Editing PS2 Emulation

=2024-present year=

The PlayStation 2 emulation on the PlayStation 4 is handled with little difference to the PlayStation 3. Some issues on PS3's PS2 emulator were left unfixed on the PlayStation 4. Consequently, the PS4 inherited some of PS3's emulation glitches, and even introduced new ones. The new issues came as a result of PS4's weakness with emulation, which likely resulted in the trade of accuracy for performance.

Each PS2onPS4 package file (.pkg) includes the PS2 emulator itself. The PS4 does not have a native built-in emulator in its System Software, but it does have a few PS2 emulator specific features, like the usermode functions sceLncUtilIsPs2Emu, sceShellCoreUtilGetImposeMenuFlagForPs2Emu, sceSystemServiceAddLocalProcessForPs2Emu, and sceSystemServiceShowImposeMenuForPs2Emu.

The PS2 emulators are programmed to be accurate for the games they were designed to emulate. In turn, this makes accurate emulation challenging, as emulators include per-title patches and different default settings. The PlayStation 2 hardware is a complex system to emulate; any small change in configuration can make or break a game. Currently, the most commonly used emulator is "Jakv2" from Jak games, as it is the most compatible one out there. When game crashes are concerned in an emulated PS2 game, then the Resident Evil Code: Veronica X's emulator is typically used, whereas for VU accuracy it is Rogue Galaxy's.

The PS2 BIOS is included in the game's pkg file. It is the same PS2 BIOS as used in the ps2_netemu in PS3 Operating System. The lack of encryption mechanism for PS2 game disc image files is also huge. This time, it is a plain ISO file, but the emulator still supports LIMG sector that allows the usage of non-2048 sector size discs. Moreover, memory cards are also unencrypted, but the emulator checks the CRC of some of the regions to ensure that the memory card was not modified. The PS2 emulator on PS4 also supports LUA scripting by a "Lua Bridge", which is a powerful interface that may prove exceptionally handy in fixing games.

This is first time that Sony has ever cared about floats in its emulators. Compared to the PS3 CPU, x86-64 is doing a worse job in mimicking FPU and VU floats calculation behavior. This is resolved by clamping, accurate math used per memory offset, and/or converting floats to double precision. The emulator uses OpenCL for GS, and likely other components as well.

The rest of the CLI and LUA commands can all be found inside of any emulator's '''Eboot.bin''' file.

| --emulog-file || 0, 1 ||  Creates a log file with information that is rarely useful || --emulog-file=1

| --path-emulog || dir/folder || Sets the directory of the emulog file. Requires --emulog-file=1 command  || --path-emulog="/tmp/recordings"

| --framelimit-scalar ||  scalar/float || || --framelimit-scalar=3.2

| --ps2-lang || system  || Sets language. Might seem like a useless command, but a tiny number of games refuse to boot without a selected language. || --ps2-lang=system  

| --verbose-cpu-cycles=1 || 0,1 || ? || --verbose-cpu-cycles=1

| --host-gamepads || 0, 1 || || --host-gamepads=1

| --pad-record ||  0, 1  ||  Enables logging pad info in emulog ||

| --pad-analog-to-digital || 0, 1 || Eternel ring emu uses the value 0 || --pad-analog-to-digital=0

| --mtap1 || Disabled, Always, ByHost || Multitap switch. The values are correct but the multitap only works in certain games. || --mtap1=always

| --mtap2 || Disabled, Always, ByHost || Multitap switch. Some games require multitap to be selected only on second port (1, 2-a, 2-b, 2-c layout). To achieve that, we need to disable the first mtap. In some cases, the emulator can be flawed in how it handles multi-tap when it involves "--mtap2". All functions for 2A, 2B, 2C, 2D must be redirected to ports on multi-tap 1 by using memory patches. || --mtap2=always

<pre>These commands work in config-emu-ps4.txt, but the official way to use them is through the CLI file (It is also needed for multi-disc packages). Additionally, some commands are still not discovered, and known commands might be missing their values.

All commands are pre-made. Creating new CLI commands is impossible.

The values of the commands are also pre-defined by Sony and/or have a set of possible values or ranges.

Some CLI commands have no effect, as they were added to the eboot.bin without being programmed. Some useful commands were removed or added in or from different emu revisions, or their effects were changed. In some emulators, CLI configurations can be found in the ram at SceLibcHeap[0]-3-881004000.</pre>

=====[https://www.psdevwiki.com/ps2/Emotion_Engine EE]=====

<pre>Commands for the emulated Emotion Engine. These commands may improve performance, prevent crashes, or provide better core sync.</pre>

|  || Speedhacks / Desync Fixes|| ||

| --ee-cycle-scalar ||Slowest cycles speed: 5.00<br>Normal cycles speed: 1.000<br>Fastest cycles speed: 0.100 ||  Can benefit core sync or performance (lowering the cycle rate may affect FMVS). The default value for most emulators is ~1.0. It is highly sensitive, as any slight change may break or fix the game. (Possibly divides 100 with the entered value and sets the result as EE's clock speed percentage % .) || --ee-cycle-scalar=1.0  

| --ee-context-switch-cycles || Accelerate cycles (360 => 0) <br>Decelerate cycles (361 => ∞ ) || Its effects are mysterious. Possibly configures how many CPU cycles the emulator should run before switching between different tasks or threads. Does not work on Jak emulators. Default value is 360. Seems to be useful for Bee Movie, Metal Gear Solid 3, as well as others with IPU issues.|| --ee-context-switch-cycles=2700?

