dolphin/Source/Core/VideoCommon/CommandProcessor.cpp
degasus d31bed8b79 Fifo: Rewrite SyncGpu
The new implementation has 3 options:
 SyncGpuMaxDistance
 SyncGpuMinDistance
 SyncGpuOverclock

The MaxDistance controlls how many CPU cycles the CPU is allowed to be in front
of the GPU. Too low values will slow down extremly, too high values are as
unsynchronized and half of the games will crash.
The -MinDistance (negative) set how many cycles the GPU is allowed to be in
front of the CPU. As we are used to emulate an infinitiv fast GPU, this may be
set to any high (negative) number.

The last parameter is to hack a faster (>1.0) or slower(<1.0) GPU. As we don't
emulate GPU timing very well (eg skip the timings of the pixel stage completely),
an overclock factor of ~0.5 is often much more accurate than 1.0
2015-06-08 23:16:24 +02:00

547 lines
15 KiB
C++

// Copyright 2008 Dolphin Emulator Project
// Licensed under GPLv2+
// Refer to the license.txt file included.
#include <atomic>
#include "Common/Atomic.h"
#include "Common/ChunkFile.h"
#include "Common/CommonTypes.h"
#include "Common/MathUtil.h"
#include "Common/Thread.h"
#include "Core/ConfigManager.h"
#include "Core/Core.h"
#include "Core/CoreTiming.h"
#include "Core/HW/GPFifo.h"
#include "Core/HW/Memmap.h"
#include "Core/HW/MMIO.h"
#include "Core/HW/ProcessorInterface.h"
#include "VideoCommon/CommandProcessor.h"
#include "VideoCommon/Fifo.h"
#include "VideoCommon/PixelEngine.h"
#include "VideoCommon/VideoCommon.h"
#include "VideoCommon/VideoConfig.h"
namespace CommandProcessor
{
static int et_UpdateInterrupts;
// TODO(ector): Warn on bbox read/write
// STATE_TO_SAVE
SCPFifoStruct fifo;
static UCPStatusReg m_CPStatusReg;
static UCPCtrlReg m_CPCtrlReg;
static UCPClearReg m_CPClearReg;
static u16 m_bboxleft;
static u16 m_bboxtop;
static u16 m_bboxright;
static u16 m_bboxbottom;
static u16 m_tokenReg;
static std::atomic<bool> s_interrupt_set;
static std::atomic<bool> s_interrupt_waiting;
static std::atomic<bool> s_interrupt_token_waiting;
static std::atomic<bool> s_interrupt_finish_waiting;
static bool IsOnThread()
{
return SConfig::GetInstance().m_LocalCoreStartupParameter.bCPUThread;
}
static void UpdateInterrupts_Wrapper(u64 userdata, int cyclesLate)
{
UpdateInterrupts(userdata);
}
void DoState(PointerWrap &p)
{
p.DoPOD(m_CPStatusReg);
p.DoPOD(m_CPCtrlReg);
p.DoPOD(m_CPClearReg);
p.Do(m_bboxleft);
p.Do(m_bboxtop);
p.Do(m_bboxright);
p.Do(m_bboxbottom);
p.Do(m_tokenReg);
p.Do(fifo);
p.Do(s_interrupt_set);
p.Do(s_interrupt_waiting);
p.Do(s_interrupt_token_waiting);
p.Do(s_interrupt_finish_waiting);
}
static inline void WriteLow(volatile u32& _reg, u16 lowbits)
{
Common::AtomicStore(_reg, (_reg & 0xFFFF0000) | lowbits);
}
static inline void WriteHigh(volatile u32& _reg, u16 highbits)
{
Common::AtomicStore(_reg, (_reg & 0x0000FFFF) | ((u32)highbits << 16));
}
