From 0c13304722ead138d2d5548297e1f195dac57bba Mon Sep 17 00:00:00 2001 From: JPikachu Date: Sun, 22 Jun 2025 13:52:17 +0100 Subject: [PATCH] texture_cache: Improve SD865 Means games like Minecraft Dungeons, Sea of Stars, Luigi Mansion 2, Astroneer, etc are now playable. It also cleans up the recent abi.cpp and bindless texture commits a bit. Everything is in #ifdef ANDROID - The biggest change is CACHING_PAGEBITS = 12. Without that the way the buffercache grows and joins buffers can cause Android to run out of memory (as you end up with just one big buffer that needs to be copied every time it grows) --- .../dynarmic/src/dynarmic/backend/x64/abi.cpp | 25 +++---------------- src/core/hle/service/sockets/sfdnsres.cpp | 14 +++++++++-- .../maxwell/translate/impl/warp_shuffle.cpp | 20 ++++++++++++++- src/shader_recompiler/ir_opt/texture_pass.cpp | 23 ++++------------- src/video_core/buffer_cache/buffer_cache.h | 7 +++++- .../buffer_cache/buffer_cache_base.h | 20 ++++++++++++++- src/video_core/cdma_pusher.cpp | 6 +++++ .../renderer_vulkan/present/layer.cpp | 4 +++ src/video_core/texture_cache/texture_cache.h | 8 ++++++ .../texture_cache/texture_cache_base.h | 14 +++++++++++ .../vulkan_common/vulkan_memory_allocator.cpp | 8 +++++- 11 files changed, 104 insertions(+), 45 deletions(-) diff --git a/externals/dynarmic/src/dynarmic/backend/x64/abi.cpp b/externals/dynarmic/src/dynarmic/backend/x64/abi.cpp index 7cbe92aaf3..c97d98d9ae 100644 --- a/externals/dynarmic/src/dynarmic/backend/x64/abi.cpp +++ b/externals/dynarmic/src/dynarmic/backend/x64/abi.cpp @@ -116,18 +116,13 @@ void ABI_PopCallerSaveRegistersAndAdjustStack(BlockOfCode& code, const std::size ABI_PopRegistersAndAdjustStack(code, frame_size, ABI_ALL_CALLER_SAVE); } -static consteval size_t ABI_AllCallerSaveSize() noexcept { - return ABI_ALL_CALLER_SAVE.max_size(); -} -static consteval std::array ABI_AllCallerSaveExcept(const std::size_t except) noexcept { - std::array arr; - for(std::size_t i = 0; i < arr.size(); ++i) { - arr[i] = static_cast(i + (i >= except ? 1 : 0)); - } +static std::vector ABI_AllCallerSaveExcept(const std::size_t except) noexcept { + std::vector arr; + std::remove_copy(ABI_ALL_CALLER_SAVE.begin(), ABI_ALL_CALLER_SAVE.end(), std::back_inserter(arr), static_cast(except)); return arr; } -alignas(64) static constinit std::array ABI_CALLER_SAVED_EXCEPT_TABLE[32] = { +alignas(64) static std::vector ABI_CALLER_SAVED_EXCEPT_TABLE[32] = { ABI_AllCallerSaveExcept(0), ABI_AllCallerSaveExcept(1), ABI_AllCallerSaveExcept(2), @@ -163,24 +158,12 @@ alignas(64) static constinit std::array AB }; void ABI_PushCallerSaveRegistersAndAdjustStackExcept(BlockOfCode& code, const HostLoc exception) { -#ifdef _MSC_VER - std::vector regs; - std::remove_copy(ABI_ALL_CALLER_SAVE.begin(), ABI_ALL_CALLER_SAVE.end(), std::back_inserter(regs), exception); - ABI_PushRegistersAndAdjustStack(code, 0, regs); -#else ASSUME(size_t(exception) < 32); ABI_PushRegistersAndAdjustStack(code, 0, ABI_CALLER_SAVED_EXCEPT_TABLE[size_t(exception)]); -#endif } void ABI_PopCallerSaveRegistersAndAdjustStackExcept(BlockOfCode& code, const HostLoc exception) { -#ifdef _MSC_VER - std::vector regs; - std::remove_copy(ABI_ALL_CALLER_SAVE.begin(), ABI_ALL_CALLER_SAVE.end(), std::back_inserter(regs), exception); - ABI_PopRegistersAndAdjustStack(code, 0, regs); -#else ASSUME(size_t(exception) < 32); ABI_PopRegistersAndAdjustStack(code, 0, ABI_CALLER_SAVED_EXCEPT_TABLE[size_t(exception)]); -#endif } } // namespace Dynarmic::Backend::X64 diff --git a/src/core/hle/service/sockets/sfdnsres.cpp b/src/core/hle/service/sockets/sfdnsres.cpp index c657c4efd8..034a4b013c 100644 --- a/src/core/hle/service/sockets/sfdnsres.cpp +++ b/src/core/hle/service/sockets/sfdnsres.cpp @@ -151,7 +151,12 @@ static std::pair GetHostByNameRequestImpl(HLERequestConte // For now, ignore options, which are in input buffer 1 for GetHostByNameRequestWithOptions. // Prevent