| [[--ee-hook|--ee-hook]] || AdvanceClock<br>FastForwardClock<br>Mfifodrain || Runs the specified functionality each time a selected EE offset is reached. FastForwardClock and AdvanceClock help fix core sync issues, but can be used as a speedhack. It is believed that AdvanceClock stalls the EE for a specified amount of cycles, while FastForwardClock behaves similarly. They are commonly used on affected COP2 instructions. Mfifodrain is still unknown. [[--ee-hook|For more examples]] ||--ee-hook=0x0025A9F2,AdvanceClock,,500 <br>--ee-hook=0x0019F0AD,FastForwardClock

| --ee-jit-pagefault-threshold || 0-254 || Ignore JIT pagefaults, usually caused by data placed right after code. JIT try to recompile it, and obviously hit some bad opcodes. This command reduces the occurrence of crashes the higher it is. || --ee-jit-pagefault-threshold=40

| --ee-const-folding || None,Gpr,Fpu,All || EE constant folding. "All" seems to be the default option. || --ee-const-folding=none  

| --ee-ignore-segfault || none, read, write, readwrite || Ignore segmentation fault. Ignore rw from/to unmapped memory, known in pcsx2 as TLB miss (load.store) || --ee-ignore-segfault=readwrite

| --ee-native-function || memset, memcpy, strlen, strcmp, strcasecmp, litodp, dptoli, floatdidf, fptoui, fptodp, dptofp, fabs, fabsf, ieee754_atan2f, ieee754_asinf, ieee754_acosf, ieee754_sinf, ieee754_cosf, ieee754_sqrtf, asinf, acosf, sinf, cosf, sqrtf. || Hook mips function with HLE x64 native version. Much faster, but tricky to keep correct timings on ps2 VM. || --ee-native-function=memcpy,0x11e328

|-style="background-color:#FFAA00"

| --ee-jit-disasm || 0 or 1 for mips and 2 for x86 || Emotion engine Just in Time disassembler mode (Likely a debug log option) || --ee-jit-disasm=1

| --ee-ignore-break || 0, 1 || Turns a BREAK instruction into a NOP Instruction. ||  

| --ee-break-as-nop || 0, 1 || Turns a BREAK instruction into a NOP Instruction. || --ee-break-as-nop=1

| --ee-pc-coherency || 0, 1 || Flush PC register to memory on every mips opcode. Result in more accurate EE PC reg reads. ||  

| --ee-kernel-hle ||  0, 1 || High-level emulation kernel. Apply HLE patches (used in mast1c0re part 2 Vulnerability 2). Apply EE kernel patches which run previously prepared x64 code instead of recompiling real code part. ||  

| --ee-injection-kernel || 0, 1 || Just like --ee-kernel-hle, except using a different name. ||

| --ee-injection-title || 0, 1 || Apply HLE patches (used in mast1c0re part 2 Vulnerability 2). Apply per title patches which run previously prepared x64 code instead of recompiling real code part. ||

| --vtune-ee || || ||

| --ee-cache-breaks-block || 0,1 || End recompiler block on r5900 cache opcode. ||

| --ee-evt-check-full || 0, 1 || Perform more accurate EE event check ||  

| --ee-peephole || 0, 1? || ||

| --ee-penalize-short-blocks || 0, 1 || Add cycles to short recompiled blocks? ||

| --ee-inline-limit-full || insn_count/integer || ||

| --detect-idle-ee || 0, 1 || || --detect-idle-ee=1

=====FPU=====

<pre>The FPU is a fast single-precision unit that typically serves the EE by calculating in-game physics. It mostly controls the accuracy of the movement of in-game objects, or the game's gravity as a whole. In some cases, it is best to use COP2 commands with the FPU's to ensure the accuracy of the game's geometrical calculations.</pre>

| --fpu-no-clamping || 0, 1 || Changes the clamping behavior of the FPU. It's set to 1 by default. Might fix freezing issues, or misplaced characters/items (Inaccuracies with in-game physics) || --fpu-no-clamping=0

| --fpu-custom-min-max || 0, 1 || Custom Max/Mini logic for denormals. Enabled by default. Disabling it might break some games || --fpu-custom-min-max=0

| --fpu-accurate-range || start,end offset (0x0 - 0x1FFFFFF) || Accurate Mul/Div/Add/Sub Math instructions for the FPU by using software floats rather than hardware floats. Can be used for full EE/FPU Mathematical accuracy in the specified range. || --fpu-accurate-range=0x1acce0,0x2acce0

| --fpu-no-clamp-range || start,end offset (0x0 - 0x1FFFFFF) || Per-range clamping for the FPU in the selected memory range. || --fpu-no-clamp-range=0x1acce0,0x2acce0

