diff options
Diffstat (limited to 'src')
34 files changed, 737 insertions, 1215 deletions
diff --git a/src/common/assert.h b/src/common/assert.h index b3ba35c0f..33060d865 100644 --- a/src/common/assert.h +++ b/src/common/assert.h @@ -52,8 +52,12 @@ assert_noinline_call(const Fn& fn) { #define DEBUG_ASSERT(_a_) ASSERT(_a_) #define DEBUG_ASSERT_MSG(_a_, ...) ASSERT_MSG(_a_, __VA_ARGS__) #else // not debug -#define DEBUG_ASSERT(_a_) -#define DEBUG_ASSERT_MSG(_a_, _desc_, ...) +#define DEBUG_ASSERT(_a_) \ + do { \ + } while (0) +#define DEBUG_ASSERT_MSG(_a_, _desc_, ...) \ + do { \ + } while (0) #endif #define UNIMPLEMENTED() ASSERT_MSG(false, "Unimplemented code!") diff --git a/src/common/uuid.cpp b/src/common/uuid.cpp index 18303a1e3..d7435a6e9 100644 --- a/src/common/uuid.cpp +++ b/src/common/uuid.cpp @@ -6,10 +6,64 @@ #include <fmt/format.h> +#include "common/assert.h" #include "common/uuid.h" namespace Common { +namespace { + +bool IsHexDigit(char c) { + return (c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F'); +} + +u8 HexCharToByte(char c) { + if (c >= '0' && c <= '9') { + return static_cast<u8>(c - '0'); + } + if (c >= 'a' && c <= 'f') { + return static_cast<u8>(c - 'a' + 10); + } + if (c >= 'A' && c <= 'F') { + return static_cast<u8>(c - 'A' + 10); + } + ASSERT_MSG(false, "{} is not a hexadecimal digit!", c); + return u8{0}; +} + +} // Anonymous namespace + +u128 HexStringToU128(std::string_view hex_string) { + const size_t length = hex_string.length(); + + // Detect "0x" prefix. + const bool has_0x_prefix = length > 2 && hex_string[0] == '0' && hex_string[1] == 'x'; + const size_t offset = has_0x_prefix ? 2 : 0; + + // Check length. + if (length > 32 + offset) { + ASSERT_MSG(false, "hex_string has more than 32 hexadecimal characters!"); + return INVALID_UUID; + } + + u64 lo = 0; + u64 hi = 0; + for (size_t i = 0; i < length - offset; ++i) { + const char c = hex_string[length - 1 - i]; + if (!IsHexDigit(c)) { + ASSERT_MSG(false, "{} is not a hexadecimal digit!", c); + return INVALID_UUID; + } + if (i < 16) { + lo |= u64{HexCharToByte(c)} << (i * 4); + } + if (i >= 16) { + hi |= u64{HexCharToByte(c)} << ((i - 16) * 4); + } + } + return u128{lo, hi}; +} + UUID UUID::Generate() { std::random_device device; std::mt19937 gen(device()); diff --git a/src/common/uuid.h b/src/common/uuid.h index 0ffa37e7c..2353179d8 100644 --- a/src/common/uuid.h +++ b/src/common/uuid.h @@ -5,6 +5,7 @@ #pragma once #include <string> +#include <string_view> #include "common/common_types.h" @@ -12,12 +13,30 @@ namespace Common { constexpr u128 INVALID_UUID{{0, 0}}; +/** + * Converts a hex string to a 128-bit unsigned integer. + * + * The hex string can be formatted in lowercase or uppercase, with or without the "0x" prefix. + * + * This function will assert and return INVALID_UUID under the following conditions: + * - If the hex string is more than 32 characters long + * - If the hex string contains non-hexadecimal characters + * + * @param hex_string Hexadecimal string + * + * @returns A 128-bit unsigned integer if successfully converted, INVALID_UUID otherwise. + */ +[[nodiscard]] u128 HexStringToU128(std::string_view hex_string); + struct UUID { // UUIDs which are 0 are considered invalid! u128 uuid; UUID() = default; constexpr explicit UUID(const u128& id) : uuid{id} {} constexpr explicit UUID(const u64 lo, const u64 hi) : uuid{{lo, hi}} {} + explicit UUID(std::string_view hex_string) { + uuid = HexStringToU128(hex_string); + } [[nodiscard]] constexpr explicit operator bool() const { return uuid != INVALID_UUID; @@ -50,3 +69,14 @@ struct UUID { static_assert(sizeof(UUID) == 16, "UUID is an invalid size!"); } // namespace Common + +namespace std { + +template <> +struct hash<Common::UUID> { + size_t operator()(const Common::UUID& uuid) const noexcept { + return uuid.uuid[1] ^ uuid.uuid[0]; + } +}; + +} // namespace std diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp index 1403a39d0..845de724d 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp @@ -166,8 +166,6 @@ NvResult nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vecto LOG_ERROR(Service_NVDRV, "failed to map size={}", object->size); } else { cmd_buffer.map_address = object->dma_map_addr; - AddBufferMap(object->dma_map_addr, object->size, object->addr, - object->status == nvmap::Object::Status::Allocated); } } std::memcpy(output.data(), ¶ms, sizeof(IoctlMapBuffer)); @@ -178,30 +176,11 @@ NvResult nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vecto } NvResult nvhost_nvdec_common::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output) { - IoctlMapBuffer params{}; - std::memcpy(¶ms, input.data(), sizeof(IoctlMapBuffer)); - std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries); - SliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer)); - - auto& gpu = system.GPU(); - - for (auto& cmd_buffer : cmd_buffer_handles) { - const auto object{nvmap_dev->GetObject(cmd_buffer.map_handle)}; - if (!object) { - LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmd_buffer.map_handle); - std::memcpy(output.data(), ¶ms, output.size()); - return NvResult::InvalidState; - } - if (const auto size{RemoveBufferMap(object->dma_map_addr)}; size) { - gpu.MemoryManager().Unmap(object->dma_map_addr, *size); - } else { - // This occurs quite frequently, however does not seem to impact functionality - LOG_DEBUG(Service_NVDRV, "invalid offset=0x{:X} dma=0x{:X}", object->addr, - object->dma_map_addr); - } - object->dma_map_addr = 0; - } + // This is intntionally stubbed. + // Skip unmapping buffers here, as to not break the continuity of the VP9 reference frame + // addresses, and risk invalidating data before the async GPU thread is done with it std::memset(output.data(), 0, output.size()); + LOG_DEBUG(Service_NVDRV, "(STUBBED) called"); return NvResult::Success; } @@ -212,33 +191,4 @@ NvResult nvhost_nvdec_common::SetSubmitTimeout(const std::vector<u8>& input, return NvResult::Success; } -std::optional<nvhost_nvdec_common::BufferMap> nvhost_nvdec_common::FindBufferMap( - GPUVAddr gpu_addr) const { - const auto it = std::find_if( - buffer_mappings.begin(), buffer_mappings.upper_bound(gpu_addr), [&](const auto& entry) { - return (gpu_addr >= entry.second.StartAddr() && gpu_addr < entry.second.EndAddr()); - }); - - ASSERT(it != buffer_mappings.end()); - return it->second; -} - -void nvhost_nvdec_common::AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr, - bool is_allocated) { - buffer_mappings.insert_or_assign(gpu_addr, BufferMap{gpu_addr, size, cpu_addr, is_allocated}); -} - -std::optional<std::size_t> nvhost_nvdec_common::RemoveBufferMap(GPUVAddr gpu_addr) { - const auto iter{buffer_mappings.find(gpu_addr)}; - if (iter == buffer_mappings.end()) { - return std::nullopt; - } - std::size_t size = 0; - if (iter->second.IsAllocated()) { - size = iter->second.Size(); - } - buffer_mappings.erase(iter); - return size; -} - } // namespace Service::Nvidia::Devices diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h index da10f5f41..af59f00d2 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h +++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h @@ -23,45 +23,6 @@ public: ~nvhost_nvdec_common() override; protected: - class BufferMap final { - public: - constexpr BufferMap() = default; - - constexpr BufferMap(GPUVAddr start_addr_, std::size_t size_) - : start_addr{start_addr_}, end_addr{start_addr_ + size_} {} - - constexpr BufferMap(GPUVAddr start_addr_, std::size_t size_, VAddr cpu_addr_, - bool is_allocated_) - : start_addr{start_addr_}, end_addr{start_addr_ + size_}, cpu_addr{cpu_addr_}, - is_allocated{is_allocated_} {} - - constexpr VAddr StartAddr() const { - return start_addr; - } - - constexpr VAddr EndAddr() const { - return end_addr; - } - - constexpr std::size_t Size() const { - return end_addr - start_addr; - } - - constexpr VAddr CpuAddr() const { - return cpu_addr; - } - - constexpr bool IsAllocated() const { - return is_allocated; - } - - private: - GPUVAddr start_addr{}; - GPUVAddr end_addr{}; - VAddr cpu_addr{}; - bool is_allocated{}; - }; - struct IoctlSetNvmapFD { s32_le nvmap_fd{}; }; @@ -154,17 +115,11 @@ protected: NvResult UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output); NvResult SetSubmitTimeout(const std::vector<u8>& input, std::vector<u8>& output); - std::optional<BufferMap> FindBufferMap(GPUVAddr gpu_addr) const; - void AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr, bool is_allocated); - std::optional<std::size_t> RemoveBufferMap(GPUVAddr gpu_addr); - s32_le nvmap_fd{}; u32_le submit_timeout{}; std::shared_ptr<nvmap> nvmap_dev; SyncpointManager& syncpoint_manager; std::array<u32, MaxSyncPoints> device_syncpoints{}; - // This is expected to be ordered, therefore we must use a map, not unordered_map - std::map<GPUVAddr, BufferMap> buffer_mappings; }; }; // namespace Devices } // namespace Service::Nvidia diff --git a/src/core/memory.cpp b/src/core/memory.cpp index f285c6f63..51c4dea26 100644 --- a/src/core/memory.cpp +++ b/src/core/memory.cpp @@ -4,8 +4,6 @@ #include <algorithm> #include <cstring> -#include <optional> -#include <utility> #include "common/assert.h" #include "common/atomic_ops.h" @@ -14,12 +12,10 @@ #include "common/page_table.h" #include "common/settings.h" #include "common/swap.h" -#include "core/arm/arm_interface.h" #include "core/core.h" #include "core/device_memory.h" #include "core/hle/kernel/k_page_table.h" #include "core/hle/kernel/k_process.h" -#include "core/hle/kernel/physical_memory.h" #include "core/memory.h" #include "video_core/gpu.h" @@ -62,17 +58,7 @@ struct Memory::Impl { } } - bool IsValidVirtualAddress(const Kernel::KProcess& process, const VAddr vaddr) const { - const auto& page_table = process.PageTable().PageTableImpl(); - const auto [pointer, type] = page_table.pointers[vaddr >> PAGE_BITS].PointerType(); - return pointer != nullptr || type == Common::PageType::RasterizerCachedMemory; - } - - bool IsValidVirtualAddress(VAddr vaddr) const { - return IsValidVirtualAddress(*system.CurrentProcess(), vaddr); - } - - u8* GetPointerFromRasterizerCachedMemory(VAddr vaddr) const { + [[nodiscard]] u8* GetPointerFromRasterizerCachedMemory(VAddr vaddr) const { const PAddr paddr{current_page_table->backing_addr[vaddr >> PAGE_BITS]}; if (!paddr) { @@ -82,18 +68,6 @@ struct Memory::Impl { return system.DeviceMemory().GetPointer(paddr) + vaddr; } - u8* GetPointer(const VAddr vaddr) const { - const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw(); - if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) { - return pointer + vaddr; - } - const auto type = Common::PageTable::PageInfo::ExtractType(raw_pointer); - if (type == Common::PageType::RasterizerCachedMemory) { - return GetPointerFromRasterizerCachedMemory(vaddr); - } - return nullptr; - } - u8 Read8(const VAddr addr) { return Read<u8>(addr); } @@ -179,7 +153,7 @@ struct Memory::Impl { std::string string; string.reserve(max_length); for (std::size_t i = 0; i < max_length; ++i) { - const char c = Read8(vaddr); + const char c = Read<s8>(vaddr); if (c == '\0') { break; } @@ -190,15 +164,14 @@ struct Memory::Impl { return string; } - void ReadBlock(const Kernel::KProcess& process, const VAddr src_addr, void* dest_buffer, - const std::size_t size) { + void WalkBlock(const Kernel::KProcess& process, const VAddr addr, const std::size_t size, + auto on_unmapped, auto on_memory, auto on_rasterizer, auto increment) { const auto& page_table = process.PageTable().PageTableImpl(); - std::size_t remaining_size = size; - std::size_t page_index = src_addr >> PAGE_BITS; - std::size_t page_offset = src_addr & PAGE_MASK; + std::size_t page_index = addr >> PAGE_BITS; + std::size_t page_offset = addr & PAGE_MASK; - while (remaining_size > 0) { + while (remaining_size) { const std::size_t copy_amount = std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size); const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset); @@ -206,22 +179,18 @@ struct Memory::Impl { const auto [pointer, type] = page_table.pointers[page_index].PointerType(); switch (type) { case Common::PageType::Unmapped: { - LOG_ERROR(HW_Memory, - "Unmapped ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})", - current_vaddr, src_addr, size); - std::memset(dest_buffer, 0, copy_amount); + on_unmapped(copy_amount, current_vaddr); break; } case Common::PageType::Memory: { DEBUG_ASSERT(pointer); - const u8* const src_ptr = pointer + page_offset + (page_index << PAGE_BITS); - std::memcpy(dest_buffer, src_ptr, copy_amount); + u8* mem_ptr = pointer + page_offset + (page_index << PAGE_BITS); + on_memory(copy_amount, mem_ptr); break; } case Common::PageType::RasterizerCachedMemory: { - const u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)}; - system.GPU().FlushRegion(current_vaddr, copy_amount); - std::memcpy(dest_buffer, host_ptr, copy_amount); + u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)}; + on_rasterizer(current_vaddr, copy_amount, host_ptr); break; } default: @@ -230,248 +199,122 @@ struct Memory::Impl { page_index++; page_offset = 0; - dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount; + increment(copy_amount); remaining_size -= copy_amount; } } - void ReadBlockUnsafe(const Kernel::KProcess& process, const VAddr src_addr, void* dest_buffer, - const std::size_t size) { - const auto& page_table = process.PageTable().PageTableImpl(); - - std::size_t remaining_size = size; - std::size_t page_index = src_addr >> PAGE_BITS; - std::size_t page_offset = src_addr & PAGE_MASK; - - while (remaining_size > 0) { - const std::size_t copy_amount = - std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size); - const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset); - - const auto [pointer, type] = page_table.pointers[page_index].PointerType(); - switch (type) { - case Common::PageType::Unmapped: { + template <bool UNSAFE> + void ReadBlockImpl(const Kernel::KProcess& process, const VAddr src_addr, void* dest_buffer, + const std::size_t size) { + WalkBlock( + process, src_addr, size, + [src_addr, size, &dest_buffer](const std::size_t copy_amount, + const VAddr current_vaddr) { LOG_ERROR(HW_Memory, "Unmapped ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})", current_vaddr, src_addr, size); std::memset(dest_buffer, 0, copy_amount); - break; - } - case Common::PageType::Memory: { - DEBUG_ASSERT(pointer); - const u8* const src_ptr = pointer + page_offset + (page_index << PAGE_BITS); + }, + [&dest_buffer](const std::size_t copy_amount, const u8* const src_ptr) { std::memcpy(dest_buffer, src_ptr, copy_amount); - break; - } - case Common::PageType::RasterizerCachedMemory: { - const