static inline u16 ReadLow(u32 _reg)
{
return (u16)(_reg & 0xFFFF);
}
static inline u16 ReadHigh(u32 _reg)
{
return (u16)(_reg >> 16);
}
void Init()
{
m_CPStatusReg.Hex = 0;
m_CPStatusReg.CommandIdle = 1;
m_CPStatusReg.ReadIdle = 1;
m_CPCtrlReg.Hex = 0;
m_CPClearReg.Hex = 0;
m_bboxleft = 0;
m_bboxtop = 0;
m_bboxright = 640;
m_bboxbottom = 480;
m_tokenReg = 0;
memset(&fifo,0,sizeof(fifo));
fifo.bFF_Breakpoint = 0;
fifo.bFF_HiWatermark = 0;
fifo.bFF_HiWatermarkInt = 0;
fifo.bFF_LoWatermark = 0;
fifo.bFF_LoWatermarkInt = 0;
s_interrupt_set.store(false);
s_interrupt_waiting.store(false);
s_interrupt_finish_waiting.store(false);
s_interrupt_token_waiting.store(false);
et_UpdateInterrupts = CoreTiming::RegisterEvent("CPInterrupt", UpdateInterrupts_Wrapper);
}
void RegisterMMIO(MMIO::Mapping* mmio, u32 base)
{
struct {
u32 addr;
u16* ptr;
bool readonly;
bool writes_align_to_32_bytes;
} directly_mapped_vars[] = {
{ FIFO_TOKEN_REGISTER, &m_tokenReg },
// Bounding box registers are read only.
{ FIFO_BOUNDING_BOX_LEFT, &m_bboxleft, true },
{ FIFO_BOUNDING_BOX_RIGHT, &m_bboxright, true },
{ FIFO_BOUNDING_BOX_TOP, &m_bboxtop, true },
{ FIFO_BOUNDING_BOX_BOTTOM, &m_bboxbottom, true },
// Some FIFO addresses need to be aligned on 32 bytes on write - only
// the high part can be written directly without a mask.
{ FIFO_BASE_LO, MMIO::Utils::LowPart(&fifo.CPBase), false, true },
{ FIFO_BASE_HI, MMIO::Utils::HighPart(&fifo.CPBase) },
{ FIFO_END_LO, MMIO::Utils::LowPart(&fifo.CPEnd), false, true },
{ FIFO_END_HI, MMIO::Utils::HighPart(&fifo.CPEnd) },
{ FIFO_HI_WATERMARK_LO, MMIO::Utils::LowPart(&fifo.CPHiWatermark) },
{ FIFO_HI_WATERMARK_HI, MMIO::Utils::HighPart(&fifo.CPHiWatermark) },
{ FIFO_LO_WATERMARK_LO, MMIO::Utils::LowPart(&fifo.CPLoWatermark) },
{ FIFO_LO_WATERMARK_HI, MMIO::Utils::HighPart(&fifo.CPLoWatermark) },
// FIFO_RW_DISTANCE has some complex read code different for
// single/dual core.
{ FIFO_WRITE_POINTER_LO, MMIO::Utils::LowPart(&fifo.CPWritePointer), false, true },
{ FIFO_WRITE_POINTER_HI, MMIO::Utils::HighPart(&fifo.CPWritePointer) },
// FIFO_READ_POINTER has different code for single/dual core.
};
for (auto& mapped_var : directly_mapped_vars)
{
u16 wmask = mapped_var.writes_align_to_32_bytes ? 0xFFE0 : 0xFFFF;
mmio->Register(base | mapped_var.addr,
MMIO::DirectRead<u16>(mapped_var.ptr),
mapped_var.readonly
? MMIO::InvalidWrite<u16>()
: MMIO::DirectWrite<u16>(mapped_var.ptr, wmask)
);
}
mmio->Register(base | FIFO_BP_LO,
MMIO::DirectRead<u16>(MMIO::Utils::LowPart(&fifo.CPBreakpoint)),
MMIO::ComplexWrite<u16>([](u32, u16 val) {
WriteLow(fifo.CPBreakpoint, val & 0xffe0);
})
);
mmio->Register(base | FIFO_BP_HI,
MMIO::DirectRead<u16>(MMIO::Utils::HighPart(&fifo.CPBreakpoint)),
MMIO::ComplexWrite<u16>([](u32, u16 val) {
WriteHigh(fifo.CPBreakpoint, val);
})
);
// Timing and metrics MMIOs are stubbed with fixed values.
