mirror of
https://git.triforceproject.com/triforce/source
synced 2026-11-20 08:39:58 -04:00
Triforce v0.2.0
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
64
nx2/census.py
Normal file
64
nx2/census.py
Normal file
@@ -0,0 +1,64 @@
|
||||
"""Instruction census: every distinct word in each module's .text, through the lifter's classify() and suyu's
|
||||
translator (a64emit), counted by capstone mnemonic. Counts occurrences, so data in .text and padding are included;
|
||||
the per-function numbers come from lift.py (discovered code only).
|
||||
|
||||
Usage: python -I nx2/census.py [--exefs <dir>] [--modules main,sdk] [--top 40] [--nx2] (with the Switch 2 forms)
|
||||
"""
|
||||
import argparse
|
||||
import collections
|
||||
import os
|
||||
import struct
|
||||
import sys
|
||||
|
||||
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__))) # (python -I leaves the script folder off)
|
||||
import kit
|
||||
import nso2 # noqa: F401 (DT_RELR)
|
||||
import nso
|
||||
import a64emit
|
||||
import a64_lift
|
||||
import capstone
|
||||
|
||||
|
||||
def main():
|
||||
p = argparse.ArgumentParser()
|
||||
p.add_argument("--exefs", default=kit.EXEFS)
|
||||
p.add_argument("--modules", default="")
|
||||
p.add_argument("--top", type=int, default=40)
|
||||
p.add_argument("--nx2", action="store_true", help="apply the Switch 2 instruction extension first")
|
||||
a = p.parse_args()
|
||||
a.exefs = a.exefs or kit.need_exefs()
|
||||
if a.nx2:
|
||||
import insn
|
||||
insn.install()
|
||||
md = capstone.Cs(capstone.CS_ARCH_ARM64, capstone.CS_MODE_ARM)
|
||||
want = set(a.modules.split(",")) if a.modules else None
|
||||
total = collections.Counter()
|
||||
for m in nso.load_exefs(a.exefs):
|
||||
if want and m.name not in want:
|
||||
continue
|
||||
lo, hi = m.code_range()
|
||||
words = struct.unpack_from("<%dI" % ((hi - lo) // 4), m.image, lo)
|
||||
count = collections.Counter(words)
|
||||
ops = [w for w in count if a64_lift.classify(w, 0x10000)[0] == "op"]
|
||||
_, flags = a64emit.translate([(0x10000, w) for w in ops])
|
||||
bad = collections.Counter()
|
||||
bad_words = collections.Counter()
|
||||
for w, f in zip(ops, flags):
|
||||
if f & a64emit.UNHANDLED:
|
||||
ins = next(md.disasm(struct.pack("<I", w), 0x10000), None)
|
||||
key = ins.mnemonic if ins else "(not an instruction)"
|
||||
bad[key] += count[w]
|
||||
bad_words[key] += 1
|
||||
nbad = sum(bad.values())
|
||||
print("== %s: %d words, %d distinct, %d distinct translatable-kind; untranslated %d words (%.3f%%) in %d distinct"
|
||||
% (m.name, len(words), len(count), len(ops), nbad, 100.0 * nbad / len(words), sum(bad_words.values())))
|
||||
for k, v in bad.most_common(a.top):
|
||||
print(" %9d %-22s (%d distinct)" % (v, k, bad_words[k]))
|
||||
total.update(bad)
|
||||
print("== total untranslated by mnemonic")
|
||||
for k, v in total.most_common(a.top):
|
||||
print(" %9d %s" % (v, k))
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
Reference in New Issue
Block a user