u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)}; + }, + [&system = system, &dest_buffer](const VAddr current_vaddr, + const std::size_t copy_amount, + const u8* const host_ptr) { + if constexpr (!UNSAFE) { + system.GPU().FlushRegion(current_vaddr, copy_amount); + } std::memcpy(dest_buffer, host_ptr, copy_amount); - break; - } - default: - UNREACHABLE(); - } - - page_index++; - page_offset = 0; - dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount; - remaining_size -= copy_amount; - } + }, + [&dest_buffer](const std::size_t copy_amount) { + dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount; + }); } void ReadBlock(const VAddr src_addr, void* dest_buffer, const std::size_t size) { - ReadBlock(*system.CurrentProcess(), src_addr, dest_buffer, size); + ReadBlockImpl<false>(*system.CurrentProcess(), src_addr, dest_buffer, size); } void ReadBlockUnsafe(const VAddr src_addr, void* dest_buffer, const std::size_t size) { - ReadBlockUnsafe(*system.CurrentProcess(), src_addr, dest_buffer, size); + ReadBlockImpl<true>(*system.CurrentProcess(), src_addr, dest_buffer, size); } - void WriteBlock(const Kernel::KProcess& process, const VAddr dest_addr, const void* src_buffer, - const std::size_t size) { - const auto& page_table = process.PageTable().PageTableImpl(); - std::size_t remaining_size = size; - std::size_t page_index = dest_addr >> PAGE_BITS; - std::size_t page_offset = dest_addr & PAGE_MASK; - - while (remaining_size > 0) { - const std::size_t copy_amount = - std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size); - const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset); - - const auto [pointer, type] = page_table.pointers[page_index].PointerType(); - switch (type) { - case Common::PageType::Unmapped: { + template <bool UNSAFE> + void WriteBlockImpl(const Kernel::KProcess& process, const VAddr dest_addr, + const void* src_buffer, const std::size_t size) { + WalkBlock( + process, dest_addr, size, + [dest_addr, size](const std::size_t copy_amount, const VAddr current_vaddr) { LOG_ERROR(HW_Memory, "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})", current_vaddr, dest_addr, size); - break; - } - case Common::PageType::Memory: { - DEBUG_ASSERT(pointer); - u8* const dest_ptr = pointer + page_offset + (page_index << PAGE_BITS); + }, + [&src_buffer](const std::size_t copy_amount, u8* const dest_ptr) { std::memcpy(dest_ptr, src_buffer, copy_amount); - break; - } - case Common::PageType::RasterizerCachedMemory: { - u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)}; - system.GPU().InvalidateRegion(current_vaddr, copy_amount); - std::memcpy(host_ptr, src_buffer, copy_amount); - break; - } - default: - UNREACHABLE(); - } - - page_index++; - page_offset = 0; - src_buffer = static_cast<const u8*>(src_buffer) + copy_amount; - remaining_size -= copy_amount; - } - } - - void WriteBlockUnsafe(const Kernel::KProcess& process, const VAddr dest_addr, - const void* src_buffer, const std::size_t size) { - const auto& page_table = process.PageTable().PageTableImpl(); - std::size_t remaining_size = size; - std::size_t page_index = dest_addr >> PAGE_BITS; - std::size_t page_offset = dest_addr & PAGE_MASK; - - while (remaining_size > 0) { - const std::size_t copy_amount = - std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size); - const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset); - - const auto [pointer, type] = page_table.pointers[page_index].PointerType(); - switch (type) { - case Common::PageType::Unmapped: { - LOG_ERROR(HW_Memory, - "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})", - current_vaddr, dest_addr, size); - break; - } - case Common::PageType::Memory: { - DEBUG_ASSERT(pointer); - u8* const dest_ptr = pointer + page_offset + (page_index << PAGE_BITS); - std::memcpy(dest_ptr, src_buffer, copy_amount); - break; - } - case Common::PageType::RasterizerCachedMemory: { - u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)}; + }, + [&system = system, &src_buffer](const VAddr current_vaddr, + const std::size_t copy_amount, u8* const host_ptr) { + if constexpr (!UNSAFE) { + system.GPU().InvalidateRegion(current_vaddr, copy_amount); + } std::memcpy(host_ptr, src_buffer, copy_amount); - break; - } - default: - UNREACHABLE(); - } - - page_index++; - page_offset = 0; - src_buffer = static_cast<const u8*>(src_buffer) + copy_amount; - remaining_size -= copy_amount; - } + }, + [&src_buffer](const std::size_t copy_amount) { + src_buffer = static_cast<const u8*>(src_buffer) + copy_amount; + }); } void WriteBlock(const VAddr dest_addr, const void* src_buffer, const std::size_t size) { - WriteBlock(*system.CurrentProcess(), dest_addr, src_buffer, size); + WriteBlockImpl<false>(*system.CurrentProcess(), dest_addr, src_buffer, size); } void WriteBlockUnsafe(const VAddr dest_addr, const void* src_buffer, const std::size_t size) { - WriteBlockUnsafe(*system.CurrentProcess(), dest_addr, src_buffer, size); + WriteBlockImpl<true>(*system.CurrentProcess(), dest_addr, src_buffer, size); } void ZeroBlock(const Kernel::KProcess& process, const VAddr dest_addr, const std::size_t size) { - const auto& page_table = process.PageTable().PageTableImpl(); - std::size_t remaining_size = size; - std::size_t page_index = dest_addr >> PAGE_BITS; - std::size_t page_offset = dest_addr & PAGE_MASK; - - while (remaining_size > 0) { - const std::size_t copy_amount = - std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size); - const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset); - - const auto [pointer, type] = page_table.pointers[page_index].PointerType(); - switch (type) { - case Common::PageType::Unmapped: { + WalkBlock( + process, dest_addr, size, + [dest_addr, size](const std::size_t copy_amount, const VAddr current_vaddr) { LOG_ERROR(HW_Memory, "Unmapped ZeroBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})", current_vaddr, dest_addr, size); - break; - } - case Common::PageType::Memory: { - DEBUG_ASSERT(pointer); - u8* const dest_ptr = pointer + page_offset + (page_index << PAGE_BITS); + }, + [](const std::size_t copy_amount, u8* const dest_ptr) { std::memset(dest_ptr, 0, copy_amount); - break; - } - case Common::PageType::RasterizerCachedMemory: { - u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)}; + }, + [&system = system](const VAddr current_vaddr, const std::size_t copy_amount, + u8* const host_ptr) { system.GPU().InvalidateRegion(current_vaddr, copy_amount); std::memset(host_ptr, 0, copy_amount); - break; - } - default: - UNREACHABLE(); - } - - page_index++; - page_offset = 0; - remaining_size -= copy_amount; - } - } - - void ZeroBlock(const VAddr dest_addr, const std::size_t size) { - ZeroBlock(*system.CurrentProcess(), dest_addr, size); + }, + [](const std::size_t copy_amount) {}); } void CopyBlock(const Kernel::KProcess& process, VAddr dest_addr, VAddr src_addr, const std::size_t size) { - const auto& page_table = process.PageTable().PageTableImpl(); - std::size_t remaining_size = size; - std::size_t page_index = src_addr >> PAGE_BITS; - std::size_t page_offset = src_addr & PAGE_MASK; - - while (remaining_size > 0) { - const std::size_t copy_amount = - std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size); - const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset); - - const auto [pointer, type] = page_table.pointers[page_index].PointerType(); - switch (type) { - case Common::PageType::Unmapped: { + WalkBlock( + process, dest_addr, size, + [this, &process, &dest_addr, &src_addr, size](const std::size_t copy_amount, + const VAddr current_vaddr) { LOG_ERROR(HW_Memory, "Unmapped CopyBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})", current_vaddr, src_addr, size); ZeroBlock(process, dest_addr, copy_amount); - break; - } - case Common::PageType::Memory: { - DEBUG_ASSERT(pointer); - const u8* src_ptr = pointer + page_offset + (page_index << PAGE_BITS); - WriteBlock(process, dest_addr, src_ptr, copy_amount); - break; - } - case Common::PageType::RasterizerCachedMemory: { - const u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)}; + }, + [this, &process, &dest_addr](const std::size_t copy_amount, const u8* const src_ptr) { + WriteBlockImpl<false>(process, dest_addr, src_ptr, copy_amount); + }, + [this, &system = system, &process, &dest_addr]( + const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) { system.GPU().FlushRegion(current_vaddr, copy_amount); - WriteBlock(process, dest_addr, host_ptr, copy_amount); - break; - } - default: - UNREACHABLE(); - } - - page_index++; - page_offset = 0; - dest_addr += static_cast<VAddr>(copy_amount); - src_addr += static_cast<VAddr>(copy_amount); - remaining_size -= copy_amount; - } - } - - void CopyBlock(VAddr dest_addr, VAddr src_addr, std::size_t size) { - return CopyBlock(*system.CurrentProcess(), dest_addr, src_addr, size); + WriteBlockImpl<false>(process, dest_addr, host_ptr, copy_amount); + }, + [&dest_addr, &src_addr](const std::size_t copy_amount) { + dest_addr += static_cast<VAddr>(copy_amount); + src_addr += static_cast<VAddr>(copy_amount); + }); } void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) { @@ -514,7 +357,7 @@ struct Memory::Impl { } else { // Switch page type to uncached if now uncached switch (page_type) { - case Common::PageType::Unmapped: + case Common::PageType::Unmapped: // NOLINT(bugprone-branch-clone) // It is not necessary for a process to have this region mapped into its address // space, for example, a system module need not have a VRAM mapping. break; @@ -597,52 +440,68 @@ struct Memory::Impl { } } - /** - * Reads a particular data type out of memory at the given virtual address. - * - * @param vaddr The virtual address to read the data type from. - * - * @tparam T The data type to read out of memory. This type *must* be - * trivially copyable, otherwise the behavior of this function - * is undefined. - * - * @returns The instance of T read from the specified virtual address. - */ - template <typename T> - T Read(VAddr vaddr) { + [[nodiscard]] u8* GetPointerImpl(VAddr vaddr, auto on_unmapped, auto on_rasterizer) const { // AARCH64 masks the upper 16 bit of all memory accesses vaddr &= 0xffffffffffffLL; if (vaddr >= 1uLL << current_page_table->GetAddressSpaceBits()) { - LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:08X}", sizeof(T) * 8, vaddr); - return 0; + on_unmapped(); + return nullptr; } // Avoid adding any extra logic to this fast-path block const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw(); - if (const u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) { - T value; - std::memcpy(&value, &pointer[vaddr], sizeof(T)); - return value; + if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) { + return &pointer[vaddr]; } switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) { case Common::PageType::Unmapped: - LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:08X}", sizeof(T) * 8, vaddr); - return 0; + on_unmapped(); + return nullptr; case Common::PageType::Memory: - ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr); - break; + ASSERT_MSG(false, "Mapped memory page without a pointer @ 0x{:016X}", vaddr); + return nullptr; case Common::PageType::RasterizerCachedMemory: { - const u8* const host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)}; - system.GPU().FlushRegion(vaddr, sizeof(T)); - T value; - std::memcpy(&value, host_ptr, sizeof(T)); - return value; + u8* const host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)}; + on_rasterizer(); + return host_ptr; } default: UNREACHABLE(); } - return {}; + return nullptr; + } + + [[nodiscard]] u8* GetPointer(const VAddr vaddr) const { + return GetPointerImpl( + vaddr, [vaddr]() { LOG_ERROR(HW_Memory, "Unmapped GetPointer @ 0x{:016X}", vaddr); }, + []() {}); + } + + /** + * Reads a particular data type out of memory at the given virtual address. + * + * @param vaddr The virtual address to read the data type from. + * + * @tparam T The data type to read out of memory. This type *must* be + * trivially copyable, otherwise the behavior of this function + * is undefined. + * + * @returns The instance of T read from the specified virtual address. + */ + template <typename T> + T Read(VAddr vaddr) { + T result = 0; + const u8* const ptr = GetPointerImpl( + vaddr, + [vaddr]() { + LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:016X}", sizeof(T) * 8, vaddr); + }, + [&system = system, vaddr]() { system.GPU().FlushRegion(vaddr, sizeof(T)); }); + if (ptr) { + std::memcpy(&result, ptr, sizeof(T)); + } + return result; } /** @@ -656,110 +515,46 @@ struct Memory::Impl { */ template <typename T> void Write(VAddr vaddr, const T data) { - // AARCH64 masks the upper 16 bit of all memory accesses - vaddr &= 0xffffffffffffLL; - - if (vaddr >= 1uLL << current_page_table->GetAddressSpaceBits()) { - LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8, - static_cast<u32>(data), vaddr); - return; - } - - // Avoid adding any extra logic to this fast-path block - const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw(); - if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) { - std::memcpy(&pointer[vaddr], &data, sizeof(T)); - return; - } - switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) { - case Common::PageType::Unmapped: - LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8, - static_cast<u32>(data), vaddr); - return; - case Common::PageType::Memory: - ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr); - break; - case Common::PageType::RasterizerCachedMemory: { - u8* const host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)}; - system.GPU().InvalidateRegion(vaddr, sizeof(T)); - std::memcpy(host_ptr, &data, sizeof(T)); - break; - } - default: - UNREACHABLE(); + u8* const ptr = GetPointerImpl( + vaddr, + [vaddr, data]() { + LOG_ERROR(HW_Memory, "Unmapped Write{} @ 0x{:016X} = 0x{:016X}", sizeof(T) * 8, + vaddr, static_cast<u64>(data)); + }, + [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); }); + if (ptr) { + std::memcpy(ptr, &data, sizeof(T)); } } template <typename T> bool WriteExclusive(VAddr vaddr, const T data, const T expected) { - // AARCH64 masks the upper 16 bit of all memory accesses - vaddr &= 0xffffffffffffLL; - - if (vaddr >= 1uLL << current_page_table->GetAddressSpaceBits()) { - LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8, - static_cast<u32>(data), vaddr); - return true; - } - - const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw(); - if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) { - // NOTE: Avoid adding any extra logic to this fast-path block - const auto volatile_pointer = reinterpret_cast<volatile