diff options
-rw-r--r-- | src/core/arm/dynarmic/arm_dynarmic.cpp | 1 | ||||
-rw-r--r-- | src/core/hle/kernel/svc.cpp | 51 | ||||
-rw-r--r-- | src/core/hle/kernel/thread.cpp | 3 | ||||
-rw-r--r-- | src/core/hle/kernel/vm_manager.cpp | 103 | ||||
-rw-r--r-- | src/core/hle/kernel/vm_manager.h | 139 | ||||
-rw-r--r-- | src/core/hle/service/am/am.cpp | 8 | ||||
-rw-r--r-- | src/core/hle/service/am/applets/applets.cpp | 6 | ||||
-rw-r--r-- | src/core/hle/service/am/applets/software_keyboard.cpp | 3 | ||||
-rw-r--r-- | src/core/hle/service/nfp/nfp.cpp | 4 | ||||
-rw-r--r-- | src/core/hle/service/sm/sm.cpp | 9 | ||||
-rw-r--r-- | src/video_core/engines/shader_bytecode.h | 4 | ||||
-rw-r--r-- | src/video_core/renderer_opengl/gl_shader_decompiler.cpp | 54 |
12 files changed, 335 insertions, 50 deletions
diff --git a/src/core/arm/dynarmic/arm_dynarmic.cpp b/src/core/arm/dynarmic/arm_dynarmic.cpp index 4d2491870..afbda8d8b 100644 --- a/src/core/arm/dynarmic/arm_dynarmic.cpp +++ b/src/core/arm/dynarmic/arm_dynarmic.cpp @@ -151,6 +151,7 @@ std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit() const { config.tpidr_el0 = &cb->tpidr_el0; config.dczid_el0 = 4; config.ctr_el0 = 0x8444c004; + config.cntfrq_el0 = 19200000; // Value from fusee. // Unpredictable instructions config.define_unpredictable_behaviour = true; diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp index 030333077..28268e112 100644 --- a/src/core/hle/kernel/svc.cpp +++ b/src/core/hle/kernel/svc.cpp @@ -254,11 +254,52 @@ static ResultCode SetMemoryPermission(VAddr addr, u64 size, u32 prot) { return vm_manager.ReprotectRange(addr, size, converted_permissions); } -static ResultCode SetMemoryAttribute(VAddr addr, u64 size, u32 state0, u32 state1) { - LOG_WARNING(Kernel_SVC, - "(STUBBED) called, addr=0x{:X}, size=0x{:X}, state0=0x{:X}, state1=0x{:X}", addr, - size, state0, state1); - return RESULT_SUCCESS; +static ResultCode SetMemoryAttribute(VAddr address, u64 size, u32 mask, u32 attribute) { + LOG_DEBUG(Kernel_SVC, + "called, address=0x{:016X}, size=0x{:X}, mask=0x{:08X}, attribute=0x{:08X}", address, + size, mask, attribute); + + if (!Common::Is4KBAligned(address)) { + LOG_ERROR(Kernel_SVC, "Address not page aligned (0x{:016X})", address); + return ERR_INVALID_ADDRESS; + } + + if (size == 0 || !Common::Is4KBAligned(size)) { + LOG_ERROR(Kernel_SVC, "Invalid size (0x{:X}). Size must be non-zero and page aligned.", + size); + return ERR_INVALID_ADDRESS; + } + + if (!IsValidAddressRange(address, size)) { + LOG_ERROR(Kernel_SVC, "Address range overflowed (Address: 0x{:016X}, Size: 0x{:016X})", + address, size); + return ERR_INVALID_ADDRESS_STATE; + } + + const auto mem_attribute = static_cast<MemoryAttribute>(attribute); + const auto mem_mask = static_cast<MemoryAttribute>(mask); + const auto attribute_with_mask = mem_attribute | mem_mask; + + if (attribute_with_mask != mem_mask) { + LOG_ERROR(Kernel_SVC, + "Memory attribute doesn't match the given mask (Attribute: 0x{:X}, Mask: {:X}", + attribute, mask); + return ERR_INVALID_COMBINATION; + } + + if ((attribute_with_mask | MemoryAttribute::Uncached) != MemoryAttribute::Uncached) { + LOG_ERROR(Kernel_SVC, "Specified attribute isn't equal to MemoryAttributeUncached (8)."); + return ERR_INVALID_COMBINATION; + } + + auto& vm_manager = Core::CurrentProcess()->VMManager(); + if (!IsInsideAddressSpace(vm_manager, address, size)) { + LOG_ERROR(Kernel_SVC, + "Given address (0x{:016X}) is outside the bounds of the address space.", address); + return ERR_INVALID_ADDRESS_STATE; + } + + return vm_manager.SetMemoryAttribute(address, size, mem_mask, mem_attribute); } /// Maps a memory range into