Compare commits

..

2 Commits

Author SHA1 Message Date
lizzie 4464c3dbcb Revert "[android] Add Enable GPU Buffer Readback option to Android (#4132)"
This reverts commit d142b5dd6a.
2026-06-29 04:17:07 +00:00
lizzie ebefa6ea19 Revert? 2026-06-29 04:17:07 +00:00
64 changed files with 603 additions and 545 deletions
@@ -31,7 +31,6 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"),
ENABLE_BUFFER_HISTORY("enable_buffer_history"),
USE_OPTIMIZED_VERTEX_BUFFERS("use_optimized_vertex_buffers"),
ENABLE_GPU_BUFFER_READBACK("enable_gpu_buffer_readback"),
SYNC_MEMORY_OPERATIONS("sync_memory_operations"),
BUFFER_REORDER_DISABLE("disable_buffer_reorder"),
RENDERER_DEBUG("debug"),
@@ -806,13 +806,6 @@ abstract class SettingsItem(
descriptionId = R.string.enable_buffer_history_description
)
)
put(
SwitchSetting(
BooleanSetting.ENABLE_GPU_BUFFER_READBACK,
titleId = R.string.enable_gpu_buffer_readback,
descriptionId = R.string.enable_gpu_buffer_readback_description
)
)
put(
SwitchSetting(
BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS,
@@ -292,7 +292,6 @@ class SettingsFragmentPresenter(
add(BooleanSetting.RENDERER_FORCE_MAX_CLOCK.key)
add(BooleanSetting.RENDERER_REACTIVE_FLUSHING.key)
add(BooleanSetting.ENABLE_BUFFER_HISTORY.key)
add(BooleanSetting.ENABLE_GPU_BUFFER_READBACK.key)
add(BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS.key)
add(HeaderSetting(R.string.hacks))
@@ -503,8 +503,6 @@
<string name="renderer_reactive_flushing_description">Improves rendering accuracy in some games at the cost of performance.</string>
<string name="enable_buffer_history">Enable buffer history</string>
<string name="enable_buffer_history_description">Enables access to previous buffer states. This option may improve rendering quality and performance consistency in some games.</string>
<string name="enable_gpu_buffer_readback">Enable GPU Buffer Readback</string>
<string name="enable_gpu_buffer_readback_description">Preserves GPU-modified buffer data by reading it back before uploads. Some games require this to render certain effects properly. May cause issues if the hardware cannot handle the additional workload.</string>
<string name="use_optimized_vertex_buffers">Optimized Vertex Buffers</string>
<string name="use_optimized_vertex_buffers_description">Enables optimized vertex buffer binding for improved performance. Requires Mesa 26.0+ Turnip drivers/ QCOM drivers. Will crash on older Turnip drivers (25.3 and below).</string>
+4
View File
@@ -5,7 +5,11 @@
// SPDX-License-Identifier: GPL-2.0-or-later
#include <string>
#ifndef _WIN32
#include <cstring>
#endif
#include <utility>
#include <fmt/ranges.h>
+6 -6
View File
@@ -729,7 +729,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface(ctx, ensure_token_id);
rb.PushIpcInterface(ensure_token_id);
}
void LoadIdTokenCacheDeprecated(HLERequestContext& ctx) {
@@ -921,7 +921,7 @@ void Module::Interface::GetProfile(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IProfile>(ctx, system, user_id, *profile_manager);
rb.PushIpcInterface<IProfile>(system, user_id, *profile_manager);
}
void Module::Interface::IsUserRegistrationRequestPermitted(HLERequestContext& ctx) {
@@ -993,7 +993,7 @@ void Module::Interface::GetBaasAccountManagerForApplication(HLERequestContext& c
LOG_DEBUG(Service_ACC, "called");
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IManagerForApplication>(ctx, system, profile_manager);
rb.PushIpcInterface<IManagerForApplication>(system, profile_manager);
}
void Module::Interface::IsUserAccountSwitchLocked(HLERequestContext& ctx) {
@@ -1089,7 +1089,7 @@ void Module::Interface::GetProfileEditor(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IProfileEditor>(ctx, system, user_id, *profile_manager);
rb.PushIpcInterface<IProfileEditor>(system, user_id, *profile_manager);
}
void Module::Interface::GetBaasAccountAdministrator(HLERequestContext &ctx) {
@@ -1100,7 +1100,7 @@ void Module::Interface::GetBaasAccountAdministrator(HLERequestContext &ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IAdministrator>(ctx, system, uuid);
rb.PushIpcInterface<IAdministrator>(system, uuid);
}
void Module::Interface::ListQualifiedUsers(HLERequestContext& ctx) {
@@ -1143,7 +1143,7 @@ void Module::Interface::GetBaasAccountManagerForSystemService(HLERequestContext&
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IManagerForSystemService>(ctx, system, uuid);
rb.PushIpcInterface<IManagerForSystemService>(system, uuid);
}
void Module::Interface::StoreSaveDataThumbnailSystem(HLERequestContext& ctx) {
+1 -1
View File
@@ -35,7 +35,7 @@ void IAsyncContext::GetSystemEvent(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, completion_event->GetReadableEvent());
rb.PushCopyObjects(completion_event->GetReadableEvent());
}
void IAsyncContext::Cancel(HLERequestContext& ctx) {
+2 -2
View File
@@ -82,7 +82,7 @@ void APM::OpenSession(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ISession>(ctx, system, controller);
rb.PushIpcInterface<ISession>(system, controller);
}
void APM::GetPerformanceMode(HLERequestContext& ctx) {
@@ -125,7 +125,7 @@ void APM_Sys::GetPerformanceEvent(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ISession>(ctx, system, controller);
rb.PushIpcInterface<ISession>(system, controller);
}
void APM_Sys::SetCpuBoostMode(HLERequestContext& ctx) {
+1 -4
View File
@@ -1,6 +1,3 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
@@ -48,7 +45,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IRequest>(ctx, system);
rb.PushIpcInterface<IRequest>(system);
}
};
+7 -7
View File
@@ -184,7 +184,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(readable_event.Signal(system.Kernel()));
rb.PushCopyObjects(ctx, readable_event);
rb.PushCopyObjects(readable_event);
}
void Cancel(HLERequestContext& ctx) {
@@ -400,7 +400,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, notification_event->GetReadableEvent());
rb.PushCopyObjects(notification_event->GetReadableEvent());
}
void Clear(HLERequestContext& ctx) {
@@ -476,7 +476,7 @@ private:
void Module::Interface::CreateFriendService(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IFriendService>(ctx, system);
rb.PushIpcInterface<IFriendService>(system);
LOG_DEBUG(Service_Friend, "called");
}
@@ -488,12 +488,12 @@ void Module::Interface::CreateNotificationService(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<INotificationService>(ctx, system, uuid);
rb.PushIpcInterface<INotificationService>(system, uuid);
}
Module::Interface::Interface(std::shared_ptr<Module> module_, Core::System& system_, const char* name)
: ServiceFramework{system_, name}, module{std::move(module_)}
{}
Module::Interface::Interface(std::shared_ptr<Module> module_, Core::System& system_,
const char* name)
: ServiceFramework{system_, name}, module{std::move(module_)} {}
Module::Interface::~Interface() = default;
+1 -1
View File
@@ -279,7 +279,7 @@ void ARP_W::AcquireRegistrar(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface(ctx, registrar);
rb.PushIpcInterface(registrar);
}
void ARP_W::UnregisterApplicationInstance(HLERequestContext& ctx) {
+1 -1
View File
@@ -28,7 +28,7 @@ void BGTC_T::OpenTaskService(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ITaskService>(ctx, system);
rb.PushIpcInterface<ITaskService>(system);
}
ITaskService::ITaskService(Core::System& system_) : ServiceFramework{system_, "ITaskService"} {
+1 -4
View File
@@ -1,6 +1,3 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
@@ -59,7 +56,7 @@ ECTX_AW::~ECTX_AW() = default;
void ECTX_AW::CreateContextRegistrar(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IContextRegistrar>(ctx, system);
rb.PushIpcInterface<IContextRegistrar>(std::make_shared<IContextRegistrar>(system));
}
} // namespace Service::Glue
@@ -726,7 +726,7 @@ void IHidSystemServer::AcquireConnectionTriggerTimeoutEvent(HLERequestContext& c
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, acquire_device_registered_event->GetReadableEvent());
rb.PushCopyObjects(acquire_device_registered_event->GetReadableEvent());
}
void IHidSystemServer::AcquireDeviceRegisteredEventForControllerSupport(HLERequestContext& ctx) {
@@ -734,7 +734,7 @@ void IHidSystemServer::AcquireDeviceRegisteredEventForControllerSupport(HLEReque
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, acquire_device_registered_event->GetReadableEvent());
rb.PushCopyObjects(acquire_device_registered_event->GetReadableEvent());
}
void IHidSystemServer::GetRegisteredDevices(HLERequestContext& ctx) {
@@ -759,7 +759,7 @@ void IHidSystemServer::AcquireUniquePadConnectionEventHandle(HLERequestContext&
LOG_WARNING(Service_HID, "(STUBBED) called");
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.PushCopyObjects(ctx, unique_pad_connection_event->GetReadableEvent());
rb.PushCopyObjects(unique_pad_connection_event->GetReadableEvent());
rb.Push(ResultSuccess);
}
@@ -776,7 +776,7 @@ void IHidSystemServer::AcquireJoyDetachOnBluetoothOffEventHandle(HLERequestConte
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, joy_detach_event->GetReadableEvent());
rb.PushCopyObjects(joy_detach_event->GetReadableEvent());
}
void IHidSystemServer::IsUsbFullKeyControllerEnabled(HLERequestContext& ctx) {
+2 -2
View File
@@ -31,7 +31,7 @@ class Memory;
}
namespace IPC {
struct ResponseBuilder;
class ResponseBuilder;
}
namespace Service {
@@ -392,7 +392,7 @@ public:
}
private:
friend struct IPC::ResponseBuilder;
friend class IPC::ResponseBuilder;
void ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming);
+153 -91
View File
@@ -24,7 +24,45 @@ namespace IPC {
constexpr Result ResultSessionClosed{ErrorModule::HIPC, 301};
struct ResponseBuilder {
class RequestHelperBase {
protected:
Service::HLERequestContext* context = nullptr;
u32* cmdbuf;
u32 index = 0;
public:
explicit RequestHelperBase(u32* command_buffer) : cmdbuf(command_buffer) {}
explicit RequestHelperBase(Service::HLERequestContext& ctx)
: context(&ctx), cmdbuf(ctx.CommandBuffer()) {}
void Skip(u32 size_in_words, bool set_to_null) {
if (set_to_null) {
memset(cmdbuf + index, 0, size_in_words * sizeof(u32));
}
index += size_in_words;
}
/**
* Aligns the current position forward to a 16-byte boundary, padding with zeros.
*/
void AlignWithPadding() {
if (index & 3) {
Skip(static_cast<u32>(4 - (index & 3)), true);
}
}
u32 GetCurrentOffset() const {
return index;
}
void SetCurrentOffset(u32 offset) {
index = offset;
}
};
class ResponseBuilder : public RequestHelperBase {
public:
/// Flags used for customizing the behavior of ResponseBuilder
enum class Flags : u32 {
None = 0,
@@ -33,13 +71,14 @@ struct ResponseBuilder {
AlwaysMoveHandles = 1,
};
inline explicit ResponseBuilder(Service::HLERequestContext& ctx, u32 normal_params_size_, u32 num_handles_to_copy_ = 0, u32 num_objects_to_move_ = 0, Flags flags = Flags::None)
: cmdbuf(ctx.CommandBuffer())
, normal_params_size(normal_params_size_)
, num_handles_to_copy(num_handles_to_copy_)
, num_objects_to_move(num_objects_to_move_)
{
std::memset(cmdbuf, 0, sizeof(u32) * IPC::COMMAND_BUFFER_LENGTH);
explicit ResponseBuilder(Service::HLERequestContext& ctx, u32 normal_params_size_,
u32 num_handles_to_copy_ = 0, u32 num_objects_to_move_ = 0,
Flags flags = Flags::None)
: RequestHelperBase(ctx), normal_params_size(normal_params_size_),
num_handles_to_copy(num_handles_to_copy_),
num_objects_to_move(num_objects_to_move_), kernel{ctx.kernel} {
memset(cmdbuf, 0, sizeof(u32) * IPC::COMMAND_BUFFER_LENGTH);
IPC::CommandHeader header{};
auto const mgr = ctx.GetManager().get();
@@ -78,7 +117,9 @@ struct ResponseBuilder {
handle_descriptor_header.num_handles_to_copy.Assign(num_handles_to_copy_);
handle_descriptor_header.num_handles_to_move.Assign(num_handles_to_move);
PushRaw(handle_descriptor_header);
ctx.handles_offset = index;
Skip(num_handles_to_copy + num_handles_to_move, true);
}
@@ -90,6 +131,7 @@ struct ResponseBuilder {
domain_header.num_objects = num_domain_objects;
PushRaw(domain_header);
}
IPC::DataPayloadHeader data_payload_header{};
data_payload_header.magic = Common::MakeMagic('S', 'F', 'C', 'O');
PushRaw(data_payload_header);
@@ -99,39 +141,34 @@ struct ResponseBuilder {
ctx.data_payload_offset = index;
ctx.write_size += index;
ctx.domain_offset = u32(index + raw_data_size / sizeof(u32));
ctx.domain_offset = static_cast<u32>(index + raw_data_size / sizeof(u32));
}
inline void Skip(u32 size_in_words, bool set_to_null) {
if (set_to_null) std::memset(cmdbuf + index, 0, size_in_words * sizeof(u32));
index += size_in_words;
}
/// @brief Aligns the current position forward to a 16-byte boundary, padding with zeros.