T*>(&pointer[vaddr]); + u8* const ptr = GetPointerImpl( + vaddr, + [vaddr, data]() { + LOG_ERROR(HW_Memory, "Unmapped WriteExclusive{} @ 0x{:016X} = 0x{:016X}", + sizeof(T) * 8, vaddr, static_cast<u64>(data)); + }, + [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); }); + if (ptr) { + const auto volatile_pointer = reinterpret_cast<volatile T*>(ptr); return Common::AtomicCompareAndSwap(volatile_pointer, data, expected); } - switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) { - case Common::PageType::Unmapped: - LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8, - static_cast<u32>(data), vaddr); - return true; - case Common::PageType::Memory: - ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr); - break; - case Common::PageType::RasterizerCachedMemory: { - u8* host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)}; - system.GPU().InvalidateRegion(vaddr, sizeof(T)); - auto* pointer = reinterpret_cast<volatile T*>(&host_ptr); - return Common::AtomicCompareAndSwap(pointer, data, expected); - } - default: - UNREACHABLE(); - } return true; } bool WriteExclusive128(VAddr vaddr, const u128 data, const u128 expected) { - // AARCH64 masks the upper 16 bit of all memory accesses - vaddr &= 0xffffffffffffLL; - - if (vaddr >= 1uLL << current_page_table->GetAddressSpaceBits()) { - LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8, - static_cast<u32>(data[0]), vaddr); - return true; - } - - const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw(); - if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) { - // NOTE: Avoid adding any extra logic to this fast-path block - const auto volatile_pointer = reinterpret_cast<volatile u64*>(&pointer[vaddr]); + u8* const ptr = GetPointerImpl( + vaddr, + [vaddr, data]() { + LOG_ERROR(HW_Memory, "Unmapped WriteExclusive128 @ 0x{:016X} = 0x{:016X}{:016X}", + vaddr, static_cast<u64>(data[1]), static_cast<u64>(data[0])); + }, + [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(u128)); }); + if (ptr) { + const auto volatile_pointer = reinterpret_cast<volatile u64*>(ptr); return Common::AtomicCompareAndSwap(volatile_pointer, data, expected); } - switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) { - case Common::PageType::Unmapped: - LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}{:016X}", sizeof(data) * 8, - static_cast<u64>(data[1]), static_cast<u64>(data[0]), vaddr); - return true; - case Common::PageType::Memory: - ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr); - break; - case Common::PageType::RasterizerCachedMemory: { - u8* host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)}; - system.GPU().InvalidateRegion(vaddr, sizeof(u128)); - auto* pointer = reinterpret_cast<volatile u64*>(&host_ptr); - return Common::AtomicCompareAndSwap(pointer, data, expected); - } - default: - UNREACHABLE(); - } return true; } @@ -789,12 +584,11 @@ void Memory::UnmapRegion(Common::PageTable& page_table, VAddr base, u64 size) { impl->UnmapRegion(page_table, base, size); } -bool Memory::IsValidVirtualAddress(const Kernel::KProcess& process, const VAddr vaddr) const { - return impl->IsValidVirtualAddress(process, vaddr); -} - bool Memory::IsValidVirtualAddress(const VAddr vaddr) const { - return impl->IsValidVirtualAddress(vaddr); + const Kernel::KProcess& process = *system.CurrentProcess(); + const auto& page_table = process.PageTable().PageTableImpl(); + const auto [pointer, type] = page_table.pointers[vaddr >> PAGE_BITS].PointerType(); + return pointer != nullptr || type == Common::PageType::RasterizerCachedMemory; } u8* Memory::GetPointer(VAddr vaddr) { @@ -863,64 +657,38 @@ std::string Memory::ReadCString(VAddr vaddr, std::size_t max_length) { void Memory::ReadBlock(const Kernel::KProcess& process, const VAddr src_addr, void* dest_buffer, const std::size_t size) { - impl->ReadBlock(process, src_addr, dest_buffer, size); + impl->ReadBlockImpl<false>(process, src_addr, dest_buffer, size); } void Memory::ReadBlock(const VAddr src_addr, void* dest_buffer, const std::size_t size) { impl->ReadBlock(src_addr, dest_buffer, size); } -void Memory::ReadBlockUnsafe(const Kernel::KProcess& process, const VAddr src_addr, - void* dest_buffer, const std::size_t size) { - impl->ReadBlockUnsafe(process, src_addr, dest_buffer, size); -} - void Memory::ReadBlockUnsafe(const VAddr src_addr, void* dest_buffer, const std::size_t size) { impl->ReadBlockUnsafe(src_addr, dest_buffer, size); } void Memory::WriteBlock(const Kernel::KProcess& process, VAddr dest_addr, const void* src_buffer, std::size_t size) { - impl->WriteBlock(process, dest_addr, src_buffer, size); + impl->WriteBlockImpl<false>(process, dest_addr, src_buffer, size); } void Memory::WriteBlock(const VAddr dest_addr, const void* src_buffer, const std::size_t size) { impl->WriteBlock(dest_addr, src_buffer, size); } -void Memory::WriteBlockUnsafe(const Kernel::KProcess& process, VAddr dest_addr, - const void* src_buffer, std::size_t size) { - impl->WriteBlockUnsafe(process, dest_addr, src_buffer, size); -} - void Memory::WriteBlockUnsafe(const VAddr dest_addr, const void* src_buffer, const std::size_t size) { impl->WriteBlockUnsafe(dest_addr, src_buffer, size); } -void Memory::ZeroBlock(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size) { - impl->ZeroBlock(process, dest_addr, size); -} - -void Memory::ZeroBlock(VAddr dest_addr, std::size_t size) { - impl->ZeroBlock(dest_addr, size); -} - void Memory::CopyBlock(const Kernel::KProcess& process, VAddr dest_addr, VAddr src_addr, const std::size_t size) { impl->CopyBlock(process, dest_addr, src_addr, size); } -void Memory::CopyBlock(VAddr dest_addr, VAddr src_addr, std::size_t size) { - impl->CopyBlock(dest_addr, src_addr, size); -} - void Memory::RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) { impl->RasterizerMarkRegionCached(vaddr, size, cached); } -bool IsKernelVirtualAddress(const VAddr vaddr) { - return KERNEL_REGION_VADDR <= vaddr && vaddr < KERNEL_REGION_END; -} - } // namespace Core::Memory diff --git a/src/core/memory.h b/src/core/memory.h index c91eeced9..b5721b740 100644 --- a/src/core/memory.h +++ b/src/core/memory.h @@ -39,11 +39,6 @@ enum : VAddr { /// Application stack DEFAULT_STACK_SIZE = 0x100000, - - /// Kernel Virtual Address Range - KERNEL_REGION_VADDR = 0xFFFFFF8000000000, - KERNEL_REGION_SIZE = 0x7FFFE00000, - KERNEL_REGION_END = KERNEL_REGION_VADDR + KERNEL_REGION_SIZE, }; /// Central class that handles all memory operations and state. @@ -56,7 +51,7 @@ public: Memory& operator=(const Memory&) = delete; Memory(Memory&&) = default; - Memory& operator=(Memory&&) = default; + Memory& operator=(Memory&&) = delete; /** * Resets the state of the Memory system. @@ -92,24 +87,13 @@ public: /** * Checks whether or not the supplied address is a valid virtual - * address for the given process. - * - * @param process The emulated process to check the address against. - * @param vaddr The virtual address to check the validity of. - * - * @returns True if the given virtual address is valid, false otherwise. - */ - bool IsValidVirtualAddress(const Kernel::KProcess& process, VAddr vaddr) const; - - /** - * Checks whether or not the supplied address is a valid virtual * address for the current process. * * @param vaddr The virtual address to check the validity of. * * @returns True if the given virtual address is valid, false otherwise. */ - bool IsValidVirtualAddress(VAddr vaddr) const; + [[nodiscard]] bool IsValidVirtualAddress(VAddr vaddr) const; /** * Gets a pointer to the given address. @@ -134,7 +118,7 @@ public: * @returns The pointer to the given address, if the address is valid. * If the address is not valid, nullptr will be returned. */ - const u8* GetPointer(VAddr vaddr) const; + [[nodiscard]] const u8* GetPointer(VAddr vaddr) const; template <typename T> const T* GetPointer(VAddr vaddr) const { @@ -328,27 +312,6 @@ public: std::size_t size); /** - * Reads a contiguous block of bytes from a specified process' address space. - * This unsafe version does not trigger GPU flushing. - * - * @param process The process to read the data from. - * @param src_addr The virtual address to begin reading from. - * @param dest_buffer The buffer to place the read bytes into. - * @param size The amount of data to read, in bytes. - * - * @note If a size of 0 is specified, then this function reads nothing and - * no attempts to access memory are made at all. - * - * @pre dest_buffer must be at least size bytes in length, otherwise a - * buffer overrun will occur. - * - * @post The range [dest_buffer, size) contains the read bytes from the - * process' address space. - */ - void ReadBlockUnsafe(const Kernel::KProcess& process, VAddr src_addr, void* dest_buffer, - std::size_t size); - - /** * Reads a contiguous block of bytes from the current process' address space. * * @param src_addr The virtual address to begin reading from. @@ -409,26 +372,6 @@ public: std::size_t size); /** - * Writes a range of bytes into a given process' address space at the specified - * virtual address. - * This unsafe version does not invalidate GPU Memory. - * - * @param process The process to write data into the address space of. - * @param dest_addr The destination virtual address to begin writing the data at. - * @param src_buffer The data to write into the process' address space. - * @param size The size of the data to write, in bytes. - * - * @post The address range [dest_addr, size) in the process' address space - * contains the data that was within src_buffer. - * - * @post If an attempt is made to write into an unmapped region of memory, the writes - * will be ignored and an error will be logged. - * - */ - void WriteBlockUnsafe(const Kernel::KProcess& process, VAddr dest_addr, const void* src_buffer, - std::size_t size); - - /** * Writes a range of bytes into the current process' address space at the specified * virtual address. * @@ -468,29 +411,6 @@ public: void WriteBlockUnsafe(VAddr dest_addr, const void* src_buffer, std::size_t size); /** - * Fills the specified address range within a process' address space with zeroes. - * - * @param process The process that will have a portion of its memory zeroed out. - * @param dest_addr The starting virtual address of the range to zero out. - * @param size The size of the address range to zero out, in bytes. - * - * @post The range [dest_addr, size) within the process' address space is - * filled with zeroes. - */ - void ZeroBlock(const Kernel::KProcess& process, VAddr dest_addr, std::size_t size); - - /** - * Fills the specified address range within the current process' address space with zeroes. - * - * @param dest_addr The starting virtual address of the range to zero out. - * @param size The size of the address range to zero out, in bytes. - * - * @post The range [dest_addr, size) within the current process' address space is - * filled with zeroes. - */ - void ZeroBlock(VAddr dest_addr, std::size_t size); - - /** * Copies data within a process' address space to another location within the * same address space. * @@ -506,19 +426,6 @@ public: std::size_t size); /** - * Copies data within the current process' address space to another location within the - * same address space. - * - * @param dest_addr The destination virtual address to begin copying the data into. - * @param src_addr The source virtual address to begin copying the data from. - * @param size The size of the data to copy, in bytes. - * - * @post The range [dest_addr, size) within the current process' address space - * contains the same data within the range [src_addr, size). - */ - void CopyBlock(VAddr dest_addr, VAddr src_addr, std::size_t size); - - /** * Marks each page within the specified address range as cached or uncached. * * @param vaddr The virtual address indicating the start of the address range. @@ -535,7 +442,4 @@ private: std::unique_ptr<Impl> impl; }; -/// Determines if the given VAddr is a kernel address -bool IsKernelVirtualAddress(VAddr vaddr); - } // namespace Core::Memory diff --git a/src/input_common/main.cpp b/src/input_common/main.cpp index 8de3d4520..ff23230f0 100644 --- a/src/input_common/main.cpp +++ b/src/input_common/main.cpp @@ -304,10 +304,10 @@ std::vector<std::unique_ptr<Polling::DevicePoller>> InputSubsystem::GetPollers([ } std::string GenerateKeyboardParam(int key_code) { - Common::ParamPackage param{ - {"engine", "keyboard"}, - {"code", std::to_string(key_code)}, - }; + Common::ParamPackage param; + param.Set("engine", "keyboard"); + param.Set("code", key_code); + param.Set("toggle", false); return param.Serialize(); } diff --git a/src/input_common/mouse/mouse_poller.cpp b/src/input_common/mouse/mouse_poller.cpp index efcdd85d2..090b26972 100644 --- a/src/input_common/mouse/mouse_poller.cpp +++ b/src/input_common/mouse/mouse_poller.cpp @@ -57,6 +57,7 @@ Common::ParamPackage MouseButtonFactory::GetNextInput() const { if (pad.button != MouseInput::MouseButton::Undefined) { params.Set("engine", "mouse"); params.Set("button", static_cast<u16>(pad.button)); + params.Set("toggle", false); return params; } } diff --git a/src/input_common/sdl/sdl_impl.cpp b/src/input_common/sdl/sdl_impl.cpp index 70a0ba09c..f1f950d8a 100644 --- a/src/input_common/sdl/sdl_impl.cpp +++ b/src/input_common/sdl/sdl_impl.cpp @@ -82,6 +82,12 @@ public: state.buttons.insert_or_assign(button, value); } + void PreSetButton(int button) { + if (!state.buttons.contains(button)) { + SetButton(button, false); + } + } + void SetMotion(SDL_ControllerSensorEvent event) { constexpr float gravity_constant = 9.80665f; std::lock_guard lock{mutex}; @@ -155,9 +161,16 @@ public: state.axes.insert_or_assign(axis, value); } - float GetAxis(int axis, float range) const { + void PreSetAxis(int axis) { + if (!state.axes.contains(axis)) { + SetAxis(axis, 0); + } + } + + float GetAxis(int axis, float range, float offset) const { std::lock_guard lock{mutex}; - return static_cast<float>(state.axes.at(axis)) / (32767.0f * range); + const float value = static_cast<float>(state.axes.at(axis)) / 32767.0f; + return (value + offset) / range; } bool RumblePlay(u16 amp_low, u16 amp_high) { @@ -174,9 +187,10 @@ public: return false; } - std::tuple<float, float> GetAnalog(int axis_x, int axis_y, float range) const { - float x = GetAxis(axis_x, range); - float y = GetAxis(axis_y, range); + std::tuple<float, float> GetAnalog(int axis_x, int axis_y, float range, float offset_x, + float offset_y) const { + float x = GetAxis(axis_x, range, offset_x); + float y = GetAxis(axis_y, range, offset_y); y = -y; // 3DS uses an y-axis inverse from SDL // Make sure the coordinates are in the unit circle, @@ -483,7 +497,7 @@ public: trigger_if_greater(trigger_if_greater_) {} bool GetStatus() const override { - const float axis_value = joystick->GetAxis(axis, 1.0f); + const float axis_value = joystick->GetAxis(axis, 1.0f, 0.0f); if (trigger_if_greater) { return axis_value > threshold; } @@ -500,12 +514,14 @@ private: class SDLAnalog final : public Input::AnalogDevice { public: explicit SDLAnalog(std::shared_ptr<SDLJoystick> joystick_, int axis_x_, int axis_y_, - bool invert_x_, bool invert_y_, float deadzone_, float range_) + bool invert_x_, bool invert_y_, float deadzone_, float range_, + float offset_x_, float offset_y_) : joystick(std::move(joystick_)), axis_x(axis_x_), axis_y(axis_y_), invert_x(invert_x_), - invert_y(invert_y_), deadzone(deadzone_), range(range_) {} + invert_y(invert_y_), deadzone(deadzone_), range(range_), offset_x(offset_x_), + offset_y(offset_y_) {} std::tuple<float, float> GetStatus() const override { - auto [x, y] = joystick->GetAnalog(axis_x, axis_y, range); + auto [x, y] = joystick->GetAnalog(axis_x, axis_y, range, offset_x, offset_y); const float r = std::sqrt((x * x) + (y * y)); if (invert_x) { x = -x; @@ -522,8 +538,8 @@ public: } std::tuple<float, float> GetRawStatus() const override { - const float x = joystick->GetAxis(axis_x, range); - const float y = joystick->GetAxis(axis_y, range); + const float x = joystick->GetAxis(axis_x, range, offset_x); + const float y = joystick->GetAxis(axis_y, range, offset_y); return {x, -y}; } @@ -555,6 +571,8 @@ private: const bool invert_y; const float deadzone; const float range; + const float offset_x; + const float offset_y; }; class SDLVibration final : public Input::VibrationDevice { @@ -621,7 +639,7 @@ public: trigger_if_greater(trigger_if_greater_) {} Input::MotionStatus GetStatus() const override { - const float axis_value = joystick->GetAxis(axis, 1.0f); + const float axis_value = joystick->GetAxis(axis, 1.0f, 0.0f); bool trigger = axis_value < threshold; if (trigger_if_greater) { trigger = axis_value > threshold; @@ -720,13 +738,13 @@ public: LOG_ERROR(Input, "Unknown direction {}", direction_name); } // This is necessary so accessing GetAxis with axis won't crash - joystick->SetAxis(axis, 0); + joystick->PreSetAxis(axis); return std::make_unique<SDLAxisButton>(joystick, axis, threshold, trigger_if_greater); } const int button = params.Get("button", 0); // This is necessary so accessing GetButton with button won't crash - joystick->SetButton(button, false); + joystick->PreSetButton(button); return std::make_unique<SDLButton>(joystick, button, toggle); } @@ -757,13 +775,15 @@ public: const std::string invert_y_value = params.Get("invert_y", "+"); const bool invert_x = invert_x_value == "-"; const bool invert_y = invert_y_value == "-"; + const float offset_x = params.Get("offset_x", 0.0f); + const float offset_y = params.Get("offset_y", 0.0f); auto joystick = state.GetSDLJoystickByGUID(guid, port); // This is necessary so accessing GetAxis with axis_x and axis_y won't crash - joystick->SetAxis(axis_x, 0); - joystick->SetAxis(axis_y, 0); + joystick->PreSetAxis(axis_x); + joystick->PreSetAxis(axis_y); return std::make_unique<SDLAnalog>(joystick, axis_x, axis_y, invert_x, invert_y, deadzone, - range); + range, offset_x, offset_y); } private: @@ -844,13 +864,13 @@ public: LOG_ERROR(Input, "Unknown direction {}", direction_name); } // This is necessary so accessing GetAxis with axis won't crash - joystick->SetAxis(axis, 0); + joystick->PreSetAxis(axis); return std::make_unique<SDLAxisMotion>(joystick, axis, threshold, trigger_if_greater); } const int button = params.Get("button", 0); // This is necessary so accessing GetButton with button won't crash - joystick->SetButton(button, false); + joystick->PreSetButton(button); return std::make_unique<SDLButtonMotion>(joystick, button); } @@ -995,6 +1015,7 @@ Common::ParamPackage BuildButtonParamPackageForButton(int port, std::string guid params.Set("port", port); params.Set("guid", std::move(guid)); params.Set("button", button); + params.Set("toggle", false); return params; } @@ -1134,13 +1155,15 @@ Common::ParamPackage BuildParamPackageForBinding(int port, const std::string& gu } Common::ParamPackage BuildParamPackageForAnalog(int port, const std::string& guid, int axis_x, - int axis_y) { + int axis_y, float offset_x, float offset_y) { Common::ParamPackage params; params.Set("engine", "sdl"); params.Set("port", port); params.Set("guid", guid); params.Set("axis_x", axis_x); params.Set("axis_y", axis_y); + params.Set("offset_x", offset_x); + params.Set("offset_y", offset_y); params.Set("invert_x", "+"); params.Set("invert_y", "+"); return params; @@ -1342,24 +1365,39 @@ AnalogMapping SDLState::GetAnalogMappingForDevice(const Common::ParamPackage& pa const auto& binding_left_y = SDL_GameControllerGetBindForAxis(controller, SDL_CONTROLLER_AXIS_LEFTY); if (params.Has("guid2")) { + joystick2->PreSetAxis(binding_left_x.value.axis); + joystick2->PreSetAxis(binding_left_y.value.axis); + const auto left_offset_x = -joystick2->GetAxis(binding_left_x.value.axis, 1.0f, 0); + const auto left_offset_y = -joystick2->GetAxis(binding_left_y.value.axis, 1.0f, 0); mapping.insert_or_assign( Settings::NativeAnalog::LStick, BuildParamPackageForAnalog(joystick2->GetPort(), joystick2->GetGUID(), - binding_left_x.value.axis, binding_left_y.value.axis)); + binding_left_x.value.axis, binding_left_y.value.axis, + left_offset_x, left_offset_y)); } else { + joystick->PreSetAxis(binding_left_x.value.axis); + joystick->PreSetAxis(binding_left_y.value.axis); + const auto left_offset_x = -joystick->GetAxis(binding_left_x.value.axis, 1.0f, 0); + const auto left_offset_y = -joystick->GetAxis(binding_left_y.value.axis, 1.0f, 0); mapping.insert_or_assign( Settings::NativeAnalog::LStick, BuildParamPackageForAnalog(joystick->GetPort(), joystick->GetGUID(), - binding_left_x.value.axis, binding_left_y.value.axis)); + binding_left_x.value.axis, binding_left_y.value.axis, + left_offset_x, left_offset_y)); } const auto& binding_right_x = SDL_GameControllerGetBindForAxis(controller, SDL_CONTROLLER_AXIS_RIGHTX); const auto& binding_right_y = SDL_GameControllerGetBindForAxis(controller, SDL_CONTROLLER_AXIS_RIGHTY); + joystick->PreSetAxis(binding_right_x.value.axis); + joystick->PreSetAxis(binding_right_y.value.axis); + const auto right_offset_x = -joystick->GetAxis(binding_right_x.value.axis, 1.0f, 0); + const auto right_offset_y = -joystick->GetAxis(binding_right_y.value.axis, 1.0f, 0); mapping.insert_or_assign(Settings::NativeAnalog::RStick, BuildParamPackageForAnalog(joystick->GetPort(), joystick->GetGUID(), binding_right_x.value.axis, - binding_right_y.value.axis)); + binding_right_y.value.axis, right_offset_x, + right_offset_y)); return mapping; } @@ -1563,8 +1601,9 @@ public: } if (const auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which)) { + // Set offset to zero since the joystick is not on center auto params = BuildParamPackageForAnalog(joystick->GetPort(), joystick->GetGUID(), - first_axis, axis); + first_axis, axis, 0, 0); first_axis = -1; return params; } diff --git a/src/video_core/command_classes/codecs/vp9.cpp b/src/video_core/command_classes/codecs/vp9.cpp index 902bc2a98..70030066a 100644 --- a/src/video_core/command_classes/codecs/vp9.cpp +++ b/src/video_core/command_classes/codecs/vp9.cpp @@ -11,6 +11,9 @@ namespace Tegra::Decoder { namespace { +constexpr u32 diff_update_probability = 252; +constexpr u32 frame_sync_code = 0x498342; + // Default compressed header probabilities once frame context resets constexpr Vp9EntropyProbs default_probs{ .y_mode_prob{ @@ -361,8 +364,7 @@ Vp9PictureInfo VP9::GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state) InsertEntropy(state.vp9_entropy_probs_offset, vp9_info.entropy); // surface_luma_offset[0:3] contains the address of the reference frame offsets in the following - // order: last, golden, altref, current. It may be worthwhile to track the updates done here - // to avoid buffering frame data needed for reference frame updating in the header composition. + // order: last, golden, altref, current. std::copy(state.surface_luma_offset.begin(), state.surface_luma_offset.begin() + 4, vp9_info.frame_offsets.begin()); @@ -384,40 +386,25 @@ Vp9FrameContainer VP9::GetCurrentFrame(const NvdecCommon::NvdecRegisters& state) gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset, current_frame.bit_stream.data(), current_frame.info.bitstream_size); } - // Buffer two frames, saving the last show frame info - if (!next_next_frame.bit_stream.empty()) { + if (!next_frame.bit_stream.empty()) { Vp9FrameContainer temp{ .info = current_frame.info, .bit_stream = std::move(current_frame.bit_stream), }; - next_next_frame.info.show_frame = current_frame.info.last_frame_shown; - current_frame.info = next_next_frame.info; - current_frame.bit_stream = std::move(next_next_frame.bit_stream); - next_next_frame = std::move(temp); - - if (!next_frame.bit_stream.empty()) { - Vp9FrameContainer temp2{ - .info = current_frame.info, - .bit_stream = std::move(current_frame.bit_stream), - }; - next_frame.info.show_frame = current_frame.info.last_frame_shown; - current_frame.info = next_frame.info; - current_frame.bit_stream = std::move(next_frame.bit_stream); - next_frame = std::move(temp2); - } else { - next_frame.info = current_frame.info; - next_frame.bit_stream = std::move(current_frame.bit_stream); - } + next_frame.info.show_frame = current_frame.info.last_frame_shown; + current_frame.info = next_frame.info; + current_frame.bit_stream = std::move(next_frame.bit_stream); + next_frame = std::move(temp); } else { - next_next_frame.info = current_frame.info; - next_next_frame.bit_stream = std::move(current_frame.bit_stream); + next_frame.info = current_frame.info; + next_frame.bit_stream = current_frame.bit_stream; } return current_frame; } std::vector<u8> VP9::ComposeCompressedHeader() { VpxRangeEncoder writer{}; - const bool update_probs = current_frame_info.show_frame && !current_frame_info.is_key_frame; + const bool update_probs = !current_frame_info.is_key_frame && current_frame_info.show_frame; if (!current_frame_info.lossless) { if (static_cast<u32>(current_frame_info.transform_mode) >= 3) { writer.Write(3, 2); @@ -613,86 +600,64 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() { // Reset context prev_frame_probs = default_probs; - swap_next_golden = false; + swap_ref_indices = false; loop_filter_ref_deltas.fill(0); loop_filter_mode_deltas.fill(0); - - // allow frames offsets to stabilize before checking for golden frames - grace_period = 4; - - // On key frames, all frame slots are set to the current frame, - // so the value of the selected slot doesn't really matter. - frame_ctxs.fill({current_frame_number, false, default_probs}); + frame_ctxs.fill(default_probs); // intra only, meaning the frame can be recreated with no other references current_frame_info.intra_only = true; - } else { - if (!current_frame_info.show_frame) { uncomp_writer.WriteBit(current_frame_info.intra_only); - if (!current_frame_info.last_frame_was_key) { - swap_next_golden = !swap_next_golden; - } } else { current_frame_info.intra_only = false; } if (!current_frame_info.error_resilient_mode) { uncomp_writer.WriteU(0, 2); // Reset frame context. } - - // Last, Golden, Altref frames - std::array<s32, 3> ref_frame_index{0, 1, 2}; - - // Set when next frame is hidden - // altref and golden references are swapped - if (swap_next_golden) { - ref_frame_index = std::array<s32, 3>{0, 2, 1}; + const auto& curr_offsets = current_frame_info.frame_offsets; + const auto& next_offsets = next_frame.info.frame_offsets; + const bool ref_frames_different = curr_offsets[1] != curr_offsets[2]; + const bool next_references_swap = + (next_offsets[1] == curr_offsets[2]) || (next_offsets[2] == curr_offsets[1]); + const bool needs_ref_swap = ref_frames_different && next_references_swap; + if (needs_ref_swap) { + swap_ref_indices = !swap_ref_indices; } - - // update Last Frame - u64 refresh_frame_flags = 1; - - // golden frame may refresh, determined if the next golden frame offset is changed - bool golden_refresh = false; - if (grace_period <= 0) { - for (s32 index = 1; index < 3; ++index) { - if (current_frame_info.frame_offsets[index] != - next_frame.info.frame_offsets[index]) { - current_frame_info.refresh_frame[index] = true; - golden_refresh = true; - grace_period = 3; - } + union { + u32 raw; + BitField<0, 1, u32> refresh_last; + BitField<1, 2, u32> refresh_golden; + BitField<2, 1, u32> refresh_alt; + } refresh_frame_flags; + + refresh_frame_flags.raw = 0; + for (u32 index = 0; index < 3; ++index) { + // Refresh indices that use the current frame as an index + if (curr_offsets[3] == next_offsets[index]) { + refresh_frame_flags.raw |= 1u << index; } } - - if (current_frame_info.show_frame && - (!next_frame.info.show_frame || next_frame.info.is_key_frame)) { - // Update golden frame - refresh_frame_flags = swap_next_golden ? 2 : 4; - } - - if (!current_frame_info.show_frame) { - // Update altref - refresh_frame_flags = swap_next_golden ? 