a different range. diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp index 4ffb76818..63f8923fd 100644 --- a/src/core/hle/kernel/thread.cpp +++ b/src/core/hle/kernel/thread.cpp @@ -158,6 +158,9 @@ static void ResetThreadContext(Core::ARM_Interface::ThreadContext& context, VAdd context.cpu_registers[0] = arg; context.pc = entry_point; context.sp = stack_top; + // TODO(merry): Perform a hardware test to determine the below value. + // AHP = 0, DN = 1, FTZ = 1, RMode = Round towards zero + context.fpcr = 0x03C00000; } ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name, VAddr entry_point, diff --git a/src/core/hle/kernel/vm_manager.cpp b/src/core/hle/kernel/vm_manager.cpp index d3b55a51e..f39e096ca 100644 --- a/src/core/hle/kernel/vm_manager.cpp +++ b/src/core/hle/kernel/vm_manager.cpp @@ -37,7 +37,7 @@ static const char* GetMemoryStateName(MemoryState state) { bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const { ASSERT(base + size == next.base); - if (permissions != next.permissions || meminfo_state != next.meminfo_state || + if (permissions != next.permissions || state != next.state || attribute != next.attribute || type != next.type) { return false; } @@ -115,7 +115,7 @@ ResultVal<VMManager::VMAHandle> VMManager::MapMemoryBlock(VAddr target, final_vma.type = VMAType::AllocatedMemoryBlock; final_vma.permissions = VMAPermission::ReadWrite; - final_vma.meminfo_state = state; + final_vma.state = state; final_vma.backing_block = std::move(block); final_vma.offset = offset; UpdatePageTableForVMA(final_vma); @@ -140,7 +140,7 @@ ResultVal<VMManager::VMAHandle> VMManager::MapBackingMemory(VAddr target, u8* me final_vma.type = VMAType::BackingMemory; final_vma.permissions = VMAPermission::ReadWrite; - final_vma.meminfo_state = state; + final_vma.state = state; final_vma.backing_memory = memory; UpdatePageTableForVMA(final_vma); @@ -177,7 +177,7 @@ ResultVal<VMManager::VMAHandle> VMManager::MapMMIO(VAddr target, PAddr paddr, u6 final_vma.type = VMAType::MMIO; final_vma.permissions = VMAPermission::ReadWrite; - final_vma.meminfo_state = state; + final_vma.state = state; final_vma.paddr = paddr; final_vma.mmio_handler = std::move(mmio_handler); UpdatePageTableForVMA(final_vma); @@ -189,7 +189,7 @@ VMManager::VMAIter VMManager::Unmap(VMAIter vma_handle) { VirtualMemoryArea& vma = vma_handle->second; vma.type = VMAType::Free; vma.permissions = VMAPermission::None; - vma.meminfo_state = MemoryState::Unmapped; + vma.state = MemoryState::Unmapped; vma.backing_block = nullptr; vma.offset = 0; @@ -308,9 +308,10 @@ MemoryInfo VMManager::QueryMemory(VAddr address) const { if (IsValidHandle(vma)) { memory_info.base_address = vma->second.base; + memory_info.attributes = ToSvcMemoryAttribute(vma->second.attribute); memory_info.permission = static_cast<u32>(vma->second.permissions); memory_info.size = vma->second.size; - memory_info.state = ToSvcMemoryState(vma->second.meminfo_state); + memory_info.state = ToSvcMemoryState(vma->second.state); } else { memory_info.base_address = address_space_end; memory_info.permission = static_cast<u32>(VMAPermission::None); @@ -321,6 +322,34 @@ MemoryInfo VMManager::QueryMemory(VAddr address) const { return memory_info; } +ResultCode VMManager::SetMemoryAttribute(VAddr address, u64 size, MemoryAttribute mask, + MemoryAttribute attribute) { + constexpr auto ignore_mask = MemoryAttribute::Uncached | MemoryAttribute::DeviceMapped; + constexpr auto attribute_mask = ~ignore_mask; + + const auto result = CheckRangeState( + address, size, MemoryState::FlagUncached, MemoryState::FlagUncached, VMAPermission::None, + VMAPermission::None, attribute_mask, MemoryAttribute::None, ignore_mask); + + if (result.Failed()) { + return result.Code(); + } + + const auto [prev_state, prev_permissions, prev_attributes] = *result; + const auto new_attribute = (prev_attributes & ~mask) | (mask & attribute); + + const auto carve_result = CarveVMARange(address, size); + if (carve_result.Failed()) { + return carve_result.Code(); + } + + auto vma_iter = *carve_result; + vma_iter->second.attribute = new_attribute; + + MergeAdjacent(vma_iter); + return RESULT_SUCCESS; +} + ResultCode VMManager::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size, MemoryState state) { const auto vma = FindVMA(src_addr); @@ -364,7 +393,7 @@ void VMManager::LogLayout() const { (u8)vma.permissions & (u8)VMAPermission::Read ? 'R' : '-', (u8)vma.permissions & (u8)VMAPermission::Write ? 'W' : '-', (u8)vma.permissions & (u8)VMAPermission::Execute ? 'X' : '-', - GetMemoryStateName(vma.meminfo_state)); + GetMemoryStateName(vma.state)); } } @@ -591,6 +620,66 @@ void VMManager::ClearPageTable() { Memory::PageType::Unmapped); } +VMManager::CheckResults VMManager::CheckRangeState(VAddr address, u64 size, MemoryState state_mask, + MemoryState state, VMAPermission permission_mask, + VMAPermission permissions, + MemoryAttribute attribute_mask, + MemoryAttribute attribute, + MemoryAttribute ignore_mask) const { + auto iter = FindVMA(address); + + // If we don't have a valid VMA handle at this point, then it means this is + // being called with an address outside of the address space, which is definitely + // indicative of a bug, as this function only operates on mapped memory regions. + DEBUG_ASSERT(IsValidHandle(iter)); + + const VAddr end_address = address + size - 1; + const MemoryAttribute initial_attributes = iter->second.attribute; + const VMAPermission initial_permissions = iter->second.permissions; + const MemoryState initial_state = iter->second.state; + + while (true) { + // The iterator should be valid throughout the traversal. Hitting the end of + // the mapped VMA regions is unquestionably indicative of a bug. + DEBUG_ASSERT(IsValidHandle(iter)); + + const auto& vma = iter->second; + + if (vma.state != initial_state) { + return ERR_INVALID_ADDRESS_STATE; + } + + if ((vma.state & state_mask) != state) { + return ERR_INVALID_ADDRESS_STATE; + } + + if (vma.permissions != initial_permissions) { + return ERR_INVALID_ADDRESS_STATE; + } + + if ((vma.permissions & permission_mask) != permissions) { + return ERR_INVALID_ADDRESS_STATE; + } + + if ((vma.attribute | ignore_mask) != (initial_attributes | ignore_mask)) { + return ERR_INVALID_ADDRESS_STATE; + } + + if ((vma.attribute & attribute_mask) != attribute) { + return ERR_INVALID_ADDRESS_STATE; + } + + if (end_address <= vma.EndAddress()) { + break; + } + + ++iter; + } + + return MakeResult( + std::make_tuple(initial_state, initial_permissions, initial_attributes & ~ignore_mask)); +} + u64 VMManager::GetTotalMemoryUsage() const { LOG_WARNING(Kernel, "(STUBBED) called"); return 0xF8000000; diff --git a/src/core/hle/kernel/vm_manager.h b/src/core/hle/kernel/vm_manager.h index 10bacac3e..6091533bc 100644 --- a/src/core/hle/kernel/vm_manager.h +++ b/src/core/hle/kernel/vm_manager.h @@ -6,6 +6,7 @@ #include <map> #include <memory> +#include <tuple> #include <vector> #include "common/common_types.h" #include "core/hle/result.h" @@ -43,6 +44,88 @@ enum class VMAPermission : u8 { ReadWriteExecute = Read | Write | Execute, }; +constexpr VMAPermission operator|(VMAPermission lhs, VMAPermission rhs) { + return static_cast<VMAPermission>(u32(lhs) | u32(rhs)); +} + +constexpr VMAPermission operator&(VMAPermission lhs, VMAPermission rhs) { + return