inline void AlignWithPadding() { if (index & 3) Skip(u32(4 - (index & 3)), true); }
inline u32 GetCurrentOffset() const { return index; }
inline void SetCurrentOffset(u32 offset) { index = offset; }
template <class T>
void PushIpcInterface(std::shared_ptr<T> iface) {
auto manager{context->GetManager()};
template <class T> inline void PushIpcInterface(Service::HLERequestContext& ctx, std::shared_ptr<T> iface) {
auto manager = ctx.GetManager();
if (manager->IsDomain()) {
ctx.AddDomainObject(std::move(iface));
context->AddDomainObject(std::move(iface));
} else {
ASSERT(Kernel::GetCurrentProcess(ctx.kernel).GetResourceLimit()->Reserve(ctx.kernel, Kernel::LimitableResource::SessionCountMax, 1));
ASSERT(Kernel::GetCurrentProcess(kernel).GetResourceLimit()->Reserve(kernel, Kernel::LimitableResource::SessionCountMax, 1));
auto* session = Kernel::KSession::Create(ctx.kernel);
session->Initialize(ctx.kernel, nullptr, 0);
Kernel::KSession::Register(ctx.kernel, session);
auto* session = Kernel::KSession::Create(kernel);
session->Initialize(kernel, nullptr, 0);
Kernel::KSession::Register(kernel, session);
auto next_manager = std::make_shared<Service::SessionRequestManager>(ctx.kernel, manager->GetServerManager());
auto next_manager = std::make_shared<Service::SessionRequestManager>(
kernel, manager->GetServerManager());
next_manager->SetSessionHandler(iface);
manager->GetServerManager().RegisterSession(&session->GetServerSession(), next_manager);
ctx.AddMoveObject(&session->GetClientSession());
context->AddMoveObject(&session->GetClientSession());
}
}
template <class T, class... Args> inline void PushIpcInterface(Service::HLERequestContext& ctx, Args&&... args) {
PushIpcInterface<T>(ctx, std::make_shared<T>(std::forward<Args>(args)...));
template <class T, class... Args>
void PushIpcInterface(Args&&... args) {
PushIpcInterface<T>(std::make_shared<T>(std::forward<Args>(args)...));
}
void PushImpl(s8 value);
@@ -147,38 +184,59 @@ struct ResponseBuilder {
void PushImpl(bool value);
void PushImpl(Result value);
template <typename T> inline void Push(T value) {
template <typename T>
void Push(T value) {
return PushImpl(value);
}
template <typename First, typename... Other>
void Push(const First& first_value, const Other&... other_values);
/// @brief Helper function for pushing strongly-typed enumeration values.
/// @tparam Enum The enumeration type to be pushed
/// @param value The value to push.
/// @note The underlying size of the enumeration type is the size of the data that gets pushed.
/// e.g. "enum class SomeEnum : u16" will push a u16-sized amount of data.
template <typename Enum> inline void PushEnum(Enum value) {
/**
* Helper function for pushing strongly-typed enumeration values.
*
* @tparam Enum The enumeration type to be pushed
*
* @param value The value to push.
*
* @note The underlying size of the enumeration type is the size of the
* data that gets pushed. e.g. "enum class SomeEnum : u16" will
* push a u16-sized amount of data.
*/
template <typename Enum>
void PushEnum(Enum value) {
static_assert(std::is_enum_v<Enum>, "T must be an enum type within a PushEnum call.");
static_assert(!std::is_convertible_v<Enum, int>, "enum type in PushEnum must be a strongly typed enum.");
static_assert(!std::is_convertible_v<Enum, int>,
"enum type in PushEnum must be a strongly typed enum.");
Push(static_cast<std::underlying_type_t<Enum>>(value));
}
/// @brief Copies the content of the given trivially copyable class to the buffer as a normal param
/// @note: The input class must be correctly packed/padded to fit hardware layout.
template <typename T> void PushRaw(const T& value);
template <typename... O> void PushMoveObjects(Service::HLERequestContext& ctx, O*... pointers);
template <typename... O> void PushMoveObjects(Service::HLERequestContext& ctx, O&... pointers);
template <typename... O> void PushCopyObjects(Service::HLERequestContext& ctx, O*... pointers);
template <typename... O> void PushCopyObjects(Service::HLERequestContext& ctx, O&... pointers);
/**
* @brief Copies the content of the given trivially copyable class to the buffer as a normal
* param
* @note: The input class must be correctly packed/padded to fit hardware layout.
*/
template <typename T>
void PushRaw(const T& value);
u32* cmdbuf;
u32 index = 0;
template <typename... O>
void PushMoveObjects(O*... pointers);
template <typename... O>
void PushMoveObjects(O&... pointers);
template <typename... O>
void PushCopyObjects(O*... pointers);
template <typename... O>
void PushCopyObjects(O&... pointers);
private:
u32 normal_params_size{};
u32 num_handles_to_copy{};
u32 num_objects_to_move{}; ///< Domain objects or move handles, context dependent
u32 data_payload_index{};
Kernel::KernelCore& kernel;
};
/// Push ///
@@ -193,7 +251,8 @@ inline void ResponseBuilder::PushImpl(u32 value) {
template <typename T>
void ResponseBuilder::PushRaw(const T& value) {
static_assert(std::is_trivially_copyable_v<T>, "It's undefined behavior to use memcpy with non-trivially copyable objects");
static_assert(std::is_trivially_copyable_v<T>,
"It's undefined behavior to use memcpy with non-trivially copyable objects");
std::memcpy(cmdbuf + index, &value, sizeof(T));
index += (sizeof(T) + 3) / 4; // round up to word length
}
@@ -213,8 +272,8 @@ inline void ResponseBuilder::PushImpl(s16 value) {
}
inline void ResponseBuilder::PushImpl(s64 value) {
PushImpl(u32(value));
PushImpl(u32(value >> 32));
PushImpl(static_cast<u32>(value));
PushImpl(static_cast<u32>(value >> 32));
}
inline void ResponseBuilder::PushImpl(u8 value) {
@@ -226,8 +285,8 @@ inline void ResponseBuilder::PushImpl(u16 value) {
}
inline void ResponseBuilder::PushImpl(u64 value) {
PushImpl(u32(value));
PushImpl(u32(value >> 32));
PushImpl(static_cast<u32>(value));
PushImpl(static_cast<u32>(value >> 32));
}
inline void ResponseBuilder::PushImpl(float value) {
@@ -253,88 +312,90 @@ void ResponseBuilder::Push(const First& first_value, const Other&... other_value
}
template <typename... O>
inline void ResponseBuilder::PushCopyObjects(Service::HLERequestContext& ctx, O*... pointers) {
inline void ResponseBuilder::PushCopyObjects(O*... pointers) {
auto objects = {pointers...};
for (auto& object : objects) {
ctx.AddCopyObject(object);
context->AddCopyObject(object);
}
}
template <typename... O>
inline void ResponseBuilder::PushCopyObjects(Service::HLERequestContext& ctx, O&... pointers) {
inline void ResponseBuilder::PushCopyObjects(O&... pointers) {
auto objects = {&pointers...};
for (auto& object : objects) {
ctx.AddCopyObject(object);
context->AddCopyObject(object);
}
}
template <typename... O>
inline void ResponseBuilder::PushMoveObjects(Service::HLERequestContext& ctx, O*... pointers) {
inline void ResponseBuilder::PushMoveObjects(O*... pointers) {
auto objects = {pointers...};
for (auto& object : objects) {
ctx.AddMoveObject(object);
context->AddMoveObject(object);
}
}
template <typename... O>
inline void ResponseBuilder::PushMoveObjects(Service::HLERequestContext& ctx, O&... pointers) {
inline void ResponseBuilder::PushMoveObjects(O&... pointers) {
auto objects = {&pointers...};
for (auto& object : objects) {
ctx.AddMoveObject(object);
context->AddMoveObject(object);
}
}
struct RequestParser {
inline explicit RequestParser(u32* command_buffer) : cmdbuf(command_buffer) {}
inline explicit RequestParser(Service::HLERequestContext& ctx)
: cmdbuf(ctx.CommandBuffer())
{
class RequestParser : public RequestHelperBase {
public:
explicit RequestParser(u32* command_buffer) : RequestHelperBase(command_buffer) {}
explicit RequestParser(Service::HLERequestContext& ctx) : RequestHelperBase(ctx) {
// TIPC does not have data payload offset
if (!ctx.IsTipc()) {
ASSERT_MSG(ctx.GetDataPayloadOffset(), "context is incomplete");
Skip(ctx.GetDataPayloadOffset(), false);
}
// Skip the u64 command id, it's already stored in the context
static constexpr u32 CommandIdSize = 2;
Skip(CommandIdSize, false);
}
inline void Skip(u32 size_in_words, bool set_to_null) {
if (set_to_null) std::memset(cmdbuf + index, 0, size_in_words * sizeof(u32));
index += size_in_words;
}
/// @brief Aligns the current position forward to a 16-byte boundary, padding with zeros.
inline void AlignWithPadding() { if (index & 3) Skip(u32(4 - (index & 3)), true); }
inline u32 GetCurrentOffset() const { return index; }
inline void SetCurrentOffset(u32 offset) { index = offset; }
template <typename T>
T Pop();
template <typename T> T Pop();
template <typename T> void Pop(T& value);
template <typename First, typename... Other> void Pop(First& first_value, Other&... other_values);
template <typename T>
void Pop(T& value);
template <typename First, typename... Other>
void Pop(First& first_value, Other&... other_values);
template <typename T>
T PopEnum() {
static_assert(std::is_enum_v<T>, "T must be an enum type within a PopEnum call.");
static_assert(!std::is_convertible_v<T, int>, "enum type in PopEnum must be a strongly typed enum.");
return T(Pop<std::underlying_type_t<T>>());
static_assert(!std::is_convertible_v<T, int>,
"enum type in PopEnum must be a strongly typed enum.");
return static_cast<T>(Pop<std::underlying_type_t<T>>());
}
/// @brief Reads the next normal parameters as a struct, by copying it
/// @note: The output class must be correctly packed/padded to fit hardware layout.
template <typename T> void PopRaw(T& value);
/**
* @brief Reads the next normal parameters as a struct, by copying it
* @note: The output class must be correctly packed/padded to fit hardware layout.
*/
template <typename T>
void PopRaw(T& value);
/// @brief Reads the next normal parameters as a struct, by copying it into a new value
/// @note: The output class must be correctly packed/padded to fit hardware layout.
template <typename T> T PopRaw();
/**
* @brief Reads the next normal parameters as a struct, by copying it into a new value
* @note: The output class must be correctly packed/padded to fit hardware layout.
*/
template <typename T>
T PopRaw();
template <class T> [[nodiscard]] std::weak_ptr<T> PopIpcInterface(Service::HLERequestContext& ctx) {
ASSERT(ctx.GetManager()->IsDomain());
ASSERT(ctx.GetDomainMessageHeader().input_object_count > 0);
return ctx.GetDomainHandler<T>(Pop<u32>() - 1);
template <class T>
std::weak_ptr<T> PopIpcInterface() {
ASSERT(context->GetManager()->IsDomain());
ASSERT(context->GetDomainMessageHeader().input_object_count > 0);
return context->GetDomainHandler<T>(Pop<u32>() - 1);
}
u32* cmdbuf;
u32 index = 0;
};
/// Pop ///
@@ -346,7 +407,7 @@ inline u32 RequestParser::Pop() {
template <>
inline s32 RequestParser::Pop() {
return s32(Pop<u32>());
return static_cast<s32>(Pop<u32>());
}
// Ignore the -Wclass-memaccess warning on memcpy for non-trivially default constructible objects.