struct {
u32 addr;
u16 value;
} metrics_mmios[] = {
{ XF_RASBUSY_L, 0 },
{ XF_RASBUSY_H, 0 },
{ XF_CLKS_L, 0 },
{ XF_CLKS_H, 0 },
{ XF_WAIT_IN_L, 0 },
{ XF_WAIT_IN_H, 0 },
{ XF_WAIT_OUT_L, 0 },
{ XF_WAIT_OUT_H, 0 },
{ VCACHE_METRIC_CHECK_L, 0 },
{ VCACHE_METRIC_CHECK_H, 0 },
{ VCACHE_METRIC_MISS_L, 0 },
{ VCACHE_METRIC_MISS_H, 0 },
{ VCACHE_METRIC_STALL_L, 0 },
{ VCACHE_METRIC_STALL_H, 0 },
{ CLKS_PER_VTX_OUT, 4 },
};
for (auto& metrics_mmio : metrics_mmios)
{
mmio->Register(base | metrics_mmio.addr,
MMIO::Constant<u16>(metrics_mmio.value),
MMIO::InvalidWrite<u16>()
);
}
mmio->Register(base | STATUS_REGISTER,
MMIO::ComplexRead<u16>([](u32) {
SetCpStatusRegister();
return m_CPStatusReg.Hex;
}),
MMIO::InvalidWrite<u16>()
);
mmio->Register(base | CTRL_REGISTER,
MMIO::DirectRead<u16>(&m_CPCtrlReg.Hex),
MMIO::ComplexWrite<u16>([](u32, u16 val) {
UCPCtrlReg tmp(val);
m_CPCtrlReg.Hex = tmp.Hex;
SetCpControlRegister();
RunGpu();
})
);
mmio->Register(base | CLEAR_REGISTER,
MMIO::DirectRead<u16>(&m_CPClearReg.Hex),
MMIO::ComplexWrite<u16>([](u32, u16 val) {
UCPClearReg tmp(val);
m_CPClearReg.Hex = tmp.Hex;
SetCpClearRegister();
RunGpu();
})
);
mmio->Register(base | PERF_SELECT,
MMIO::InvalidRead<u16>(),
MMIO::Nop<u16>()
);
// Some MMIOs have different handlers for single core vs. dual core mode.
mmio->Register(base | FIFO_RW_DISTANCE_LO,
IsOnThread()
? MMIO::ComplexRead<u16>([](u32) {
if (fifo.CPWritePointer >= fifo.SafeCPReadPointer)
return ReadLow(fifo.CPWritePointer - fifo.SafeCPReadPointer);
else
return ReadLow(fifo.CPEnd - fifo.SafeCPReadPointer + fifo.CPWritePointer - fifo.CPBase + 32);
})
: MMIO::DirectRead<u16>(MMIO::Utils::LowPart(&fifo.CPReadWriteDistance)),
MMIO::DirectWrite<u16>(MMIO::Utils::LowPart(&fifo.CPReadWriteDistance), 0xFFE0)
);
mmio->Register(base | FIFO_RW_DISTANCE_HI,
IsOnThread()
? MMIO::ComplexRead<u16>([](u32) {
if (fifo.CPWritePointer >= fifo.SafeCPReadPointer)
return ReadHigh(fifo.CPWritePointer - fifo.SafeCPReadPointer);
else
return ReadHigh(fifo.CPEnd - fifo.SafeCPReadPointer + fifo.CPWritePointer - fifo.CPBase + 32);