static_cast<VMAPermission>(u32(lhs) & u32(rhs)); +} + +constexpr VMAPermission operator^(VMAPermission lhs, VMAPermission rhs) { + return static_cast<VMAPermission>(u32(lhs) ^ u32(rhs)); +} + +constexpr VMAPermission operator~(VMAPermission permission) { + return static_cast<VMAPermission>(~u32(permission)); +} + +constexpr VMAPermission& operator|=(VMAPermission& lhs, VMAPermission rhs) { + lhs = lhs | rhs; + return lhs; +} + +constexpr VMAPermission& operator&=(VMAPermission& lhs, VMAPermission rhs) { + lhs = lhs & rhs; + return lhs; +} + +constexpr VMAPermission& operator^=(VMAPermission& lhs, VMAPermission rhs) { + lhs = lhs ^ rhs; + return lhs; +} + +/// Attribute flags that can be applied to a VMA +enum class MemoryAttribute : u32 { + Mask = 0xFF, + + /// No particular qualities + None = 0, + /// Memory locked/borrowed for use. e.g. This would be used by transfer memory. + Locked = 1, + /// Memory locked for use by IPC-related internals. + LockedForIPC = 2, + /// Mapped as part of the device address space. + DeviceMapped = 4, + /// Uncached memory + Uncached = 8, +}; + +constexpr MemoryAttribute operator|(MemoryAttribute lhs, MemoryAttribute rhs) { + return static_cast<MemoryAttribute>(u32(lhs) | u32(rhs)); +} + +constexpr MemoryAttribute operator&(MemoryAttribute lhs, MemoryAttribute rhs) { + return static_cast<MemoryAttribute>(u32(lhs) & u32(rhs)); +} + +constexpr MemoryAttribute operator^(MemoryAttribute lhs, MemoryAttribute rhs) { + return static_cast<MemoryAttribute>(u32(lhs) ^ u32(rhs)); +} + +constexpr MemoryAttribute operator~(MemoryAttribute attribute) { + return static_cast<MemoryAttribute>(~u32(attribute)); +} + +constexpr MemoryAttribute& operator|=(MemoryAttribute& lhs, MemoryAttribute rhs) { + lhs = lhs | rhs; + return lhs; +} + +constexpr MemoryAttribute& operator&=(MemoryAttribute& lhs, MemoryAttribute rhs) { + lhs = lhs & rhs; + return lhs; +} + +constexpr MemoryAttribute& operator^=(MemoryAttribute& lhs, MemoryAttribute rhs) { + lhs = lhs ^ rhs; + return lhs; +} + +constexpr u32 ToSvcMemoryAttribute(MemoryAttribute attribute) { + return static_cast<u32>(attribute & MemoryAttribute::Mask); +} + // clang-format off /// Represents memory states and any relevant flags, as used by the kernel. /// svcQueryMemory interprets these by masking away all but the first eight @@ -174,6 +257,16 @@ struct PageInfo { * also backed by a single host memory allocation. */ struct VirtualMemoryArea { + /// Gets the starting (base) address of this VMA. + VAddr StartAddress() const { + return base; + } + + /// Gets the ending address of this VMA. + VAddr EndAddress() const { + return base + size - 1; + } + /// Virtual base address of the region. VAddr base = 0; /// Size of the region. @@ -181,8 +274,8 @@ struct VirtualMemoryArea { VMAType type = VMAType::Free; VMAPermission permissions = VMAPermission::None; - /// Tag returned by svcQueryMemory. Not otherwise used. - MemoryState meminfo_state = MemoryState::Unmapped; + MemoryState state = MemoryState::Unmapped; + MemoryAttribute attribute = MemoryAttribute::None; // Settings for type = AllocatedMemoryBlock /// Memory block backing this VMA. @@ -299,6 +392,19 @@ public: /// MemoryInfo QueryMemory(VAddr address) const; + /// Sets an attribute across the given address range. + /// + /// @param address The starting address + /// @param size The size of the range to set the attribute on. + /// @param mask The attribute mask + /// @param attribute The attribute to set across the given address range + /// + /// @returns RESULT_SUCCESS if successful + /// @returns ERR_INVALID_ADDRESS_STATE if the attribute could not be set. + /// + ResultCode