video_core: NVDEC Implementation

This commit aims to implement the NVDEC (Nvidia Decoder) functionality, with video frame decoding being handled by the FFmpeg library.

The process begins with Ioctl commands being sent to the NVDEC and VIC (Video Image Composer) emulated devices. These allocate the necessary GPU buffers for the frame data, along with providing information on the incoming video data. A Submit command then signals the GPU to process and decode the frame data.

To decode the frame, the respective codec's header must be manually composed from the information provided by NVDEC, then sent with the raw frame data to the ffmpeg library.

Currently, H264 and VP9 are supported, with VP9 having some minor artifacting issues related mainly to the reference frame composition in its uncompressed header.

Async GPU is not properly implemented at the moment.

Co-Authored-By: David <25727384+ogniK5377@users.noreply.github.com>
This commit is contained in:
ameerj 2020-10-26 23:07:36 -04:00
commit eb67a45ca8
53 changed files with 4033 additions and 310 deletions

View file

@ -263,6 +263,7 @@ if (CONAN_REQUIRED_LIBS)
libzip:with_openssl=False
libzip:enable_windows_crypto=False
)
conan_check(VERSION 1.24.0 REQUIRED)
# Add the bincrafters remote
conan_add_remote(NAME bincrafters
@ -354,6 +355,19 @@ if (NOT LIBUSB_FOUND)
set(LIBUSB_LIBRARIES usb)
endif()
# Use system installed ffmpeg.
if (NOT MSVC)
find_package(FFmpeg REQUIRED)
else()
set(FFMPEG_EXT_NAME "ffmpeg-4.2.1")
set(FFMPEG_PATH "${CMAKE_BINARY_DIR}/externals/${FFMPEG_EXT_NAME}")
download_bundled_external("ffmpeg/" ${FFMPEG_EXT_NAME} "")
set(FFMPEG_FOUND YES)
set(FFMPEG_INCLUDE_DIR "${FFMPEG_PATH}/include" CACHE PATH "Path to FFmpeg headers" FORCE)
set(FFMPEG_LIBRARY_DIR "${FFMPEG_PATH}/bin" CACHE PATH "Path to FFmpeg library" FORCE)
set(FFMPEG_DLL_DIR "${FFMPEG_PATH}/bin" CACHE PATH "Path to FFmpeg dll's" FORCE)
endif()
# Prefer the -pthread flag on Linux.
set(THREADS_PREFER_PTHREAD_FLAG ON)
find_package(Threads REQUIRED)

View file

@ -0,0 +1,10 @@
function(copy_yuzu_FFmpeg_deps target_dir)
include(WindowsCopyFiles)
set(DLL_DEST "${CMAKE_BINARY_DIR}/bin/$<CONFIG>/")
windows_copy_files(${target_dir} ${FFMPEG_DLL_DIR} ${DLL_DEST}
avcodec-58.dll
avutil-56.dll
swresample-3.dll
swscale-5.dll
)
endfunction(copy_yuzu_FFmpeg_deps)

100
externals/find-modules/FindFFmpeg.cmake vendored Normal file
View file

@ -0,0 +1,100 @@
# - Try to find ffmpeg libraries (libavcodec, libavformat and libavutil)
# Once done this will define
#
# FFMPEG_FOUND - system has ffmpeg or libav
# FFMPEG_INCLUDE_DIR - the ffmpeg include directory
# FFMPEG_LIBRARIES - Link these to use ffmpeg
# FFMPEG_LIBAVCODEC
# FFMPEG_LIBAVFORMAT
# FFMPEG_LIBAVUTIL
#
# Copyright (c) 2008 Andreas Schneider <mail@cynapses.org>
# Modified for other libraries by Lasse Kärkkäinen <tronic>
# Modified for Hedgewars by Stepik777
# Modified for FFmpeg-example Tuukka Pasanen 2018
# Modified for yuzu toastUnlimted 2020
#
# Redistribution and use is allowed according to the terms of the New
# BSD license.
#
include(FindPackageHandleStandardArgs)
find_package_handle_standard_args(FFMPEG
FOUND_VAR FFMPEG_FOUND
REQUIRED_VARS
FFMPEG_LIBRARY
FFMPEG_INCLUDE_DIR
VERSION_VAR FFMPEG_VERSION
)
if(FFMPEG_LIBRARIES AND FFMPEG_INCLUDE_DIR)
# in cache already
set(FFMPEG_FOUND TRUE)
else()
# use pkg-config to get the directories and then use these values
# in the FIND_PATH() and FIND_LIBRARY() calls
find_package(PkgConfig)
if(PKG_CONFIG_FOUND)
pkg_check_modules(_FFMPEG_AVCODEC libavcodec)
pkg_check_modules(_FFMPEG_AVUTIL libavutil)
pkg_check_modules(_FFMPEG_SWSCALE libswscale)
endif()
find_path(FFMPEG_AVCODEC_INCLUDE_DIR
NAMES libavcodec/avcodec.h
PATHS ${_FFMPEG_AVCODEC_INCLUDE_DIRS}
/usr/include
/usr/local/include
/opt/local/include
/sw/include
PATH_SUFFIXES ffmpeg libav)
find_library(FFMPEG_LIBAVCODEC
NAMES avcodec
PATHS ${_FFMPEG_AVCODEC_LIBRARY_DIRS}
/usr/lib
/usr/local/lib
/opt/local/lib
/sw/lib)
find_library(FFMPEG_LIBAVUTIL
NAMES avutil
PATHS ${_FFMPEG_AVUTIL_LIBRARY_DIRS}
/usr/lib
/usr/local/lib
/opt/local/lib
/sw/lib)
find_library(FFMPEG_LIBSWSCALE
NAMES swscale
PATHS ${_FFMPEG_SWSCALE_LIBRARY_DIRS}
/usr/lib
/usr/local/lib
/opt/local/lib
/sw/lib)
if(FFMPEG_LIBAVCODEC AND FFMPEG_LIBAVUTIL AND FFMPEG_LIBSWSCALE)
set(FFMPEG_FOUND TRUE)
endif()
if(FFMPEG_FOUND)
set(FFMPEG_INCLUDE_DIR ${FFMPEG_AVCODEC_INCLUDE_DIR})
set(FFMPEG_LIBRARIES
${FFMPEG_LIBAVCODEC}
${FFMPEG_LIBAVUTIL}
${FFMPEG_LIBSWSCALE})
endif()
if(FFMPEG_FOUND)
if(NOT FFMPEG_FIND_QUIETLY)
message(STATUS
"Found FFMPEG or Libav: ${FFMPEG_LIBRARIES}, ${FFMPEG_INCLUDE_DIR}")
endif()
else()
if(FFMPEG_FIND_REQUIRED)
message(FATAL_ERROR
"Could not find libavcodec or libavutil or libswscale")
endif()
endif()
endif()