})
: MMIO::DirectRead<u16>(MMIO::Utils::HighPart(&fifo.CPReadWriteDistance)),
MMIO::ComplexWrite<u16>([](u32, u16 val) {
WriteHigh(fifo.CPReadWriteDistance, val);
SyncGPU(SYNC_GPU_OTHER);
if (fifo.CPReadWriteDistance == 0)
{
GPFifo::ResetGatherPipe();
ResetVideoBuffer();
}
else
{
ResetVideoBuffer();
}
RunGpu();
})
);
mmio->Register(base | FIFO_READ_POINTER_LO,
IsOnThread()
? MMIO::DirectRead<u16>(MMIO::Utils::LowPart(&fifo.SafeCPReadPointer))
: MMIO::DirectRead<u16>(MMIO::Utils::LowPart(&fifo.CPReadPointer)),
MMIO::DirectWrite<u16>(MMIO::Utils::LowPart(&fifo.CPReadPointer), 0xFFE0)
);
mmio->Register(base | FIFO_READ_POINTER_HI,
IsOnThread()
? MMIO::DirectRead<u16>(MMIO::Utils::HighPart(&fifo.SafeCPReadPointer))
: MMIO::DirectRead<u16>(MMIO::Utils::HighPart(&fifo.CPReadPointer)),
IsOnThread()
? MMIO::ComplexWrite<u16>([](u32, u16 val) {
WriteHigh(fifo.CPReadPointer, val);
fifo.SafeCPReadPointer = fifo.CPReadPointer;
})
: MMIO::DirectWrite<u16>(MMIO::Utils::HighPart(&fifo.CPReadPointer))
);
}
void GatherPipeBursted()
{
if (IsOnThread())
SetCPStatusFromCPU();
ProcessFifoEvents();
// if we aren't linked, we don't care about gather pipe data
if (!m_CPCtrlReg.GPLinkEnable)
{
if (IsOnThread() && !g_use_deterministic_gpu_thread)
{
// In multibuffer mode is not allowed write in the same FIFO attached to the GPU.
// Fix Pokemon XD in DC mode.
if ((ProcessorInterface::Fifo_CPUEnd == fifo.CPEnd) &&
(ProcessorInterface::Fifo_CPUBase == fifo.CPBase) &&
fifo.CPReadWriteDistance > 0)
{
FlushGpu();
}
}
RunGpu();
return;
}
// update the fifo pointer
if (fifo.CPWritePointer == fifo.CPEnd)
fifo.CPWritePointer = fifo.CPBase;
else
fifo.CPWritePointer += GATHER_PIPE_SIZE;
if (m_CPCtrlReg.GPReadEnable && m_CPCtrlReg.GPLinkEnable)
{
ProcessorInterface::Fifo_CPUWritePointer = fifo.CPWritePointer;
ProcessorInterface::Fifo_CPUBase = fifo.CPBase;
ProcessorInterface::Fifo_CPUEnd = fifo.CPEnd;
}
// If the game is running close to overflowing, make the exception checking more frequent.