| --fpu-accurate-muldiv  || 0 = Disabled<br>1 = Enabled || Allows the FPU to accurately Divide/Multiply for all offsets by using software floats rather than hardware floats. (Extremely slow) || --fpu-accurate-muldiv=1

| --fpu-accurate-addsub || 0 = Disabled<br>1 = Enabled || Allows the FPU to accurately Add/Subtract for all offsets by using software floats rather than hardware floats. (Extremely slow) || --fpu-accurate-addsub=1

| --fpu-accurate-muldiv-range || start,end offset (0x0 - 0x1FFFFFF) || Allows the FPU to accurately Divide/Multiply in the specified ee memory range by using software floats rather than hardware floats. || --fpu-accurate-muldiv-range=0x1acce0,0x2acce0

| --fpu-accurate-addsub-range || start,end offset (0x0 - 0x1FFFFFF) || Allows the FPU to accurately Add/Subtract in the specified ee memory range by using software floats rather than hardware floats. || --fpu-accurate-addsub-range=0x1acce0,0x2acce0

| --fpu-rsqrt-fast-estimate || 0, 1 || Decreases RSQRT's accuracy. Enabled by default, disabling it helps with achieving full FPU accuracy. || --fpu-rsqrt-fast-estimate=1

| --fpu-accurate-mul-fast || 0, 1 || Speedhack for CLI FPU muldiv commands.  || --fpu-accurate-mul-fast=1

|-style="background-color:#FFAA00"

| --fpu-custom-fused-madd || 0, 1 || Custom FMA (fused multiply-add). Disabled by default. || --fpu-custom-fused-madd=0

| --fpu-to-double || 0, 1 || Converts floats to doubles (Accurate). Limited testing shows that infinity is still wrong at 7F7FFFFF, rather than 7FFFFFFF with it enabled. || --fpu-to-double=1

| --fpu-clamp-operands || 0, 1 || Affects the fpu's clamping behavior || --fpu-clamp-operands=1

| --fpu-clamp-results || 0, 1 || Affects the fpu's clamping behavior || --fpu-clamp-results=1

===== COP2 =====

 

<pre>VU0 Macro-mode. These commands assist the FPU.</pre>

| --cop2-opt-flags || 0, 1 ,2 || Optimize flags upon recompilation (1 is safe, 2 update sticky, and mac flag on every status flag update) setting it to 0 can in some cases improve FPU accuracy. [https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L280 More Info] || --cop2-opt-flags=1

| --cop2-opt-vf00 || 0, 1, 2 || Optimization for the vf00 register. || --cop2-opt-vf00=1

| --cop2-const-prop || 0, 1 || Disables or enables constant propagation. Can Be Used as speedhack for COP2 hungry games when set to 1 [https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L312 More info] || --cop2-const-prop=1

| --cop2-di-bits || 0, 1 || 0 to skip setting invalid, and div by Zero flags in status register, and therefore improve performance. Can be used as a speedhack to skip costly calculations. || --cop2-di-bits=0

| --cop2-no-clamping || 0, 1 || Changes the clamping behavior of the COP2. The default value is set to 1.  || --cop2-no-clamping=0

| --cop2-accurate-range || start,end offset (0x0 - 0x1FFFFFF) ||Accurate Mul/Div/Add/Sub math instructions for the COP2 by using software floats rather than hardware floats. Can be used for full COP2 mathematical accuracy in the specified range. || --cop2-accurate-range=0x123456,0x134567

| --cop2-accurate-mul-range || start,end offset (0x0 - 0x1FFFFFF) || Accurate multiplication instructions for the COP2 in the specified range by using software floats rather than hardware floats (Offset range seems not to directly match EE's) || --cop2-accurate-mul-range=0x123456,0x134567

| --cop2-accurate-mul || 1 = Enabled<br>0 = Disabled || Allows the COP2 to accurately multiply for all offsets by using software floats rather than hardware floats. (Extremely slow) || --cop2-accurate-mul=1

|-

| --cop2-accurate-addsub ||  1 = Enabled<br>0 = Disabled || Allows the COP2 to accurately Add/Subtract for all offsets by using software floats rather than hardware floats. (Extremely slow) || --cop2-accurate-addsub=1

| --cop2-accurate-addsub-range || start,end offset (0x0 - 0x1FFFFFF) || Allows the COP2 to accurately Add/Sub in the specified ee memory range by using software floats rather than hardware floats. (Offset range seems not to directly match EE's. It is possibly 0x95400~ behind(?) || --cop2-accurate-addsub-range=0x123456,0x134567

|-style="background-color:#FFAA00"

| || Other || ||

| --cop2-regalloc || 0, 1 ? || Optimized Register Allocation(?). PCSX2 seems to have this enabled by default. Seems to slow down COP2 when disabled. ||

| --cop2-inst-q || 0, 1 || instant Q, no stalling on VWAITQ, or instances of Q. Crashes some games when enabled. ||

| --cop2-inst-p || 0, 1 || instant P, no stalling on VWAITP, or instances of P. ||

| --cop2-use-rcp || 0, 1 || Enables SSE RCP. Disabled by default ||

| --cop2-use-rsqrt || 0, 1 || Enables SSE RSQRT || --cop2-use-rsqrt=0

| --cop2-clamp-operands || 0, 1 || Affects the COP2's clamping behavior|| --cop2-clamp-operands=1

| --cop2-clamp-results || 0, 1 || Affects the COP2's clamping behavior  || --cop2-clamp-results=1

|-  

 