resolution of Nintendo servers - if (host.find("srv.nintendo.net") != std::string::npos) { + if (host.find("srv.nintendo.net") != std::string::npos || + host.find("battle.net") != std::string::npos || + host.find("microsoft.com") != std::string::npos || + host.find("mojang.com") != std::string::npos || + host.find("xboxlive.com") != std::string::npos || + host.find("minecraftservices.com") != std::string::npos) { LOG_WARNING(Network, "Resolution of hostname {} requested, returning EAI_AGAIN", host); return {0, GetAddrInfoError::AGAIN}; } @@ -268,7 +273,12 @@ static std::pair GetAddrInfoRequestImpl(HLERequestContext const std::string host = Common::StringFromBuffer(host_buffer); // Prevent resolution of Nintendo servers - if (host.find("srv.nintendo.net") != std::string::npos) { + if (host.find("srv.nintendo.net") != std::string::npos || + host.find("battle.net") != std::string::npos || + host.find("microsoft.com") != std::string::npos || + host.find("mojang.com") != std::string::npos || + host.find("xboxlive.com") != std::string::npos || + host.find("minecraftservices.com") != std::string::npos) { LOG_WARNING(Network, "Resolution of hostname {} requested, returning EAI_AGAIN", host); return {0, GetAddrInfoError::AGAIN}; } diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/warp_shuffle.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/warp_shuffle.cpp index f0436994b2..b68d7ed7ad 100644 --- a/src/shader_recompiler/frontend/maxwell/translate/impl/warp_shuffle.cpp +++ b/src/shader_recompiler/frontend/maxwell/translate/impl/warp_shuffle.cpp @@ -1,9 +1,13 @@ +// SPDX-FileCopyrightText: 2025 Eden Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later #include "common/bit_field.h" #include "common/common_types.h" #include "shader_recompiler/frontend/maxwell/translate/impl/impl.h" +#include namespace Shader::Maxwell { namespace { @@ -33,6 +37,17 @@ enum class ShuffleMode : u64 { } } +bool IsKONA() { + std::ifstream machineFile("/sys/devices/soc0/machine"); + if (machineFile.is_open()) { + std::string line; + std::getline(machineFile, line); + if (line == "KONA") + return true; + } + return false; +} + void Shuffle(TranslatorVisitor& v, u64 insn, const IR::U32& index, const IR::U32& mask) { union { u64 insn; @@ -44,7 +59,10 @@ void Shuffle(TranslatorVisitor& v, u64 insn, const IR::U32& index, const IR::U32 const IR::U32 result{ShuffleOperation(v.ir, v.X(shfl.src_reg), index, mask, shfl.mode)}; v.ir.SetPred(shfl.pred, v.ir.GetInBoundsFromOp(result)); - v.X(shfl.dest_reg, result); + if (IsKONA()) + v.X(shfl.dest_reg, v.ir.Imm32(0xffffffff)); // This fixes the freeze for Retroid / Snapdragon SD865 + else + v.X(shfl.dest_reg, result); } } // Anonymous namespace diff --git a/src/shader_recompiler/ir_opt/texture_pass.cpp b/src/shader_recompiler/ir_opt/texture_pass.cpp index ef1ef63500..4d1f5588df 100644 --- a/src/shader_recompiler/ir_opt/texture_pass.cpp +++ b/src/shader_recompiler/ir_opt/texture_pass.cpp @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2025 Eden Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later @@ -327,30 +330,14 @@ std::optional TryGetConstBuffer(const IR::Inst* inst, Environme }; } -// TODO:xbzk: shall be dropped when Track method cover all bindless stuff -static ConstBufferAddr last_valid_addr = ConstBufferAddr{ - .index = 0, - .offset = 0, - .shift_left = 0, - .secondary_index = 0, - .secondary_offset = 0, - .secondary_shift_left = 0, - .dynamic_offset = {}, - .count = 1, - .has_secondary = false, -}; - TextureInst MakeInst(Environment& env, IR::Block* block, IR::Inst& inst) { - ConstBufferAddr addr; + ConstBufferAddr addr{}; if (IsBindless(inst)) { const std::optional track_addr{Track(inst.Arg(0), env)}; - if (!track_addr) { - //throw NotImplementedException("Failed to track bindless texture constant buffer"); - addr = last_valid_addr; // TODO:xbzk: shall be dropped when Track method cover all bindless stuff + LOG_WARNING(Shader, "Failed to track bindless texture constant buffer"); } else { addr = *track_addr; - last_valid_addr = addr; // TODO:xbzk: shall be dropped when Track method cover all bindless stuff } } else { addr = ConstBufferAddr{ diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index af237703d5..57e7fda059 100644 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2025 Eden Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project // SPDX-License-Identifier: GPL-3.0-or-later @@ -42,6 +45,8 @@ BufferCache

