This noticeably includes GL_ARB_get_program_binary, which was previously
thought unsupported on OS X. Well, actually, the OS X implementation is
trivial and reports 0 binary formats (as of 10.10; this is hardcoded in
GLEngine, by the way), but at least it'll work if it's fixed someday.
It now affects the GPU determinism mode as well as some miscellaneous
things that were calling IsNetPlayRunning. Probably incomplete.
Notably, this can change while paused, if the user starts recording a
movie. The movie code appears to have been missing locking between
setting g_playMode and doing other things, which probably had a small
chance of causing crashes or even desynced movies; fix that with
PauseAndLock.
The next commit will add a hidden config variable to override GPU
determinism mode.
It's a relatively big commit (less big with -w), but it's hard to test
any of this separately...
The basic problem is that in netplay or movies, the state of the CPU
must be deterministic, including when the game receives notification
that the GPU has processed FIFO data. Dual core mode notifies the game
whenever the GPU thread actually gets around to doing the work, so it
isn't deterministic. Single core mode is because it notifies the game
'instantly' (after processing the data synchronously), but it's too slow
for many systems and games.
My old dc-netplay branch worked as follows: everything worked as normal
except the state of the CP registers was a lie, and the CPU thread only
delivered results when idle detection triggered (waiting for the GPU if
they weren't ready at that point). Usually, a game is idle iff all the
work for the frame has been done, except for a small amount of work
depending on the GPU result, so neither the CPU or the GPU waiting on
the other affected performance much. However, it's possible that the
game could be waiting for some earlier interrupt, and any of several
games which, for whatever reason, never went into a detectable idle
(even when I tried to improve the detection) would never receive results
at all. (The current method should have better compatibility, but it
also has slightly higher overhead and breaks some other things, so I
want to reimplement this, hopefully with less impact on the code, in the
future.)
With this commit, the basic idea is that the CPU thread acts as if the
work has been done instantly, like single core mode, but actually hands
it off asynchronously to the GPU thread (after backing up some data that
the game might change in memory before it's actually done). Since the
work isn't done, any feedback from the GPU to the CPU, such as real
XFB/EFB copies (virtual are OK), EFB pokes, performance queries, etc. is
broken; but most games work with these options disabled, and there is no
need to try to detect what the CPU thread is doing.
Technically: when the flag g_use_deterministic_gpu_thread (currently
stuck on) is on, the CPU thread calls RunGpu like in single core mode.
This function synchronously copies the data from the FIFO to the
internal video buffer and updates the CP registers, interrupts, etc.
However, instead of the regular ReadDataFromFifo followed by running the
opcode decoder, it runs ReadDataFromFifoOnCPU ->
OpcodeDecoder_Preprocess, which relatively quickly scans through the
FIFO data, detects SetFinish calls etc., which are immediately fired,
and saves certain associated data from memory (e.g. display lists) in
AuxBuffers (a parallel stream to the main FIFO, which is a bit slow at
the moment), before handing the data off to the GPU thread to actually
render. That makes up the bulk of this commit.
In various circumstances, including the aforementioned EFB pokes and
performance queries as well as swap requests (i.e. the end of a frame -
we don't want the CPU potentially pumping out frames too quickly and the
GPU falling behind*), SyncGPU is called to wait for actual completion.
The overhead mainly comes from OpcodeDecoder_Preprocess (which is,
again, synchronous), as well as the actual copying.
Currently, display lists and such are escrowed from main memory even
though they usually won't change over the course of a frame, and
textures are not even though they might, resulting in a small chance of
graphical glitches. When the texture locking (i.e. fault on write) code
lands, I can make this all correct and maybe a little faster.
* This suggests an alternate determinism method of just delaying results
until a short time before the end of each frame. For all I know this
might mostly work - I haven't tried it - but if any significant work
hinges on the competion of render to texture etc., the frame will be
missed.
videoBuffer -> s_video_buffer
size -> s_video_buffer_write_ptr
g_pVideoData -> g_video_buffer_read_ptr (impl moved to Fifo.cpp)
This eradicates the wonderful use of 'size' as a global name, and makes
it clear that s_video_buffer_write_ptr and g_video_buffer_read_ptr are
the two ends of the FIFO buffer s_video_buffer.