 

===== VU =====

 

 

{| cellspacing="0" cellpadding="2" border="1" class="wikitable" style="text-align: center;"

| --vu-xgkick-delay  || Limit is between (0 => 31) || Delay xgkick instruction execution. PCSX2 uses 6 cycles, PS3 uses 2-8 cycles depending on the game. A game with missing graphics will require a low value such as 0, while one that crashes might require a high value such as 7.|| --vu-xgkick-delay=8

| --vu-custom-min-max  || 0, 1 || Custom Max/Mini logic for denormals. Disabled on pcsx2 by default but enabled by default on a couple of emus such as jakv2. Disabling it could potentially restore missing graphics.|| --vu-custom-min-max=0

|-style="background-color:#D7EF54"

| || Game fixes || ||

| --vu-hack-triace  || 0, 1 || A hack designed specifically for games developed by Tri Ace. Corrects the VU's addi calculation. Without it, the result will be one bit off, which is enough to break the game's decryption algo. || --vu-hack-triace=1

| --vu-branch-hazard || 0, 1 || This command relates to how the emulator handles branch hazards in the Vector Units (VU). Setting it to 0 is the compatible option, while 1 might improve performance. Seems to be enabled by default. Disabling it managed to prevent "'''Batman - Rise of Sin Tzu'''" from crashing|| --vu-branch-hazard=0

| --vu-evil-branches || 0, 1 || Take in count branch in delay slot, or not. [https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L306 More Info]||

| --vu-to-double || 0, 1 || Converts floats to double. Enabling it is the accurate option. || --vu-to-double=1

| --vu-custom-fused-madd  || 0, 1 || Custom FMA (fused multiply-add). Disabled by default || --vu-custom-fused-madd=1

|-style="background-color:#7698FF"

|  || Speedhacks || ||

| --vu-opt-sf-check  || 0, 1 || Updates status flags only on blocks which will read them. Known in pcsx2 as mVU flag hack. Won't work with most emulators, but will work on Arc Twilight of the Spirits™ || --vu-opt-sf-check=1

| --vu-opt-jr-caching  || 0, 1 || Optimize Jump Register caching. (vi15). Won't work with most emulators but will work on Arc Twilight of the Spirits™ || --vu-opt-jr-caching=1

| --vu-d-bit  || 0, 1 || Debug break. Halts the VU and sends an interrupt to the EE. || --vu-d-bit=0

| --vu-t-bit || 0, 1 || Debug halt. Acts similarly to D-bit || --vu-t-bit=0

| --vu-inst-mflag  || 0, 1 || Instant macflag? It's enabled by default. When disabled, some texture chunks pop and fade. || --vu-inst-mflag=1

| --vu-inst-cflag || 0, 1 || Instant cflag? || --vu-inst-cflag=1

| --vtune-vu || ? || ? ||

| --vu-jit-disasm  || 0, 1, 2 || ? || --vu-jit-disasm=1

| --vu-range-merge || vu_inst_cnt || ? ||  

===== VU1 =====

 

<pre>VU1 is the Graphics Synthesizer’s alternate processing unit. Commands here typically improve performance, prevent crashes, and fix SPS (graphical glitches).</pre>

|-style="background-color:#7698FF"

| || Speedhacks || ||

| --vu1-mpg-cycles || Decimals<br>0 - 65535 || Set initial speed for VU1 Micro-programs. 100 is the default value. If VU1 requirements were a bottleneck, increasing it will result in better performance while decreasing it will result in the opposite. Increasing it can be beneficial for MTVU sensitive games.|| --vu1-mpg-cycles=1000<br><pre> Symbols (They change values close to VU1-mpg-cycles' offset):

 

$ goes to the specified succeeding offsets. $0 will go 2 bytes forward and change them. $1 will go 4 bytes forward and change the 2 bytes specified. For example, $0,$1 will change the 4 bytes ahead as the specified 2 bytes' value.

 

 

'-' or '+' Changes the specified preceding or succeeding 2 bytes of VU1 mpg cycles' offset according to the added number. Adding -1, for example will change the 2 bytes before it. Likewise, adding -2 will go 4 bytes away and change the specified 2 bytes accordingly. It even can change EE cycle scalar itself if you go far enough. These numbers are in HEX. Infinite amounts of '-' and '+' symbols seem to be supported.</pre>

| --vu1-di-bits || 0, 1 || 0 Skips setting invalid, and Divides by zero flags in status register. Can be used as a speedhack as it can skip costly calculations. But at the same time, it can cause issues such as broken geometry. || --vu1-di-bits=0

| --vu1-const-prop || 0, 1|| Enabled by default. 0 will disable it and cause performance issues but with higher compatability.[https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L312 More info] || --vu1-const-prop=1

| --vu1-opt-flags || 0, 1, 2 || Optimize flags upon recompilation (1 is safe, 2 update sticky, and mac flag on every status flag update) [https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L280 More Info] || --vu1-opt-flags=0