View file

@ -150,6 +150,8 @@ add_library(common STATIC
scope_exit.h
spin_lock.cpp
spin_lock.h
stream.cpp
stream.h
string_util.cpp
string_util.h
swap.h

47
src/common/stream.cpp Normal file
View file

@ -0,0 +1,47 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <stdexcept>
#include "common/common_types.h"
#include "common/stream.h"
namespace Common {
Stream::Stream() = default;
Stream::~Stream() = default;
void Stream::Seek(s32 offset, SeekOrigin origin) {
if (origin == SeekOrigin::SetOrigin) {
if (offset < 0) {
position = 0;
} else if (position >= buffer.size()) {
position = buffer.size();
} else {
position = offset;
}
} else if (origin == SeekOrigin::FromCurrentPos) {
Seek(static_cast<s32>(position) + offset, SeekOrigin::SetOrigin);
} else if (origin == SeekOrigin::FromEnd) {
Seek(static_cast<s32>(buffer.size()) - offset, SeekOrigin::SetOrigin);
}
}
u8 Stream::ReadByte() {
if (position < buffer.size()) {
return buffer[position++];
} else {
throw std::out_of_range("Attempting to read a byte not within the buffer range");
}
}
void Stream::WriteByte(u8 byte) {
if (position == buffer.size()) {
buffer.push_back(byte);
position++;
} else {
buffer.insert(buffer.begin() + position, byte);
}
}
} // namespace Common

50
src/common/stream.h Normal file
View file

@ -0,0 +1,50 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <vector>
#include "common/common_types.h"
namespace Common {
enum class SeekOrigin {
SetOrigin,
FromCurrentPos,
FromEnd,
};
class Stream {
public:
/// Stream creates a bitstream and provides common functionality on the stream.
explicit Stream();
~Stream();
/// Reposition bitstream "cursor" to the specified offset from origin
void Seek(s32 offset, SeekOrigin origin);
/// Reads next byte in the stream buffer and increments position
u8 ReadByte();
/// Writes byte at current position
void WriteByte(u8 byte);
std::size_t GetPosition() const {
return position;
}
std::vector<u8>& GetBuffer() {
return buffer;
}
const std::vector<u8>& GetBuffer() const {
return buffer;
}
private:
std::vector<u8> buffer;
std::size_t position{0};
};
} // namespace Common

View file

@ -439,6 +439,8 @@ add_library(core STATIC
hle/service/nvdrv/devices/nvhost_gpu.h
hle/service/nvdrv/devices/nvhost_nvdec.cpp
hle/service/nvdrv/devices/nvhost_nvdec.h
hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
hle/service/nvdrv/devices/nvhost_nvdec_common.h
hle/service/nvdrv/devices/nvhost_nvjpg.cpp
hle/service/nvdrv/devices/nvhost_nvjpg.h
hle/service/nvdrv/devices/nvhost_vic.cpp

View file

@ -2,15 +2,17 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <cstring>
#include "common/assert.h"
#include "common/logging/log.h"
#include "core/core.h"
#include "core/hle/service/nvdrv/devices/nvhost_nvdec.h"
#include "video_core/memory_manager.h"
#include "video_core/renderer_base.h"
namespace Service::Nvidia::Devices {
nvhost_nvdec::nvhost_nvdec(Core::System& system) : nvdevice(system) {}
nvhost_nvdec::nvhost_nvdec(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
: nvhost_nvdec_common(system, std::move(nvmap_dev)) {}
nvhost_nvdec::~nvhost_nvdec() = default;
u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@ -21,7 +23,7 @@ u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::
switch (static_cast<IoctlCommand>(command.raw)) {
case IoctlCommand::IocSetNVMAPfdCommand:
return SetNVMAPfd(input, output);
return SetNVMAPfd(input);
case IoctlCommand::IocSubmit:
return Submit(input, output);
case IoctlCommand::IocGetSyncpoint:
@ -29,79 +31,29 @@ u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, const std::
case IoctlCommand::IocGetWaitbase:
return GetWaitbase(input, output);
case IoctlCommand::IocMapBuffer:
return MapBuffer(input, output);
case IoctlCommand::IocMapBuffer2:
case IoctlCommand::IocMapBuffer3:
case IoctlCommand::IocMapBufferEx:
return MapBufferEx(input, output);
case IoctlCommand::IocUnmapBufferEx:
return UnmapBufferEx(input, output);
return MapBuffer(input, output);
case IoctlCommand::IocUnmapBufferEx: {
// This command is sent when the video stream has ended, flush all video contexts
// This is usually sent in the folowing order: vic, nvdec, vic.
// Inform the GPU to clear any remaining nvdec buffers when this is detected.
LOG_INFO(Service_NVDRV, "NVDEC video stream ended");
Tegra::ChCommandHeaderList cmdlist(1);
cmdlist[0] = Tegra::ChCommandHeader{0xDEADB33F};
system.GPU().PushCommandBuffer(cmdlist);
[[fallthrough]]; // fallthrough to unmap buffers
};
case IoctlCommand::IocUnmapBuffer:
case IoctlCommand::IocUnmapBuffer2:
case IoctlCommand::IocUnmapBuffer3:
return UnmapBuffer(input, output);
case IoctlCommand::IocSetSubmitTimeout:
return SetSubmitTimeout(input, output);
}
UNIMPLEMENTED_MSG("Unimplemented ioctl");
return 0;
}
u32 nvhost_nvdec::SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlSetNvmapFD params{};
std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
nvmap_fd = params.nvmap_fd;
return 0;
}
u32 nvhost_nvdec::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlSubmit params{};
std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
LOG_WARNING(Service_NVDRV, "(STUBBED) called");
std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
return 0;
}
u32 nvhost_nvdec::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlGetSyncpoint params{};
std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
params.value = 0; // Seems to be hard coded at 0
std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
return 0;
}
u32 nvhost_nvdec::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlGetWaitbase params{};
std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
params.value = 0; // Seems to be hard coded at 0
std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
return 0;
}
u32 nvhost_nvdec::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlMapBuffer params{};
std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
params.address_1);
params.address_1 = 0;
params.address_2 = 0;
std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
return 0;
}
u32 nvhost_nvdec::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlMapBufferEx params{};
std::memcpy(&params, input.data(), sizeof(IoctlMapBufferEx));
LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
params.address_1);
params.address_1 = 0;
params.address_2 = 0;
std::memcpy(output.data(), &params, sizeof(IoctlMapBufferEx));
return 0;
}
u32 nvhost_nvdec::UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlUnmapBufferEx params{};
std::memcpy(&params, input.data(), sizeof(IoctlUnmapBufferEx));
LOG_WARNING(Service_NVDRV, "(STUBBED) called");
std::memcpy(output.data(), &params, sizeof(IoctlUnmapBufferEx));
UNIMPLEMENTED_MSG("Unimplemented ioctl 0x{:X}", command.raw);
return 0;
}