::BufferCache(Tegra::MaxwellDeviceMemoryManager& device_memory_, R critical_memory = static_cast( std::max(std::min(device_local_memory - min_vacancy_critical, min_spacing_critical), DEFAULT_CRITICAL_MEMORY)); + + LOG_INFO(Render_Vulkan, "BufferCache - device_local_memory: {} minimum_memory: {} critical_memory: {}", device_local_memory, minimum_memory, critical_memory); } template @@ -50,7 +55,7 @@ BufferCache

::~BufferCache() = default; template void BufferCache

::RunGarbageCollector() { const bool aggressive_gc = total_used_memory >= critical_memory; - const u64 ticks_to_destroy = aggressive_gc ? 60 : 120; + const u64 ticks_to_destroy = aggressive_gc ? 60 : 150; int num_iterations = aggressive_gc ? 64 : 32; const auto clean_up = [this, &num_iterations](BufferId buffer_id) { if (num_iterations == 0) { diff --git a/src/video_core/buffer_cache/buffer_cache_base.h b/src/video_core/buffer_cache/buffer_cache_base.h index 240e9f0150..e799f89d39 100644 --- a/src/video_core/buffer_cache/buffer_cache_base.h +++ b/src/video_core/buffer_cache/buffer_cache_base.h @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2025 Eden Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project // SPDX-License-Identifier: GPL-3.0-or-later @@ -156,7 +159,11 @@ template class BufferCache : public VideoCommon::ChannelSetupCaches { // Page size for caching purposes. // This is unrelated to the CPU page size and it can be changed as it seems optimal. +#ifdef ANDROID + static constexpr u32 CACHING_PAGEBITS = 12; +#else static constexpr u32 CACHING_PAGEBITS = 16; +#endif static constexpr u64 CACHING_PAGESIZE = u64{1} << CACHING_PAGEBITS; static constexpr bool IS_OPENGL = P::IS_OPENGL; @@ -170,9 +177,15 @@ class BufferCache : public VideoCommon::ChannelSetupCaches slot_buffers; - DelayedDestructionRing delayed_destruction_ring; +#ifdef ANDROID + static constexpr size_t TICKS_TO_DESTROY = 6; +#else + static constexpr size_t TICKS_TO_DESTROY = 8; +#endif + DelayedDestructionRing delayed_destruction_ring; const Tegra::Engines::DrawManager::IndirectParams* current_draw_indirect{}; diff --git a/src/video_core/cdma_pusher.cpp b/src/video_core/cdma_pusher.cpp index 3bcf1b0664..bd73311c8f 100644 --- a/src/video_core/cdma_pusher.cpp +++ b/src/video_core/cdma_pusher.cpp @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2025 Eden Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + // SPDX-FileCopyrightText: Ryujinx Team and Contributors // SPDX-License-Identifier: MIT @@ -20,6 +23,9 @@ CDmaPusher::CDmaPusher(Host1x::Host1x& host1x_, s32 id) : host1x{host1x_}, memory_manager{host1x.GMMU()}, host_processor{std::make_unique(host1x_)}, current_class{ static_cast(id)} { +#ifdef ANDROID + std::this_thread::sleep_for(std::chrono::milliseconds{500}); // HACK: Fix for Astroneer - doesn't always start without this delay. Happens on Windows too (but rarer) +#endif thread = std::jthread([this](std::stop_token stop_token) { ProcessEntries(stop_token); }); } diff --git a/src/video_core/renderer_vulkan/present/layer.cpp b/src/video_core/renderer_vulkan/present/layer.cpp index 4e41afe5b4..f97015df35 100644 --- a/src/video_core/renderer_vulkan/present/layer.cpp +++ b/src/video_core/renderer_vulkan/present/layer.cpp @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2025 Eden Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + // SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later @@ -137,6 +140,7 @@ void Layer::CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer) { .pQueueFamilyIndices = nullptr, }; + buffer.reset(); buffer = memory_allocator.CreateBuffer(ci, MemoryUsage::Upload); } diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index feb1c575eb..e335a69b04 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2025 Eden Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + // SPDX-FileCopyrightText: 2023 yuzu Emulator Project // SPDX-License-Identifier: GPL-3.0-or-later @@ -63,6 +66,8 @@ TextureCache