2 : 4; - } else if (golden_refresh) { - refresh_frame_flags = 3; + if (swap_ref_indices) { + const u32 temp = refresh_frame_flags.refresh_golden; + refresh_frame_flags.refresh_golden.Assign(refresh_frame_flags.refresh_alt.Value()); + refresh_frame_flags.refresh_alt.Assign(temp); } - if (current_frame_info.intra_only) { uncomp_writer.WriteU(frame_sync_code, 24); - uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8); + uncomp_writer.WriteU(refresh_frame_flags.raw, 8); uncomp_writer.WriteU(current_frame_info.frame_size.width - 1, 16); uncomp_writer.WriteU(current_frame_info.frame_size.height - 1, 16); uncomp_writer.WriteBit(false); // Render and frame size different. } else { - uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8); - - for (s32 index = 1; index < 4; index++) { + const bool swap_indices = needs_ref_swap ^ swap_ref_indices; + const auto ref_frame_index = swap_indices ? std::array{0, 2, 1} : std::array{0, 1, 2}; + uncomp_writer.WriteU(refresh_frame_flags.raw, 8); + for (size_t index = 1; index < 4; index++) { uncomp_writer.WriteU(ref_frame_index[index - 1], 3); uncomp_writer.WriteU(current_frame_info.ref_frame_sign_bias[index], 1); } - uncomp_writer.WriteBit(true); // Frame size with refs. uncomp_writer.WriteBit(false); // Render and frame size different. uncomp_writer.WriteBit(current_frame_info.allow_high_precision_mv); @@ -714,10 +679,9 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() { frame_ctx_idx = 1; } - uncomp_writer.WriteU(frame_ctx_idx, 2); // Frame context index. - prev_frame_probs = - frame_ctxs[frame_ctx_idx].probs; // reference probabilities for compressed header - frame_ctxs[frame_ctx_idx] = {current_frame_number, false, current_frame_info.entropy}; + uncomp_writer.WriteU(frame_ctx_idx, 2); // Frame context index. + prev_frame_probs = frame_ctxs[frame_ctx_idx]; // reference probabilities for compressed header + frame_ctxs[frame_ctx_idx] = current_frame_info.entropy; uncomp_writer.WriteU(current_frame_info.first_level, 6); uncomp_writer.WriteU(current_frame_info.sharpness_level, 3); @@ -812,7 +776,6 @@ const std::vector<u8>& VP9::ComposeFrameHeader(const NvdecCommon::NvdecRegisters current_frame_info = curr_frame.info; bitstream = std::move(curr_frame.bit_stream); } - // The uncompressed header routine sets PrevProb parameters needed for the compressed header auto uncomp_writer = ComposeUncompressedHeader(); std::vector<u8> compressed_header = ComposeCompressedHeader(); @@ -828,13 +791,6 @@ const std::vector<u8>& VP9::ComposeFrameHeader(const NvdecCommon::NvdecRegisters frame.begin() + uncompressed_header.size()); std::copy(bitstream.begin(), bitstream.end(), frame.begin() + uncompressed_header.size() + compressed_header.size()); - - // keep track of frame number - current_frame_number++; - grace_period--; - - // don't display hidden frames - hidden = !current_frame_info.show_frame; return frame; } diff --git a/src/video_core/command_classes/codecs/vp9.h b/src/video_core/command_classes/codecs/vp9.h index 8396c8105..e6e9fc17e 100644 --- a/src/video_core/command_classes/codecs/vp9.h +++ b/src/video_core/command_classes/codecs/vp9.h @@ -14,7 +14,6 @@ namespace Tegra { class GPU; -enum class FrameType { KeyFrame = 0, InterFrame = 1 }; namespace Decoder { /// The VpxRangeEncoder, and VpxBitStreamWriter classes are used to compose the @@ -124,7 +123,7 @@ public: /// Returns true if the most recent frame was a hidden frame. [[nodiscard]] bool WasFrameHidden() const { - return hidden; + return !current_frame_info.show_frame; } private: @@ -178,19 +177,12 @@ private: std::array<s8, 4> loop_filter_ref_deltas{}; std::array<s8, 2> loop_filter_mode_deltas{}; - bool hidden = false; - s64 current_frame_number = -2; // since we buffer 2 frames - s32 grace_period = 6; // frame offsets need to stabilize - std::array<FrameContexts, 4> frame_ctxs{}; Vp9FrameContainer next_frame{}; - Vp9FrameContainer next_next_frame{}; - bool swap_next_golden{}; + std::array<Vp9EntropyProbs, 4> frame_ctxs{}; + bool swap_ref_indices{}; Vp9PictureInfo current_frame_info{}; Vp9EntropyProbs prev_frame_probs{}; - - s32 diff_update_probability = 252; - s32 frame_sync_code = 0x498342; }; } // namespace Decoder diff --git a/src/video_core/command_classes/codecs/vp9_types.h b/src/video_core/command_classes/codecs/vp9_types.h index 2da14f3ca..87eafdb03 100644 --- a/src/video_core/command_classes/codecs/vp9_types.h +++ b/src/video_core/command_classes/codecs/vp9_types.h @@ -176,7 +176,7 @@ struct PictureInfo { .frame_size_changed = (vp9_flags & FrameFlags::FrameSizeChanged) != 0, .error_resilient_mode = (vp9_flags & FrameFlags::ErrorResilientMode) != 0, .last_frame_shown = (vp9_flags & FrameFlags::LastShowFrame) != 0, - .show_frame = false, + .show_frame = true, .ref_frame_sign_bias = ref_frame_sign_bias, .base_q_index = base_q_index, .y_dc_delta_q = y_dc_delta_q, @@ -296,12 +296,6 @@ struct RefPoolElement { bool refresh{}; }; -struct FrameContexts { - s64 from; - bool adapted; - Vp9EntropyProbs probs; -}; - #define ASSERT_POSITION(field_name, position) \ static_assert(offsetof(Vp9EntropyProbs, field_name) == position, \ "Field " #field_name " has invalid position") diff --git a/src/video_core/command_classes/nvdec.cpp b/src/video_core/command_classes/nvdec.cpp index b5e3b70fc..b5c55f14a 100644 --- a/src/video_core/command_classes/nvdec.cpp +++ b/src/video_core/command_classes/nvdec.cpp @@ -39,7 +39,7 @@ void Nvdec::Execute() { codec->Decode(); break; default: - UNIMPLEMENTED_MSG("Unknown codec {}", static_cast<u32>(codec->GetCurrentCodec())); + UNIMPLEMENTED_MSG("Codec {}", codec->GetCurrentCodecName()); break; } } diff --git a/src/video_core/host_shaders/astc_decoder.comp b/src/video_core/host_shaders/astc_decoder.comp index c37f15bfd..f34c5f5d9 100644 --- a/src/video_core/host_shaders/astc_decoder.comp +++ b/src/video_core/host_shaders/astc_decoder.comp @@ -10,33 +10,27 @@ #define END_PUSH_CONSTANTS }; #define UNIFORM(n) #define BINDING_INPUT_BUFFER 0 -#define BINDING_ENC_BUFFER 1 -#define BINDING_SWIZZLE_BUFFER 2 -#define BINDING_OUTPUT_IMAGE 3 +#define BINDING_OUTPUT_IMAGE 1 #else // ^^^ Vulkan ^^^ // vvv OpenGL vvv #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_ENC_BUFFER 2 +#define BINDING_INPUT_BUFFER 0 #define BINDING_OUTPUT_IMAGE 0 #endif -layout(local_size_x = 32, local_size_y = 32, local_size_z = 1) in; +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; BEGIN_PUSH_CONSTANTS UNIFORM(1) uvec2 block_dims; - -UNIFORM(2) uint bytes_per_block_log2; -UNIFORM(3) uint layer_stride; -UNIFORM(4) uint block_size; -UNIFORM(5) uint x_shift; -UNIFORM(6) uint block_height; -UNIFORM(7) uint block_height_mask; +UNIFORM(2) uint layer_stride; +UNIFORM(3) uint block_size; +UNIFORM(4) uint x_shift; +UNIFORM(5) uint block_height; +UNIFORM(6) uint block_height_mask; END_PUSH_CONSTANTS struct EncodingData { @@ -55,45 +49,35 @@ struct TexelWeightParams { bool void_extent_hdr; }; -// Swizzle data -layout(binding = BINDING_SWIZZLE_BUFFER, std430) readonly buffer SwizzleTable { - uint swizzle_table[]; -}; - layout(binding = BINDING_INPUT_BUFFER, std430) readonly buffer InputBufferU32 { - uint astc_data[]; -}; - -// ASTC Encodings data -layout(binding = BINDING_ENC_BUFFER, std430) readonly buffer EncodingsValues { - EncodingData encoding_values[]; + uvec4 astc_data[]; }; layout(binding = BINDING_OUTPUT_IMAGE, rgba8) uniform writeonly image2DArray dest_image; -const uint GOB_SIZE_X = 64; -const uint GOB_SIZE_Y = 8; -const uint GOB_SIZE_Z = 1; -const uint GOB_SIZE = GOB_SIZE_X * GOB_SIZE_Y * GOB_SIZE_Z; - const uint GOB_SIZE_X_SHIFT = 6; const uint GOB_SIZE_Y_SHIFT = 3; -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 - 1, GOB_SIZE_Y - 1); +const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT; -const int BLOCK_SIZE_IN_BYTES = 16; - -const int BLOCK_INFO_ERROR = 0; -const int BLOCK_INFO_VOID_EXTENT_HDR = 1; -const int BLOCK_INFO_VOID_EXTENT_LDR = 2; -const int BLOCK_INFO_NORMAL = 3; +const uint BYTES_PER_BLOCK_LOG2 = 4; const int JUST_BITS = 0; const int QUINT = 1; const int TRIT = 2; +// ASTC Encodings data, sorted in ascending order based on their BitLength value +// (see GetBitLength() function) +EncodingData encoding_values[22] = EncodingData[]( + EncodingData(JUST_BITS, 0, 0, 0), EncodingData(JUST_BITS, 1, 0, 0), EncodingData(TRIT, 0, 0, 0), + EncodingData(JUST_BITS, 2, 0, 0), EncodingData(QUINT, 0, 0, 0), EncodingData(TRIT, 1, 0, 0), + EncodingData(JUST_BITS, 3, 0, 0), EncodingData(QUINT, 1, 0, 0), EncodingData(TRIT, 2, 0, 0), + EncodingData(JUST_BITS, 4, 0, 0), EncodingData(QUINT, 2, 0, 0), EncodingData(TRIT, 3, 0, 0), + EncodingData(JUST_BITS, 5, 0, 0), EncodingData(QUINT, 3, 0, 0), EncodingData(TRIT, 4, 0, 0), + EncodingData(JUST_BITS, 6, 0, 0), EncodingData(QUINT, 4, 0, 0), EncodingData(TRIT, 5, 0, 0), + EncodingData(JUST_BITS, 7, 0, 0), EncodingData(QUINT, 5, 0, 0), EncodingData(TRIT, 6, 0, 0), + EncodingData(JUST_BITS, 8, 0, 0) +); + // The following constants are expanded variants of the Replicate() // function calls corresponding to the following arguments: // value: index into the generated table @@ -135,44 +119,37 @@ const uint REPLICATE_7_BIT_TO_8_TABLE[128] = // Input ASTC texture globals uint current_index = 0; int bitsread = 0; -uint total_bitsread = 0; -uint local_buff[16]; +int total_bitsread = 0; +uvec4 local_buff; // Color data globals -uint color_endpoint_data[16]; +uvec4 color_endpoint_data; int color_bitsread = 0; -uint total_color_bitsread = 0; -int color_index = 0; // Four values, two endpoints, four maximum paritions uint color_values[32]; int colvals_index = 0; // Weight data globals -uint texel_weight_data[16]; +uvec4 texel_weight_data; int texel_bitsread = 0; -uint total_texel_bitsread = 0; -int texel_index = 0; bool texel_flag = false; // Global "vectors" to be pushed into when decoding -EncodingData result_vector[100]; +EncodingData result_vector[144]; int result_index = 0; -EncodingData texel_vector[100]; +EncodingData texel_vector[144]; int texel_vector_index = 0; uint unquantized_texel_weights[2][144]; uint SwizzleOffset(uvec2 pos) { - pos = pos & SWIZZLE_MASK; - return swizzle_table[pos.y * 64 + pos.x]; -} - -uint ReadTexel(uint offset) { - // extract the 8-bit value from the 32-bit packed data. - return bitfieldExtract(astc_data[offset / 4], int((offset * 8) & 24), 8); + uint x = pos.x; + uint y = pos.y; + return ((x % 64) / 32) * 256 + ((y % 8) / 2) * 64 + ((x % 32) / 16) * 32 + + (y % 2) * 16 + (x % 16); } // Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)] @@ -278,14 +255,10 @@ uint Hash52(uint p) { return p; } -uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bool small_block) { - if (partition_count == 1) { - return 0; - } +uint Select2DPartition(uint seed, uint x, uint y, uint partition_count, bool small_block) { if (small_block) { x <<= 1; y <<= 1; - z <<= 1; } seed += (partition_count - 1) * 1024; @@ -299,10 +272,6 @@ uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bo uint seed6 = uint((rnum >> 20) & 0xF); uint seed7 = uint((rnum >> 24) & 0xF); uint seed8 = uint((rnum >> 28) & 0xF); - uint seed9 = uint((rnum >> 18) & 0xF); - uint seed10 = uint((rnum >> 22) & 0xF); - uint seed11 = uint((rnum >> 26) & 0xF); - uint seed12 = uint(((rnum >> 30) | (rnum << 2)) & 0xF); seed1 = (seed1 * seed1); seed2 = (seed2 * seed2); @@ -312,12 +281,8 @@ uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bo seed6 = (seed6 * seed6); seed7 = (seed7 * seed7); seed8 = (seed8 * seed8); - seed9 = (seed9 * seed9); - seed10 = (seed10 * seed10); - seed11 = (seed11 * seed11); - seed12 = (seed12 * seed12); - int sh1, sh2, sh3; + uint sh1, sh2; if ((seed & 1) > 0) { sh1 = (seed & 2) > 0 ? 4 : 5; sh2 = (partition_count == 3) ? 6 : 5; @@ -325,25 +290,19 @@ uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bo sh1 = (partition_count == 3) ? 6 : 5; sh2 = (seed & 2) > 0 ? 4 : 5; } - sh3 = (seed & 0x10) > 0 ? sh1 : sh2; - - seed1 = (seed1 >> sh1); - seed2 = (seed2 >> sh2); - seed3 = (seed3 >> sh1); - seed4 = (seed4 >> sh2); - seed5 = (seed5 >> sh1); - seed6 = (seed6 >> sh2); - seed7 = (seed7 >> sh1); - seed8 = (seed8 >> sh2); - seed9 = (seed9 >> sh3); - seed10 = (seed10 >> sh3); - seed11 = (seed11 >> sh3); - seed12 = (seed12 >> sh3); - - uint a = seed1 * x + seed2 * y + seed11 * z + (rnum >> 14); - uint b = seed3 * x + seed4 * y + seed12 * z + (rnum >> 10); - uint c = seed5 * x + seed6 * y + seed9 * z + (rnum >> 6); - uint d = seed7 * x + seed8 * y + seed10 * z + (rnum >> 2); + seed1 >>= sh1; + seed2 >>= sh2; + seed3 >>= sh1; + seed4 >>= sh2; + seed5 >>= sh1; + seed6 >>= sh2; + seed7 >>= sh1; + seed8 >>= sh2; + + uint a = seed1 * x + seed2 * y + (rnum >> 14); + uint b = seed3 * x + seed4 * y + (rnum >> 10); + uint c = seed5 * x + seed6 * y + (rnum >> 6); + uint d = seed7 * x + seed8 * y + (rnum >> 2); a &= 0x3F; b &= 0x3F; @@ -368,58 +327,37 @@ uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bo } } -uint Select2DPartition(uint seed, uint x, uint y, uint partition_count, bool small_block) { - return SelectPartition(seed, x, y, 0, partition_count, small_block); -} - -uint ReadBit() { - if (current_index >= local_buff.length()) { +uint ExtractBits(uvec4 payload, int offset, int bits) { + if (bits <= 0) { return 0; } - uint bit = bitfieldExtract(local_buff[current_index], bitsread, 1); - ++bitsread; - ++total_bitsread; - if (bitsread == 8) { - ++current_index; - bitsread = 0; + int last_offset = offset + bits - 1; + int shifted_offset = offset >> 5; + if ((last_offset >> 5) == shifted_offset) { + return bitfieldExtract(payload[shifted_offset], offset & 31, bits); } - return bit; + int first_bits = 32 - (offset & 31); + int result_first = int(bitfieldExtract(payload[shifted_offset], offset & 31, first_bits)); + int result_second = int(bitfieldExtract(payload[shifted_offset + 1], 0, bits - first_bits)); + return result_first | (result_second << first_bits); } uint StreamBits(uint num_bits) { - uint ret = 0; - for (uint i = 0; i < num_bits; i++) { - ret |= ((ReadBit() & 1) << i); - } + int int_bits = int(num_bits); + uint ret = ExtractBits(local_buff, total_bitsread, int_bits); + total_bitsread += int_bits; return ret; } -uint ReadColorBit() { - uint bit = 0; - if (texel_flag) { - bit = bitfieldExtract(texel_weight_data[texel_index], texel_bitsread, 1); - ++texel_bitsread; - ++total_texel_bitsread; - if (texel_bitsread == 8) { - ++texel_index; - texel_bitsread = 0; - } - } else { - bit = bitfieldExtract(color_endpoint_data[color_index], color_bitsread, 1); - ++color_bitsread; - ++total_color_bitsread; - if (color_bitsread == 8) { - ++color_index; - color_bitsread = 0; - } - } - return bit; -} - uint StreamColorBits(uint num_bits) { uint ret = 0; - for (uint i = 0; i < num_bits; i++) { - ret |= ((ReadColorBit() & 1) << i); + int int_bits = int(num_bits); + if (texel_flag) { + ret = ExtractBits(texel_weight_data, texel_bitsread, int_bits); + texel_bitsread += int_bits; + } else { + ret = ExtractBits(color_endpoint_data, color_bitsread, int_bits); + color_bitsread += int_bits; } return ret; } @@ -596,22 +534,16 @@ void DecodeColorValues(uvec4 modes, uint num_partitions, uint color_data_bits) { for (uint i = 0; i < num_partitions; i++) { num_values += ((modes[i] >> 2) + 1) << 1; } - int range = 256; - while (--range > 0) { - EncodingData val = encoding_values[range]; + // Find the largest encoding that's within color_data_bits + // TODO(ameerj): profile with binary search + int range = 0; + while (++range < encoding_values.length()) { uint bit_length = GetBitLength(num_values, range); - if (bit_length <= color_data_bits) { - while (--range > 0) { - EncodingData newval = encoding_values[range]; - if (newval.encoding != val.encoding && newval.num_bits != val.num_bits) { - break; - } - } - ++range; + if (bit_length > color_data_bits) { break; } } - DecodeIntegerSequence(range, num_values); + DecodeIntegerSequence(range - 1, num_values); uint out_index = 0; for (int itr = 0; itr < result_index; ++itr) { if (out_index >= num_values) { @@ -1028,7 +960,7 @@ int FindLayout(uint mode) { return 5; } -TexelWeightParams DecodeBlockInfo(uint block_index) { +TexelWeightParams DecodeBlockInfo() { TexelWeightParams params = TexelWeightParams(uvec2(0), 0, false, false, false, false); uint mode = StreamBits(11); if ((mode & 0x1ff) == 0x1fc) { @@ -1110,10 +1042,10 @@ TexelWeightParams DecodeBlockInfo(uint block_index) { } weight_index -= 2; if ((mode_layout != 9) && ((mode & 0x200) != 0)) { - const int max_weights[6] = int[6](9, 11, 15, 19, 23, 31); + const int max_weights[6] = int[6](7, 8, 9, 10, 11, 12); params.max_weight = max_weights[weight_index]; } else { - const int max_weights[6] = int[6](1, 2, 3, 4, 5, 7); + const int max_weights[6] = int[6](1, 2, 3, 4, 5, 6); params.max_weight = max_weights[weight_index]; } return params; @@ -1144,8 +1076,8 @@ void FillVoidExtentLDR(ivec3 coord) { } } -void DecompressBlock(ivec3 coord, uint block_index) { - TexelWeightParams params = DecodeBlockInfo(block_index); +void DecompressBlock(ivec3 coord) { + TexelWeightParams params = DecodeBlockInfo(); if (params.error_state) { FillError(coord); return; @@ -1212,7 +1144,7 @@ void DecompressBlock(ivec3 coord, uint block_index) { // Read color data... uint color_data_bits = remaining_bits; while (remaining_bits > 0) { - int nb = int(min(remaining_bits, 8U)); + int nb = int(min(remaining_bits, 32U)); uint b = StreamBits(nb); color_endpoint_data[ced_pointer] = uint(bitfieldExtract(b, 0, nb)); ++ced_pointer; @@ -1254,25 +1186,20 @@ void DecompressBlock(ivec3 coord, uint block_index) { ComputeEndpoints(endpoints[i][0], endpoints[i][1], color_endpoint_mode[i]); } - for (uint i = 0; i < 16; i++) { - texel_weight_data[i] = local_buff[i]; - } - for (uint i = 0; i < 8; i++) { -#define REVERSE_BYTE(b) ((b * 0x0802U & 0x22110U) | (b * 0x8020U & 0x88440U)) * 0x10101U >> 16 - uint a = REVERSE_BYTE(texel_weight_data[i]); - uint b = REVERSE_BYTE(texel_weight_data[15 - i]); -#undef REVERSE_BYTE - texel_weight_data[i] = uint(bitfieldExtract(b, 0, 8)); - texel_weight_data[15 - i] = uint(bitfieldExtract(a, 0, 8)); - } + texel_weight_data = local_buff; + texel_weight_data = bitfieldReverse(texel_weight_data).wzyx; uint clear_byte_start = (GetPackedBitSize(params.size, params.dual_plane, params.max_weight) >> 3) + 1; - texel_weight_data[clear_byte_start - 1] = - texel_weight_data[clear_byte_start - 1] & + + uint byte_insert = ExtractBits(texel_weight_data, int(clear_byte_start - 1) * 8, 8) & uint( ((1 << (GetPackedBitSize(params.size, params.dual_plane, params.max_weight) % 8)) - 1)); - for (uint i = 0; i < 16 - clear_byte_start; i++) { - texel_weight_data[clear_byte_start + i] = 0U; + uint vec_index = (clear_byte_start - 1) >> 2; + texel_weight_data[vec_index] = + bitfieldInsert(texel_weight_data[vec_index], byte_insert, int((clear_byte_start - 1) % 4) * 8, 8); + for (uint i = clear_byte_start; i < 16; ++i) { + uint idx = i >> 2; + texel_weight_data[idx] = bitfieldInsert(texel_weight_data[idx], 0, int(i % 4) * 8, 8); } texel_flag = true; // use texel "vector" and bit stream in integer decoding DecodeIntegerSequence(params.max_weight, GetNumWeightValues(params.size, params.dual_plane)); @@ -1281,8 +1208,11 @@ void DecompressBlock(ivec3 coord, uint block_index) { for (uint j = 0; j < block_dims.y; j++) { for (uint i = 0; i < block_dims.x; i++) { - uint local_partition = Select2DPartition(partition_index, i, j, num_partitions, + uint local_partition = 0; + if (num_partitions > 1) { + local_partition = Select2DPartition(partition_index, i, j, num_partitions, (block_dims.y * block_dims.x) < 32); + } vec4 p; uvec4 C0 = ReplicateByteTo16(endpoints[local_partition][0]); uvec4 C1 = ReplicateByteTo16(endpoints[local_partition][1]); @@ -1303,7 +1233,7 @@ void DecompressBlock(ivec3 coord, uint block_index) { void main() { uvec3 pos = gl_GlobalInvocationID; - pos.x <<= bytes_per_block_log2; + pos.x <<= BYTES_PER_BLOCK_LOG2; // Read as soon as possible due to its latency const uint swizzle = SwizzleOffset(pos.xy); @@ -1321,13 +1251,8 @@ void main() { if (any(greaterThanEqual(coord, imageSize(dest_image)))) { return; } - uint block_index = - pos.z * gl_WorkGroupSize.x * gl_WorkGroupSize.y + pos.y * gl_WorkGroupSize.x + pos.x; - current_index = 0; bitsread = 0; - for (int i = 0; i < 16; i++) { - local_buff[i] = ReadTexel(offset + i); - } - DecompressBlock(coord, block_index); + local_buff = astc_data[offset / 16]; + DecompressBlock(coord); } diff --git a/src/video_core/renderer_opengl/util_shaders.cpp b/src/video_core/renderer_opengl/util_shaders.cpp index 37a4d1d9d..333f35a1c 100644 --- a/src/video_core/renderer_opengl/util_shaders.cpp +++ b/src/video_core/renderer_opengl/util_shaders.cpp @@ -60,19 +60,14 @@ UtilShaders::UtilShaders(ProgramManager& program_manager_) copy_bc4_program(MakeProgram(OPENGL_COPY_BC4_COMP)) { const auto swizzle_table = Tegra::Texture::MakeSwizzleTable(); swizzle_table_buffer.Create(); - astc_buffer.Create(); glNamedBufferStorage(swizzle_table_buffer.handle, sizeof(swizzle_table), &swizzle_table, 0); - glNamedBufferStorage(astc_buffer.handle, sizeof(ASTC_ENCODINGS_VALUES), &ASTC_ENCODINGS_VALUES, - 0); } UtilShaders::~UtilShaders() = default; void UtilShaders::ASTCDecode(Image& image, const ImageBufferMap& map, std::span<const VideoCommon::SwizzleParameters> swizzles) { - static constexpr GLuint BINDING_SWIZZLE_BUFFER = 0; - static constexpr GLuint BINDING_INPUT_BUFFER = 1; - static constexpr GLuint BINDING_ENC_BUFFER = 2; + static constexpr GLuint BINDING_INPUT_BUFFER = 0; static constexpr GLuint BINDING_OUTPUT_IMAGE = 0; const Extent2D tile_size{ @@ -80,34 +75,32 @@ void UtilShaders::ASTCDecode(Image& image, const ImageBufferMap& map, .height = VideoCore::Surface::DefaultBlockHeight(image.info.format), }; program_manager.BindComputeProgram(astc_decoder_program.handle); - glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_SWIZZLE_BUFFER, swizzle_table_buffer.handle); - glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_ENC_BUFFER, astc_buffer.handle); - glFlushMappedNamedBufferRange(map.buffer, map.offset, image.guest_size_bytes); glUniform2ui(1, tile_size.width, tile_size.height); + // Ensure buffer data is valid before dispatching glFlush(); for (const SwizzleParameters& swizzle : swizzles) { const size_t input_offset = swizzle.buffer_offset + map.offset; - const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 32U); - const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 32U); + const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 8U); + const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 8U); const auto params = MakeBlockLinearSwizzle2DParams(swizzle, image.info); ASSERT(params.origin == (std::array<u32, 3>{0, 0, 0})); ASSERT(params.destination == (std::array<s32, 3>{0, 0, 0})); + ASSERT(params.bytes_per_block_log2 == 4); - glUniform1ui(2, params.bytes_per_block_log2); - glUniform1ui(3, params.layer_stride); - glUniform1ui(4, params.block_size); - glUniform1ui(5, params.x_shift); - glUniform1ui(6, params.block_height); - glUniform1ui(7, params.block_height_mask); + glUniform1ui(2, params.layer_stride); + glUniform1ui(3, params.block_size); + glUniform1ui(4, params.x_shift); + glUniform1ui(5, params.block_height); + glUniform1ui(6, params.block_height_mask); - glBindImageTexture(BINDING_OUTPUT_IMAGE, image.StorageHandle(), swizzle.level, GL_TRUE, 0, - GL_WRITE_ONLY, GL_RGBA8); // ASTC texture data glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_INPUT_BUFFER, map.buffer, input_offset, image.guest_size_bytes - swizzle.buffer_offset); + glBindImageTexture(BINDING_OUTPUT_IMAGE, image.StorageHandle(), swizzle.level, GL_TRUE, 0, + GL_WRITE_ONLY, GL_RGBA8); glDispatchCompute(num_dispatches_x, num_dispatches_y, image.info.resources.layers); } diff --git a/src/video_core/renderer_opengl/util_shaders.h b/src/video_core/renderer_opengl/util_shaders.h index 53d65f368..ef881e35f 100644 --- a/src/video_core/renderer_opengl/util_shaders.h +++ b/src/video_core/renderer_opengl/util_shaders.h @@ -62,7 +62,6 @@ private: ProgramManager& program_manager; OGLBuffer swizzle_table_buffer; - OGLBuffer astc_buffer; OGLProgram astc_decoder_program; OGLProgram block_linear_unswizzle_2d_program; diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.cpp b/src/video_core/renderer_vulkan/vk_compute_pass.cpp index 561cf5e11..3e96c0f60 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pass.cpp @@ -30,16 +30,12 @@ namespace Vulkan { using Tegra::Texture::SWIZZLE_TABLE; -using Tegra::Texture::ASTC::ASTC_ENCODINGS_VALUES; -using namespace Tegra::Texture::ASTC; namespace { constexpr u32 ASTC_BINDING_INPUT_BUFFER = 0; -constexpr u32 ASTC_BINDING_ENC_BUFFER = 1; -constexpr u32 ASTC_BINDING_SWIZZLE_BUFFER = 2; -constexpr u32 ASTC_BINDING_OUTPUT_IMAGE = 3; -constexpr size_t ASTC_NUM_BINDINGS = 4; +constexpr u32 ASTC_BINDING_OUTPUT_IMAGE = 1; +constexpr size_t ASTC_NUM_BINDINGS = 2; template <size_t size> inline constexpr VkPushConstantRange COMPUTE_PUSH_CONSTANT_RANGE{ @@ -75,7 +71,7 @@ constexpr DescriptorBankInfo INPUT_OUTPUT_BANK_INFO{ .score = 2, }; -constexpr std::array<VkDescriptorSetLayoutBinding, 4> ASTC_DESCRIPTOR_SET_BINDINGS{{ +constexpr std::array<VkDescriptorSetLayoutBinding, ASTC_NUM_BINDINGS> ASTC_DESCRIPTOR_SET_BINDINGS{{ { .binding = ASTC_BINDING_INPUT_BUFFER, .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, @@ -84,20 +80,6 @@ constexpr std::array<VkDescriptorSetLayoutBinding, 4> ASTC_DESCRIPTOR_SET_BINDIN .pImmutableSamplers = nullptr, }, { - .binding = ASTC_BINDING_ENC_BUFFER, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .descriptorCount = 1, - .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT, - .pImmutableSamplers = nullptr, - }, - { - .binding = ASTC_BINDING_SWIZZLE_BUFFER, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .descriptorCount = 1, - .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT, - .pImmutableSamplers = nullptr, - }, - { .binding = ASTC_BINDING_OUTPUT_IMAGE, .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, .descriptorCount = 1, @@ -108,12 +90,12 @@ constexpr std::array<VkDescriptorSetLayoutBinding, 4> ASTC_DESCRIPTOR_SET_BINDIN constexpr DescriptorBankInfo ASTC_BANK_INFO{ .uniform_buffers = 0, - .storage_buffers = 3, + .storage_buffers = 1, .texture_buffers = 0, .image_buffers = 0, .textures = 0, .images = 1, - .score = 4, + .score = 2, }; constexpr VkDescriptorUpdateTemplateEntryKHR INPUT_OUTPUT_DESCRIPTOR_UPDATE_TEMPLATE{ @@ -136,22 +118,6 @@ constexpr std::array<VkDescriptorUpdateTemplateEntryKHR, ASTC_NUM_BINDINGS> .stride = sizeof(DescriptorUpdateEntry), }, { - .dstBinding = ASTC_BINDING_ENC_BUFFER, - .dstArrayElement = 0, - .descriptorCount = 1, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .offset = ASTC_BINDING_ENC_BUFFER * sizeof(DescriptorUpdateEntry), - .stride = sizeof(DescriptorUpdateEntry), - }, - { - .dstBinding = ASTC_BINDING_SWIZZLE_BUFFER, - .dstArrayElement = 0, - .descriptorCount = 1, - .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, - .offset = ASTC_BINDING_SWIZZLE_BUFFER * sizeof(DescriptorUpdateEntry), - .stride = sizeof(DescriptorUpdateEntry), - }, - { .dstBinding = ASTC_BINDING_OUTPUT_IMAGE, .dstArrayElement = 0, .descriptorCount = 1, @@ -163,7 +129,6 @@ constexpr std::array<VkDescriptorUpdateTemplateEntryKHR, ASTC_NUM_BINDINGS> struct AstcPushConstants { std::array<u32, 2> blocks_dims; - u32 bytes_per_block_log2; u32 layer_stride; u32 block_size; u32 x_shift; @@ -354,46 +319,6 @@ ASTCDecoderPass::ASTCDecoderPass(const Device& device_, VKScheduler& scheduler_, ASTCDecoderPass::~ASTCDecoderPass() = default; -void ASTCDecoderPass::MakeDataBuffer() { - constexpr size_t TOTAL_BUFFER_SIZE = sizeof(ASTC_ENCODINGS_VALUES) + sizeof(SWIZZLE_TABLE); - data_buffer = device.GetLogical().CreateBuffer(VkBufferCreateInfo{ - .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO, - .pNext = nullptr, - .flags = 0, - .size = TOTAL_BUFFER_SIZE, - .usage = VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT, - .sharingMode = VK_SHARING_MODE_EXCLUSIVE, - .queueFamilyIndexCount = 0, - .pQueueFamilyIndices = nullptr, - }); - data_buffer_commit = memory_allocator.Commit(data_buffer, MemoryUsage::Upload); - - const auto staging_ref = staging_buffer_pool.Request(TOTAL_BUFFER_SIZE, MemoryUsage::Upload); - std::memcpy(staging_ref.mapped_span.data(), &ASTC_ENCODINGS_VALUES, - sizeof(ASTC_ENCODINGS_VALUES)); - // Tack on the swizzle table at the end of the buffer - std::memcpy(staging_ref.mapped_span.data() + sizeof(ASTC_ENCODINGS_VALUES), &SWIZZLE_TABLE, - sizeof(SWIZZLE_TABLE)); - - scheduler.Record([src = staging_ref.buffer, offset = staging_ref.offset, dst = *data_buffer, - TOTAL_BUFFER_SIZE](vk::CommandBuffer cmdbuf) { - static constexpr VkMemoryBarrier write_barrier{ - .