@@ -356,7 +417,8 @@ inline s32 RequestParser::Pop() {
#endif
template <typename T>
void RequestParser::PopRaw(T& value) {
static_assert(std::is_trivially_copyable_v<T>, "It's undefined behavior to use memcpy with non-trivially copyable objects");
static_assert(std::is_trivially_copyable_v<T>,
"It's undefined behavior to use memcpy with non-trivially copyable objects");
std::memcpy(&value, cmdbuf + index, sizeof(T));
index += (sizeof(T) + 3) / 4; // round up to word length
}
+1 -1
View File
@@ -353,7 +353,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ILogger>(ctx, system);
rb.PushIpcInterface<ILogger>(system);
}
};
+4 -4
View File
@@ -151,7 +151,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IAm>(ctx, system);
rb.PushIpcInterface<IAm>(system);
}
};
@@ -173,7 +173,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<MFIUser>(ctx, system);
rb.PushIpcInterface<MFIUser>(system);
}
};
@@ -195,7 +195,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IUser>(ctx, system);
rb.PushIpcInterface<IUser>(system);
}
};
@@ -217,7 +217,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ISystem>(ctx, system);
rb.PushIpcInterface<ISystem>(system);
}
};
+3 -3
View File
@@ -143,7 +143,7 @@ void NfcInterface::AttachAvailabilityChangeEvent(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, GetManager()->AttachAvailabilityChangeEvent());
rb.PushCopyObjects(GetManager()->AttachAvailabilityChangeEvent());
}
void NfcInterface::StartDetection(HLERequestContext& ctx) {
@@ -203,7 +203,7 @@ void NfcInterface::AttachActivateEvent(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(result);
rb.PushCopyObjects(ctx, out_event);
rb.PushCopyObjects(out_event);
}
void NfcInterface::AttachDeactivateEvent(HLERequestContext& ctx) {
@@ -217,7 +217,7 @@ void NfcInterface::AttachDeactivateEvent(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(result);
rb.PushCopyObjects(ctx, out_event);
rb.PushCopyObjects(out_event);
}
void NfcInterface::SetNfcEnabled(HLERequestContext& ctx) {
+3 -3
View File
@@ -157,7 +157,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IUser>(ctx, system);
rb.PushIpcInterface<IUser>(system);
}
};
@@ -179,7 +179,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ISystem>(ctx, system);
rb.PushIpcInterface<ISystem>(system);
}
};
@@ -201,7 +201,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IDebug>(ctx, system);
rb.PushIpcInterface<IDebug>(system);
}
};
+8 -8
View File
@@ -307,7 +307,7 @@ private:
void GetSystemEventReadableHandle(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 2};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, evt_scan_complete->GetReadableEvent(),
rb.PushCopyObjects(evt_scan_complete->GetReadableEvent(),
evt_processing->GetReadableEvent());
}
@@ -452,7 +452,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 2};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, event1->GetReadableEvent(), event2->GetReadableEvent());
rb.PushCopyObjects(event1->GetReadableEvent(), event2->GetReadableEvent());
}
void Cancel(HLERequestContext& ctx) {
@@ -528,7 +528,7 @@ void IGeneralService::CreateScanRequest(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IScanRequest>(ctx, system);
rb.PushIpcInterface<IScanRequest>(system);
}
void IGeneralService::CreateRequest(HLERequestContext& ctx) {
@@ -537,7 +537,7 @@ void IGeneralService::CreateRequest(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IRequest>(ctx, system);
rb.PushIpcInterface<IRequest>(system);
}
void IGeneralService::GetCurrentNetworkProfile(HLERequestContext& ctx) {
@@ -716,7 +716,7 @@ void IGeneralService::GetNetworkProfile(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2};
rb.Push(ResultSuccess);
rb.PushIpcInterface<INetworkProfile>(ctx, system);
rb.PushIpcInterface<INetworkProfile>(system);
}
void IGeneralService::SetNetworkProfile(HLERequestContext& ctx) {
@@ -869,7 +869,7 @@ void IGeneralService::CreateTemporaryNetworkProfile(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 6, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<INetworkProfile>(ctx, system);
rb.PushIpcInterface<INetworkProfile>(system);
rb.PushRaw<u128>(uuid);
}
@@ -1124,7 +1124,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IGeneralService>(ctx, system);
rb.PushIpcInterface<IGeneralService>(system);
}
void CreateGeneralService(HLERequestContext& ctx) {
@@ -1132,7 +1132,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IGeneralService>(ctx, system);
rb.PushIpcInterface<IGeneralService>(system);
}
};
+6 -6
View File
@@ -53,7 +53,7 @@ private:
LOG_WARNING(Service_NIM, "(STUBBED) called");
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IShopServiceAsync>(ctx, system);
rb.PushIpcInterface<IShopServiceAsync>(system);
}
};
@@ -75,7 +75,7 @@ private:
LOG_WARNING(Service_NIM, "(STUBBED) called");
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IShopServiceAccessor>(ctx, system);
rb.PushIpcInterface<IShopServiceAccessor>(system);
}
};
@@ -336,7 +336,7 @@ private:
LOG_DEBUG(Service_NIM, "(STUBBED) called");
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IShopServiceAccessServer>(ctx, system);
rb.PushIpcInterface<IShopServiceAccessServer>(system);
}
void IsLargeResourceAvailable(HLERequestContext& ctx) {
@@ -356,7 +356,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IShopServiceAccessServer>(ctx, system);
rb.PushIpcInterface<IShopServiceAccessServer>(system);
}
};
@@ -439,7 +439,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, finished_event->GetReadableEvent());
rb.PushCopyObjects(finished_event->GetReadableEvent());
}
void GetResult(HLERequestContext& ctx) {
@@ -500,7 +500,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IEnsureNetworkClockAvailabilityService>(ctx, system);
rb.PushIpcInterface<IEnsureNetworkClockAvailabilityService>(system);
}
// TODO(ogniK): Do we need these?
@@ -359,8 +359,8 @@ void IReadOnlyApplicationControlDataInterface::ListApplicationTitle(HLERequestCo
IPC::ResponseBuilder rb{ctx, 2, 1, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, async_value->ReadableEvent());
rb.PushIpcInterface(ctx, std::move(async_value));
rb.PushCopyObjects(async_value->ReadableEvent());
rb.PushIpcInterface(std::move(async_value));
}
Result IReadOnlyApplicationControlDataInterface::GetApplicationControlData3(
@@ -199,7 +199,7 @@ void NVDRV::QueryEvent(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 3, 1};
rb.Push(ResultSuccess);
auto& readable_event = event->GetReadableEvent();
rb.PushCopyObjects(ctx, readable_event);
rb.PushCopyObjects(readable_event);
rb.PushEnum(NvResult::Success);
} else {
LOG_ERROR(Service_NVDRV, "Invalid event request!");
+1 -4
View File
@@ -1,6 +1,3 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
@@ -127,7 +124,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IClkrstSession>(ctx, system, device_code);
rb.PushIpcInterface<IClkrstSession>(system, device_code);
}
};
+1 -1
View File
@@ -161,7 +161,7 @@ private:
IPC::ResponseBuilder rb{ctx, 10, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, *process);
rb.PushCopyObjects(*process);
rb.PushRaw(program_location);
rb.PushRaw(override_status);
}
+3 -3
View File
@@ -147,7 +147,7 @@ void IAlarmService::CreateWakeupAlarm(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ISteadyClockAlarm>(ctx, system, m_alarms, AlarmType::WakeupAlarm);
rb.PushIpcInterface<ISteadyClockAlarm>(system, m_alarms, AlarmType::WakeupAlarm);
}
void IAlarmService::CreateBackgroundTaskAlarm(HLERequestContext& ctx) {
@@ -155,7 +155,7 @@ void IAlarmService::CreateBackgroundTaskAlarm(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ISteadyClockAlarm>(ctx, system, m_alarms, AlarmType::BackgroundTaskAlarm);
rb.PushIpcInterface<ISteadyClockAlarm>(system, m_alarms, AlarmType::BackgroundTaskAlarm);
}
ISteadyClockAlarm::ISteadyClockAlarm(Core::System& system_, Alarms& alarms, AlarmType type)
@@ -179,7 +179,7 @@ void ISteadyClockAlarm::GetAlarmEvent(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, m_alarm.GetEventHandle());
rb.PushCopyObjects(m_alarm.GetEventHandle());
}
void ISteadyClockAlarm::Enable(HLERequestContext& ctx) {
+2 -2
View File
@@ -67,7 +67,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 1};
rb.Push(ResultSuccess);
rb.PushCopyObjects(ctx, state_change_event->GetReadableEvent());
rb.PushCopyObjects(state_change_event->GetReadableEvent());
}
void UnbindStateChangeEvent(HLERequestContext& ctx) {
@@ -186,7 +186,7 @@ void PSM::OpenSession(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IPsmSession>(ctx, system);
rb.PushIpcInterface<IPsmSession>(system);
}
void PSM::GetBatteryVoltageState(HLERequestContext& ctx) {
+1 -4
View File
@@ -1,6 +1,3 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
@@ -85,7 +82,7 @@ void TS::OpenSession(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ISession>(ctx, system);
rb.PushIpcInterface<ISession>(system);
}
} // namespace Service::PTM
+5 -4
View File
@@ -140,7 +140,7 @@ void SM::GetServiceCmif(HLERequestContext& ctx) {
if (result == ResultSuccess) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1, IPC::ResponseBuilder::Flags::AlwaysMoveHandles};
rb.Push(result);
rb.PushMoveObjects(ctx, client_session);
rb.PushMoveObjects(client_session);
} else {
IPC::ResponseBuilder rb{ctx, 2};
rb.Push(result);
@@ -157,7 +157,7 @@ void SM::GetServiceTipc(HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1, IPC::ResponseBuilder::Flags::AlwaysMoveHandles};
rb.Push(result);
rb.PushMoveObjects(ctx, result == ResultSuccess ? client_session : nullptr);
rb.PushMoveObjects(result == ResultSuccess ? client_session : nullptr);
}
static std::string PopServiceName(IPC::RequestParser& rp) {
@@ -230,7 +230,8 @@ void SM::RegisterServiceImpl(HLERequestContext& ctx, std::string name, u32 max_s
max_session_count, is_light);
Kernel::KServerPort* server_port{};
if (const auto result = service_manager.RegisterService(std::addressof(server_port), name, max_session_count, nullptr);
if (const auto result = service_manager.RegisterService(std::addressof(server_port), name,
max_session_count, nullptr);
result.IsError()) {
LOG_ERROR(Service_SM, "failed to register service with error_code={:08X}", result.raw);
IPC::ResponseBuilder rb{ctx, 2};
@@ -240,7 +241,7 @@ void SM::RegisterServiceImpl(HLERequestContext& ctx, std::string name, u32 max_s
IPC::ResponseBuilder rb{ctx, 2, 0, 1, IPC::ResponseBuilder::Flags::AlwaysMoveHandles};
rb.Push(ResultSuccess);
rb.PushMoveObjects(ctx, server_port);
rb.PushMoveObjects(server_port);
}
void SM::UnregisterService(HLERequestContext& ctx) {
+1 -1
View File
@@ -60,7 +60,7 @@ void Controller::CloneCurrentObject(HLERequestContext& ctx) {
// We succeeded.