View file

@ -4,16 +4,14 @@
#pragma once
#include <vector>
#include "common/common_types.h"
#include "common/swap.h"
#include "core/hle/service/nvdrv/devices/nvdevice.h"
#include <memory>
#include "core/hle/service/nvdrv/devices/nvhost_nvdec_common.h"
namespace Service::Nvidia::Devices {
class nvhost_nvdec final : public nvdevice {
class nvhost_nvdec final : public nvhost_nvdec_common {
public:
explicit nvhost_nvdec(Core::System& system);
explicit nvhost_nvdec(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
~nvhost_nvdec() override;
u32 ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@ -27,62 +25,15 @@ private:
IocGetSyncpoint = 0xC0080002,
IocGetWaitbase = 0xC0080003,
IocMapBuffer = 0xC01C0009,
IocMapBuffer2 = 0xC16C0009,
IocMapBuffer3 = 0xC15C0009,
IocMapBufferEx = 0xC0A40009,
IocUnmapBufferEx = 0xC0A4000A,
IocUnmapBuffer = 0xC0A4000A,
IocUnmapBuffer2 = 0xC16C000A,
IocUnmapBufferEx = 0xC01C000A,
IocUnmapBuffer3 = 0xC15C000A,
IocSetSubmitTimeout = 0x40040007,
};
struct IoctlSetNvmapFD {
u32_le nvmap_fd;
};
static_assert(sizeof(IoctlSetNvmapFD) == 0x4, "IoctlSetNvmapFD is incorrect size");
struct IoctlSubmit {
INSERT_PADDING_BYTES(0x40); // TODO(DarkLordZach): RE this structure
};
static_assert(sizeof(IoctlSubmit) == 0x40, "IoctlSubmit has incorrect size");
struct IoctlGetSyncpoint {
u32 unknown; // seems to be ignored? Nintendo added this
u32 value;
};
static_assert(sizeof(IoctlGetSyncpoint) == 0x08, "IoctlGetSyncpoint has incorrect size");
struct IoctlGetWaitbase {
u32 unknown; // seems to be ignored? Nintendo added this
u32 value;
};
static_assert(sizeof(IoctlGetWaitbase) == 0x08, "IoctlGetWaitbase has incorrect size");
struct IoctlMapBuffer {
u32 unknown;
u32 address_1;
u32 address_2;
INSERT_PADDING_BYTES(0x10); // TODO(DarkLordZach): RE this structure
};
static_assert(sizeof(IoctlMapBuffer) == 0x1C, "IoctlMapBuffer is incorrect size");
struct IoctlMapBufferEx {
u32 unknown;
u32 address_1;
u32 address_2;
INSERT_PADDING_BYTES(0x98); // TODO(DarkLordZach): RE this structure
};
static_assert(sizeof(IoctlMapBufferEx) == 0xA4, "IoctlMapBufferEx has incorrect size");
struct IoctlUnmapBufferEx {
INSERT_PADDING_BYTES(0xA4); // TODO(DarkLordZach): RE this structure
};
static_assert(sizeof(IoctlUnmapBufferEx) == 0xA4, "IoctlUnmapBufferEx has incorrect size");
u32_le nvmap_fd{};
u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
u32 MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
u32 UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
};
} // namespace Service::Nvidia::Devices