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT, - .dstAccessMask = VK_ACCESS_SHADER_READ_BIT, - }; - const VkBufferCopy copy{ - .srcOffset = offset, - .dstOffset = 0, - .size = TOTAL_BUFFER_SIZE, - }; - cmdbuf.CopyBuffer(src, dst, copy); - cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, - 0, write_barrier); - }); -} - void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map, std::span<const VideoCommon::SwizzleParameters> swizzles) { using namespace VideoCommon::Accelerated; @@ -402,9 +327,6 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map, VideoCore::Surface::DefaultBlockHeight(image.info.format), }; scheduler.RequestOutsideRenderPassOperationContext(); - if (!data_buffer) { - MakeDataBuffer(); - } const VkPipeline vk_pipeline = *pipeline; const VkImageAspectFlags aspect_mask = image.AspectMask(); const VkImage vk_image = image.Handle(); @@ -436,16 +358,13 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map, }); for (const VideoCommon::SwizzleParameters& swizzle : swizzles) { const size_t input_offset = swizzle.buffer_offset + map.offset; - const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 32U); - const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 32U); + const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 8U); + const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 8U); const u32 num_dispatches_z = image.info.resources.layers; update_descriptor_queue.Acquire(); update_descriptor_queue.AddBuffer(map.buffer, input_offset, image.guest_size_bytes - swizzle.buffer_offset); - update_descriptor_queue.AddBuffer(*data_buffer, 0, sizeof(ASTC_ENCODINGS_VALUES)); - update_descriptor_queue.AddBuffer(*data_buffer, sizeof(ASTC_ENCODINGS_VALUES), - sizeof(SWIZZLE_TABLE)); update_descriptor_queue.AddImage(image.StorageImageView(swizzle.level)); const void* const descriptor_data{update_descriptor_queue.UpdateData()}; @@ -453,11 +372,11 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map, const auto params = MakeBlockLinearSwizzle2DParams(swizzle, image.info); ASSERT(params.origin == (std::array<u32, 3>{0, 0, 0})); ASSERT(params.destination == (std::array<s32, 3>{0, 0, 0})); + ASSERT(params.bytes_per_block_log2 == 4); scheduler.Record([this, num_dispatches_x, num_dispatches_y, num_dispatches_z, block_dims, params, descriptor_data](vk::CommandBuffer cmdbuf) { const AstcPushConstants uniforms{ .blocks_dims = block_dims, - .bytes_per_block_log2 = params.bytes_per_block_log2, .layer_stride = params.layer_stride, .block_size = params.block_size, .x_shift = params.x_shift, diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.h b/src/video_core/renderer_vulkan/vk_compute_pass.h index 114aef2bd..c7b92cce0 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.h +++ b/src/video_core/renderer_vulkan/vk_compute_pass.h @@ -96,15 +96,10 @@ public: std::span<const VideoCommon::SwizzleParameters> swizzles); private: - void MakeDataBuffer(); - VKScheduler& scheduler; StagingBufferPool& staging_buffer_pool; VKUpdateDescriptorQueue& update_descriptor_queue; MemoryAllocator& memory_allocator; - - vk::Buffer data_buffer; - MemoryCommit data_buffer_commit; }; } // namespace Vulkan diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp index 3ab500760..25161df1f 100644 --- a/src/video_core/textures/astc.cpp +++ b/src/video_core/textures/astc.cpp @@ -151,6 +151,76 @@ private: const IntType& m_Bits; }; +enum class IntegerEncoding { JustBits, Quint, Trit }; + +struct IntegerEncodedValue { + constexpr IntegerEncodedValue() = default; + + constexpr IntegerEncodedValue(IntegerEncoding encoding_, u32 num_bits_) + : encoding{encoding_}, num_bits{num_bits_} {} + + constexpr bool MatchesEncoding(const IntegerEncodedValue& other) const { + return encoding == other.encoding && num_bits == other.num_bits; + } + + // Returns the number of bits required to encode num_vals values. + u32 GetBitLength(u32 num_vals) const { + u32 total_bits = num_bits * num_vals; + if (encoding == IntegerEncoding::Trit) { + total_bits += (num_vals * 8 + 4) / 5; + } else if (encoding == IntegerEncoding::Quint) { + total_bits += (num_vals * 7 + 2) / 3; + } + return total_bits; + } + + IntegerEncoding encoding{}; + u32 num_bits = 0; + u32 bit_value = 0; + union { + u32 quint_value = 0; + u32 trit_value; + }; +}; + +// Returns a new instance of this struct that corresponds to the +// can take no more than mav_value values +static constexpr IntegerEncodedValue CreateEncoding(u32 mav_value) { + while (mav_value > 0) { + u32 check = mav_value + 1; + + // Is mav_value a power of two? + if (!(check & (check - 1))) { + return IntegerEncodedValue(IntegerEncoding::JustBits, std::popcount(mav_value)); + } + + // Is mav_value of the type 3*2^n - 1? + if ((check % 3 == 0) && !((check / 3) & ((check / 3) - 1))) { + return IntegerEncodedValue(IntegerEncoding::Trit, std::popcount(check / 3 - 1)); + } + + // Is mav_value of the type 5*2^n - 1? + if ((check % 5 == 0) && !((check / 5) & ((check / 5) - 1))) { + return IntegerEncodedValue(IntegerEncoding::Quint, std::popcount(check / 5 - 1)); + } + + // Apparently it can't be represented with a bounded integer sequence... + // just iterate. + mav_value--; + } + return IntegerEncodedValue(IntegerEncoding::JustBits, 0); +} + +static constexpr std::array<IntegerEncodedValue, 256> MakeEncodedValues() { + std::array<IntegerEncodedValue, 256> encodings{}; + for (std::size_t i = 0; i < encodings.size(); ++i) { + encodings[i] = CreateEncoding(static_cast<u32>(i)); + } + return encodings; +} + +static constexpr std::array<IntegerEncodedValue, 256> ASTC_ENCODINGS_VALUES = MakeEncodedValues(); + namespace Tegra::Texture::ASTC { using IntegerEncodedVector = boost::container::static_vector< IntegerEncodedValue, 256, @@ -521,35 +591,41 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) { return params; } -static void FillVoidExtentLDR(InputBitStream& strm, std::span<u32> outBuf, u32 blockWidth, - u32 blockHeight) { - // Don't actually care about the void extent, just read the bits... - for (s32 i = 0; i < 4; ++i) { - strm.ReadBits<13>(); +// Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)] +// is the same as [(num_bits - 1):0] and repeats all the way down. +template <typename IntType> +static constexpr IntType Replicate(IntType val, u32 num_bits, u32 to_bit) { + if (num_bits == 0 || to_bit == 0) { + return 0; } - - // Decode the RGBA components and renormalize them to the range [0, 255] - u16 r = static_cast<u16>(strm.ReadBits<16>()); - u16 g = static_cast<u16>(strm.ReadBits<16>()); - u16 b = static_cast<u16>(strm.ReadBits<16>()); - u16 a = static_cast<u16>(strm.ReadBits<16>()); - - u32 rgba = (r >> 8) | (g & 0xFF00) | (static_cast<u32>(b) & 0xFF00) << 8 | - (static_cast<u32>(a) & 0xFF00) << 16; - - for (u32 j = 0; j < blockHeight; j++) { - for (u32 i = 0; i < blockWidth; i++) { - outBuf[j * blockWidth + i] = rgba; + const IntType v = val & static_cast<IntType>((1 << num_bits) - 1); + IntType res = v; + u32 reslen = num_bits; + while (reslen < to_bit) { + u32 comp = 0; + if (num_bits > to_bit - reslen) { + u32 newshift = to_bit - reslen; + comp = num_bits - newshift; + num_bits = newshift; } + res = static_cast<IntType>(res << num_bits); + res = static_cast<IntType>(res | (v >> comp)); + reslen += num_bits; } + return res; } -static void FillError(std::span<u32> outBuf, u32 blockWidth, u32 blockHeight) { - for (u32 j = 0; j < blockHeight; j++) { - for (u32 i = 0; i < blockWidth; i++) { - outBuf[j * blockWidth + i] = 0xFFFF00FF; - } +static constexpr std::size_t NumReplicateEntries(u32 num_bits) { + return std::size_t(1) << num_bits; +} + +template <typename IntType, u32 num_bits, u32 to_bit> +static constexpr auto MakeReplicateTable() { + std::array<IntType, NumReplicateEntries(num_bits)> table{}; + for (IntType value = 0; value < static_cast<IntType>(std::size(table)); ++value) { + table[value] = Replicate(value, num_bits, to_bit); } + return table; } static constexpr auto REPLICATE_BYTE_TO_16_TABLE = MakeReplicateTable<u32, 8, 16>(); @@ -572,6 +648,9 @@ static constexpr auto REPLICATE_2_BIT_TO_8_TABLE = MakeReplicateTable<u32, 2, 8> static constexpr auto REPLICATE_3_BIT_TO_8_TABLE = MakeReplicateTable<u32, 3, 8>(); static constexpr auto REPLICATE_4_BIT_TO_8_TABLE = MakeReplicateTable<u32, 4, 8>(); static constexpr auto REPLICATE_5_BIT_TO_8_TABLE = MakeReplicateTable<u32, 5, 8>(); +static constexpr auto REPLICATE_6_BIT_TO_8_TABLE = MakeReplicateTable<u32, 6, 8>(); +static constexpr auto REPLICATE_7_BIT_TO_8_TABLE = MakeReplicateTable<u32, 7, 8>(); +static constexpr auto REPLICATE_8_BIT_TO_8_TABLE = MakeReplicateTable<u32, 8, 8>(); /// Use a precompiled table with the most common usages, if it's not in the expected range, fallback /// to the runtime implementation static constexpr u32 FastReplicateTo8(u32 value, u32 num_bits) { @@ -1316,6 +1395,37 @@ static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const u32*& colorValues, #undef READ_INT_VALUES } +static void FillVoidExtentLDR(InputBitStream& strm, std::span<u32> outBuf, u32 blockWidth, + u32 blockHeight) { + // Don't actually care about the void extent, just read the bits... + for (s32 i = 0; i < 4; ++i) { + strm.ReadBits<13>(); + } + + // Decode the RGBA components and renormalize them to the range [0, 255] + u16 r = static_cast<u16>(strm.ReadBits<16>()); + u16 g = static_cast<u16>(strm.ReadBits<16>()); + u16 b = static_cast<u16>(strm.ReadBits<16>()); + u16 a = static_cast<u16>(strm.ReadBits<16>()); + + u32 rgba = (r >> 8) | (g & 0xFF00) | (static_cast<u32>(b) & 0xFF00) << 8 | + (static_cast<u32>(a) & 0xFF00) << 16; + + for (u32 j = 0; j < blockHeight; j++) { + for (u32 i = 0; i < blockWidth; i++) { + outBuf[j * blockWidth + i] = rgba; + } + } +} + +static void FillError(std::span<u32> outBuf, u32 blockWidth, u32 blockHeight) { + for (u32 j = 0; j < blockHeight; j++) { + for (u32 i = 0; i < blockWidth; i++) { + outBuf[j * blockWidth + i] = 0xFFFF00FF; + } + } +} + static void DecompressBlock(std::span<const u8, 16> inBuf, const u32 blockWidth, const u32 blockHeight, std::span<u32, 12 * 12> outBuf) { InputBitStream strm(inBuf); diff --git a/src/video_core/textures/astc.h b/src/video_core/textures/astc.h index 0229ae122..14d2beec0 100644 --- a/src/video_core/textures/astc.h +++ b/src/video_core/textures/astc.h @@ -9,117 +9,6 @@ namespace Tegra::Texture::ASTC { -enum class IntegerEncoding { JustBits, Quint, Trit }; - -struct IntegerEncodedValue { - constexpr IntegerEncodedValue() = default; - - constexpr IntegerEncodedValue(IntegerEncoding encoding_, u32 num_bits_) - : encoding{encoding_}, num_bits{num_bits_} {} - - constexpr bool MatchesEncoding(const IntegerEncodedValue& other) const { - return encoding == other.encoding && num_bits == other.num_bits; - } - - // Returns the number of bits required to encode num_vals values. - u32 GetBitLength(u32 num_vals) const { - u32 total_bits = num_bits * num_vals; - if (encoding == IntegerEncoding::Trit) { - total_bits += (num_vals * 8 + 4) / 5; - } else if (encoding == IntegerEncoding::Quint) { - total_bits += (num_vals * 7 + 2) / 3; - } - return total_bits; - } - - IntegerEncoding encoding{}; - u32 num_bits = 0; - u32 bit_value = 0; - union { - u32 quint_value = 0; - u32 trit_value; - }; -}; - -// Returns a new instance of this struct that corresponds to the -// can take no more than mav_value values -constexpr IntegerEncodedValue CreateEncoding(u32 mav_value) { - while (mav_value > 0) { - u32 check = mav_value + 1; - - // Is mav_value a power of two? - if (!(check & (check - 1))) { - return IntegerEncodedValue(IntegerEncoding::JustBits, std::popcount(mav_value)); - } - - // Is mav_value of the type 3*2^n - 1? - if ((check % 3 == 0) && !((check / 3) & ((check / 3) - 1))) { - return IntegerEncodedValue(IntegerEncoding::Trit, std::popcount(check / 3 - 1)); - } - - // Is mav_value of the type 5*2^n - 1? - if ((check % 5 == 0) && !((check / 5) & ((check / 5) - 1))) { - return IntegerEncodedValue(IntegerEncoding::Quint, std::popcount(check / 5 - 1)); - } - - // Apparently it can't be represented with a bounded integer sequence... - // just iterate. - mav_value--; - } - return IntegerEncodedValue(IntegerEncoding::JustBits, 0); -} - -constexpr std::array<IntegerEncodedValue, 256> MakeEncodedValues() { - std::array<IntegerEncodedValue, 256> encodings{}; - for (std::size_t i = 0; i < encodings.size(); ++i) { - encodings[i] = CreateEncoding(static_cast<u32>(i)); - } - return encodings; -} - -constexpr std::array<IntegerEncodedValue, 256> ASTC_ENCODINGS_VALUES = MakeEncodedValues(); - -// Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)] -// is the same as [(num_bits - 1):0] and repeats all the way down. -template <typename IntType> -constexpr IntType Replicate(IntType val, u32 num_bits, u32 to_bit) { - if (num_bits == 0 || to_bit == 0) { - return 0; - } - const IntType v = val & static_cast<IntType>((1 << num_bits) - 1); - IntType res = v; - u32 reslen = num_bits; - while (reslen < to_bit) { - u32 comp = 0; - if (num_bits > to_bit - reslen) { - u32 newshift = to_bit - reslen; - comp = num_bits - newshift; - num_bits = newshift; - } - res = static_cast<IntType>(res << num_bits); - res = static_cast<IntType>(res | (v >> comp)); - reslen += num_bits; - } - return res; -} - -constexpr std::size_t NumReplicateEntries(u32 num_bits) { - return std::size_t(1) << num_bits; -} - -template <typename IntType, u32 num_bits, u32 to_bit> -constexpr auto MakeReplicateTable() { - std::array<IntType, NumReplicateEntries(num_bits)> table{}; - for (IntType value = 0; value < static_cast<IntType>(std::size(table)); ++value) { - table[value] = Replicate(value, num_bits, to_bit); - } - return table; -} - -constexpr auto REPLICATE_6_BIT_TO_8_TABLE = MakeReplicateTable<u32, 6, 8>(); -constexpr auto REPLICATE_7_BIT_TO_8_TABLE = MakeReplicateTable<u32, 7, 8>(); -constexpr auto REPLICATE_8_BIT_TO_8_TABLE = MakeReplicateTable<u32, 8, 8>(); - void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth, uint32_t block_width, uint32_t block_height, std::span<uint8_t> output); diff --git a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp index aa173d19e..300a61205 100644 --- a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp +++ b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp @@ -228,7 +228,9 @@ void MemoryCommit::Release() { MemoryAllocator::MemoryAllocator(const Device& device_, bool export_allocations_) : device{device_}, properties{device_.GetPhysical().GetMemoryProperties()}, - export_allocations{export_allocations_} {} + export_allocations{export_allocations_}, + buffer_image_granularity{ + device_.GetPhysical().GetProperties().limits.bufferImageGranularity} {} MemoryAllocator::~MemoryAllocator() = default; @@ -258,7 +260,9 @@ MemoryCommit MemoryAllocator::Commit(const vk::Buffer& buffer, MemoryUsage usage } MemoryCommit MemoryAllocator::Commit(const vk::Image& image, MemoryUsage usage) { - auto commit = Commit(device.GetLogical().GetImageMemoryRequirements(*image), usage); + VkMemoryRequirements requirements = device.GetLogical().GetImageMemoryRequirements(*image); + requirements.size = Common::AlignUp(requirements.size, buffer_image_granularity); + auto commit = Commit(requirements, usage); image.BindMemory(commit.Memory(), commit.Offset()); return commit; } diff --git a/src/video_core/vulkan_common/vulkan_memory_allocator.h b/src/video_core/vulkan_common/vulkan_memory_allocator.h index b61e931e0..86e8ed119 100644 --- a/src/video_core/vulkan_common/vulkan_memory_allocator.h +++ b/src/video_core/vulkan_common/vulkan_memory_allocator.h @@ -123,6 +123,8 @@ private: const VkPhysicalDeviceMemoryProperties properties; ///< Physical device properties. const bool export_allocations; ///< True when memory allocations have to be exported. std::vector<std::unique_ptr<MemoryAllocation>> allocations; ///< Current allocations. + VkDeviceSize buffer_image_granularity; // The granularity for adjacent offsets between buffers + // and optimal images }; /// Returns true when a memory usage is guaranteed to be host visible. diff --git a/src/yuzu/configuration/configure_general.ui b/src/yuzu/configuration/configure_general.ui index 8ce97edec..69b6c2d66 100644 --- a/src/yuzu/configuration/configure_general.ui +++ b/src/yuzu/configuration/configure_general.ui @@ -25,6 +25,36 @@ <item> <layout class="QVBoxLayout" name="GeneralVerticalLayout"> <item> + <layout class="QHBoxLayout" name="horizontalLayout_2"> + <item> + <widget class="QLabel" name="fps_cap_label"> + <property name="text"> + <string>Framerate Cap</string> + </property> + <property name="toolTip"> + <string>Requires the use of the FPS Limiter Toggle hotkey to take effect.