if (fifo.bFF_HiWatermark)
CoreTiming::ForceExceptionCheck(0);
Common::AtomicAdd(fifo.CPReadWriteDistance, GATHER_PIPE_SIZE);
RunGpu();
_assert_msg_(COMMANDPROCESSOR, fifo.CPReadWriteDistance <= fifo.CPEnd - fifo.CPBase,
"FIFO is overflowed by GatherPipe !\nCPU thread is too fast!");
// check if we are in sync
_assert_msg_(COMMANDPROCESSOR, fifo.CPWritePointer == ProcessorInterface::Fifo_CPUWritePointer, "FIFOs linked but out of sync");
_assert_msg_(COMMANDPROCESSOR, fifo.CPBase == ProcessorInterface::Fifo_CPUBase, "FIFOs linked but out of sync");
_assert_msg_(COMMANDPROCESSOR, fifo.CPEnd == ProcessorInterface::Fifo_CPUEnd, "FIFOs linked but out of sync");
}
void UpdateInterrupts(u64 userdata)
{
if (userdata)
{
s_interrupt_set.store(true);
INFO_LOG(COMMANDPROCESSOR,"Interrupt set");
ProcessorInterface::SetInterrupt(INT_CAUSE_CP, true);
}
else
{
s_interrupt_set.store(false);
INFO_LOG(COMMANDPROCESSOR,"Interrupt cleared");
ProcessorInterface::SetInterrupt(INT_CAUSE_CP, false);
}
CoreTiming::ForceExceptionCheck(0);
s_interrupt_waiting.store(false);
RunGpu();
}
void UpdateInterruptsFromVideoBackend(u64 userdata)
{
if (!g_use_deterministic_gpu_thread)
CoreTiming::ScheduleEvent_Threadsafe(0, et_UpdateInterrupts, userdata);
}
bool IsInterruptWaiting()
{
return s_interrupt_waiting.load();
}
void SetInterruptTokenWaiting(bool waiting)
{
s_interrupt_token_waiting.store(waiting);
}
void SetInterruptFinishWaiting(bool waiting)
{
s_interrupt_finish_waiting.store(waiting);
}
void SetCPStatusFromGPU()
{
// breakpoint
if (fifo.bFF_BPEnable)
{
if (fifo.CPBreakpoint == fifo.CPReadPointer)
{
if (!fifo.bFF_Breakpoint)
{
INFO_LOG(COMMANDPROCESSOR, "Hit breakpoint at %i", fifo.CPReadPointer);
fifo.bFF_Breakpoint = true;
}
}
else
{
if (fifo.bFF_Breakpoint)
INFO_LOG(COMMANDPROCESSOR, "Cleared breakpoint at %i", fifo.CPReadPointer);
fifo.bFF_Breakpoint = false;
}
}
else
{
if (fifo.bFF_Breakpoint)
INFO_LOG(COMMANDPROCESSOR, "Cleared breakpoint at %i", fifo.CPReadPointer);
fifo.bFF_Breakpoint = false;
}
// overflow & underflow check
fifo.bFF_HiWatermark = (fifo.CPReadWriteDistance > fifo.CPHiWatermark);
fifo.bFF_LoWatermark = (fifo.CPReadWriteDistance < fifo.CPLoWatermark);
bool bpInt = fifo.bFF_Breakpoint && fifo.bFF_BPInt;
bool ovfInt = fifo.bFF_HiWatermark && fifo.bFF_HiWatermarkInt;
bool undfInt = fifo.bFF_LoWatermark && fifo.bFF_LoWatermarkInt;
bool interrupt = (bpInt || ovfInt || undfInt) && m_CPCtrlReg.GPReadEnable;
if (interrupt != s_interrupt_set.load() && !s_interrupt_waiting.load())
{
u64 userdata = interrupt ? 1 : 0;
if (IsOnThread())
{
if (!interrupt || bpInt || undfInt || ovfInt)
{
// Schedule the interrupt asynchronously
s_interrupt_waiting.store(true);
CommandProcessor::UpdateInterruptsFromVideoBackend(userdata);
}
}
else
{
CommandProcessor::UpdateInterrupts(userdata);
}
}
}
void SetCPStatusFromCPU()
{
// overflow & underflow check
