Lewati ke konten
FOURFECT

Box64 — Env Vars

Box64 itu CPU translator x86→ARM64. Dipake Winlator + semua fork (brunodev85, Ludashi/StevenMXZ, REF4IK, Cmod, Frost, Pipetto).

Sumber: data/kb/box64-envs.md COPUX KB. Cross-check: ptitSeb/box64 USAGE.md. Tag [VERIFIED] = dari testing komunitas, [THEORETICAL] = estimasi mekanisme.

JIT — tuning compiler

Section titled “JIT — tuning compiler”
BOX64_DYNAREC=1 VERIFIED

Apa: Nyalain JIT (Dynamic Recompiler). 0=interpreter, 1=JIT.
Good: JIT compile blok x86 jadi native ARM64 — 10-30x lebih cepet dari interpreter.
Bad: Off cuma buat debug. Production wajib 1.
Default: 1. Jangan pernah 0 kecuali troubleshooting.

BOX64_DYNAREC_BIGBLOCK=2 VERIFIED

Apa: Ukuran blok JIT. 0=off, 1=Normal, 2=BigBlock, 3=VeryBigBlock.
Good: Blok gede = JIT optimize lintas branch, register alloc lebih bagus. Speedup 5-15% di game CPU-bound (Unity, UE4/5).
Bad: Compile time first-load lebih lama (stutter 2-5 detik awal). Blok kegedeen kadang trigger crash multi-thread.
Default: 2 umum. 3 buat UE4/5/Unity heavy. 1 kalau crash random.

BOX64_DYNAREC_STRONGMEM=1 VERIFIED

Apa: Emulasi strong memory ordering x86 (TSO). 0=weak (ARM native), 1=strong.
Good: x86 punya Total Store Ordering — store antar core selalu observable in-order. ARM weak. Game multithread yang rely TSO crash/desync kalau ga di-emulate.
Bad: 1 = aman tapi lambat (~10-20% slower, tambah memory barrier). 0 = cepet tapi risk crash multithread.
Default: 1 buat game AAA modern. 0 buat game tua single-thread.

BOX64_DYNAREC_CALLRET=1 VERIFIED

Apa: Optimasi CALL/RET via jump table.
Good: Normalnya CALL+RET butuh 2 lookup JIT cache. Jump table = langsung jump, ~3-8% speedup.
Bad: Hampir ga ada trade-off. Aman.
Default: 1.

BOX64_DYNAREC_FASTNAN=1 VERIFIED

Apa: Fast NaN handling SSE/AVX. 0=strict IEEE-754, 1=fast (skip propagation).
Good: Skip strict NaN = faster float math.
Bad: 1 kadang bikin glitch visual di game yang rely ke NaN (rare).
Default: 1.

BOX64_DYNAREC_SAFEFLAGS=1 VERIFIED

Apa: Keamanan emulasi flag x86 untuk instruksi RET.
Good: Tanpa ini, kombinasi flag setelah RET bisa salah di-track → crash random.
Bad: Dikit overhead.
Default: 1.

BOX64_DYNAREC_WEAKBARRIER=1 VERIFIED

Apa: Memory barrier ringan (dmb ishld) bukan full barrier.
Good: Full barrier = stall pipeline. Weak barrier cukup buat banyak case.
Bad: Kalau STRONGMEM=1, weak barrier kadang ga cukup. Kalau crash → set 0.
Default: 1. Pair: STRONGMEM=1 + WEAKBARRIER=1 = cocok buat banyak game.

BOX64_DYNAREC_BLEEDING_EDGE=1 VERIFIED

Apa: Optimasi khusus Unity — pattern memory access spesifik.
Good: Aman + speedup buat game Unity.
Bad: Kadang merusak game non-Unity.
Default: 1 buat Unity. 0 buat selainnya.

Memory & address space

Section titled “Memory & address space”
BOX64_MMAP32=1 VERIFIED

Apa: Batas address mapping memory 32-bit. 0=off, 1=on.
Good: Game 32-bit butuh alamat <4GB. Tanpa ini, mmap kasih alamat >4GB → crash.
Bad: Game 64-bit ga butuh. BUKAN soal vendor GPU — berlaku di Adreno & Mali.
Default: 1 buat game 32-bit. 0 buat 64-bit.
Catatan: Mali kalau error vkMapMemory -5 → coba MMAP32=0.

BOX64_SSE_FLUSHTO0=1 VERIFIED

Apa: Flush denormal SSE register ke nol.
Good: Denormal number bikin SSE 20-100x lebih lambat di banyak CPU. Flush = perform optimization.
Bad: Dikit loss precision (irrelevant buat gaming).
Default: 1.

BOX64_DYNACACHE=1 VERIFIED

Apa: Simpan hasil JIT ke disk antar sesi.
Good: First launch JIT semua blok = stutter. Cache di-load run berikutnya = launch lebih cepet & smoother.
Bad: Makan storage 50-200MB per game.
Default: 1.

CPU capability spoofing

Section titled “CPU capability spoofing”
BOX64_AVX=2 VERIFIED

Apa: Ekspos kapabilitas AVX ke game. 0=off, 1=AVX, 2=AVX2/BMI2.
Good: Game modern (sejak ~2020) sering require AVX. Tanpa ekspos = refuse to start.
Bad: Emulasi AVX di ARM ada overhead. AVX2 lebih berat.
Default: 2 buat game baru, 1 buat 2015-2020, 0 buat lawas.

Anti-cheat & DRM workaround (darurat)

Section titled “Anti-cheat & DRM workaround (darurat)”
BOX64_NOSIGSEGV=0 THEORETICAL

Apa: Abaikan Segmentation Fault signal.
Good: Game anti-cheat lawas ngarep segfault sebagai legit signal. Tanpa skip = instant crash.
Bad: Bisa nutupin bug asli. Cuma buat game spesifik.
Default: 0. Naikin ke 1 kalau game crash di intro tanpa error jelas (DRM lawas).

BOX64_NOSIGILL=0 THEORETICAL

Apa: Abaikan Illegal Instruction signal.
Good: Anti-tamper lawas (StarForce, SecuROM) sengaja inject INT3/UD2 → segfault. Skip = bypass.
Bad: Sama kayak NOSIGSEGV — bisa hide bug nyata.
Default: 0. Naikin ke 1 buat game DRM lawas yang crash di startup.

Preset shortcut

Section titled “Preset shortcut”
BOX64_PROFILE=intermediate VERIFIED

Apa: Preset env var Box64. Value: default / compatibility / intermediate / performance.
Good: Shortcut buat ga set 30+ var manual. Preset ngatur DYNAREC_*, STRONGMEM, BIGBLOCK, dll sekaligus.
Bad: Preset performance kadang trigger crash di game tertentu. Preset nama sama ≠ behavior sama antar app — GameHub vs Ludashi 3.1 punya 7 var beda meski sama-sama "Performance".
Default: intermediate. Naik ke performance kalau stabil; turun ke compatibility kalau crash. WAJIB sebut app/fork tiap kasih value preset.

Lanjut

Section titled “Lanjut”