IPC::ResponseBuilder rb{ctx, 2, 0, 1, IPC::ResponseBuilder::Flags::AlwaysMoveHandles};
rb.Push(ResultSuccess);
rb.PushMoveObjects(ctx, session->GetClientSession());
rb.PushMoveObjects(session->GetClientSession());
}
void Controller::CloneCurrentObjectEx(HLERequestContext& ctx) {
+5 -3
View File
@@ -543,7 +543,8 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(res);
if (res == ResultSuccess) {
rb.PushIpcInterface<ISslConnection>(ctx, system, ssl_version, shared_data, std::move(backend));
rb.PushIpcInterface<ISslConnection>(system, ssl_version, shared_data,
std::move(backend));
}
}
@@ -623,11 +624,12 @@ private:
IPC::RequestParser rp{ctx};
const auto parameters = rp.PopRaw<Parameters>();
LOG_WARNING(Service_SSL, "(STUBBED) called, api_version={}, pid_placeholder={}", parameters.ssl_version.api_version, parameters.pid_placeholder);
LOG_WARNING(Service_SSL, "(STUBBED) called, api_version={}, pid_placeholder={}",
parameters.ssl_version.api_version, parameters.pid_placeholder);
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<ISslContext>(ctx, system, parameters.ssl_version);
rb.PushIpcInterface<ISslContext>(system, parameters.ssl_version);
}
void SetInterfaceVersion(HLERequestContext& ctx) {
+2 -2
View File
@@ -155,7 +155,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IPdSession>(ctx, system);
rb.PushIpcInterface<IPdSession>(system);
}
};
@@ -199,7 +199,7 @@ private:
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(ResultSuccess);
rb.PushIpcInterface<IPdCradleSession>(ctx, system);
rb.PushIpcInterface<IPdCradleSession>(system);
}
};
+2 -2
View File
@@ -50,8 +50,8 @@ NPad::NPad(Core::HID::HIDCore& hid_core_, KernelHelpers::ServiceContext& service
auto& controller = controller_data[aruid_index][i];
controller.device = hid_core.GetEmulatedControllerByIndex(i);
Core::HID::ControllerUpdateCallback engine_callback{
.on_change = [this, i, kernel = &hid_core.kernel](Core::HID::ControllerTriggerType type) {
ControllerUpdate(*kernel, type, i);
.on_change = [this, i](Core::HID::ControllerTriggerType type) {
ControllerUpdate(hid_core.kernel, type, i);
},
.is_npad_service = true,
};
@@ -236,11 +236,8 @@ void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInf
IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
Environment& env, Flow::CFG& cfg, const HostTranslateInfo& host_info) {
HostTranslateInfo normalized_host_info{host_info};
normalized_host_info.ApplyDescriptorLimitPolicy();
IR::Program program;
program.syntax_list = BuildASL(inst_pool, block_pool, env, cfg, normalized_host_info);
program.syntax_list = BuildASL(inst_pool, block_pool, env, cfg, host_info);
program.blocks = GenerateBlocks(program.syntax_list);
program.post_order_blocks = PostOrder(program.syntax_list.front());
program.stage = env.ShaderStage();
@@ -263,9 +260,9 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
program.info.passthrough.mask[i] = ((mask[i / 32] >> (i % 32)) & 1) == 0;
}
if (!normalized_host_info.support_geometry_shader_passthrough) {
if (!host_info.support_geometry_shader_passthrough) {
program.output_vertices = GetOutputTopologyVertices(program.output_topology);
LowerGeometryPassthrough(program, normalized_host_info);
LowerGeometryPassthrough(program, host_info);
}
}
break;
@@ -280,16 +277,16 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
RemoveUnreachableBlocks(program);
// Replace instructions before the SSA rewrite
if (!normalized_host_info.support_float64) {
if (!host_info.support_float64) {
Optimization::LowerFp64ToFp32(program);
}
if (!normalized_host_info.support_float16) {
if (!host_info.support_float16) {
Optimization::LowerFp16ToFp32(program);
}
if (!normalized_host_info.support_int64) {
if (!host_info.support_int64) {
Optimization::LowerInt64ToInt32(program);
}
if (!normalized_host_info.support_conditional_barrier) {
if (!host_info.support_conditional_barrier) {
Optimization::ConditionalBarrierPass(program);
}
Optimization::SsaRewritePass(program);
@@ -298,8 +295,8 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
Optimization::PositionPass(env, program);
Optimization::GlobalMemoryToStorageBufferPass(program, normalized_host_info);
Optimization::TexturePass(env, program, normalized_host_info);
Optimization::GlobalMemoryToStorageBufferPass(program, host_info);
Optimization::TexturePass(env, program, host_info);
if (Settings::values.resolution_info.active || Settings::values.rescale_hack.GetValue()) {
Optimization::RescalingPass(program);
@@ -309,7 +306,7 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
Optimization::VerificationPass(program);
}
Optimization::CollectShaderInfoPass(env, program);
Optimization::LayerPass(program, normalized_host_info);
Optimization::LayerPass(program, host_info);
Optimization::VendorWorkaroundPass(program);
CollectInterpolationInfo(env, program);
+4 -38
View File
@@ -6,8 +6,6 @@
#pragma once
#include "common/common_types.h"
namespace Shader {
// Try to keep entries here to a minimum
@@ -15,52 +13,20 @@ namespace Shader {
/// Misc information about the host
struct HostTranslateInfo {
static constexpr u32 DEFAULT_DESCRIPTOR_LIMIT = 1024;
u64 min_ssbo_alignment{}; ///< Minimum alignment supported by the device for SSBOs
u32 max_per_stage_descriptor_sampled_images{}; ///< maximum sampled descriptors per stage
u32 max_per_stage_resources{}; ///< maximum resources per stage
u32 max_descriptor_set_samplers{};
u32 max_descriptor_set_uniform_buffers{};
u32 max_descriptor_set_uniform_buffers_dynamic{};
u32 max_descriptor_set_storage_buffers{};
u32 max_descriptor_set_storage_buffers_dynamic{};
u32 max_descriptor_set_sampled_images{};
u32 max_descriptor_set_storage_images{};
u32 max_descriptor_set_input_attachements{};
bool support_float64{}; ///< True when the device supports 64-bit floats
bool support_float16{}; ///< True when the device supports 16-bit floats
bool support_int64{}; ///< True when the device supports 64-bit integers
bool needs_demote_reorder{}; ///< True when the device needs DemoteToHelperInvocation reordered
bool support_snorm_render_buffer{}; ///< True when the device supports SNORM render buffers
bool support_viewport_index_layer{}; ///< True when the device supports gl_Layer in VS
u32 min_ssbo_alignment{}; ///< Minimum alignment supported by the device for SSBOs
u32 max_per_stage_descriptor_sampled_images{1024}; ///< maximum sampled descriptors per stage
u32 max_per_stage_resources{4096}; ///< maximum resources per stage
u32 max_descriptor_set_sampled_images{1024}; ///< maximum sampled descriptors per set
bool support_geometry_shader_passthrough{}; ///< True when the device supports geometry
///< passthrough shaders
bool support_conditional_barrier{}; ///< True when the device supports barriers in conditional
///< control flow
void ApplyDescriptorLimitPolicy() noexcept {
if (min_ssbo_alignment == 0) {
min_ssbo_alignment = 1;
}
ApplyDescriptorLimitFallback(max_per_stage_descriptor_sampled_images);
ApplyDescriptorLimitFallback(max_per_stage_resources);
ApplyDescriptorLimitFallback(max_descriptor_set_samplers);
ApplyDescriptorLimitFallback(max_descriptor_set_uniform_buffers);
ApplyDescriptorLimitFallback(max_descriptor_set_uniform_buffers_dynamic);
ApplyDescriptorLimitFallback(max_descriptor_set_storage_buffers);
ApplyDescriptorLimitFallback(max_descriptor_set_storage_buffers_dynamic);
ApplyDescriptorLimitFallback(max_descriptor_set_sampled_images);
ApplyDescriptorLimitFallback(max_descriptor_set_storage_images);
ApplyDescriptorLimitFallback(max_descriptor_set_input_attachements);
}
private:
static void ApplyDescriptorLimitFallback(u32& limit) noexcept {
if (limit == 0) {
limit = DEFAULT_DESCRIPTOR_LIMIT;
}
}
};
} // namespace Shader
@@ -1,6 +1,3 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
@@ -548,7 +545,7 @@ void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateIn
IR::Block* const block{storage_inst.block};
IR::Inst* const inst{storage_inst.inst};
const IR::U32 offset{
StorageOffset(*block, *inst, storage_buffer, u32(host_info.min_ssbo_alignment))};
StorageOffset(*block, *inst, storage_buffer, host_info.min_ssbo_alignment)};
Replace(*block, *inst, index, offset);
}
}
+71 -65
View File
@@ -32,62 +32,71 @@ struct TextureInst {
using TextureInstVector = boost::container::small_vector<TextureInst, 24>;
constexpr u32 DESCRIPTOR_SIZE = 8;
constexpr u32 DESCRIPTOR_SIZE_SHIFT = u32(std::countr_zero(DESCRIPTOR_SIZE));
constexpr u32 DESCRIPTOR_MAX_COUNT = 1024;
constexpr u32 DESCRIPTOR_SIZE_SHIFT = static_cast<u32>(std::countr_zero(DESCRIPTOR_SIZE));
constexpr u32 DYNAMIC_DESCRIPTOR_CBUF_BYTES = 16 * 1024;
constexpr u32 MAX_DYNAMIC_DESCRIPTOR_COUNT = 1024;
u32 DynamicDescriptorSizeShift(const IR::U32& dynamic_offset) {
const IR::Inst* const inst = dynamic_offset.InstRecursive();
if (!inst || inst->GetOpcode() != IR::Opcode::ShiftLeftLogical32)
const IR::Inst* const inst{dynamic_offset.InstRecursive()};
if (!inst || inst->GetOpcode() != IR::Opcode::ShiftLeftLogical32) {
return DESCRIPTOR_SIZE_SHIFT;
const IR::Value shift = inst->Arg(1);
if (!shift.IsImmediate())
}
const IR::Value shift{inst->Arg(1)};
if (!shift.IsImmediate()) {
return DESCRIPTOR_SIZE_SHIFT;
const u32 size_shift = shift.U32();
return size_shift >= DESCRIPTOR_SIZE_SHIFT && size_shift < 31 ? size_shift : DESCRIPTOR_SIZE_SHIFT;
}
const u32 size_shift{shift.U32()};
return size_shift >= DESCRIPTOR_SIZE_SHIFT && size_shift < 31 ? size_shift
: DESCRIPTOR_SIZE_SHIFT;
}
u32 DynamicDescriptorCount(u32 base_offset, u32 size_shift, u32 max_descriptors) {
auto const descriptor_limit = (std::max)(1U, max_descriptors);
auto const max_cbuf_bytes = 16 * descriptor_limit;
if (size_shift >= 31 || base_offset >= max_cbuf_bytes)
u32 DynamicDescriptorCount(u32 base_offset, u32 size_shift) {
if (size_shift >= 31 || base_offset >= DYNAMIC_DESCRIPTOR_CBUF_BYTES) {
return 1;
auto const stride = 1U << size_shift;
auto const available = max_cbuf_bytes - base_offset;
if (available < DESCRIPTOR_SIZE)
}
const u32 stride{1U << size_shift};
const u32 available{DYNAMIC_DESCRIPTOR_CBUF_BYTES - base_offset};
if (available < DESCRIPTOR_SIZE) {
return 1;
auto const available_count = 1U + (available - DESCRIPTOR_SIZE) / stride;
return std::min(descriptor_limit, available_count);
}
const u32 available_count{1U + (available - DESCRIPTOR_SIZE) / stride};
return std::min(MAX_DYNAMIC_DESCRIPTOR_COUNT, available_count);
}
u32 SaturatingSub(u32 lhs, u32 rhs) {
return lhs > rhs ? lhs - rhs : 0;
}
template <typename T>
[[nodiscard]] u32 StaticDescriptorCount(T const& descriptors) noexcept {
return std::accumulate(descriptors.cbegin(), descriptors.cend(), 0U, [](auto const& acc, auto const& e) {
return acc + (e.count <= 1 ? e.count : 0);
});
template <typename Descriptors>
u32 StaticDescriptorCount(const Descriptors& descriptors) {
u32 count{};
for (const auto& desc : descriptors) {
if (desc.count <= 1) {
count += desc.count;
}
}
return count;
}
u32 DynamicSampledTextureCap(const Info& info, const HostTranslateInfo& host_info, u32 dynamic_arrays) {
auto const sampled_limit = (std::max)(1U, std::min(host_info.max_per_stage_descriptor_sampled_images,
host_info.max_descriptor_set_sampled_images));