SetMemoryAttribute(VAddr address, u64 size, MemoryAttribute mask, + MemoryAttribute attribute); + /** * Scans all VMAs and updates the page table range of any that use the given vector as backing * memory. This should be called after any operation that causes reallocation of the vector. @@ -435,6 +541,35 @@ private: /// Clears out the page table void ClearPageTable(); + using CheckResults = ResultVal<std::tuple<MemoryState, VMAPermission, MemoryAttribute>>; + + /// Checks if an address range adheres to the specified states provided. + /// + /// @param address The starting address of the address range. + /// @param size The size of the address range. + /// @param state_mask The memory state mask. + /// @param state The state to compare the individual VMA states against, + /// which is done in the form of: (vma.state & state_mask) != state. + /// @param permission_mask The memory permissions mask. + /// @param permissions The permission to compare the individual VMA permissions against, + /// which is done in the form of: + /// (vma.permission & permission_mask) != permission. + /// @param attribute_mask The memory attribute mask. + /// @param attribute The memory attributes to compare the individual VMA attributes + /// against, which is done in the form of: + /// (vma.attributes & attribute_mask) != attribute. + /// @param ignore_mask The memory attributes to ignore during the check. + /// + /// @returns If successful, returns a tuple containing the memory attributes + /// (with ignored bits specified by ignore_mask unset), memory permissions, and + /// memory state across the memory range. + /// @returns If not successful, returns ERR_INVALID_ADDRESS_STATE. + /// + CheckResults CheckRangeState(VAddr address, u64 size, MemoryState state_mask, MemoryState state, + VMAPermission permission_mask, VMAPermission permissions, + MemoryAttribute attribute_mask, MemoryAttribute attribute, + MemoryAttribute ignore_mask) const; + /** * A map covering the entirety of the managed address space, keyed by the `base` field of each * VMA. It must always be modified by splitting or merging VMAs, so that the invariant diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp index 3a7b6da84..27c31aad2 100644 --- a/src/core/hle/service/am/am.cpp +++ b/src/core/hle/service/am/am.cpp @@ -71,10 +71,13 @@ IWindowController::IWindowController() : ServiceFramework("IWindowController") { IWindowController::~IWindowController() = default; void IWindowController::GetAppletResourceUserId(Kernel::HLERequestContext& ctx) { - LOG_WARNING(Service_AM, "(STUBBED) called"); + const u64 process_id = Core::System::GetInstance().Kernel().CurrentProcess()->GetProcessID(); + + LOG_DEBUG(Service_AM, "called. Process ID=0x{:016X}", process_id); + IPC::ResponseBuilder rb{ctx, 4}; rb.Push(RESULT_SUCCESS); - rb.Push<u64>(0); + rb.Push<u64>(process_id); } void IWindowController::AcquireForegroundRights(Kernel::HLERequestContext& ctx) { @@ -565,7 +568,6 @@ private: void GetAppletStateChangedEvent(Kernel::HLERequestContext& ctx) { LOG_DEBUG(Service_AM, "called"); - applet->GetBroker().SignalStateChanged(); const auto event = applet->GetBroker().GetStateChangedEvent(); IPC::ResponseBuilder rb{ctx, 2, 1}; diff --git a/src/core/hle/service/am/applets/applets.cpp b/src/core/hle/service/am/applets/applets.cpp index 47da35537..7698ca819 100644 --- a/src/core/hle/service/am/applets/applets.cpp +++ b/src/core/hle/service/am/applets/applets.cpp @@ -16,11 +16,11 @@ namespace Service::AM::Applets { AppletDataBroker::AppletDataBroker() { auto& kernel = Core::System::GetInstance().Kernel(); state_changed_event = Kernel::WritableEvent::CreateEventPair( - kernel, Kernel::ResetType::OneShot, "ILibraryAppletAccessor:StateChangedEvent"); + kernel, Kernel::ResetType::Sticky, "ILibraryAppletAccessor:StateChangedEvent"); pop_out_data_event = Kernel::WritableEvent::CreateEventPair( - kernel, Kernel::ResetType::OneShot, "ILibraryAppletAccessor:PopDataOutEvent"); + kernel, Kernel::ResetType::Sticky, "ILibraryAppletAccessor:PopDataOutEvent"); pop_interactive_out_data_event = Kernel::WritableEvent::CreateEventPair( - kernel, Kernel::ResetType::OneShot, "ILibraryAppletAccessor:PopInteractiveDataOutEvent"); + kernel, Kernel::ResetType::Sticky, "ILibraryAppletAccessor:PopInteractiveDataOutEvent"); } AppletDataBroker::~AppletDataBroker() = default; diff --git a/src/core/hle/service/am/applets/software_keyboard.cpp b/src/core/hle/service/am/applets/software_keyboard.cpp index 981bdec51..f255f74b5 100644 --- a/src/core/hle/service/am/applets/software_keyboard.cpp +++ b/src/core/hle/service/am/applets/software_keyboard.cpp @@ -146,11 +146,10 @@ void SoftwareKeyboard::WriteText(std::optional<std::u16string> text) { if (complete) { broker.PushNormalDataFromApplet(IStorage{output_main}); + broker.SignalStateChanged(); } else { broker.PushInteractiveDataFromApplet(IStorage{output_sub}); } - - broker.SignalStateChanged(); } else { output_main[0] = 1; complete = true; diff --git a/src/core/hle/service/nfp/nfp.cpp b/src/core/hle/service/nfp/nfp.cpp index d5df112a0..a7bed0040 100644 --- a/src/core/hle/service/nfp/nfp.cpp +++ b/src/core/hle/service/nfp/nfp.cpp @@ -317,8 +317,8 @@ private: } bool has_attached_handle{}; - const u64 device_handle{Common::MakeMagic('Y', 'U', 'Z', 'U')}; - const u32 npad_id{0}; // Player 1 controller + const u64 device_handle{0}; // Npad device 1 + const u32 npad_id{0}; // Player 1 controller State state{State::NonInitialized}; DeviceState device_state{DeviceState::Initialized}; Kernel::EventPair deactivate_event; diff --git a/src/core/hle/service/sm/sm.cpp b/src/core/hle/service/sm/sm.cpp index d73530086..142929124 100644 --- a/src/core/hle/service/sm/sm.cpp +++ b/src/core/hle/service/sm/sm.cpp @@ -145,12 +145,13 @@ void SM::RegisterService(Kernel::HLERequestContext& ctx) { const std::string name(name_buf.begin(), end); - const auto unk_bool = static_cast<bool>(rp.PopRaw<u32>()); - const auto session_count = rp.PopRaw<u32>(); + const auto is_light = static_cast<bool>(rp.PopRaw<u32>()); + const auto max_session_count = rp.PopRaw<u32>(); - LOG_DEBUG(Service_SM, "called with unk_bool={}", unk_bool); + LOG_DEBUG(Service_SM, "called with name={}, max_session_count={}, is_light={}", name, + max_session_count, is_light); - auto handle = service_manager->RegisterService(name, session_count); + auto handle = service_manager->RegisterService(name, max_session_count); if (handle.Failed()) { LOG_ERROR(Service_SM, "failed to register service with error_code={:08X}", handle.Code().raw); diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h index 5ea094e64..2efeb6e1a 100644 --- a/src/video_core/engines/shader_bytecode.h +++ b/src/video_core/engines/shader_bytecode.h @@ -575,7 +575,7 @@ union Instruction { union { BitField<39, 2, u64> tab5cb8_2; - BitField<41, 3, u64> tab5c68_1; + BitField<41, 3, u64> postfactor; BitField<44, 2, u64> tab5c68_0; BitField<48, 1, u64> negate_b; } fmul; @@ -609,7 +609,7 @@ union Instruction { BitField<31, 1, u64> negate_b; BitField<30, 1, u64> abs_b; - BitField<47, 2, HalfType> type_b; + BitField<28, 2, HalfType> type_b; BitField<35, 2, HalfType> type_c; } alu_half; diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index a5cfa0070..a7d8c404d 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -1591,23 +1591,21 @@ private: process_mode == Tegra::Shader::TextureProcessMode::LL || process_mode == Tegra::Shader::TextureProcessMode::LLA; + // LOD selection (either via bias or explicit textureLod) not supported in GL for + // sampler2DArrayShadow and samplerCubeArrayShadow. const bool gl_lod_supported = !