woutermans
0fe3a28cb7
Per-layer staged-image engine for the eleven translated denoiser Swin layers: descriptors (verified offsets per phase), cut recipes that splice an isolated native phase into NVIDIA's PTX, wave templates, epilogue notes and GPU probes. The isolated phase-8 splice produces a byte-identical final 4K frame but is slower than the translated body (14.90 vs 6.59 ms) - a correctness experiment, not a shipping improvement; docs/rr-engine.md records the design and the next direct-converter win (B operands built from transposed accumulators).
30 lines
1.0 KiB
Python
Executable File
30 lines
1.0 KiB
Python
Executable File
#!/usr/bin/env python3
|
|
"""Dump the mma statements of a swin entry's phase range with their A/B/C/D registers.
|
|
|
|
Usage: swin_mma.py LAYER LO_STMT [HI_STMT]
|
|
LAYER is a rr_layer_spec alias (enc0, ...). One line per mma, with the
|
|
statement index, the physical PTX line and the register operands in order.
|
|
"""
|
|
import os
|
|
import re
|
|
import sys
|
|
|
|
sys.path.insert(0, os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "rr"))
|
|
import rr_layer_spec as R # noqa: E402
|
|
|
|
|
|
def main():
|
|
layer = sys.argv[1] if len(sys.argv) > 1 else "enc0"
|
|
lo = int(sys.argv[2]) if len(sys.argv) > 2 else 0
|
|
hi = int(sys.argv[3]) if len(sys.argv) > 3 else 10 ** 9
|
|
path, entry = R.resolve(layer, None, None)
|
|
lines = open(path).read().split("\n")
|
|
first, last = next((a, b) for (nm, a, b) in R.find_entries(lines) if nm == entry)
|
|
stmts, _ = R.linearise(lines, first, last)
|
|
for s in stmts:
|
|
if s.is_mma and lo <= s.idx <= hi:
|
|
print(f"s{s.idx} L{s.line}: " + " ".join(re.findall(r"%r\d+", s.text)))
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main() |