auto const resource_limit = (std::max)(1U, host_info.max_per_stage_resources);
if (dynamic_arrays > 0) {
auto const sampled_static_count = StaticDescriptorCount(info.texture_buffer_descriptors) + StaticDescriptorCount(info.texture_descriptors);
auto const resource_static_count =
NumDescriptors(info.constant_buffer_descriptors)
+ NumDescriptors(info.storage_buffers_descriptors)
+ sampled_static_count + NumDescriptors(info.image_buffer_descriptors)
+ NumDescriptors(info.image_descriptors);
auto const sampled_budget = SaturatingSub(sampled_limit, sampled_static_count);
auto const resource_budget = SaturatingSub(resource_limit, resource_static_count);
auto const sampled_cap = sampled_budget / dynamic_arrays;
auto const resource_cap = resource_budget / dynamic_arrays;
return (std::max)(1U, (std::min)(sampled_cap, resource_cap));
u32 DynamicSampledTextureCap(const Info& info, const HostTranslateInfo& host_info,
u32 dynamic_arrays) {
if (dynamic_arrays == 0) {
return MAX_DYNAMIC_DESCRIPTOR_COUNT;
}
return (std::min)({DESCRIPTOR_MAX_COUNT, sampled_limit, resource_limit});
const u32 sampled_static_count{StaticDescriptorCount(info.texture_buffer_descriptors) +
StaticDescriptorCount(info.texture_descriptors)};
const u32 resource_static_count{
NumDescriptors(info.constant_buffer_descriptors) +
NumDescriptors(info.storage_buffers_descriptors) + sampled_static_count +
NumDescriptors(info.image_buffer_descriptors) + NumDescriptors(info.image_descriptors)};
const u32 sampled_limit{std::min(host_info.max_per_stage_descriptor_sampled_images,
host_info.max_descriptor_set_sampled_images)};
const u32 sampled_budget{SaturatingSub(sampled_limit, sampled_static_count)};
const u32 resource_budget{SaturatingSub(host_info.max_per_stage_resources,
resource_static_count)};
const u32 sampled_cap{sampled_budget / dynamic_arrays};
const u32 resource_cap{resource_budget / dynamic_arrays};
return std::max(1U, std::min({MAX_DYNAMIC_DESCRIPTOR_COUNT, sampled_cap, resource_cap}));
}
IR::Opcode IndexedInstruction(const IR::Inst& inst) {
@@ -295,23 +304,21 @@ static inline bool IsTexturePixelFormatIntegerCached(Environment& env,
}
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env, const HostTranslateInfo& host_info);
static inline std::optional<ConstBufferAddr> TrackCached(const IR::Value& v, Environment& env, const HostTranslateInfo& host_info) {
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env);
static inline std::optional<ConstBufferAddr> TrackCached(const IR::Value& v, Environment& env) {
if (const IR::Inst* key = v.InstRecursive()) {
if (auto it = env.track_cache.find(key); it != env.track_cache.end()) return it->second;
auto found = Track(v, env, host_info);
auto found = Track(v, env);
if (found) env.track_cache.emplace(key, *found);
return found;
}
return Track(v, env, host_info);
return Track(v, env);
}
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env, const HostTranslateInfo& host_info);
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env);
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env, const HostTranslateInfo& host_info) {
return IR::BreadthFirstSearch(value, [&env, &host_info](const IR::Inst* inst) {
return TryGetConstBuffer(inst, env, host_info);
});
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env) {
return IR::BreadthFirstSearch(value, [&env](const IR::Inst* inst) { return TryGetConstBuffer(inst, env); });
}
std::optional<u32> TryGetConstant(IR::Value& value, Environment& env) {
@@ -335,13 +342,13 @@ std::optional<u32> TryGetConstant(IR::Value& value, Environment& env) {
return ReadCbufCached(env, index_number, offset_number);
}
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env, const HostTranslateInfo& host_info) {
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env) {
switch (inst->GetOpcode()) {
default:
return std::nullopt;
case IR::Opcode::BitwiseOr32: {
std::optional lhs{TrackCached(inst->Arg(0), env, host_info)};
std::optional rhs{TrackCached(inst->Arg(1), env, host_info)};
std::optional lhs{TrackCached(inst->Arg(0), env)};
std::optional rhs{TrackCached(inst->Arg(1), env)};
if (!lhs || !rhs) {
return std::nullopt;
}
@@ -371,11 +378,12 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
if (!shift.IsImmediate()) {
return std::nullopt;
}
std::optional lhs{TrackCached(inst->Arg(0), env, host_info)};
std::optional lhs{TrackCached(inst->Arg(0), env)};
if (lhs) {
lhs->shift_left = shift.U32();
}
return lhs;
break;
}
case IR::Opcode::BitwiseAnd32: {
IR::Value op1{inst->Arg(0)};
@@ -399,7 +407,7 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
return std::nullopt;
} while (false);
}
std::optional lhs{TrackCached(op1, env, host_info)};
std::optional lhs{TrackCached(op1, env)};
if (lhs) {
lhs->shift_left = static_cast<u32>(std::countr_zero(op2.U32()));
}
@@ -445,10 +453,7 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
} else {
return std::nullopt;
}
auto const size_shift = DynamicDescriptorSizeShift(dynamic_offset);
auto const sampled_limit = (std::max)(1U, (std::min)(host_info.max_per_stage_descriptor_sampled_images,
host_info.max_descriptor_set_sampled_images));
auto const resource_limit = (std::max)(1U, host_info.max_per_stage_resources);
const u32 size_shift{DynamicDescriptorSizeShift(dynamic_offset)};
return ConstBufferAddr{
.index = index.U32(),
.offset = base_offset,
@@ -457,15 +462,15 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
.secondary_offset = 0,
.secondary_shift_left = 0,
.dynamic_offset = dynamic_offset,
.count = DynamicDescriptorCount(base_offset, size_shift, (std::min)({DESCRIPTOR_MAX_COUNT, sampled_limit, resource_limit})),
.count = DynamicDescriptorCount(base_offset, size_shift),
.has_secondary = false,
};
}
TextureInst MakeInst(Environment& env, IR::Block* block, IR::Inst& inst, const HostTranslateInfo& host_info) {
TextureInst MakeInst(Environment& env, IR::Block* block, IR::Inst& inst) {
ConstBufferAddr addr;
if (IsBindless(inst)) {
const std::optional<ConstBufferAddr> track_addr{TrackCached(inst.Arg(0), env, host_info)};
const std::optional<ConstBufferAddr> track_addr{TrackCached(inst.Arg(0), env)};
if (!track_addr) {
throw NotImplementedException("Failed to track bindless texture constant buffer");
@@ -501,15 +506,15 @@ u32 GetTextureHandle(Environment& env, const ConstBufferAddr& cbuf) {
return lhs_raw | rhs_raw;
}
[[maybe_unused]] TextureType ReadTextureType(Environment& env, const ConstBufferAddr& cbuf) {
[[maybe_unused]]TextureType ReadTextureType(Environment& env, const ConstBufferAddr& cbuf) {
return env.ReadTextureType(GetTextureHandle(env, cbuf));
}
[[maybe_unused]] TexturePixelFormat ReadTexturePixelFormat(Environment& env, const ConstBufferAddr& cbuf) {
[[maybe_unused]]TexturePixelFormat ReadTexturePixelFormat(Environment& env, const ConstBufferAddr& cbuf) {
return env.ReadTexturePixelFormat(GetTextureHandle(env, cbuf));
}
[[maybe_unused]] bool IsTexturePixelFormatInteger(Environment& env, const ConstBufferAddr& cbuf) {
[[maybe_unused]]bool IsTexturePixelFormatInteger(Environment& env, const ConstBufferAddr& cbuf) {
return env.IsTexturePixelFormatInteger(GetTextureHandle(env, cbuf));
}
@@ -670,7 +675,7 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
if (!IsTextureInstruction(inst)) {
continue;
}
to_replace.push_back(MakeInst(env, block, inst, host_info));
to_replace.push_back(MakeInst(env, block, inst));
}
}
// Sort instructions to visit textures by constant buffer index, then by offset
@@ -684,7 +689,8 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
program.info.texture_descriptors,
program.info.image_descriptors,
};
const u32 sampled_dynamic_cap = DynamicSampledTextureCap(program.info, host_info, DynamicSampledTextureArrayCount(to_replace));
const u32 sampled_dynamic_cap{
DynamicSampledTextureCap(program.info, host_info, DynamicSampledTextureArrayCount(to_replace))};
for (TextureInst& texture_inst : to_replace) {
// TODO: Handle arrays
IR::Inst* const inst{texture_inst.inst};
+1
View File
@@ -92,6 +92,7 @@ struct Profile {
bool has_broken_robust{};
u64 min_ssbo_alignment{};
u32 max_user_clip_distances{};
};
+26 -24
View File
@@ -1619,35 +1619,36 @@ void BufferCache<P>::TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept {
template <class P>
bool BufferCache<P>::SynchronizeBuffer(Buffer& buffer, DAddr device_addr, u32 size) {
upload_copies.clear();
u64 total_size_bytes = 0;
u64 staging_offset = 0;
u64 largest_copy = 0;
const DAddr buffer_start = buffer.cpu_addr_cached;
memory_tracker.ForEachUploadRange(device_addr, size, [&](u64 device_addr_out, u64 range_size) {
upload_copies.push_back(BufferCopy{
.src_offset = total_size_bytes,
.dst_offset = device_addr_out - buffer_start,
.size = range_size,
DAddr buffer_start = buffer.CpuAddr();
auto push = [&](u64 start, u64 end) {
if (start >= end) {
return;
}
u64 sz = end - start;
upload_copies.push_back({
.src_offset = staging_offset,
.dst_offset = start - buffer_start,
.size = sz
});
total_size_bytes += range_size;
largest_copy = (std::max)(largest_copy, range_size);
staging_offset += sz;
largest_copy = (std::max)(largest_copy, sz);
};
memory_tracker.ForEachUploadRange(device_addr, size, [&](u64 addr, u64 range_size) {
u64 start = addr;
u64 end = addr + range_size;
gpu_modified_ranges.ForEachInRange(start, range_size, [&](u64 gstart, u64 gsize) {
u64 gend = gstart + gsize;
push(start, gstart);
start = (std::max)(start, gend);
});
push(start, end);
});
if (total_size_bytes == 0) {
if (upload_copies.empty()) {
return true;
}
if (Settings::values.enable_gpu_buffer_readback.GetValue()) {
u64 min_offset = (std::numeric_limits<u64>::max)();
u64 max_offset = 0;
for (const auto& copy : upload_copies) {
min_offset = (std::min)(min_offset, copy.dst_offset);
max_offset = (std::max)(max_offset, copy.dst_offset + copy.size);
}
const DAddr sync_addr = buffer.CpuAddr() + min_offset;
const u64 sync_size = max_offset - min_offset;
DownloadBufferMemory(buffer, sync_addr, sync_size);
}
const std::span<BufferCopy> copies_span(upload_copies.data(), upload_copies.size());
UploadMemory(buffer, total_size_bytes, largest_copy, copies_span);
any_buffer_uploaded = true;
UploadMemory(buffer, staging_offset, largest_copy, std::span(upload_copies));
return false;
}
@@ -1698,6 +1699,7 @@ void BufferCache<P>::MappedUploadMemory([[maybe_unused]] Buffer& buffer,
u8* const src_pointer = staging_pointer.data() + copy.src_offset;
const DAddr device_addr = buffer.CpuAddr() + copy.dst_offset;
device_memory.ReadBlockUnsafe(device_addr, src_pointer, copy.size);
// Apply the staging offset
copy.src_offset += upload_staging.offset;
}
@@ -11,8 +11,9 @@
#define BEGIN_PUSH_CONSTANTS layout(push_constant) uniform PushConstants {
#define END_PUSH_CONSTANTS };
#define UNIFORM(n)
#define BINDING_INPUT_BUFFER 0
#define BINDING_OUTPUT_IMAGE 1
#define BINDING_SWIZZLE_BUFFER 0
#define BINDING_INPUT_BUFFER 1
#define BINDING_OUTPUT_IMAGE 2
#else // ^^^ Vulkan ^^^ // vvv OpenGL vvv
@@ -25,7 +26,8 @@
#define BEGIN_PUSH_CONSTANTS
#define END_PUSH_CONSTANTS
#define UNIFORM(n) layout (location = n) uniform
#define BINDING_INPUT_BUFFER 0
#define BINDING_SWIZZLE_BUFFER 0
#define BINDING_INPUT_BUFFER 1
#define BINDING_OUTPUT_IMAGE 0