fifo.bFF_HiWatermark = (fifo.CPReadWriteDistance > fifo.CPHiWatermark);
fifo.bFF_LoWatermark = (fifo.CPReadWriteDistance < fifo.CPLoWatermark);
bool bpInt = fifo.bFF_Breakpoint && fifo.bFF_BPInt;
bool ovfInt = fifo.bFF_HiWatermark && fifo.bFF_HiWatermarkInt;
bool undfInt = fifo.bFF_LoWatermark && fifo.bFF_LoWatermarkInt;
bool interrupt = (bpInt || ovfInt || undfInt) && m_CPCtrlReg.GPReadEnable;
if (interrupt != s_interrupt_set.load() && !s_interrupt_waiting.load())
{
u64 userdata = interrupt ? 1 : 0;
if (IsOnThread())
{
if (!interrupt || bpInt || undfInt || ovfInt)
{
s_interrupt_set.store(interrupt);
INFO_LOG(COMMANDPROCESSOR,"Interrupt set");
ProcessorInterface::SetInterrupt(INT_CAUSE_CP, interrupt);
}
}
else
{
CommandProcessor::UpdateInterrupts(userdata);
}
}
}
void ProcessFifoEvents()
{
if (IsOnThread() && (s_interrupt_waiting.load() || s_interrupt_finish_waiting.load() || s_interrupt_token_waiting.load()))
CoreTiming::ProcessFifoWaitEvents();
}
void Shutdown()
{
}
void SetCpStatusRegister()
{
// Here always there is one fifo attached to the GPU
m_CPStatusReg.Breakpoint = fifo.bFF_Breakpoint;
m_CPStatusReg.ReadIdle = !fifo.CPReadWriteDistance || (fifo.CPReadPointer == fifo.CPWritePointer);
m_CPStatusReg.CommandIdle = !fifo.CPReadWriteDistance || AtBreakpoint() || !fifo.bFF_GPReadEnable;
m_CPStatusReg.UnderflowLoWatermark = fifo.bFF_LoWatermark;
m_CPStatusReg.OverflowHiWatermark = fifo.bFF_HiWatermark;
INFO_LOG(COMMANDPROCESSOR,"\t Read from STATUS_REGISTER : %04x", m_CPStatusReg.Hex);
DEBUG_LOG(COMMANDPROCESSOR, "(r) status: iBP %s | fReadIdle %s | fCmdIdle %s | iOvF %s | iUndF %s"
, m_CPStatusReg.Breakpoint ? "ON" : "OFF"
, m_CPStatusReg.ReadIdle ? "ON" : "OFF"
, m_CPStatusReg.CommandIdle ? "ON" : "OFF"
, m_CPStatusReg.OverflowHiWatermark ? "ON" : "OFF"
, m_CPStatusReg.UnderflowLoWatermark ? "ON" : "OFF"
);
}
void SetCpControlRegister()
{
fifo.bFF_BPInt = m_CPCtrlReg.BPInt;
fifo.bFF_BPEnable = m_CPCtrlReg.BPEnable;
fifo.bFF_HiWatermarkInt = m_CPCtrlReg.FifoOverflowIntEnable;
fifo.bFF_LoWatermarkInt = m_CPCtrlReg.FifoUnderflowIntEnable;
fifo.bFF_GPLinkEnable = m_CPCtrlReg.GPLinkEnable;
if (fifo.bFF_GPReadEnable && !m_CPCtrlReg.GPReadEnable)
{
fifo.bFF_GPReadEnable = m_CPCtrlReg.GPReadEnable;
FlushGpu();
}
else
{
fifo.bFF_GPReadEnable = m_CPCtrlReg.GPReadEnable;
}
DEBUG_LOG(COMMANDPROCESSOR, "\t GPREAD %s | BP %s | Int %s | OvF %s | UndF %s | LINK %s"
, fifo.bFF_GPReadEnable ? "ON" : "OFF"
, fifo.bFF_BPEnable ? "ON" : "OFF"
, fifo.bFF_BPInt ? "ON" : "OFF"
, m_CPCtrlReg.FifoOverflowIntEnable ? "ON" : "OFF"
, m_CPCtrlReg.FifoUnderflowIntEnable ? "ON" : "OFF"
, m_CPCtrlReg.GPLinkEnable ? "ON" : "OFF"
);
}
// NOTE: We intentionally don't emulate this function at the moment.
// We don't emulate proper GP timing anyway at the moment, so it would just slow down emulation.
void SetCpClearRegister()
{
}
} // end of namespace CommandProcessor