| --vu1-opt-vf00 || 0, 1, 2 || Optimization for the vf00 register. || --vu1-opt-vf00=2

| --vu1-jr-cache-policy || newprog, sameprog, auto, new, same || PCSX2 uses newprog as the default setting. Setting it into newprog has been proven to fix some crashes [https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L319 Info] || --vu1-jr-cache-policy=newprog

| --vu1-jalr-cache-policy || newprog, sameprog, auto, new, same || PCSX2 uses newprog as the default setting. Setting it into newprog has been proven to fix some crashes [https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L319 Info] || --vu1-jalr-cache-policy=newprog

|-style="background-color:#D7EF54"

|  || Game fixes / Graphical fixes || ||

| --vu1 || jit-sync ||  Selector between IR/JIT and it modes. Jit-sync works similarly to disabling MTVU. jit, trans, and jit-async are unimplemented functions that do not change anything. || --vu1=jit-sync

| --vu1-no-clamping || 0,1 || Changes the clamping behavior of the VU1. Setting specific to ps4 emulators, unimplemented in pcsx2, with 0 being closest to extra, and 1 to extra + preserve sign. The default emulator's value is 1. || --vu1-no-clamping=0

| --vu1-clamp-range || vu1 memory offset start,end (0x0 - 0x7FF) || Values must not exceed 0x7FF ||

| --vu1-accurate-addsub-range || vu1 memory offset start,end (0x0 - 0x7FF) || Allows the VU1 to accurately add/subtract in the specified ee memory range by using software floats rather than hardware floats.|| --vu1-accurate-addsub-range=0x0000,0x0600

| --vu1-mul0fix-range || vu1 memory offset start,end (0x0 - 0x7FF) || Allows the VU1 to accurately multiply by 0 in the specified ee memory range by using software floats rather than hardware floats. || --vu1-mul0fix-range=0x123,0x123

| --vu1-injection || 0, 1 || unknown || --vu1-injection=1

| --vu1-native-patch || 0, 1 || Use native patches from recompiler, not compatible with jak emus. Sony uses it for their official release of Primal|| --vu1-native-patch=1

| --vu1-use-rcp || 0, 1 || Use SSE/AVX RCP. Speedhack. Likely used only in EFU pipeline, due to rare use, its really not worth to enable it. Speed improvement will be marginally/unnoticeable. Disabled by default ||

| --vu1-use-rsqrt || 0, 1 || Use SSE/AVX rsqrt instead of doing sqrt and then div. Speedhack, rather safe to use on VU1. ||

| --vu1-clamp-operands || 0, 1 || Clamp operands before mathematical operation. || --vu1-clamp-operands=1

| --vu1-clamp-results || 0, 1 || Clamp result after mathematical operation. || --vu1-clamp-results=1

| --vu1-inst-q || 0, 1 || instant Q, no stalling on WAITQ, or instances of Q. || --vu1-inst-q=1  

| --assert-path1-ad || 0, 1 || Path 1 is how the GIF takes data from VU1 via XGKICK instruction. It's unknown what this command does but it's related to VU1, and helps prevent crashes in VU1 sensitive games when enabled. || --assert-path1-ad=1

=====VU0=====

<pre>VU0 is the Emotion Engine’s alternate processing unit. It is sometimes referred to as "micro-mode". Commands here prevent crashes, and fix SPS (graphical glitches).</pre>

| --vu0-mpg-cycles || Decimals<br>0 - 65535  || Set initial cycle speed for VU0 Micro-programs. 100 is the default value. It is not available in older emulators such as War of the Monsters. Raising the value helps COP2 timing, while lowering does the opposite.|| --vu0-mpg-cycles=1000<br><pre> Symbols (They change values close to VU0-mpg-cycles' offset):

 

$ goes to the specified succeeding offsets. $0 will go 2 bytes forward and changes them. $1 will go 4 bytes forward and changes the 2 bytes specified. For example, $0,$1 will change the 4 bytes ahead as the specified 2 bytes value.

 

'0x' Is like '+' a symbol that modifies succeeding values. 0x0 goes 2 bytes ahead and changes them, 0x1 goes 4 bytes ahead and so on and so forth.

 

'-' or '+' Changes the specified preceding or succeeding 2 bytes of VU0 mpg cycles' offset according to the added number. Adding -1, for example will change the 2 bytes before it. Likewise, adding -2 will go 4 bytes away and change the specified 2 bytes accordingly. It even can change EE cycle scalar itself at -15 (Jakv2). Setting it to --vu0-mpg-cycles=1280,-15 would, for example, set ee-cycle-scalar to 5. These numbers are in HEX. Infinite amounts of '-' and '+' instances seem to be supported.</pre>

| --vu0-di-bits || 0, 1 || 0 to skip setting invalid, and Div by zero flags in status register. Can be used as speedhack as it skips costly calculations, but at the same time can cause issues like broken geometry, and weird physics behavior. Depending per game.|| --vu0-di-bits=0

| --vu0-opt-vf00  || 0, 1, 2 || Optimization for the vf00 register. || --vu0-opt-vf00=1