#endif
@@ -41,6 +43,10 @@ UNIFORM(6) uint block_height;
UNIFORM(7) uint block_height_mask;
END_PUSH_CONSTANTS
layout(binding = BINDING_SWIZZLE_BUFFER, std430) readonly buffer SwizzleTable {
uint swizzle_table[];
};
#if HAS_EXTENDED_TYPES
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU8 { uint8_t u8data[]; };
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU16 { uint16_t u16data[]; };
@@ -65,19 +71,9 @@ const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT + GOB_SIZE_Z_SHI
const uvec2 SWIZZLE_MASK = uvec2(GOB_SIZE_X - 1, GOB_SIZE_Y - 1);
uint SwizzleTable(uint pos) {
const uint t[8] = uint[](
0x12100200, 0x13110301, 0x16140604, 0x17150705,
0x1a180a08, 0x1b190b09, 0x1e1c0e0c, 0x1f1d0f0d
);
const uint i = pos >> 4;
const uint h = (t[i / 4] >> ((i % 4) * 8)) & 0xff;
return (h << 4) | (pos & 0xf);
}
uint SwizzleOffset(uvec2 pos) {
pos = pos & SWIZZLE_MASK;
return SwizzleTable(pos.y * 64 + pos.x);
return swizzle_table[pos.y * 64 + pos.x];
}
uvec4 ReadTexel(uint offset) {
@@ -11,8 +11,9 @@
#define BEGIN_PUSH_CONSTANTS layout(push_constant) uniform PushConstants {
#define END_PUSH_CONSTANTS };
#define UNIFORM(n)
#define BINDING_INPUT_BUFFER 0
#define BINDING_OUTPUT_IMAGE 1
#define BINDING_SWIZZLE_BUFFER 0
#define BINDING_INPUT_BUFFER 1
#define BINDING_OUTPUT_IMAGE 2
#else // ^^^ Vulkan ^^^ // vvv OpenGL vvv
@@ -25,7 +26,8 @@
#define BEGIN_PUSH_CONSTANTS
#define END_PUSH_CONSTANTS
#define UNIFORM(n) layout (location = n) uniform
#define BINDING_INPUT_BUFFER 0
#define BINDING_SWIZZLE_BUFFER 0
#define BINDING_INPUT_BUFFER 1
#define BINDING_OUTPUT_IMAGE 0
#endif
@@ -43,6 +45,10 @@ UNIFORM(8) uint block_depth;
UNIFORM(9) uint block_depth_mask;
END_PUSH_CONSTANTS
layout(binding = BINDING_SWIZZLE_BUFFER, std430) readonly buffer SwizzleTable {
uint swizzle_table[];
};
#if HAS_EXTENDED_TYPES
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU8 { uint8_t u8data[]; };
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU16 { uint16_t u16data[]; };
@@ -67,19 +73,9 @@ const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT + GOB_SIZE_Z_SHI
const uvec2 SWIZZLE_MASK = uvec2(GOB_SIZE_X - 1, GOB_SIZE_Y - 1);
uint SwizzleTable(uint pos) {
const uint t[8] = uint[](
0x12100200, 0x13110301, 0x16140604, 0x17150705,
0x1a180a08, 0x1b190b09, 0x1e1c0e0c, 0x1f1d0f0d
);
const uint i = pos >> 4;
const uint h = (t[i / 4] >> ((i % 4) * 8)) & 0xff;
return (h << 4) | (pos & 0xf);
}
uint SwizzleOffset(uvec2 pos) {
pos = pos & SWIZZLE_MASK;
return SwizzleTable(pos.y * 64 + pos.x);
return swizzle_table[pos.y * 64 + pos.x];
}
uvec4 ReadTexel(uint offset) {
@@ -10,8 +10,9 @@
#define BEGIN_PUSH_CONSTANTS layout(push_constant) uniform PushConstants {
#define END_PUSH_CONSTANTS };
#define UNIFORM(n)
#define BINDING_INPUT_BUFFER 0
#define BINDING_OUTPUT_BUFFER 1
#define BINDING_SWIZZLE_BUFFER 0
#define BINDING_INPUT_BUFFER 1
#define BINDING_OUTPUT_BUFFER 2
#else
#extension GL_NV_gpu_shader5 : enable
#ifdef GL_NV_gpu_shader5
@@ -22,6 +23,7 @@
#define BEGIN_PUSH_CONSTANTS
#define END_PUSH_CONSTANTS
#define UNIFORM(n) layout(location = n) uniform
#define BINDING_SWIZZLE_BUFFER 0
#define BINDING_INPUT_BUFFER 1
#define BINDING_OUTPUT_BUFFER 0
#endif
@@ -64,9 +66,13 @@ END_PUSH_CONSTANTS
#endif
// --- Buffers ---
layout(binding = BINDING_SWIZZLE_BUFFER, std430) readonly buffer SwizzleTable {
uint swizzle_table[];
};
#if HAS_EXTENDED_TYPES
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU8 { uint8_t u8data[]; };
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU16 { uint16_t u16data[]; };
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU8 { uint8_t u8data[]; };
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU16 { uint16_t u16data[]; };
#endif
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU32 { uint u32data[]; };
layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU64 { uvec2 u64data[]; };
@@ -90,20 +96,10 @@ const uint GOB_SIZE_Z_SHIFT = 0;
const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT + GOB_SIZE_Z_SHIFT;
const uvec2 SWIZZLE_MASK = uvec2(GOB_SIZE_X - 1u, GOB_SIZE_Y - 1u);
uint SwizzleTable(uint pos) {
const uint t[8] = uint[](
0x12100200, 0x13110301, 0x16140604, 0x17150705,
0x1a180a08, 0x1b190b09, 0x1e1c0e0c, 0x1f1d0f0d
);
const uint i = pos >> 4;
const uint h = (t[i / 4] >> ((i % 4) * 8)) & 0xff;
return (h << 4) | (pos & 0xf);
}
// --- Helpers ---
uint SwizzleOffset(uvec2 pos) {
pos &= SWIZZLE_MASK;
return SwizzleTable(pos.y * 64u + pos.x);
return swizzle_table[pos.y * 64u + pos.x];
}
uvec4 ReadTexel(uint offset) {
@@ -245,31 +245,16 @@ ShaderCache::ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
std::min<u32>(device.GetMaxUserClipDistances(), Maxwell::Regs::NumClipDistances),
},
host_info{
.min_ssbo_alignment = static_cast<u32>(device.GetShaderStorageBufferAlignment()),
.max_per_stage_descriptor_sampled_images =
Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.max_per_stage_resources = Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.max_descriptor_set_samplers = Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.max_descriptor_set_uniform_buffers = Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.max_descriptor_set_uniform_buffers_dynamic =
Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.max_descriptor_set_storage_buffers = Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.max_descriptor_set_storage_buffers_dynamic =
Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.max_descriptor_set_sampled_images = Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.max_descriptor_set_storage_images = Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.max_descriptor_set_input_attachements =
Shader::HostTranslateInfo::DEFAULT_DESCRIPTOR_LIMIT,
.support_float64 = true,
.support_float16 = false,
.support_int64 = device.HasShaderInt64(),
.needs_demote_reorder = device.IsAmd(),
.support_snorm_render_buffer = false,
.support_viewport_index_layer = device.HasVertexViewportLayer(),
.support_geometry_shader_passthrough = device.HasGeometryShaderPassthrough(),
.support_conditional_barrier = device.SupportsConditionalBarriers(),
.support_float64 = true,
.support_float16 = false,
.support_int64 = device.HasShaderInt64(),
.needs_demote_reorder = device.IsAmd(),
.support_snorm_render_buffer = false,
.support_viewport_index_layer = device.HasVertexViewportLayer(),
.min_ssbo_alignment = static_cast<u32>(device.GetShaderStorageBufferAlignment()),
.support_geometry_shader_passthrough = device.HasGeometryShaderPassthrough(),
.support_conditional_barrier = device.SupportsConditionalBarriers(),
} {
host_info.ApplyDescriptorLimitPolicy();
if (use_asynchronous_shaders) {
workers = CreateWorkers();
}
@@ -1,6 +1,3 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
@@ -59,8 +56,10 @@ UtilShaders::UtilShaders(ProgramManager& program_manager_)
copy_bc4_program(MakeProgram(OPENGL_COPY_BC4_COMP)),
convert_s8d24_program(MakeProgram(OPENGL_CONVERT_S8D24_COMP)),
convert_ms_to_nonms_program(MakeProgram(CONVERT_MSAA_TO_NON_MSAA_COMP)),
convert_nonms_to_ms_program(MakeProgram(CONVERT_NON_MSAA_TO_MSAA_COMP))
{
convert_nonms_to_ms_program(MakeProgram(CONVERT_NON_MSAA_TO_MSAA_COMP)) {
const auto swizzle_table = Tegra::Texture::MakeSwizzleTable();
swizzle_table_buffer.Create();
glNamedBufferStorage(swizzle_table_buffer.handle, sizeof(swizzle_table), &swizzle_table, 0);
}
UtilShaders::~UtilShaders() = default;
@@ -117,11 +116,13 @@ void UtilShaders::ASTCDecode(Image& image, const StagingBufferMap& map,
void UtilShaders::BlockLinearUpload2D(Image& image, const StagingBufferMap& map,
std::span<const SwizzleParameters> swizzles) {
static constexpr Extent3D WORKGROUP_SIZE{32, 32, 1};
static constexpr GLuint BINDING_INPUT_BUFFER = 0;
static constexpr GLuint BINDING_SWIZZLE_BUFFER = 0;
static constexpr GLuint BINDING_INPUT_BUFFER = 1;
static constexpr GLuint BINDING_OUTPUT_IMAGE = 0;
program_manager.BindComputeProgram(block_linear_unswizzle_2d_program.handle);
glFlushMappedNamedBufferRange(map.buffer, map.offset, image.guest_size_bytes);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_SWIZZLE_BUFFER, swizzle_table_buffer.handle);
const GLenum store_format = StoreFormat(BytesPerBlock(image.info.format));
for (const SwizzleParameters& swizzle : swizzles) {
@@ -152,11 +153,14 @@ void UtilShaders::BlockLinearUpload2D(Image& image, const StagingBufferMap& map,
void UtilShaders::BlockLinearUpload3D(Image& image, const StagingBufferMap& map,
std::span<const SwizzleParameters> swizzles) {
static constexpr Extent3D WORKGROUP_SIZE{16, 8, 8};
static constexpr GLuint BINDING_INPUT_BUFFER = 0;
static constexpr GLuint BINDING_SWIZZLE_BUFFER = 0;
static constexpr GLuint BINDING_INPUT_BUFFER = 1;
static constexpr GLuint BINDING_OUTPUT_IMAGE = 0;
glFlushMappedNamedBufferRange(map.buffer, map.offset, image.guest_size_bytes);
program_manager.BindComputeProgram(block_linear_unswizzle_3d_program.handle);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_SWIZZLE_BUFFER, swizzle_table_buffer.handle);
const GLenum store_format = StoreFormat(BytesPerBlock(image.info.format));
for (const SwizzleParameters& swizzle : swizzles) {
@@ -1,6 +1,3 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
@@ -48,6 +45,9 @@ public:
private:
ProgramManager& program_manager;
OGLBuffer swizzle_table_buffer;
OGLProgram astc_decoder_program;
OGLProgram block_linear_unswizzle_2d_program;
OGLProgram block_linear_unswizzle_3d_program;
@@ -22,15 +22,6 @@ namespace Vulkan {
using Shader::Backend::SPIRV::NUM_TEXTURE_AND_IMAGE_SCALING_WORDS;
[[nodiscard]] inline u32 NumDescriptorEntries(const Shader::Info& info) {
return Shader::NumDescriptors(info.constant_buffer_descriptors) +
Shader::NumDescriptors(info.storage_buffers_descriptors) +
Shader::NumDescriptors(info.texture_buffer_descriptors) +
Shader::NumDescriptors(info.image_buffer_descriptors) +
Shader::NumDescriptors(info.texture_descriptors) +
Shader::NumDescriptors(info.image_descriptors);
}
class DescriptorLayoutBuilder {
public:
DescriptorLayoutBuilder(const Device& device_) : device{&device_} {}
@@ -203,11 +194,7 @@ inline void PushImageDescriptors(TextureCache& texture_cache,
const VideoCommon::ImageViewId image_view_id{(views++)->id};
const VideoCommon::SamplerId sampler_id{*(samplers++)};
ImageView& image_view{texture_cache.GetImageView(image_view_id)};
VkImageView vk_image_view{image_view.Handle(desc.type)};
if (vk_image_view == VK_NULL_HANDLE) {
const VkImageView null_image_view{texture_cache.GetImageView(VideoCommon::NULL_IMAGE_VIEW_ID).Handle(desc.type)};
if (null_image_view != VK_NULL_HANDLE) vk_image_view = null_image_view;
}
const VkImageView vk_image_view{image_view.Handle(desc.type)};
const Sampler& sampler{texture_cache.GetSampler(sampler_id)};
const bool use_fallback_sampler{sampler.HasAddedAnisotropy() &&
!image_view.SupportsAnisotropy()};
@@ -326,7 +326,7 @@ std::pair<VkBuffer, VkDeviceSize> Uint8Pass::Assemble(u32 num_vertices, VkBuffer
const u32 staging_size = static_cast<u32>(num_vertices * sizeof(u16));
const auto staging = staging_buffer_pool.Request(staging_size, MemoryUsage::DeviceLocal);
compute_pass_descriptor_queue.Acquire(scheduler, 2);