View file

@ -0,0 +1,234 @@
// Copyright 2020 yuzu emulator team
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <algorithm>
#include <cstring>
#include "common/assert.h"
#include "common/common_types.h"
#include "common/logging/log.h"
#include "core/core.h"
#include "core/hle/service/nvdrv/devices/nvhost_nvdec_common.h"
#include "core/hle/service/nvdrv/devices/nvmap.h"
#include "core/memory.h"
#include "video_core/memory_manager.h"
#include "video_core/renderer_base.h"
namespace Service::Nvidia::Devices {
namespace {
// Splice vectors will copy count amount of type T from the input vector into the dst vector.
template <typename T>
std::size_t SpliceVectors(const std::vector<u8>& input, std::vector<T>& dst, std::size_t count,
std::size_t offset) {
std::memcpy(dst.data(), input.data() + offset, count * sizeof(T));
offset += count * sizeof(T);
return offset;
}
// Write vectors will write data to the output buffer
template <typename T>
std::size_t WriteVectors(std::vector<u8>& dst, const std::vector<T>& src, std::size_t offset) {
std::memcpy(dst.data() + offset, src.data(), src.size() * sizeof(T));
offset += src.size() * sizeof(T);
return offset;
}
} // Anonymous namespace
namespace NvErrCodes {
constexpr u32 Success{};
constexpr u32 OutOfMemory{static_cast<u32>(-12)};
constexpr u32 InvalidInput{static_cast<u32>(-22)};
} // namespace NvErrCodes
nvhost_nvdec_common::nvhost_nvdec_common(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
: nvdevice(system), nvmap_dev(std::move(nvmap_dev)) {}
nvhost_nvdec_common::~nvhost_nvdec_common() = default;
u32 nvhost_nvdec_common::SetNVMAPfd(const std::vector<u8>& input) {
IoctlSetNvmapFD params{};
std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
nvmap_fd = params.nvmap_fd;
return 0;
}
u32 nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlSubmit params{};
std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
LOG_DEBUG(Service_NVDRV, "called NVDEC Submit, cmd_buffer_count={}", params.cmd_buffer_count);
// Instantiate param buffers
std::size_t offset = sizeof(IoctlSubmit);
std::vector<CommandBuffer> command_buffers(params.cmd_buffer_count);
std::vector<Reloc> relocs(params.relocation_count);
std::vector<u32> reloc_shifts(params.relocation_count);
std::vector<SyncptIncr> syncpt_increments(params.syncpoint_count);
std::vector<SyncptIncr> wait_checks(params.syncpoint_count);
std::vector<Fence> fences(params.fence_count);
// Splice input into their respective buffers
offset = SpliceVectors(input, command_buffers, params.cmd_buffer_count, offset);
offset = SpliceVectors(input, relocs, params.relocation_count, offset);
offset = SpliceVectors(input, reloc_shifts, params.relocation_count, offset);
offset = SpliceVectors(input, syncpt_increments, params.syncpoint_count, offset);
offset = SpliceVectors(input, wait_checks, params.syncpoint_count, offset);
offset = SpliceVectors(input, fences, params.fence_count, offset);
// TODO(ameerj): For async gpu, utilize fences for syncpoint 'max' increment
auto& gpu = system.GPU();
for (const auto& cmd_buffer : command_buffers) {
auto object = nvmap_dev->GetObject(cmd_buffer.memory_id);
ASSERT_OR_EXECUTE(object, return NvErrCodes::InvalidInput;);
const auto map = FindBufferMap(object->dma_map_addr);
if (!map) {
LOG_ERROR(Service_NVDRV, "Tried to submit an invalid offset 0x{:X} dma 0x{:X}",
object->addr, object->dma_map_addr);
return 0;
}
Tegra::ChCommandHeaderList cmdlist(cmd_buffer.word_count);
gpu.MemoryManager().ReadBlock(map->StartAddr() + cmd_buffer.offset, cmdlist.data(),
cmdlist.size() * sizeof(u32));
gpu.PushCommandBuffer(cmdlist);
}
std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
// Some games expect command_buffers to be written back
offset = sizeof(IoctlSubmit);
offset = WriteVectors(output, command_buffers, offset);
offset = WriteVectors(output, relocs, offset);
offset = WriteVectors(output, reloc_shifts, offset);
offset = WriteVectors(output, syncpt_increments, offset);
offset = WriteVectors(output, wait_checks, offset);
return NvErrCodes::Success;
}
u32 nvhost_nvdec_common::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlGetSyncpoint params{};
std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
LOG_DEBUG(Service_NVDRV, "called GetSyncpoint, id={}", params.param);
// We found that implementing this causes deadlocks with async gpu, along with degraded
// performance. TODO: RE the nvdec async implementation
params.value = 0;
std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
return NvErrCodes::Success;
}
u32 nvhost_nvdec_common::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlGetWaitbase params{};
std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
params.value = 0; // Seems to be hard coded at 0
std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
return 0;
}
u32 nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlMapBuffer params{};
std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
SpliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
auto& gpu = system.GPU();
for (auto& cmf_buff : cmd_buffer_handles) {
auto object{nvmap_dev->GetObject(cmf_buff.map_handle)};
if (!object) {
LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmf_buff.map_handle);
std::memcpy(output.data(), &params, output.size());
return NvErrCodes::InvalidInput;
}
if (object->dma_map_addr == 0) {
// NVDEC and VIC memory is in the 32-bit address space
// MapAllocate32 will attempt to map a lower 32-bit value in the shared gpu memory space
const GPUVAddr low_addr = gpu.MemoryManager().MapAllocate32(object->addr, object->size);
object->dma_map_addr = static_cast<u32>(low_addr);
// Ensure that the dma_map_addr is indeed in the lower 32-bit address space.
ASSERT(object->dma_map_addr == low_addr);
}
if (!object->dma_map_addr) {
LOG_ERROR(Service_NVDRV, "failed to map size={}", object->size);
} else {
cmf_buff.map_address = object->dma_map_addr;
AddBufferMap(object->dma_map_addr, object->size, object->addr,
object->status == nvmap::Object::Status::Allocated);
}
}
std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
std::memcpy(output.data() + sizeof(IoctlMapBuffer), cmd_buffer_handles.data(),
cmd_buffer_handles.size() * sizeof(MapBufferEntry));
return NvErrCodes::Success;
}
u32 nvhost_nvdec_common::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlMapBuffer params{};
std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
SpliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
auto& gpu = system.GPU();
for (auto& cmf_buff : cmd_buffer_handles) {
const auto object{nvmap_dev->GetObject(cmf_buff.map_handle)};
if (!object) {
LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmf_buff.map_handle);
std::memcpy(output.data(), &params, output.size());
return NvErrCodes::InvalidInput;
}
if (const auto size{RemoveBufferMap(object->dma_map_addr)}; size) {
gpu.MemoryManager().Unmap(object->dma_map_addr, *size);
} else {
// This occurs quite frequently, however does not seem to impact functionality
LOG_DEBUG(Service_NVDRV, "invalid offset=0x{:X} dma=0x{:X}", object->addr,
object->dma_map_addr);
}
object->dma_map_addr = 0;
}
std::memset(output.data(), 0, output.size());
return NvErrCodes::Success;
}
u32 nvhost_nvdec_common::SetSubmitTimeout(const std::vector<u8>& input, std::vector<u8>& output) {
std::memcpy(&submit_timeout, input.data(), input.size());
LOG_WARNING(Service_NVDRV, "(STUBBED) called");
return NvErrCodes::Success;
}
std::optional<nvhost_nvdec_common::BufferMap> nvhost_nvdec_common::FindBufferMap(
GPUVAddr gpu_addr) const {
const auto it = std::find_if(
buffer_mappings.begin(), buffer_mappings.upper_bound(gpu_addr), [&](const auto& entry) {
return (gpu_addr >= entry.second.StartAddr() && gpu_addr < entry.second.EndAddr());
});
ASSERT(it != buffer_mappings.end());
return it->second;
}
void nvhost_nvdec_common::AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr,
bool is_allocated) {
buffer_mappings.insert_or_assign(gpu_addr, BufferMap{gpu_addr, size, cpu_addr, is_allocated});
}
std::optional<std::size_t> nvhost_nvdec_common::RemoveBufferMap(GPUVAddr gpu_addr) {
const auto iter{buffer_mappings.find(gpu_addr)};
if (iter == buffer_mappings.end()) {
return std::nullopt;
}
std::size_t size = 0;
if (iter->second.IsAllocated()) {
size = iter->second.Size();
}
buffer_mappings.erase(iter);
return size;
}
} // namespace Service::Nvidia::Devices

