['runs'].append({'batch': batched and not bool(args.candidate_native),
|
||||
'candidate': batched, 'allFrameMs': times,
|
||||
'medianMs': statistics.median(times[4:]), 'outputHashes': hashes})
|
||||
print(f'run={index} variant={label} warmed_cpu_interop_ms={statistics.median(times[4:]):.3f}', flush=True)
|
||||
for mode in (False, True):
|
||||
values = [t for row in report['runs'] if row['candidate'] == mode for t in row['allFrameMs'][4:]]
|
||||
key = ('candidateMedianMs' if args.candidate_native else 'batchMedianMs') if mode else 'controlMedianMs'
|
||||
report[key] = statistics.median(values)
|
||||
report['passed'] = True
|
||||
except Exception as error:
|
||||
report['error'] = str(error)
|
||||
print(str(error), flush=True)
|
||||
finally:
|
||||
(args.output / 'summary.json').write_text(json.dumps(report, indent=2) + '\n')
|
||||
return 0 if report['passed'] else 1
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
raise SystemExit(main())
|
||||
Reference in new issue
Block a user