Oh, and remove a useless namespace {}.
This state will be used to calculate sizes for skipping over commands on
a separate thread. An alternative to having these state variables would
be to have the preprocessor stash "state as we go" somewhere, but I
think that would be much uglier.
GetVertexSize now takes an extra argument to determine which state to
use, as does FifoCommandRunnable, which calls it. While I'm modifying
FifoCommandRunnable, I also change it to take a buffer and size as
parameters rather than using g_pVideoData, which will also be necessary
later. I also get rid of an unused overload.
VertexLoader::VertexLoader was setting loop_counter, a *static*
variable, to 0. This was nonsensical, but harmless until I started to
run it on a separate thread, where it had a chance of interfering with a
running vertex translator.
Switch to just using a register for the loop counter.
- Lazily create the native vertex format (which involves GL calls) from
RunVertices rather than RefreshLoader itself, freeing the latter to be
run from the CPU thread (hopefully).
- In order to avoid useless allocations while doing so, store the native
format inside the VertexLoader rather than using a cache entry.
- Wrap the s_vertex_loader_map in a lock, for similar reasons.
Renamed various commands to refer to ISO instead of GCM for consistency,
as the commands are used for both Wii and GameCube files.
CompressGCM --> CompressISO
DeleteGCM --> DeleteISO
MultiCompressGCM --> MultiCompressISO
MultiDecompressGCM --> MultiDecompressISO
SetDefaultGCM --> SetDefaultISO
Fixed COMPRESSISO
Fixed missing "COMPRESSISO"
Fixed more COMPRESSISO
Final fix for COMPRESSISO
Detects a situation where the game is writing to the dcache at the address being DMA'd. As we do not have dcache emulation, invalid data is being DMA'd causing audio glitches. The following code detects this and enables the DMA to complete instantly before the invalid data is written.
Added accurate ARAM DMA transfer timing.
Removed the addition of DSP exception checking.
Updated ARAM DMA and FIFO write exception checking to uses these types.
Conflicts:
Source/Core/Core/PowerPC/Interpreter/Interpreter_Tables.cpp
Source/Core/Core/PowerPC/PPCTables.h
This helps us avoid accidentally clobbering flags between two instructions
when the flags are expected to be maintained. Dolphin will of course crash
immediately, but at least it will crash loudly and alert us of the mistake,
instead of forcing hours of bisecting to find the subtle way in which the JIT
has managed to sneak a flag-modifying instruction where there shouldn't be one.
That commit reorganized fastmem a bit; I wrote it before the patch to
support fastmem in JitIL landed, and forgot to edit it to account for
the fact. Since JitILBase now derives from Jitx86Base, the HandleFault
override can just be removed.
Also correct behavior with regards to which bits in XER are treated as zero
based on a hwtest (probably doesn't affect any real games, but might as well
be correct).
This should dramatically reduce code size in the case of blocks with
lots of branches, and certainly doesn't hurt elsewhere either.
This can probably be improved a good bit through smarter tracking of register
usage, e.g. discarding registers that are going to be overwritten, but this
is a good start and should help reduce code size and register pressure.
Unlike that sort of change, this is a "safe" patch; it only flushes registers,
which can't affect correctness, unlike actually discarding data.
As part of this, refactor PPCAnalyst to support distinguishing between
float and integer registers (to properly handle instructions that access
both, like floating-point loads and stores).
Also update every instruction in the interpreter flags table I could find
that didn't have all the correct flags.
When a game was running and someone opened the video dialog, it would crash. This is because the preprocessor macro should have been __APPLE__ not _APPLE_
Fixes issue 7644.
These ID values would clash with the window parent IDs of all the actual debugger panes (they are in the 350 range as well).