::TextureCache(Runtime& runtime_, Tegra::MaxwellDeviceMemoryManag std::max(std::min(device_local_memory - min_vacancy_critical, min_spacing_critical), DEFAULT_CRITICAL_MEMORY)); minimum_memory = static_cast((device_local_memory - mem_threshold) / 2); + + LOG_INFO(Render_Vulkan, "TextureCache - device_local_memory:{} minimum_memory: {} critical_memory: {} expected_memory: {}", device_local_memory, minimum_memory, critical_memory, expected_memory); } else { expected_memory = DEFAULT_EXPECTED_MEMORY + 512_MiB; critical_memory = DEFAULT_CRITICAL_MEMORY + 1_GiB; @@ -78,6 +83,9 @@ void TextureCache

::RunGarbageCollector() { size_t num_iterations = 0; const auto Configure = [&](bool allow_aggressive) { +#ifdef ANDROID + high_priority_mode = true; +#endif high_priority_mode = total_used_memory >= expected_memory; aggressive_mode = allow_aggressive && total_used_memory >= critical_memory; ticks_to_destroy = aggressive_mode ? 10ULL : high_priority_mode ? 25ULL : 50ULL; diff --git a/src/video_core/texture_cache/texture_cache_base.h b/src/video_core/texture_cache/texture_cache_base.h index da98a634b5..30ae5e2acc 100644 --- a/src/video_core/texture_cache/texture_cache_base.h +++ b/src/video_core/texture_cache/texture_cache_base.h @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2025 Eden Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + // SPDX-FileCopyrightText: 2023 yuzu Emulator Project // SPDX-License-Identifier: GPL-3.0-or-later @@ -107,10 +110,17 @@ class TextureCache : public VideoCommon::ChannelSetupCaches::max()}; + #ifdef ANDROID + static constexpr s64 TARGET_THRESHOLD = 3_GiB; + static constexpr s64 DEFAULT_EXPECTED_MEMORY = 1_GiB + 125_MiB; + static constexpr s64 DEFAULT_CRITICAL_MEMORY = 1_GiB + 625_MiB; + static constexpr size_t GC_EMERGENCY_COUNTS = 2; + #else static constexpr s64 TARGET_THRESHOLD = 4_GiB; static constexpr s64 DEFAULT_EXPECTED_MEMORY = 1_GiB + 125_MiB; static constexpr s64 DEFAULT_CRITICAL_MEMORY = 1_GiB + 625_MiB; static constexpr size_t GC_EMERGENCY_COUNTS = 2; + #endif using Runtime = typename P::Runtime; using Image = typename P::Image; @@ -476,7 +486,11 @@ private: }; Common::LeastRecentlyUsedCache lru_cache; + #ifdef ANDROID + static constexpr size_t TICKS_TO_DESTROY = 5; + #else static constexpr size_t TICKS_TO_DESTROY = 8; +#endif DelayedDestructionRing sentenced_images; DelayedDestructionRing sentenced_image_view; DelayedDestructionRing sentenced_framebuffers; diff --git a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp index 54331688e3..e1abd30e75 100644 --- a/src/video_core/vulkan_common/vulkan_memory_allocator.cpp +++ b/src/video_core/vulkan_common/vulkan_memory_allocator.cpp @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2025 Eden Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later @@ -268,7 +271,10 @@ vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo& ci, MemoryUsa VmaAllocation allocation{}; VkMemoryPropertyFlags property_flags{}; - vk::Check(vmaCreateBuffer(allocator, &ci, &alloc_ci, &handle, &allocation, &alloc_info)); + VkResult result = vmaCreateBuffer(allocator, &ci, &alloc_ci, &handle, &allocation, &alloc_info); + if (result == VK_ERROR_OUT_OF_DEVICE_MEMORY) { + LOG_ERROR(Render_Vulkan, "Out of memory creating buffer (size: {})", ci.size); + } vmaGetAllocationMemoryProperties(allocator, allocation, &property_flags); u8* data = reinterpret_cast(alloc_info.pMappedData);