View file

@ -0,0 +1,168 @@
// Copyright 2020 yuzu emulator team
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <map>
#include <vector>
#include "common/common_types.h"
#include "common/swap.h"
#include "core/hle/service/nvdrv/devices/nvdevice.h"
namespace Service::Nvidia::Devices {
class nvmap;
class nvhost_nvdec_common : public nvdevice {
public:
explicit nvhost_nvdec_common(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
~nvhost_nvdec_common() override;
virtual u32 ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
std::vector<u8>& output, std::vector<u8>& output2, IoctlCtrl& ctrl,
IoctlVersion version) = 0;
protected:
class BufferMap final {
public:
constexpr BufferMap() = default;
constexpr BufferMap(GPUVAddr start_addr, std::size_t size)
: start_addr{start_addr}, end_addr{start_addr + size} {}
constexpr BufferMap(GPUVAddr start_addr, std::size_t size, VAddr cpu_addr,
bool is_allocated)
: start_addr{start_addr}, end_addr{start_addr + size}, cpu_addr{cpu_addr},
is_allocated{is_allocated} {}
constexpr VAddr StartAddr() const {
return start_addr;
}
constexpr VAddr EndAddr() const {
return end_addr;
}
constexpr std::size_t Size() const {
return end_addr - start_addr;
}
constexpr VAddr CpuAddr() const {
return cpu_addr;
}
constexpr bool IsAllocated() const {
return is_allocated;
}
private:
GPUVAddr start_addr{};
GPUVAddr end_addr{};
VAddr cpu_addr{};
bool is_allocated{};
};
struct IoctlSetNvmapFD {
u32_le nvmap_fd;
};
static_assert(sizeof(IoctlSetNvmapFD) == 4, "IoctlSetNvmapFD is incorrect size");
struct IoctlSubmitCommandBuffer {
u32_le id;
u32_le offset;
u32_le count;
};
static_assert(sizeof(IoctlSubmitCommandBuffer) == 0xC,
"IoctlSubmitCommandBuffer is incorrect size");
struct IoctlSubmit {
u32_le cmd_buffer_count;
u32_le relocation_count;
u32_le syncpoint_count;
u32_le fence_count;
};
static_assert(sizeof(IoctlSubmit) == 0x10, "IoctlSubmit has incorrect size");
struct CommandBuffer {
s32 memory_id;
u32 offset;
s32 word_count;
};
static_assert(sizeof(CommandBuffer) == 0xC, "CommandBuffer has incorrect size");
struct Reloc {
s32 cmdbuffer_memory;
s32 cmdbuffer_offset;
s32 target;
s32 target_offset;
};
static_assert(sizeof(Reloc) == 0x10, "CommandBuffer has incorrect size");
struct SyncptIncr {
u32 id;
u32 increments;
};
static_assert(sizeof(SyncptIncr) == 0x8, "CommandBuffer has incorrect size");
struct Fence {
u32 id;
u32 value;
};
static_assert(sizeof(Fence) == 0x8, "CommandBuffer has incorrect size");
struct IoctlGetSyncpoint {
// Input
u32_le param;
// Output
u32_le value;
};
static_assert(sizeof(IoctlGetSyncpoint) == 8, "IocGetIdParams has wrong size");
struct IoctlGetWaitbase {
u32_le unknown; // seems to be ignored? Nintendo added this
u32_le value;
};
static_assert(sizeof(IoctlGetWaitbase) == 0x8, "IoctlGetWaitbase is incorrect size");
struct IoctlMapBuffer {
u32_le num_entries;
u32_le data_address; // Ignored by the driver.
u32_le attach_host_ch_das;
};
static_assert(sizeof(IoctlMapBuffer) == 0x0C, "IoctlMapBuffer is incorrect size");
struct IocGetIdParams {
// Input
u32_le param;
// Output
u32_le value;
};
static_assert(sizeof(IocGetIdParams) == 8, "IocGetIdParams has wrong size");
// Used for mapping and unmapping command buffers
struct MapBufferEntry {
u32_le map_handle;
u32_le map_address;
};
static_assert(sizeof(IoctlMapBuffer) == 0x0C, "IoctlMapBuffer is incorrect size");
/// Ioctl command implementations
u32 SetNVMAPfd(const std::vector<u8>& input);
u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
u32 UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
u32 SetSubmitTimeout(const std::vector<u8>& input, std::vector<u8>& output);
std::optional<BufferMap> FindBufferMap(GPUVAddr gpu_addr) const;
void AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr, bool is_allocated);
std::optional<std::size_t> RemoveBufferMap(GPUVAddr gpu_addr);
u32_le nvmap_fd{};
u32_le submit_timeout{};
std::shared_ptr<nvmap> nvmap_dev;
// This is expected to be ordered, therefore we must use a map, not unordered_map
std::map<GPUVAddr, BufferMap> buffer_mappings;
};
}; // namespace Service::Nvidia::Devices