For example, attempting to show and then close the memory window would cause an assertion, because it would attempt to destroy the text control for searching through memory, rather than destroying the actual parent window it's attached to.
These IDs are only used locally, so their value doesn't matter.
This is inconsistent with how other containers are used (i.e. with Do()), but making std::array be used with Do() seems rather confusing when there's also a DoArray available.
I didn't realize the I and W fields were in a different place for these
variants.
This should fix Paper Mario and probably lots of other things I accidentally
broke.
fselx was the main problem, but ps_sel was wrong too (even if there were no
known reported bugs with it).
This fixes Beyond Good and Evil (at the least).
4 more instructions down.
Store ones should be pretty well-tested; load ones seem to almost never be
used. I found them in Turok Evolution, so I was able to check code generation,
but the relevant code didn't seem to be called.
My code didn't maintain correct semantics with floating-point NaNs (a < b is
not the same as "not a >= b" in float), which seems to have broken FIFA 12.
Rather than *MemTools.cpp checking whether the address is in the
emulated range itself (which, as of the next commit, doesn't cover every
kind of access the JIT might want to intercept) and doing PC
replacement, they just pass the access address and context to
jit->HandleFault, which does the rest itself.
Because SContext is now in JitInterface, I wanted JitBackpatch.h (which
defines it) to be lightweight, so I moved TrampolineCache and associated
x64{Analyzer,Emitter} dependencies into its own file. I hate adding new
files in three places, two of which are MSVC...
While I'm at it, edit a misleading comment.
When executing a BL-type instruction, push the new LR onto the stack,
then CALL the dispatcher or linked block rather than JMPing to it. When
executing BLR, compare [rsp+8] to LR, and RET if it's right, which it
usually will be unless the thread was switched out. If it's not right,
reset RSP to avoid overflow.
This both saves a trip through the dispatcher and improves branch
prediction.
There is a small possibility of stack overflow anyway, which should
be handled... *yawn*
These calls are made outside of JIT blocks, and thus previously did not
read any protection - register use is taken into account and the outer
dispatcher stack frame is sufficient. However, if data is to be stored
on the stack, these calls must reserve stack shadow space on Windows to
avoid clobbering it.
Use some SSE4 instructions in on CPUs that support them.
Use float instructions instead of int where appropriate (it's a cycle faster
on CPUs with arithmetic unit forwarding penalties).
JitIL's fastmem was stubbed out when Sonicadvance1 merged JitARMIL
into the tree. Since JitARMIL has been deleted, I simply re-arrange
the inheritance to base JitIL on Jitx86Base, so it can inherit the
backpatch function.
Povray Benchmark: 1985 seconds to 1316 seconds.
Tries as hard as possible to push carry-using operations (like addc and adde)
next to each other. Refactor the instruction reordering to be more flexible
and allow multiple passes.
353 -> 192 x86 instructions on a carry-heavy code block in Pokemon Puzzle.
12% faster overall in Pokemon Puzzle; probably less in typical games (Virtual
Console games seem to be carry-heavy for some reason; maybe a different
compiler?)
Has been broken since the flags-opt merge. The idle skipping code in
JitIL was very brittle and depended on the IL of it's inputs not
changing in any way.
flags-opt changed the IR generated by the cmp instruction, which is part
of the idle loop, causing JitIL to break in really weird ways, which
were almost impossible to track down.
This fixes various wii games crashing/not booting and the Regspill
error on (all?) gamecube mmu games.
Fixes all the current issues I've been experiencing.
Scaled back the register cache idea for now so I can actually work on some real instructions.
Tested this work with unit tests so I know it works.
Unit tests are pretty great things.
In situations where conditional continue isn't supported + if a JIT doesn't implement a instruction that has the FL_ENDBLOCK flag. This would cause an
infinite loop.
In reality all the JITs should implement every FL_ENDBLOCK instruction regardless, but JITIL doesn't implement tw/twi which are FL_ENDBLOCK
instructions.
skidau