| --vu0-opt-flags  || 0, 1, 2 || Optimize flags upon recompilation (1 is safe, 2 update sticky, and mac flag on every status flag update) [https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L280 More Info] || --vu0-opt-flags=1

| --vu0-jr-cache-policy || newprog, sameprog, auto, new, same || PCSX2 uses newprog as the default setting [https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L319 Info] || --vu0-jr-cache-policy=sameprog

| --vu0-jalr-cache-policy || newprog, sameprog, auto, new, same || PCSX2 uses newprog as the default setting [https://github.com/PCSX2/pcsx2/blob/58f5a5b915915293b90e7e8f34c33f9c0424d1e8/pcsx2/x86/microVU_Misc.h#L319 Info] || --vu0-jalr-cache-policy=sameprog

| --vu0-clamp-range || start, end offset (0x0 - 0x1FF) || per range clamping for VU0 || --vu0-clamp-range=0x100,0x120

| --vu0-no-clamping || 0,1 || Changes the clamping behavior of the VU0. Setting specific to ps4 emulators, unimplemented in pcsx2. The default emulator's value is 1. || --vu0-no-clamping=0

| --vu0-accurate-addsub-range || offset (0x0 - 0x1FF) ||Allows the VU0 to accurately Add/Subtract in the specified ee memory range by using software floats rather than hardware floats.||  

| --vu0-mul0fix-range || vu0 memory offset start,end (0x0 - 0x1FF) || Allows the VU0 to accurately Multiply by 0 in the specified ee memory range by using software floats rather than hardware floats ||

| --vu0-injection || 0, 1 || Unknown usage. || --vu0-injection=1

| --vu0-inst-p || 0, 1 || instant P, no stalling on WAITP, or instances of P|| --vu0-inst-p=1

| --vu0-use-rcp  || 0, 1 || Use SSE/AVX RCP. Speedhack. Not sure why it is even available for VU0. Disabled by default || --vu0-use-rcp=1

| --vu0-use-rsqrt || 0, 1 || Use SSE/AVX rsqrt instead of doing sqrt and then div. Speedhack, rather unsafe to use on VU0. || --vu0-use-rsqrt=1

| --vu0-clamp-operands  || 0, 1 || Clamp operands before mathematical operation. || --vu0-clamp-operands=1

| --vu0-clamp-results  || 0, 1 || Clamp result after mathematical operation. || --vu0-clamp-results=1

====DMA Channels====

=====VIF=====

<pre>The VIF decompresses vector data, uploads microprograms to the VUs, and sends graphical data to the GIF. Its commands

are often used to fix freezing.</pre>

! Command !! Values !! Notes !! Usage

|-style="background-color:#D7EF54"

|  || VIF || Game fixes ||

| --vif-ignore-invalid-cmd || 0, 1 || Ignore invalid vif commands. In some cases, setting it to 1 can fix games that freeze while showing the same frame. || --vif-ignore-invalid-cmd=1

| --vif-thread-chunk-size || 0, 1024 || Slows down both VIFS the lower it is, and speeds them up the higher it is. 1024 is the limit, with 112 being the default value.  || --vif-thread-chunk-size=100

| --detect-idle-vif || 0, 1 || Enabled by default. Unknown function || --detect-idle-vif=0

|-style="background-color:#ff8080"

| || VIF1 || Game fixes ||

| --vif1-ignore-cmd-ints || 0, 1 || Set to 1 to ignore command interrupt bit. ([https://psi-rockin.github.io/ps2tek/#vifcommands info] , explanation of int bit is right before command list). Can in some cases fix games that freeze while showing the same frame if set to 1. Games likes "Men in Black II - Alien Escape" and "Test Drive Unlimited" Will always need it. || --vif1-ignore-cmd-ints=1

| --vif1-instant-xfer || 1 = Instant VIF1<br>0 = Delayed VIF1 || Changes VIF1 timing. 0 slows its timing and is the compatible option for most games, while 1 speeds up its timing, and being more compatible with a small fraction of games. It can be used to fix graphical glitches or potentially prevent games from freezing. || --vif1-instant-xfer=0

=====SIF=====

<pre>The SIF is how the EE and IOP communicate with each other. SIF has some mailbox hardware registers that the CPUs can use to pass values to each other, which happens during SIF initialization. However, once both sides have booted, they use the SIF0 (IOP->EE) and SIF1 (EE->IOP) DMA channels to communicate. Despite common belief, these commands have no effect on performance, and are rarely ever useful.</pre>

|-style="background-color:#99bbff"

|  ||  || SIF1 (IOP) ||

| --iop-sif1-cycle-scalar || Accelerate SIF cycles: (0.99 => 0.1)<br>Decelerate SIF cycles: (1.1 => ∞ )|| Controls the speed at which the IOP receives from the EE || --iop-sif1-cycle-scalar=0.1

| --ee-sif1-cycle-scalar ||Accelerate SIF cycles: (0.99 => 0.1)<br>Decelerate SIF cycles: (1.1 => ∞ )|| Controls the speed at which the EE sends to the IOP || --ee-sif1-cycle-scalar=0.1

| --iop-sif0-cycle-scalar || Accelerate SIF cycles: (0.99 => 0.1)<br>Decelerate SIF cycles: (1.1 => ∞ )|| Controls the speed at which the IOP sends to the EE || --iop-sif0-cycle-scalar=0.1

| --ee-sif0-cycle-scalar || Accelerate SIF cycles: (0.99 => 0.1)<br>Decelerate SIF cycles: (1.1 => ∞ )|| Controls the speed at which the EE receives from the IOP || --ee-sif0-cycle-scalar=0.1