compute_pass_descriptor_queue.Acquire();
compute_pass_descriptor_queue.AddBuffer(src_buffer, src_offset, num_vertices);
compute_pass_descriptor_queue.AddBuffer(staging.buffer, staging.offset, staging_size);
const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
@@ -384,7 +384,7 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
const std::size_t staging_size = num_tri_vertices * sizeof(u32);
const auto staging = staging_buffer_pool.Request(staging_size, MemoryUsage::DeviceLocal);
compute_pass_descriptor_queue.Acquire(scheduler, 2);
compute_pass_descriptor_queue.Acquire();
compute_pass_descriptor_queue.AddBuffer(src_buffer, src_offset, input_size);
compute_pass_descriptor_queue.AddBuffer(staging.buffer, staging.offset, staging_size);
const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
@@ -429,7 +429,7 @@ void ConditionalRenderingResolvePass::Resolve(VkBuffer dst_buffer, VkBuffer src_
}
const size_t compare_size = compare_to_zero ? 8 : 24;
compute_pass_descriptor_queue.Acquire(scheduler, 2);
compute_pass_descriptor_queue.Acquire();
compute_pass_descriptor_queue.AddBuffer(src_buffer, src_offset, compare_size);
compute_pass_descriptor_queue.AddBuffer(dst_buffer, 0, sizeof(u32));
const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
@@ -498,7 +498,7 @@ void QueriesPrefixScanPass::Run(VkBuffer accumulation_buffer, VkBuffer dst_buffe
static constexpr size_t DISPATCH_SIZE = 2048U;
size_t runs_to_do = std::min<size_t>(current_runs, DISPATCH_SIZE);
current_runs -= runs_to_do;
compute_pass_descriptor_queue.Acquire(scheduler, 3);
compute_pass_descriptor_queue.Acquire();
compute_pass_descriptor_queue.AddBuffer(src_buffer, 0, number_of_sums * sizeof(u64));
compute_pass_descriptor_queue.AddBuffer(dst_buffer, 0, number_of_sums * sizeof(u64));
compute_pass_descriptor_queue.AddBuffer(accumulation_buffer, 0, sizeof(u64));
@@ -600,7 +600,7 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 8U);
const u32 num_dispatches_z = image.info.resources.layers;
compute_pass_descriptor_queue.Acquire(scheduler, 2);
compute_pass_descriptor_queue.Acquire();
compute_pass_descriptor_queue.AddBuffer(map.buffer, input_offset,
image.guest_size_bytes - swizzle.buffer_offset);
compute_pass_descriptor_queue.AddImage(image.StorageImageView(swizzle.level));
@@ -653,8 +653,71 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
scheduler.Finish();
}
constexpr u32 BL3D_BINDING_INPUT_BUFFER = 0;
constexpr u32 BL3D_BINDING_OUTPUT_BUFFER = 1;
constexpr u32 BL3D_BINDING_SWIZZLE_TABLE = 0;
constexpr u32 BL3D_BINDING_INPUT_BUFFER = 1;
constexpr u32 BL3D_BINDING_OUTPUT_BUFFER = 2;
constexpr std::array<VkDescriptorSetLayoutBinding, 3> BL3D_DESCRIPTOR_SET_BINDINGS{{
{
.binding = BL3D_BINDING_SWIZZLE_TABLE,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, // swizzle_table[]
.descriptorCount = 1,
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.pImmutableSamplers = nullptr,
},
{
.binding = BL3D_BINDING_INPUT_BUFFER,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, // block-linear input
.descriptorCount = 1,
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.pImmutableSamplers = nullptr,
},
{
.binding = BL3D_BINDING_OUTPUT_BUFFER,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.descriptorCount = 1,
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.pImmutableSamplers = nullptr,
},
}};
constexpr DescriptorBankInfo BL3D_BANK_INFO{
.uniform_buffers = 0,
.storage_buffers = 3,
.texture_buffers = 0,
.image_buffers = 0,
.textures = 0,
.images = 0,
.score = 3,
};
constexpr std::array<VkDescriptorUpdateTemplateEntry, 3>
BL3D_DESCRIPTOR_UPDATE_TEMPLATE_ENTRY{{
{
.dstBinding = BL3D_BINDING_SWIZZLE_TABLE,
.dstArrayElement = 0,
.descriptorCount = 1,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.offset = BL3D_BINDING_SWIZZLE_TABLE * sizeof(DescriptorUpdateEntry),
.stride = sizeof(DescriptorUpdateEntry),
},
{
.dstBinding = BL3D_BINDING_INPUT_BUFFER,
.dstArrayElement = 0,
.descriptorCount = 1,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.offset = BL3D_BINDING_INPUT_BUFFER * sizeof(DescriptorUpdateEntry),
.stride = sizeof(DescriptorUpdateEntry),
},
{
.dstBinding = BL3D_BINDING_OUTPUT_BUFFER,
.dstArrayElement = 0,
.descriptorCount = 1,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.offset = BL3D_BINDING_OUTPUT_BUFFER * sizeof(DescriptorUpdateEntry),
.stride = sizeof(DescriptorUpdateEntry),
}
}};
struct alignas(16) BlockLinearUnswizzle3DPushConstants {
u32 blocks_dim[3]; // Offset 0
@@ -682,50 +745,11 @@ BlockLinearUnswizzle3DPass::BlockLinearUnswizzle3DPass(
DescriptorPool& descriptor_pool_,
StagingBufferPool& staging_buffer_pool_,
ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
: ComputePass(device_, scheduler_, descriptor_pool_,
std::array<VkDescriptorSetLayoutBinding, 2>{{
{
.binding = BL3D_BINDING_INPUT_BUFFER,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, // block-linear input
.descriptorCount = 1,
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.pImmutableSamplers = nullptr,
},
{
.binding = BL3D_BINDING_OUTPUT_BUFFER,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.descriptorCount = 1,
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.pImmutableSamplers = nullptr,
},
}},
std::array<VkDescriptorUpdateTemplateEntry, 2>{{
{
.dstBinding = BL3D_BINDING_INPUT_BUFFER,
.dstArrayElement = 0,
.descriptorCount = 1,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.offset = BL3D_BINDING_INPUT_BUFFER * sizeof(DescriptorUpdateEntry),
.stride = sizeof(DescriptorUpdateEntry),
},
{
.dstBinding = BL3D_BINDING_OUTPUT_BUFFER,
.dstArrayElement = 0,
.descriptorCount = 1,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.offset = BL3D_BINDING_OUTPUT_BUFFER * sizeof(DescriptorUpdateEntry),
.stride = sizeof(DescriptorUpdateEntry),
}
}},
DescriptorBankInfo{
.uniform_buffers = 0,
.storage_buffers = 2,
.texture_buffers = 0,
.image_buffers = 0,
.textures = 0,
.images = 0,
.score = 2,
},
: ComputePass(
device_, scheduler_, descriptor_pool_,
BL3D_DESCRIPTOR_SET_BINDINGS,
BL3D_DESCRIPTOR_UPDATE_TEMPLATE_ENTRY,
BL3D_BANK_INFO,
COMPUTE_PUSH_CONSTANT_RANGE<sizeof(BlockLinearUnswizzle3DPushConstants)>,
BLOCK_LINEAR_UNSWIZZLE_3D_BCN_COMP_SPV),
scheduler{scheduler_},
@@ -797,7 +821,9 @@ void BlockLinearUnswizzle3DPass::UnswizzleChunk(
pc.blocks_dim[1] = blocks_y;
pc.blocks_dim[2] = z_count; // Only process the count
compute_pass_descriptor_queue.Acquire(scheduler, 3);
compute_pass_descriptor_queue.Acquire();
compute_pass_descriptor_queue.AddBuffer(*image.runtime->swizzle_table_buffer, 0,
image.runtime->swizzle_table_size);
compute_pass_descriptor_queue.AddBuffer(swizzled.buffer,
sw.buffer_offset + swizzled.offset,
image.guest_size_bytes - sw.buffer_offset);
@@ -963,7 +989,7 @@ void MSAACopyPass::CopyImage(Image& dst_image, Image& src_image,
ASSERT(copy.dst_subresource.base_layer == 0);
ASSERT(copy.dst_subresource.num_layers == 1);
compute_pass_descriptor_queue.Acquire(scheduler, 2);
compute_pass_descriptor_queue.Acquire();
compute_pass_descriptor_queue.AddImage(
src_image.StorageImageView(copy.src_subresource.base_level));
compute_pass_descriptor_queue.AddImage(
@@ -45,7 +45,6 @@ ComputePipeline::ComputePipeline(const Device& device_, Scheduler& scheduler, vk
}
std::copy_n(info.constant_buffer_used_sizes.begin(), uniform_buffer_sizes.size(),
uniform_buffer_sizes.begin());
num_descriptor_entries = NumDescriptorEntries(info);
auto func{[this, &scheduler, &descriptor_pool, shader_notify, pipeline_statistics] {
DescriptorLayoutBuilder builder{device};
@@ -114,7 +113,7 @@ ComputePipeline::ComputePipeline(const Device& device_, Scheduler& scheduler, vk
void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute,
Tegra::MemoryManager& gpu_memory, Scheduler& scheduler,
BufferCache& buffer_cache, TextureCache& texture_cache) {
guest_descriptor_queue.Acquire(scheduler, num_descriptor_entries);
guest_descriptor_queue.Acquire();
buffer_cache.SetComputeUniformBufferState(info.constant_buffer_mask, &uniform_buffer_sizes);
buffer_cache.UnbindComputeStorageBuffers();
@@ -53,7 +53,6 @@ private:
vk::PipelineCache& pipeline_cache;
GuestDescriptorQueue& guest_descriptor_queue;
Shader::Info info;
u32 num_descriptor_entries{};
VideoCommon::ComputeUniformBufferSizes uniform_buffer_sizes{};
@@ -268,7 +268,6 @@ GraphicsPipeline::GraphicsPipeline(
num_textures += Shader::NumDescriptors(info->texture_descriptors);
num_image_elements += Shader::NumDescriptors(info->texture_descriptors);
num_image_elements += Shader::NumDescriptors(info->image_descriptors);
num_descriptor_entries += NumDescriptorEntries(*info);
}
fragment_has_color0_output = stage_infos[NUM_STAGES - 1].stores_frag_color[0];
auto func{[this, shader_notify, &render_pass_cache, &descriptor_pool, pipeline_statistics] {
@@ -474,7 +473,7 @@ bool GraphicsPipeline::ConfigureImpl(bool is_indexed) {
buffer_cache.UpdateGraphicsBuffers(is_indexed);
buffer_cache.BindHostGeometryBuffers(is_indexed);
guest_descriptor_queue.Acquire(scheduler, num_descriptor_entries);
guest_descriptor_queue.Acquire();
RescalingPushConstant rescaling;
RenderAreaPushConstant render_area;
@@ -159,7 +159,6 @@ private:
std::array<Shader::Info, NUM_STAGES> stage_infos;
std::array<u32, 5> enabled_uniform_buffer_masks{};
VideoCommon::UniformBufferSizes uniform_buffer_sizes{};
u32 num_descriptor_entries{};
size_t num_image_elements{};
u32 num_textures{};
bool fragment_has_color0_output{};
@@ -439,21 +439,10 @@ PipelineCache::PipelineCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
.has_broken_robust =
device.IsNvidia() && device.GetNvidiaArch() <= NvidiaArchitecture::Arch_Pascal,
.min_ssbo_alignment = device.GetStorageBufferAlignment(),
.max_user_clip_distances = device.GetMaxUserClipDistances()
.max_user_clip_distances = device.GetMaxUserClipDistances(),
};
host_info = Shader::HostTranslateInfo{
.min_ssbo_alignment = device.GetStorageBufferAlignment(),
.max_per_stage_descriptor_sampled_images = device.GetMaxPerStageDescriptorSampledImages(),
.max_per_stage_resources = device.GetMaxPerStageResources(),
.max_descriptor_set_samplers = device.GetMaxDescriptorSetSamplers(),
.max_descriptor_set_uniform_buffers = device.GetMaxDescriptorSetUniformBuffers(),
.max_descriptor_set_uniform_buffers_dynamic = device.GetMaxDescriptorSetUniformBuffersDynamic(),
.max_descriptor_set_storage_buffers = device.GetMaxDescriptorSetStorageBuffers(),
.max_descriptor_set_storage_buffers_dynamic = device.GetMaxDescriptorSetStorageBuffersDynamic(),
.max_descriptor_set_sampled_images = device.GetMaxDescriptorSetSampledImages(),
.max_descriptor_set_storage_images = device.GetMaxDescriptorSetStorageImages(),
.max_descriptor_set_input_attachements = device.GetMaxDescriptorSetInputAttachments(),
.support_float64 = device.IsFloat64Supported(),
.support_float16 = device.IsFloat16Supported(),
.support_int64 = device.IsShaderInt64Supported(),
@@ -462,10 +451,13 @@ PipelineCache::PipelineCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
driver_id == VK_DRIVER_ID_SAMSUNG_PROPRIETARY,
.support_snorm_render_buffer = true,
.support_viewport_index_layer = device.IsExtShaderViewportIndexLayerSupported(),
.min_ssbo_alignment = static_cast<u32>(device.GetStorageBufferAlignment()),