View file

@ -2,15 +2,17 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <cstring>
#include "common/assert.h"
#include "common/logging/log.h"
#include "core/core.h"
#include "core/hle/service/nvdrv/devices/nvhost_vic.h"
#include "video_core/memory_manager.h"
#include "video_core/renderer_base.h"
namespace Service::Nvidia::Devices {
nvhost_vic::nvhost_vic(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
: nvhost_nvdec_common(system, std::move(nvmap_dev)) {}
nvhost_vic::nvhost_vic(Core::System& system) : nvdevice(system) {}
nvhost_vic::~nvhost_vic() = default;
u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
@ -21,7 +23,7 @@ u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::ve
switch (static_cast<IoctlCommand>(command.raw)) {
case IoctlCommand::IocSetNVMAPfdCommand:
return SetNVMAPfd(input, output);
return SetNVMAPfd(input);
case IoctlCommand::IocSubmit:
return Submit(input, output);
case IoctlCommand::IocGetSyncpoint:
@ -29,83 +31,19 @@ u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, const std::ve
case IoctlCommand::IocGetWaitbase:
return GetWaitbase(input, output);
case IoctlCommand::IocMapBuffer:
return MapBuffer(input, output);
case IoctlCommand::IocMapBuffer2:
case IoctlCommand::IocMapBuffer3:
case IoctlCommand::IocMapBuffer4:
case IoctlCommand::IocMapBufferEx:
return MapBuffer(input, output);
case IoctlCommand::IocUnmapBuffer:
case IoctlCommand::IocUnmapBuffer2:
case IoctlCommand::IocUnmapBuffer3:
case IoctlCommand::IocUnmapBufferEx:
return UnmapBufferEx(input, output);
return UnmapBuffer(input, output);
}
UNIMPLEMENTED_MSG("Unimplemented ioctl");
return 0;
}
u32 nvhost_vic::SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlSetNvmapFD params{};
std::memcpy(&params, input.data(), sizeof(IoctlSetNvmapFD));
LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);
nvmap_fd = params.nvmap_fd;
return 0;
}
u32 nvhost_vic::Submit(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlSubmit params{};
std::memcpy(&params, input.data(), sizeof(IoctlSubmit));
LOG_WARNING(Service_NVDRV, "(STUBBED) called");
// Workaround for Luigi's Mansion 3, as nvhost_vic is not implemented for asynch GPU
params.command_buffer = {};
std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
return 0;
}
u32 nvhost_vic::GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlGetSyncpoint params{};
std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
params.value = 0; // Seems to be hard coded at 0
std::memcpy(output.data(), &params, sizeof(IoctlGetSyncpoint));
return 0;
}
u32 nvhost_vic::GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlGetWaitbase params{};
std::memcpy(&params, input.data(), sizeof(IoctlGetWaitbase));
LOG_INFO(Service_NVDRV, "called, unknown=0x{:X}", params.unknown);
params.value = 0; // Seems to be hard coded at 0
std::memcpy(output.data(), &params, sizeof(IoctlGetWaitbase));
return 0;
}
u32 nvhost_vic::MapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlMapBuffer params{};
std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
params.address_1);
params.address_1 = 0;
params.address_2 = 0;
std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
return 0;
}
u32 nvhost_vic::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlMapBufferEx params{};
std::memcpy(&params, input.data(), sizeof(IoctlMapBufferEx));
LOG_WARNING(Service_NVDRV, "(STUBBED) called with address={:08X}{:08X}", params.address_2,
params.address_1);
params.address_1 = 0;
params.address_2 = 0;
std::memcpy(output.data(), &params, sizeof(IoctlMapBufferEx));
return 0;
}
u32 nvhost_vic::UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) {
IoctlUnmapBufferEx params{};
std::memcpy(&params, input.data(), sizeof(IoctlUnmapBufferEx));
LOG_WARNING(Service_NVDRV, "(STUBBED) called");
std::memcpy(output.data(), &params, sizeof(IoctlUnmapBufferEx));
UNIMPLEMENTED_MSG("Unimplemented ioctl 0x{:X}", command.raw);
return 0;
}

View file

@ -4,19 +4,15 @@
#pragma once
#include <array>
#include <vector>
#include "common/common_types.h"
#include "common/swap.h"
#include "core/hle/service/nvdrv/devices/nvdevice.h"
#include "core/hle/service/nvdrv/devices/nvhost_nvdec_common.h"
namespace Service::Nvidia::Devices {
class nvmap;
class nvhost_vic final : public nvdevice {
class nvhost_vic final : public nvhost_nvdec_common {
public:
explicit nvhost_vic(Core::System& system);
~nvhost_vic() override;
explicit nvhost_vic(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
~nvhost_vic();
u32 ioctl(Ioctl command, const std::vector<u8>& input, const std::vector<u8>& input2,
std::vector<u8>& output, std::vector<u8>& output2, IoctlCtrl& ctrl,
IoctlVersion version) override;
@ -28,74 +24,14 @@ private:
IocGetSyncpoint = 0xC0080002,
IocGetWaitbase = 0xC0080003,
IocMapBuffer = 0xC01C0009,
IocMapBuffer2 = 0xC0340009,
IocMapBuffer3 = 0xC0140009,
IocMapBuffer4 = 0xC00C0009,
IocMapBufferEx = 0xC03C0009,
IocUnmapBufferEx = 0xC03C000A,
IocUnmapBuffer = 0xC03C000A,
IocUnmapBuffer2 = 0xC034000A,
IocUnmapBuffer3 = 0xC00C000A,
IocUnmapBufferEx = 0xC01C000A,
};
struct IoctlSetNvmapFD {
u32_le nvmap_fd;
};
static_assert(sizeof(IoctlSetNvmapFD) == 4, "IoctlSetNvmapFD is incorrect size");
struct IoctlSubmitCommandBuffer {
u32 id;
u32 offset;
u32 count;
};
static_assert(sizeof(IoctlSubmitCommandBuffer) == 0xC,
"IoctlSubmitCommandBuffer is incorrect size");
struct IoctlSubmit {
u32 command_buffer_count;
u32 relocations_count;
u32 syncpt_count;
u32 wait_count;
std::array<IoctlSubmitCommandBuffer, 4> command_buffer;
};
static_assert(sizeof(IoctlSubmit) == 0x40, "IoctlSubmit is incorrect size");
struct IoctlGetSyncpoint {
u32 unknown; // seems to be ignored? Nintendo added this
u32 value;
};
static_assert(sizeof(IoctlGetSyncpoint) == 0x8, "IoctlGetSyncpoint is incorrect size");
struct IoctlGetWaitbase {
u32 unknown; // seems to be ignored? Nintendo added this
u32 value;
};
static_assert(sizeof(IoctlGetWaitbase) == 0x8, "IoctlGetWaitbase is incorrect size");
struct IoctlMapBuffer {
u32 unknown;
u32 address_1;
u32 address_2;
INSERT_PADDING_BYTES(0x10); // TODO(DarkLordZach): RE this structure
};
static_assert(sizeof(IoctlMapBuffer) == 0x1C, "IoctlMapBuffer is incorrect size");
struct IoctlMapBufferEx {
u32 unknown;
u32 address_1;
u32 address_2;
INSERT_PADDING_BYTES(0x30); // TODO(DarkLordZach): RE this structure
};
static_assert(sizeof(IoctlMapBufferEx) == 0x3C, "IoctlMapBufferEx is incorrect size");
struct IoctlUnmapBufferEx {
INSERT_PADDING_BYTES(0x3C); // TODO(DarkLordZach): RE this structure
};
static_assert(sizeof(IoctlUnmapBufferEx) == 0x3C, "IoctlUnmapBufferEx is incorrect size");
u32_le nvmap_fd{};
u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
u32 Submit(const std::vector<u8>& input, std::vector<u8>& output);
u32 GetSyncpoint(const std::vector<u8>& input, std::vector<u8>& output);
u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
u32 MapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
u32 MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
u32 UnmapBufferEx(const std::vector<u8>& input, std::vector<u8>& output);
};
} // namespace Service::Nvidia::Devices