( (texture_type == Tegra::Shader::TextureType::Texture2D && is_array && depth_compare) || - (texture_type == Tegra::Shader::TextureType::TextureCube && !is_array && - depth_compare)); + (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && depth_compare)); const std::string read_method = lod_needed && gl_lod_supported ? "textureLod(" : "texture("; std::string texture = read_method + sampler + ", coord"; - if (process_mode != Tegra::Shader::TextureProcessMode::None) { + UNIMPLEMENTED_IF(process_mode != Tegra::Shader::TextureProcessMode::None && + !gl_lod_supported); + + if (process_mode != Tegra::Shader::TextureProcessMode::None && gl_lod_supported) { if (process_mode == Tegra::Shader::TextureProcessMode::LZ) { - if (gl_lod_supported) { - texture += ", 0"; - } else { - // Lod 0 is emulated by a big negative bias - // in scenarios that are not supported by glsl - texture += ", -1000"; - } + texture += ", 0.0"; } else { // If present, lod or bias are always stored in the register indexed by the // gpr20 @@ -1645,15 +1643,15 @@ private: if (depth_compare && !is_array && texture_type == Tegra::Shader::TextureType::Texture1D) { coord += ",0.0"; } + if (is_array) { + coord += ',' + regs.GetRegisterAsInteger(array_register); + } if (depth_compare) { // Depth is always stored in the register signaled by gpr20 // or in the next register if lod or bias are used const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0); coord += ',' + regs.GetRegisterAsFloat(depth_register); } - if (is_array) { - coord += ',' + regs.GetRegisterAsInteger(array_register); - } coord += ");"; return std::make_pair( coord, GetTextureCode(instr, texture_type, process_mode, depth_compare, is_array, 0)); @@ -1686,15 +1684,15 @@ private: } } + if (is_array) { + coord += ',' + regs.GetRegisterAsInteger(array_register); + } if (depth_compare) { // Depth is always stored in the register signaled by gpr20 // or in the next register if lod or bias are used const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0); coord += ',' + regs.GetRegisterAsFloat(depth_register); } - if (is_array) { - coord += ',' + regs.GetRegisterAsInteger(array_register); - } coord += ");"; return std::make_pair(coord, @@ -1867,9 +1865,6 @@ private: UNIMPLEMENTED_IF_MSG(instr.fmul.tab5cb8_2 != 0, "FMUL tab5cb8_2({}) is not implemented", instr.fmul.tab5cb8_2.Value()); - UNIMPLEMENTED_IF_MSG(instr.fmul.tab5c68_1 != 0, - "FMUL tab5cb8_1({}) is not implemented", - instr.fmul.tab5c68_1.Value()); UNIMPLEMENTED_IF_MSG( instr.fmul.tab5c68_0 != 1, "FMUL tab5cb8_0({}) is not implemented", instr.fmul.tab5c68_0 @@ -1879,7 +1874,26 @@ private: op_b = GetOperandAbsNeg(op_b, false, instr.fmul.negate_b); - regs.SetRegisterToFloat(instr.gpr0, 0, op_a + " * " + op_b, 1, 1, + std::string postfactor_op; + if (instr.fmul.postfactor != 0) { + s8 postfactor = static_cast<s8>(instr.fmul.postfactor); + + // postfactor encoded as 3-bit 1's complement in instruction, + // interpreted with below logic. + if (postfactor >= 4) { + postfactor = 7 - postfactor; + } else { + postfactor = 0 - postfactor; + } + + if (postfactor > 0) { + postfactor_op = " * " + std::to_string(1 << postfactor); + } else { + postfactor_op = " / " + std::to_string(1 << -postfactor); + } + } + + regs.SetRegisterToFloat(instr.gpr0, 0, op_a + " * " + op_b + postfactor_op, 1, 1, instr.alu.saturate_d, 0, true); break; } |