.max_per_stage_descriptor_sampled_images = device.GetMaxPerStageDescriptorSampledImages(),
.max_per_stage_resources = device.GetMaxPerStageResources(),
.max_descriptor_set_sampled_images = device.GetMaxDescriptorSetSampledImages(),
.support_geometry_shader_passthrough = device.IsNvGeometryShaderPassthroughSupported(),
.support_conditional_barrier = device.SupportsConditionalBarriers(),
};
host_info.ApplyDescriptorLimitPolicy();
if (device.GetMaxVertexInputAttributes() < Maxwell::NumVertexAttributes) {
LOG_WARNING(Render_Vulkan, "maxVertexInputAttributes is too low: {} < {}",
@@ -203,7 +203,7 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
: gpu{gpu_}, device_memory{device_memory_}, device{device_},
memory_allocator{memory_allocator_}, state_tracker{state_tracker_}, scheduler{scheduler_},
staging_pool(device, memory_allocator, scheduler), descriptor_pool(device, scheduler),
guest_descriptor_queue(device), compute_pass_descriptor_queue(device),
guest_descriptor_queue(device, scheduler), compute_pass_descriptor_queue(device, scheduler),
blit_image(device, scheduler, state_tracker, descriptor_pool), render_pass_cache(device),
texture_cache_runtime{
device, scheduler, memory_allocator, staging_pool,
@@ -155,14 +155,15 @@ void Scheduler::WaitWorker() {
}
void Scheduler::DispatchWork() {
if (chunk && !chunk->Empty()) {
{
std::scoped_lock ql{queue_mutex};
work_queue.push(std::move(chunk));
}
event_cv.notify_all();
AcquireNewChunk();
if (chunk->Empty()) {
return;
}
{
std::scoped_lock ql{queue_mutex};
work_queue.push(std::move(chunk));
}
event_cv.notify_all();
AcquireNewChunk();
}
void Scheduler::RequestRenderpass(const Framebuffer* framebuffer) {
@@ -909,6 +909,40 @@ TextureCacheRuntime::TextureCacheRuntime(const Device& device_, Scheduler& sched
bl3d_unswizzle_pass.emplace(device, scheduler, descriptor_pool,
staging_buffer_pool, compute_pass_descriptor_queue);
}
// --- Create swizzle table buffer ---
{
auto table = Tegra::Texture::MakeSwizzleTable();
swizzle_table_size = static_cast<VkDeviceSize>(table.size() * sizeof(table[0]));
auto staging = staging_buffer_pool.Request(swizzle_table_size, MemoryUsage::Upload);
std::memcpy(staging.mapped_span.data(), table.data(), static_cast<size_t>(swizzle_table_size));
VkBufferCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.size = swizzle_table_size,
.usage = VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_TRANSFER_SRC_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
};
swizzle_table_buffer = memory_allocator.CreateBuffer(ci, MemoryUsage::DeviceLocal);
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([staging_buf = staging.buffer,
dst_buf = *swizzle_table_buffer,
size = swizzle_table_size,
src_off = staging.offset](vk::CommandBuffer cmdbuf) {
const VkBufferCopy region{
.srcOffset = src_off,
.dstOffset = 0,
.size = size,
};
cmdbuf.CopyBuffer(staging_buf, dst_buf, region);
});
}
}
void TextureCacheRuntime::Finish() {
@@ -2450,13 +2484,18 @@ void TextureCacheRuntime::AccelerateImageUpload(
if (!Settings::values.gpu_unswizzle_enabled.GetValue() || !bl3d_unswizzle_pass) {
if (IsPixelFormatBCn(image.info.format) && image.info.type == ImageType::e3D) {
ASSERT(false && "GPU unswizzle is disabled for BCn 3D texture");
ASSERT_MSG(false, "GPU unswizzle is disabled for BCn 3D texture");
}
ASSERT(false);
return;
}
if (bl3d_unswizzle_pass && IsPixelFormatBCn(image.info.format) && image.info.type == ImageType::e3D && image.info.resources.levels == 1 && image.info.resources.layers == 1) {
if (bl3d_unswizzle_pass &&
IsPixelFormatBCn(image.info.format) &&
image.info.type == ImageType::e3D &&
image.info.resources.levels == 1 &&
image.info.resources.layers == 1) {
return bl3d_unswizzle_pass->Unswizzle(image, map, swizzles, z_start, z_count);
}
@@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
@@ -130,6 +130,9 @@ public:
std::optional<ASTCDecoderPass> astc_decoder_pass;
std::optional<BlockLinearUnswizzle3DPass> bl3d_unswizzle_pass;
vk::Buffer swizzle_table_buffer;
VkDeviceSize swizzle_table_size = 0;
std::optional<MSAACopyPass> msaa_copy_pass;
const Settings::ResolutionScalingInfo& resolution;
std::array<std::vector<VkFormat>, VideoCore::Surface::MaxPixelFormat> view_formats;
@@ -7,7 +7,6 @@
#include <variant>
#include <boost/container/static_vector.hpp>
#include "common/assert.h"
#include "common/logging.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_update_descriptor.h"
@@ -16,9 +15,8 @@
namespace Vulkan {
UpdateDescriptorQueue::UpdateDescriptorQueue(const Device& device_)
: device{device_}
{
UpdateDescriptorQueue::UpdateDescriptorQueue(const Device& device_, Scheduler& scheduler_)
: device{device_}, scheduler{scheduler_} {
payload_start = payload.data();
payload_cursor = payload.data();
}
@@ -33,15 +31,13 @@ void UpdateDescriptorQueue::TickFrame() {
payload_cursor = payload_start;
}
void UpdateDescriptorQueue::Acquire(Scheduler& scheduler, size_t required_entries) {
static constexpr size_t DEFAULT_REQUIRED_ENTRIES = 0x400;
const size_t reserve = required_entries > 0 ? required_entries : DEFAULT_REQUIRED_ENTRIES;
ASSERT_MSG(reserve < FRAME_PAYLOAD_SIZE, "Descriptor reservation {} >= frame capacity {}",
reserve, FRAME_PAYLOAD_SIZE);
const size_t used = static_cast<size_t>(std::distance(payload_start, payload_cursor));
if (used + reserve >= FRAME_PAYLOAD_SIZE) {
LOG_WARNING(Render_Vulkan, "Payload overflow (used={}, reserve={}, capacity={})",
used, reserve, FRAME_PAYLOAD_SIZE);
void UpdateDescriptorQueue::Acquire() {
// Minimum number of entries required.
// This is the maximum number of entries a single draw call might use.
static constexpr size_t MIN_ENTRIES = 0x400;
if (std::distance(payload_start, payload_cursor) + MIN_ENTRIES >= FRAME_PAYLOAD_SIZE) {
LOG_WARNING(Render_Vulkan, "Payload overflow, waiting for worker thread");
scheduler.WaitWorker();
payload_cursor = payload_start;
}
@@ -34,11 +34,12 @@ class UpdateDescriptorQueue final {
static constexpr size_t PAYLOAD_SIZE = FRAME_PAYLOAD_SIZE * FRAMES_IN_FLIGHT;
public:
explicit UpdateDescriptorQueue(const Device& device_);
explicit UpdateDescriptorQueue(const Device& device_, Scheduler& scheduler_);
~UpdateDescriptorQueue();
void TickFrame();
void Acquire(Scheduler& scheduler, size_t required_entries = 0);
void Acquire();
const DescriptorUpdateEntry* UpdateData() const noexcept {
return upload_start;
@@ -74,6 +75,8 @@ public:
private:
const Device& device;
Scheduler& scheduler;
size_t frame_index{0};
DescriptorUpdateEntry* payload_cursor = nullptr;
DescriptorUpdateEntry* payload_start = nullptr;
+3 -2
View File
@@ -255,7 +255,8 @@ std::optional<u64> GenericEnvironment::TryFindSize() {
static constexpr u64 SELF_BRANCH_A = 0xE2400FFFFF87000FULL;
static constexpr u64 SELF_BRANCH_B = 0xE2400FFFFF07000FULL;
static constexpr u64 EXIT_VALUE = 0xE30000000007000FULL;
static constexpr u64 MESA_EXIT_MASK = 0xFFF00000000F001FULL;
static constexpr u64 MESA_EXIT_VALUE = (0xE30ULL << 52) | (0x7ULL << 16) | 0xFULL;
code.resize(MAXIMUM_SIZE / INST_SIZE);
@@ -270,7 +271,7 @@ std::optional<u64> GenericEnvironment::TryFindSize() {
if (inst == SELF_BRANCH_A || inst == SELF_BRANCH_B) {
return offset + index;
}
if (!is_proprietary_driver && inst == EXIT_VALUE) {
if ((inst & MESA_EXIT_MASK) == MESA_EXIT_VALUE) {
return offset + index + INST_SIZE;
}
}
+18 -3
View File
@@ -1,6 +1,3 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
@@ -26,6 +23,24 @@ constexpr u32 GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT + GOB_SIZE_Z_
constexpr u32 SWIZZLE_X_BITS = 0b100101111;
constexpr u32 SWIZZLE_Y_BITS = 0b011010000;
using SwizzleTable = std::array<std::array<u32, GOB_SIZE_X>, GOB_SIZE_Y>;
/**
* This table represents the internal swizzle of a gob, in format 16 bytes x 2 sector packing.
* Calculates the offset of an (x, y) position within a swizzled texture.
* Taken from the Tegra X1 Technical Reference Manual. pages 1187-1188
*/
constexpr SwizzleTable MakeSwizzleTable() {
SwizzleTable table{};
for (u32 y = 0; y < table.size(); ++y) {
for (u32 x = 0; x < table[0].size(); ++x) {
table[y][x] = ((x % 64) / 32) * 256 + ((y % 8) / 2) * 64 + ((x % 32) / 16) * 32 +
(y % 2) * 16 + (x % 16);
}
}
return table;
}
/// Unswizzles a block linear texture into linear memory.
void UnswizzleTexture(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel,
u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth,
+26 -17
View File
@@ -321,23 +321,32 @@ public:
return properties.properties.limits.maxPushConstantsSize;
}
#define FN_MAX_LIMIT_LIST \
FN_MAX_LIMIT_ELEM(ComputeSharedMemorySize) \
FN_MAX_LIMIT_ELEM(PerStageDescriptorSampledImages) \
FN_MAX_LIMIT_ELEM(PerStageResources) \
FN_MAX_LIMIT_ELEM(DescriptorSetSamplers) \
FN_MAX_LIMIT_ELEM(DescriptorSetUniformBuffers) \
FN_MAX_LIMIT_ELEM(DescriptorSetUniformBuffersDynamic) \
FN_MAX_LIMIT_ELEM(DescriptorSetStorageBuffers) \
FN_MAX_LIMIT_ELEM(DescriptorSetStorageBuffersDynamic) \
FN_MAX_LIMIT_ELEM(DescriptorSetSampledImages) \
FN_MAX_LIMIT_ELEM(DescriptorSetStorageImages) \
FN_MAX_LIMIT_ELEM(DescriptorSetInputAttachments)
#define FN_MAX_LIMIT_ELEM(name) \
u32 GetMax##name() const { return properties.properties.limits.max##name; }
FN_MAX_LIMIT_LIST
#undef FN_MAX_LIMIT_ELEM
#undef FN_MAX_LIMIT_LIST
/// Returns the maximum size for shared memory.
u32 GetMaxComputeSharedMemorySize() const {
return properties.properties.limits.maxComputeSharedMemorySize;
}
/// Returns the maximum number of dynamic storage buffer descriptors per set.
u32 GetMaxDescriptorSetStorageBuffersDynamic() const {
return properties.properties.limits.maxDescriptorSetStorageBuffersDynamic;
}
/// Returns the maximum number of dynamic uniform buffer descriptors per set.
u32 GetMaxDescriptorSetUniformBuffersDynamic() const {
return properties.properties.limits.maxDescriptorSetUniformBuffersDynamic;
}
u32 GetMaxPerStageDescriptorSampledImages() const {
return properties.properties.limits.maxPerStageDescriptorSampledImages;
}
u32 GetMaxPerStageResources() const {
return properties.properties.limits.maxPerStageResources;
}
u32 GetMaxDescriptorSetSampledImages() const {
return properties.properties.limits.maxDescriptorSetSampledImages;
}
/// Returns float control properties of the device.
const VkPhysicalDeviceFloatControlsPropertiesKHR& FloatControlProperties() const {
-1
View File
@@ -44,7 +44,6 @@ int IrShaderRecompilerImpl(int argc, char *argv[]) {
host_info.support_geometry_shader_passthrough = true;
host_info.support_conditional_barrier = true;
host_info.min_ssbo_alignment = 0;
host_info.ApplyDescriptorLimitPolicy();
auto program = Shader::Maxwell::TranslateProgram(inst_pool, block_pool, env, cfg, host_info);
auto const dumped_ir = Shader::IR::DumpProgram(program);
std::printf("%s\n", dumped_ir.c_str());
@@ -52,7 +52,6 @@ int SpirvShaderRecompilerImpl(int argc, char *argv[]) {
host_info.support_geometry_shader_passthrough = true;
host_info.support_conditional_barrier = true;
host_info.min_ssbo_alignment = 0;
host_info.ApplyDescriptorLimitPolicy();
auto program = Shader::Maxwell::TranslateProgram(inst_pool, block_pool, env, cfg, host_info);
// IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,