</string> + </property> + </widget> + </item> + <item> + <widget class="QSpinBox" name="fps_cap"> + <property name="suffix"> + <string>x</string> + </property> + <property name="minimum"> + <number>1</number> + </property> + <property name="maximum"> + <number>1000</number> + </property> + <property name="value"> + <number>500</number> + </property> + </widget> + </item> + </layout> + </item> + <item> <layout class="QHBoxLayout" name="horizontalLayout_2"> <item> <widget class="QCheckBox" name="toggle_speed_limit"> @@ -52,36 +82,6 @@ </layout> </item> <item> - <layout class="QHBoxLayout" name="horizontalLayout_2"> - <item> - <widget class="QLabel" name="fps_cap_label"> - <property name="text"> - <string>Framerate Cap</string> - </property> - <property name="toolTip"> - <string>Requires the use of the FPS Limiter Toggle hotkey to take effect.</string> - </property> - </widget> - </item> - <item> - <widget class="QSpinBox" name="fps_cap"> - <property name="suffix"> - <string>x</string> - </property> - <property name="minimum"> - <number>1</number> - </property> - <property name="maximum"> - <number>1000</number> - </property> - <property name="value"> - <number>500</number> - </property> - </widget> - </item> - </layout> - </item> - <item> <widget class="QCheckBox" name="use_multi_core"> <property name="text"> <string>Multicore CPU Emulation</string> diff --git a/src/yuzu/configuration/configure_graphics_advanced.ui b/src/yuzu/configuration/configure_graphics_advanced.ui index 379dc5d2e..4fe6b86ae 100644 --- a/src/yuzu/configuration/configure_graphics_advanced.ui +++ b/src/yuzu/configuration/configure_graphics_advanced.ui @@ -82,14 +82,17 @@ <string>Enables asynchronous shader compilation, which may reduce shader stutter. This feature is experimental.</string> </property> <property name="text"> - <string>Use asynchronous shader building</string> + <string>Use asynchronous shader building (hack)</string> </property> </widget> </item> <item> <widget class="QCheckBox" name="use_fast_gpu_time"> + <property name="toolTip"> + <string>Enables Fast GPU Time. This option will force most games to run at their highest native resolution.</string> + </property> <property name="text"> - <string>Use Fast GPU Time</string> + <string>Use Fast GPU Time (hack)</string> </property> </widget> </item> diff --git a/src/yuzu/configuration/configure_input_player.cpp b/src/yuzu/configuration/configure_input_player.cpp index 6b9bd05f1..7527c068b 100644 --- a/src/yuzu/configuration/configure_input_player.cpp +++ b/src/yuzu/configuration/configure_input_player.cpp @@ -309,11 +309,14 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i buttons_param[button_id].Clear(); button_map[button_id]->setText(tr("[not set]")); }); - context_menu.addAction(tr("Toggle button"), [&] { - const bool toggle_value = !buttons_param[button_id].Get("toggle", false); - buttons_param[button_id].Set("toggle", toggle_value); - button_map[button_id]->setText(ButtonToText(buttons_param[button_id])); - }); + if (buttons_param[button_id].Has("toggle")) { + context_menu.addAction(tr("Toggle button"), [&] { + const bool toggle_value = + !buttons_param[button_id].Get("toggle", false); + buttons_param[button_id].Set("toggle", toggle_value); + button_map[button_id]->setText(ButtonToText(buttons_param[button_id])); + }); + } if (buttons_param[button_id].Has("threshold")) { context_menu.addAction(tr("Set threshold"), [&] { const int button_threshold = static_cast<int>( diff --git a/src/yuzu/configuration/configure_input_player_widget.cpp b/src/yuzu/configuration/configure_input_player_widget.cpp index f50cda2f3..cd633e45f 100644 --- a/src/yuzu/configuration/configure_input_player_widget.cpp +++ b/src/yuzu/configuration/configure_input_player_widget.cpp @@ -122,6 +122,7 @@ void PlayerControlPreview::UpdateColors() { colors.slider_arrow = QColor(14, 15, 18); colors.font2 = QColor(255, 255, 255); colors.indicator = QColor(170, 238, 255); + colors.indicator2 = QColor(100, 255, 100); colors.deadzone = QColor(204, 136, 136); colors.slider_button = colors.button; } @@ -139,6 +140,7 @@ void PlayerControlPreview::UpdateColors() { colors.slider_arrow = QColor(65, 68, 73); colors.font2 = QColor(0, 0, 0); colors.indicator = QColor(0, 0, 200); + colors.indicator2 = QColor(0, 150, 0); colors.deadzone = QColor(170, 0, 0); colors.slider_button = QColor(153, 149, 149); } @@ -317,8 +319,7 @@ void PlayerControlPreview::DrawLeftController(QPainter& p, const QPointF center) using namespace Settings::NativeAnalog; DrawJoystick(p, center + QPointF(9, -69) + (axis_values[LStick].value * 8), 1.8f, button_values[Settings::NativeButton::LStick]); - DrawRawJoystick(p, center + QPointF(-140, 90), axis_values[LStick].raw_value, - axis_values[LStick].properties); + DrawRawJoystick(p, center + QPointF(-140, 90), QPointF(0, 0)); } using namespace Settings::NativeButton; @@ -432,8 +433,7 @@ void PlayerControlPreview::DrawRightController(QPainter& p, const QPointF center using namespace Settings::NativeAnalog; DrawJoystick(p, center + QPointF(-9, 11) + (axis_values[RStick].value * 8), 1.8f, button_values[Settings::NativeButton::RStick]); - DrawRawJoystick(p, center + QPointF(140, 90), axis_values[RStick].raw_value, - axis_values[RStick].properties); + DrawRawJoystick(p, QPointF(0, 0), center + QPointF(140, 90)); } using namespace Settings::NativeButton; @@ -547,8 +547,7 @@ void PlayerControlPreview::DrawDualController(QPainter& p, const QPointF center) DrawJoystick(p, center + QPointF(-65, -65) + (l_stick.value * 7), 1.62f, l_button); DrawJoystick(p, center + QPointF(65, 12) + (r_stick.value * 7), 1.62f, r_button); - DrawRawJoystick(p, center + QPointF(-180, 90), l_stick.raw_value, l_stick.properties); - DrawRawJoystick(p, center + QPointF(180, 90), r_stick.raw_value, r_stick.properties); + DrawRawJoystick(p, center + QPointF(-180, 90), center + QPointF(180, 90)); } using namespace Settings::NativeButton; @@ -634,8 +633,7 @@ void PlayerControlPreview::DrawHandheldController(QPainter& p, const QPointF cen DrawJoystick(p, center + QPointF(-171, -41) + (l_stick.value * 4), 1.0f, l_button); DrawJoystick(p, center + QPointF(171, 8) + (r_stick.value * 4), 1.0f, r_button); - DrawRawJoystick(p, center + QPointF(-50, 0), l_stick.raw_value, l_stick.properties); - DrawRawJoystick(p, center + QPointF(50, 0), r_stick.raw_value, r_stick.properties); + DrawRawJoystick(p, center + QPointF(-50, 0), center + QPointF(50, 0)); } using namespace Settings::NativeButton; @@ -728,10 +726,7 @@ void PlayerControlPreview::DrawProController(QPainter& p, const QPointF center) button_values[Settings::NativeButton::LStick]); DrawProJoystick(p, center + QPointF(51, 0), axis_values[RStick].value, 11, button_values[Settings::NativeButton::RStick]); - DrawRawJoystick(p, center + QPointF(-50, 105), axis_values[LStick].raw_value, - axis_values[LStick].properties); - DrawRawJoystick(p, center + QPointF(50, 105), axis_values[RStick].raw_value, - axis_values[RStick].properties); + DrawRawJoystick(p, center + QPointF(-50, 105), center + QPointF(50, 105)); } using namespace Settings::NativeButton; @@ -821,10 +816,7 @@ void PlayerControlPreview::DrawGCController(QPainter& p, const QPointF center) { p.setBrush(colors.font); DrawSymbol(p, center + QPointF(61, 37) + (axis_values[RStick].value * 9.5f), Symbol::C, 1.0f); - DrawRawJoystick(p, center + QPointF(-198, -125), axis_values[LStick].raw_value, - axis_values[LStick].properties); - DrawRawJoystick(p, center + QPointF(198, -125), axis_values[RStick].raw_value, - axis_values[RStick].properties); + DrawRawJoystick(p, center + QPointF(-198, -125), center + QPointF(198, -125)); } using namespace Settings::NativeButton; @@ -2358,8 +2350,33 @@ void PlayerControlPreview::DrawGCJoystick(QPainter& p, const QPointF center, boo DrawCircle(p, center, 7.5f); } -void PlayerControlPreview::DrawRawJoystick(QPainter& p, const QPointF center, const QPointF value, - const Input::AnalogProperties& properties) { +void PlayerControlPreview::DrawRawJoystick(QPainter& p, QPointF center_left, QPointF center_right) { + using namespace Settings::NativeAnalog; + if (controller_type != Settings::ControllerType::LeftJoycon) { + DrawJoystickProperties(p, center_right, axis_values[RStick].properties); + p.setPen(colors.indicator); + p.setBrush(colors.indicator); + DrawJoystickDot(p, center_right, axis_values[RStick].raw_value, + axis_values[RStick].properties); + p.setPen(colors.indicator2); + p.setBrush(colors.indicator2); + DrawJoystickDot(p, center_right, axis_values[RStick].value, axis_values[RStick].properties); + } + + if (controller_type != Settings::ControllerType::RightJoycon) { + DrawJoystickProperties(p, center_left, axis_values[LStick].properties); + p.setPen(colors.indicator); + p.setBrush(colors.indicator); + DrawJoystickDot(p, center_left, axis_values[LStick].raw_value, + axis_values[LStick].properties); + p.setPen(colors.indicator2); + p.setBrush(colors.indicator2); + DrawJoystickDot(p, center_left, axis_values[LStick].value, axis_values[LStick].properties); + } +} + +void PlayerControlPreview::DrawJoystickProperties(QPainter& p, const QPointF center, + const Input::AnalogProperties& properties) { constexpr float size = 45.0f; const float range = size * properties.range; const float deadzone = size * properties.deadzone; @@ -2376,10 +2393,14 @@ void PlayerControlPreview::DrawRawJoystick(QPainter& p, const QPointF center, co pen.setColor(colors.deadzone); p.setPen(pen); DrawCircle(p, center, deadzone); +} + +void PlayerControlPreview::DrawJoystickDot(QPainter& p, const QPointF center, const QPointF value, + const Input::AnalogProperties& properties) { + constexpr float size = 45.0f; + const float range = size * properties.range; // Dot pointer - p.setPen(colors.indicator); - p.setBrush(colors.indicator); DrawCircle(p, center + (value * range), 2); } diff --git a/src/yuzu/configuration/configure_input_player_widget.h b/src/yuzu/configuration/configure_input_player_widget.h index 5fc16d8af..f4a6a5e1b 100644 --- a/src/yuzu/configuration/configure_input_player_widget.h +++ b/src/yuzu/configuration/configure_input_player_widget.h @@ -90,6 +90,7 @@ private: QColor highlight2{}; QColor transparent{}; QColor indicator{}; + QColor indicator2{}; QColor led_on{}; QColor led_off{}; QColor slider{}; @@ -139,7 +140,10 @@ private: // Draw joystick functions void DrawJoystick(QPainter& p, QPointF center, float size, bool pressed); void DrawJoystickSideview(QPainter& p, QPointF center, float angle, float size, bool pressed); - void DrawRawJoystick(QPainter& p, QPointF center, QPointF value, + void DrawRawJoystick(QPainter& p, QPointF center_left, QPointF center_right); + void DrawJoystickProperties(QPainter& p, QPointF center, + const Input::AnalogProperties& properties); + void DrawJoystickDot(QPainter& p, QPointF center, QPointF value, const Input::AnalogProperties& properties); void DrawProJoystick(QPainter& p, QPointF center, QPointF offset, float scalar, bool pressed); void DrawGCJoystick(QPainter& p, QPointF center, bool pressed); diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp index 9544f0fb0..5940e0cfd 100644 --- a/src/yuzu/main.cpp +++ b/src/yuzu/main.cpp @@ -2814,8 +2814,6 @@ void GMainWindow::OnToggleFilterBar() { } void GMainWindow::OnCaptureScreenshot() { - OnPauseGame(); - const u64 title_id = Core::System::GetInstance().CurrentProcess()->GetTitleID(); const auto screenshot_path = QString::fromStdString(Common::FS::GetYuzuPathString(Common::FS::YuzuPath::ScreenshotsDir)); @@ -2827,23 +2825,22 @@ void GMainWindow::OnCaptureScreenshot() { .arg(date); if (!Common::FS::CreateDir(screenshot_path.toStdString())) { - OnStartGame(); return; } #ifdef _WIN32 if (UISettings::values.enable_screenshot_save_as) { + OnPauseGame(); filename = QFileDialog::getSaveFileName(this, tr("Capture Screenshot"), filename, tr("PNG Image (*.png)")); + OnStartGame(); if (filename.isEmpty()) { - OnStartGame(); return; } } #endif render_window->CaptureScreenshot(UISettings::values.screenshot_resolution_factor.GetValue(), filename); - OnStartGame(); } // TODO: Written 2020-10-01: Remove per-game config migration code when it is irrelevant diff --git a/src/yuzu_cmd/CMakeLists.txt b/src/yuzu_cmd/CMakeLists.txt index e55a19649..74fc24972 100644 --- a/src/yuzu_cmd/CMakeLists.txt +++ b/src/yuzu_cmd/CMakeLists.txt @@ -1,5 +1,6 @@ set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${PROJECT_SOURCE_DIR}/CMakeModules) +# Credits to Samantas5855 and others for this function. function(create_resource file output filename) # Read hex data from file file(READ ${file} filedata HEX) diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp index f643a4b0b..87fce0c23 100644 --- a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp +++ b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp @@ -122,6 +122,10 @@ void EmuWindow_SDL2::OnResize() { UpdateCurrentFramebufferLayout(width, height); } +void EmuWindow_SDL2::ShowCursor(bool show_cursor) { + SDL_ShowCursor(show_cursor ? SDL_ENABLE : SDL_DISABLE); +} + void EmuWindow_SDL2::Fullscreen() { switch (Settings::values.fullscreen_mode.GetValue()) { case Settings::FullscreenMode::Exclusive: @@ -228,6 +232,7 @@ void EmuWindow_SDL2::WaitEvent() { } } +// Credits to Samantas5855 and others for this function. void EmuWindow_SDL2::SetWindowIcon() { SDL_RWops* const yuzu_icon_stream = SDL_RWFromConstMem((void*)yuzu_icon, yuzu_icon_size); if (yuzu_icon_stream == nullptr) { diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2.h b/src/yuzu_cmd/emu_window/emu_window_sdl2.h index aa0d52ae4..4810f8775 100644 --- a/src/yuzu_cmd/emu_window/emu_window_sdl2.h +++ b/src/yuzu_cmd/emu_window/emu_window_sdl2.h @@ -67,6 +67,9 @@ protected: /// Called by WaitEvent when any event that may cause the window to be resized occurs void OnResize(); + /// Called when users want to hide the mouse cursor + void ShowCursor(bool show_cursor); + /// Called when user passes the fullscreen parameter flag void Fullscreen(); diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp index 5b98c255b..a075ad08a 100644 --- a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp +++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp @@ -111,6 +111,7 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(InputCommon::InputSubsystem* input_subsyste if (fullscreen) { Fullscreen(); + ShowCursor(false); } window_context = SDL_GL_CreateContext(render_window); diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp index cdda375d8..de40b76bf 100644 --- a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp +++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp @@ -45,6 +45,7 @@ EmuWindow_SDL2_VK::EmuWindow_SDL2_VK(InputCommon::InputSubsystem* input_subsyste if (fullscreen) { Fullscreen(); + ShowCursor(false); } switch (wm.subsystem) { |