View file

@ -37,6 +37,7 @@ public:
VAddr addr;
Status status;
u32 refcount;
u32 dma_map_addr;
};
std::shared_ptr<Object> GetObject(u32 handle) const {

View file

@ -51,9 +51,9 @@ Module::Module(Core::System& system) {
devices["/dev/nvmap"] = nvmap_dev;
devices["/dev/nvdisp_disp0"] = std::make_shared<Devices::nvdisp_disp0>(system, nvmap_dev);
devices["/dev/nvhost-ctrl"] = std::make_shared<Devices::nvhost_ctrl>(system, events_interface);
devices["/dev/nvhost-nvdec"] = std::make_shared<Devices::nvhost_nvdec>(system);
devices["/dev/nvhost-nvdec"] = std::make_shared<Devices::nvhost_nvdec>(system, nvmap_dev);
devices["/dev/nvhost-nvjpg"] = std::make_shared<Devices::nvhost_nvjpg>(system);
devices["/dev/nvhost-vic"] = std::make_shared<Devices::nvhost_vic>(system);
devices["/dev/nvhost-vic"] = std::make_shared<Devices::nvhost_vic>(system, nvmap_dev);
}
Module::~Module() = default;

View file

@ -63,6 +63,7 @@ void LogSettings() {
log_setting("Renderer_GPUAccuracyLevel", values.gpu_accuracy.GetValue());
log_setting("Renderer_UseAsynchronousGpuEmulation",
values.use_asynchronous_gpu_emulation.GetValue());
log_setting("Renderer_UseNvdecEmulation", values.use_nvdec_emulation.GetValue());
log_setting("Renderer_UseVsync", values.use_vsync.GetValue());
log_setting("Renderer_UseAssemblyShaders", values.use_assembly_shaders.GetValue());
log_setting("Renderer_UseAsynchronousShaders", values.use_asynchronous_shaders.GetValue());
@ -119,6 +120,7 @@ void RestoreGlobalState() {
values.use_disk_shader_cache.SetGlobal(true);
values.gpu_accuracy.SetGlobal(true);
values.use_asynchronous_gpu_emulation.SetGlobal(true);
values.use_nvdec_emulation.SetGlobal(true);
values.use_vsync.SetGlobal(true);
values.use_assembly_shaders.SetGlobal(true);
values.use_asynchronous_shaders.SetGlobal(true);

View file

@ -111,6 +111,7 @@ struct Values {
Setting<bool> use_disk_shader_cache;
Setting<GPUAccuracy> gpu_accuracy;
Setting<bool> use_asynchronous_gpu_emulation;
Setting<bool> use_nvdec_emulation;
Setting<bool> use_vsync;
Setting<bool> use_assembly_shaders;
Setting<bool> use_asynchronous_shaders;

View file

@ -206,6 +206,8 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
TranslateGPUAccuracyLevel(Settings::values.gpu_accuracy.GetValue()));
AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
Settings::values.use_asynchronous_gpu_emulation.GetValue());
AddField(field_type, "Renderer_UseNvdecEmulation",
Settings::values.use_nvdec_emulation.GetValue());
AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync.GetValue());
AddField(field_type, "Renderer_UseAssemblyShaders",
Settings::values.use_assembly_shaders.GetValue());

View file

@ -5,6 +5,24 @@ add_library(video_core STATIC
buffer_cache/buffer_cache.h
buffer_cache/map_interval.cpp
buffer_cache/map_interval.h
cdma_pusher.cpp
cdma_pusher.h
command_classes/codecs/codec.cpp
command_classes/codecs/codec.h
command_classes/codecs/h264.cpp
command_classes/codecs/h264.h
command_classes/codecs/vp9.cpp
command_classes/codecs/vp9.h
command_classes/codecs/vp9_types.h
command_classes/host1x.cpp
command_classes/host1x.h
command_classes/nvdec.cpp
command_classes/nvdec.h
command_classes/nvdec_common.h
command_classes/sync_manager.cpp
command_classes/sync_manager.h
command_classes/vic.cpp
command_classes/vic.h
compatible_formats.cpp
compatible_formats.h
dirty_flags.cpp
@ -250,6 +268,14 @@ create_target_directory_groups(video_core)
target_link_libraries(video_core PUBLIC common core)
target_link_libraries(video_core PRIVATE glad xbyak)
if (MSVC)
target_include_directories(video_core PRIVATE ${FFMPEG_INCLUDE_DIR})
target_link_libraries(video_core PUBLIC ${FFMPEG_LIBRARY_DIR}/swscale.lib ${FFMPEG_LIBRARY_DIR}/avcodec.lib ${FFMPEG_LIBRARY_DIR}/avutil.lib)
else()
target_include_directories(video_core PRIVATE ${FFMPEG_INCLUDE_DIR})
target_link_libraries(video_core PRIVATE ${FFMPEG_LIBRARIES})
endif()
add_dependencies(video_core host_shaders)
target_include_directories(video_core PRIVATE ${HOST_SHADERS_INCLUDE})