====[https://www.psdevwiki.com/ps2/Graphics_Synthesizer GS]====

<pre>"GS" stands for Graphics Synthesizer. It's the PlayStation®2's co-processor that's responsible for rendering the graphics (but is also used by some games for additional calculations and other purposes)</pre>

|-style="background-color:#7698FF"

|  || Speedhacks || ||

| --gs-optimize-30fps || 0, 1 || Speedhack for GS hungry games. || --gs-optimize-30fps=1

| --gs-adaptive-frameskip || 0, 1 || Speedhack. Skips frames when the gs demands more resources than there are available. doesn't work on all emulators.  It does however, work on jak emulators || --gs-adaptive-frameskip=1

| --gs-h2l-list-opt || 0, 1 || Setting it to 1 improves GS performance||--gs-h2l-list-opt=1

|-style="background-color:#cbddfb"

| || GS features || ||

| --gs-use-mipmap || 0, 1 || Enables mipmapping support. Can be used to fix graphics. Needs --gs-kernel-cl="mipmap" for full mipmapping support. If upscaling, however, --gs-kernel-cl-up="mipmap2x2" is also needed. It works best on War of the Monsters emulator, as the Ratchet series' mipmapping issues can only be fixed by it.|| --gs-use-mipmap=1

| --gs-use-deferred-l2h || 0, 1 || Delay option for L2H (local to host, GS to EE). || --gs-use-deferred-l2h=1

| --gs-use-clut-merge || 0, 1 || Enabling CLUT merging helps improve how the emulator processes textures that use indexed color palettes. It ensures more accurate emulation of the PS2’s color handling. It sometimes restores missing graphical elements, and fix texture flickering (prominently flickering shadows or UI elements).  It prevents Spyro's shadow from flickering in Enter the Dragonfly. Enabling it is the accurate, albeit slow option. || --gs-use-clut-merge=1

| --gs-flush-ad-xyz || always, safe, safeZwrite, off, 0 || Force a primitive flush when a framebuffer is also an input texture. This fixes some processing effects, but it can sometimes be heavy on the GS in terms of performance. GTA: SA, and Jak 3 use this command. || --gs-flush-ad-xyz=safe

| --gs-force-bilinear || 0, 1 || Forces bilinear filtering, can fix ghosting problems in some cases. Not recommended for games that use 2D images, as it could possibly break games that rely on nearest filtering.|| --gs-force-bilinear=1

|-style="background-color:#D7EF54"

|  || Gs settings / Behaviour || ||

| --gs-uprender || none,2x2 || Internal resolution upscaler. || --gs-uprender=2x2

| --gs-kernel-cl-up || "DarkCloud2" "fantavision" "h2lpool2x2", "OptRightTri", "clutmerge2x2", "mipmap2x2", "up2x2simple", "up2x2skipinterp", "up2x2tc", "up2x2", default || Kernel Variant Color lookup Upscaler (?), mipmap and clutmerge and h2l need to be enabled before their options become usable|| --gs-kernel-cl-up="clutmerge2x2"

|-

| --gs-override-small-tri-area || 0, 1 || This option is designed to modify how the emulator handles or renders small triangles that fall below a certain area threshold. Enabling this option could resolve graphical problems such as missing, flickering, or incorrectly rendered small details, and could potentially restore missing text in some games. || --gs-override-small-tri-area=1

| --gs-ignore-dirty-page-border || 0, 1 || Unknown. Used by Samurai Shodown Anthology ||--gs-ignore-dirty-page-border=1

|-

| --gs-ignore-rect-correction || 0, 1 || Setting it to 1 is known to fix flickering textures and graphical glitches in far distances. || --gs-ignore-rect-correction=1

| --gs-opt-frbuff-switch || 0, 1 || Unknown. Used by Everybody's Tennis/Hot Shots Tennis || --gs-opt-frbuff-switch=0

| --gs-kernel-cl || h2lpool, clutmerge, mipmap, DarkCloud2, fantavision, Mipmap and clutmerge and h2l need to be enabled before their options become usable ||Kernel Variant Color lookup(?). Options included here can be upscaling if --gs-kernel-cl-up were to be used along with it || --gs-kernel-cl="clutmerge"

| --force-frame-blend || 0, 1 || Enables blend (Deinterlacing?). Should be used to fix games with shaking screens. || --force-frame-blend=1

| --force-pal-60hz || 0, 1 || Enables 60hz PAL mode. || --force-pal-60hz=1

| --gs-progressive || 0, 1 || Enables progressive scan. Can fix graphical glitches/double screen issues.  ||