View file

@ -0,0 +1,171 @@
// MIT License
//
// Copyright (c) Ryujinx Team and Contributors
//
// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and
// associated documentation files (the "Software"), to deal in the Software without restriction,
// including without limitation the rights to use, copy, modify, merge, publish, distribute,
// sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in all copies or
// substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT
// NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
// DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
//
#include "command_classes/host1x.h"
#include "command_classes/nvdec.h"
#include "command_classes/vic.h"
#include "common/bit_util.h"
#include "video_core/cdma_pusher.h"
#include "video_core/command_classes/nvdec_common.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/gpu.h"
#include "video_core/memory_manager.h"
namespace Tegra {
CDmaPusher::CDmaPusher(GPU& gpu)
: gpu(gpu), nvdec_processor(std::make_shared<Nvdec>(gpu)),
vic_processor(std::make_unique<Vic>(gpu, nvdec_processor)),
host1x_processor(std::make_unique<Host1x>(gpu)),
nvdec_sync(std::make_unique<SyncptIncrManager>(gpu)),
vic_sync(std::make_unique<SyncptIncrManager>(gpu)) {}
CDmaPusher::~CDmaPusher() = default;
void CDmaPusher::Push(ChCommandHeaderList&& entries) {
cdma_queue.push(std::move(entries));
}
void CDmaPusher::DispatchCalls() {
while (!cdma_queue.empty()) {
Step();
}
}
void CDmaPusher::Step() {
const auto entries{cdma_queue.front()};
cdma_queue.pop();
std::vector<u32> values(entries.size());
std::memcpy(values.data(), entries.data(), entries.size() * sizeof(u32));
for (const u32 value : values) {
if (mask != 0) {
const u32 lbs = Common::CountTrailingZeroes32(mask);
mask &= ~(1U << lbs);
ExecuteCommand(static_cast<u32>(offset + lbs), value);
continue;
} else if (count != 0) {
--count;
ExecuteCommand(static_cast<u32>(offset), value);
if (incrementing) {
++offset;
}
continue;
}
const auto mode = static_cast<ChSubmissionMode>((value >> 28) & 0xf);
switch (mode) {
case ChSubmissionMode::SetClass: {
mask = value & 0x3f;
offset = (value >> 16) & 0xfff;
current_class = static_cast<ChClassId>((value >> 6) & 0x3ff);
break;
}
case ChSubmissionMode::Incrementing:
case ChSubmissionMode::NonIncrementing:
count = value & 0xffff;
offset = (value >> 16) & 0xfff;
incrementing = mode == ChSubmissionMode::Incrementing;
break;
case ChSubmissionMode::Mask:
mask = value & 0xffff;
offset = (value >> 16) & 0xfff;
break;
case ChSubmissionMode::Immediate: {
const u32 data = value & 0xfff;
offset = (value >> 16) & 0xfff;
ExecuteCommand(static_cast<u32>(offset), data);
break;
}
default:
UNIMPLEMENTED_MSG("ChSubmission mode {} is not implemented!", static_cast<u32>(mode));
break;
}
}
}
void CDmaPusher::ExecuteCommand(u32 offset, u32 data) {
switch (current_class) {
case ChClassId::NvDec:
ThiStateWrite(nvdec_thi_state, offset, {data});
switch (static_cast<ThiMethod>(offset)) {
case ThiMethod::IncSyncpt: {
LOG_DEBUG(Service_NVDRV, "NVDEC Class IncSyncpt Method");
const auto syncpoint_id = static_cast<u32>(data & 0xFF);
const auto cond = static_cast<u32>((data >> 8) & 0xFF);
if (cond == 0) {
nvdec_sync->Increment(syncpoint_id);
} else {
nvdec_sync->IncrementWhenDone(static_cast<u32>(current_class), syncpoint_id);
nvdec_sync->SignalDone(syncpoint_id);
}
break;
}
case ThiMethod::SetMethod1:
LOG_DEBUG(Service_NVDRV, "NVDEC method 0x{:X}",
static_cast<u32>(nvdec_thi_state.method_0));
nvdec_processor->ProcessMethod(
static_cast<Tegra::Nvdec::Method>(nvdec_thi_state.method_0), {data});
break;
default:
break;
}
break;
case ChClassId::GraphicsVic:
ThiStateWrite(vic_thi_state, static_cast<u32>(offset), {data});
switch (static_cast<ThiMethod>(offset)) {
case ThiMethod::IncSyncpt: {
LOG_DEBUG(Service_NVDRV, "VIC Class IncSyncpt Method");
const auto syncpoint_id = static_cast<u32>(data & 0xFF);
const auto cond = static_cast<u32>((data >> 8) & 0xFF);
if (cond == 0) {
vic_sync->Increment(syncpoint_id);
} else {
vic_sync->IncrementWhenDone(static_cast<u32>(current_class), syncpoint_id);
vic_sync->SignalDone(syncpoint_id);
}
break;
}
case ThiMethod::SetMethod1:
LOG_DEBUG(Service_NVDRV, "VIC method 0x{:X}, Args=({})",
static_cast<u32>(vic_thi_state.method_0));
vic_processor->ProcessMethod(static_cast<Tegra::Vic::Method>(vic_thi_state.method_0),
{data});
break;
default:
break;
}
break;
case ChClassId::Host1x:
// This device is mainly for syncpoint synchronization
LOG_DEBUG(Service_NVDRV, "Host1X Class Method");
host1x_processor->ProcessMethod(static_cast<Tegra::Host1x::Method>(offset), {data});
break;
default:
UNIMPLEMENTED_MSG("Current class not implemented {:X}", static_cast<u32>(current_class));
break;
}
}
void CDmaPusher::ThiStateWrite(ThiRegisters& state, u32 offset, const std::vector<u32>& arguments) {
u8* const state_offset = reinterpret_cast<u8*>(&state) + sizeof(u32) * offset;
std::memcpy(state_offset, arguments.data(), sizeof(u32) * arguments.size());
}
} // namespace Tegra

Some files were not shown because too many files have changed in this diff Show more