| --gs-vert-precision || 8, 16 || This option allows the user to choose between 8-bit and 16-bit precision for the vertex calculations in 3D graphics rendering. It reduces the likelihood of visual distortions such as jagged edges, improper alignment of polygons, or model "warping." The emu expects the values of 8 or 16, any other values are not permitted. || --gs-vert-precision=8

| --gs-skip-dirty-flush-on-mipmap || 0, 1 || Skip flushing textures marked as dirty from cache when enabled. It does, however, require mipmap to be enabled first. || --gs-skip-dirty-flush-on-mipmap=1

| --gs-fieldswap-delay || 0, 254 || Wait longer than usual to change field. In essence, it blurs/sharpens image. ||

|- 

| --gs-uv-shift-pointsampling || 0, 1 || Can be used on games like manhunt that have issues with [https://github.com/PCSX2/pcsx2/issues/257#issuecomment-1140338680 UV light rendering] || --gs-uv-shift-pointsampling=1

| --gs-render-tile-threshold || 0, 3000000 || Unknown. Used by Psychonauts|| --gs-render-tile-threshold=300000

| --threaded-gs || 0, 1 || Multi-threaded-gs ? || --threaded-gs=1

| --gs-aspect-ratio || aspect/float (default=0.81) || || --gs-aspect-ratio=0.65

| --gs-frontend-opt-mode || 0, 1, 2 || GS optimization mode? Unknown effect. Used on JakX || --gs-frontend-opt-mode=1

| --gs-motion-factor || 25, 50 ||  Unknown. Used by Wild ARMS 3|| --gs-motion-factor=25

| --gs-scanout-delay || 0, 200 || || --gs-scanout-delay=200

| --gs-check-trans-rejection || 0, 1 || Check transfer rejection? Unknown. Used by Grand Theft Auto: Vice City || --gs-check-trans-rejection=1

|-  

| --l2h-2d-params || TRXREG,BITBLTBUF,height || || --l2h-2d-params=0x0000000800000001,0x000000003a0a2300,512,2

| --gs-h2l-accurate-hash || 0, 1 || ||--gs-h2l-accurate-hash=1

|-style="background-color:#9042f5"

| || Aspect ratio || ||

| --gs-scanout-offsetx || relative offset/ignored || A zoom in setting in the x direction || --gs-scanout-offsetx=27

| --safe-area-min || area/float (range 0.9 to 1.0) any other values outside of that range will be rejected by the emulator || An overscan setting|| --safe-area-min=0.9

|-style="background-color:#FEA15C"

| || Other || ||

| --framelimit-mode || slowest,slower,slow,normal,fast,fastest,turbo ||Standalone Framelimiter|| --framelimit-mode=fast

| --framelimiter || 0, 1 || Enable or disable Frame limiting || --framelimiter=1

====IOP====

<pre>The emulated I/O Processor (IOP) settings. Commands here are rarely useful, as very few games will require them. The IOP controls the emulated DEV9, SPU2, USB, Memory cards, CDVD, Firewire, along with other input/output devices</pre>

! Command !! Values !! Notes !! Usage

|-style="background-color:#D7EF54"

| || Game fixes || ||

| --iop-cycle-scalar ||<br>Accelerate IOP cycles: (0.999 => 0.100)<br>Decelerate IOP cycles: (1.001 => 5.000)|| '''Not a performance speedhack'''. It impacts the speed of loading times and input latency. The default value for most emulators is 0.5|| --iop-cycle-scalar=1.0

| --iop-const-folding || None,Gpr,All || IOP constant folding [https://en.wikipedia.org/wiki/Constant_folding More info] || --iop-const-folding=All

| --iop-tight-slice-count || 0, 2000 || Seems to slow down the iop the higher the value. Can be used to fix CDVD timing issues. || --iop-tight-slice-count=12

| --iop-hook || AdvanceClock<br>FastForwardClock<br>Mfifodrain || IOP native hook, Sony uses it on their official release of Red dead revolver. It helps fix timing issues || --iop-hook=0x0086ac,FastForwardClock

| --iop-block-validation || IsC, ShortHash, Hash || Way of validating that block been modified, and require recompilation. Other Possible values include PageProt, None, PageProtection || --iop-block-validation=IsC

|-style="background-color:#FEA15C"

|  || Other || ||

| --iop-validate-kernel || 0, 1 || ||  

| --iop-shorthash-len || inst_count/integer || Length of shorthash in instruction/opcode count. Require --iop-block-validation=ShortHash ||

| --iop-pc-coherency || 0, 1||  ||  

| --iop-inst-marking || 0, 1? || ||  

| --detect-idle-iop || 0, 1 || Enabled by default. || --detect-idle-iop=0

| --iop-jit-disasm || || ||  

| --iop-evt-check-full || 0, 1 ?|| ||  

|}

 

 

{| cellspacing="0" cellpadding="2" border="1" class="wikitable" style="text-align: center;"

|-style="background-color:#e9f562"

|  || General commands || ||

| --max-disc-num || 1-5  || numbers of discs in package (maximum=5)  || --max-disc-num=1

| --boot-disc-id || 1-5  || sets boot disc for multi-disc pkg || --boot-disc-id=0