summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--.ci/templates/build-single.yml14
-rw-r--r--.ci/templates/build-standard.yml3
-rw-r--r--.ci/templates/build-testing.yml3
-rw-r--r--.ci/yuzu-mainline.yml2
-rw-r--r--.ci/yuzu-verify.yml4
-rw-r--r--CMakeModules/GenerateSCMRev.cmake1
-rw-r--r--src/common/CMakeLists.txt1
-rw-r--r--src/common/alignment.h60
-rw-r--r--src/core/CMakeLists.txt3
-rw-r--r--src/core/core.cpp12
-rw-r--r--src/core/core.h10
-rw-r--r--src/core/hardware_interrupt_manager.cpp30
-rw-r--r--src/core/hardware_interrupt_manager.h31
-rw-r--r--src/core/hle/kernel/code_set.h3
-rw-r--r--src/core/hle/kernel/physical_memory.h19
-rw-r--r--src/core/hle/kernel/process.cpp6
-rw-r--r--src/core/hle/kernel/shared_memory.cpp6
-rw-r--r--src/core/hle/kernel/shared_memory.h13
-rw-r--r--src/core/hle/kernel/transfer_memory.cpp2
-rw-r--r--src/core/hle/kernel/transfer_memory.h3
-rw-r--r--src/core/hle/kernel/vm_manager.cpp15
-rw-r--r--src/core/hle/kernel/vm_manager.h9
-rw-r--r--src/core/hle/service/audio/audio.cpp6
-rw-r--r--src/core/hle/service/audio/audio.h6
-rw-r--r--src/core/hle/service/audio/audout_u.cpp36
-rw-r--r--src/core/hle/service/audio/audout_u.h12
-rw-r--r--src/core/hle/service/audio/audren_u.cpp200
-rw-r--r--src/core/hle/service/audio/audren_u.h25
-rw-r--r--src/core/hle/service/ns/pl_u.cpp12
-rw-r--r--src/core/hle/service/nvdrv/devices/nvdevice.h13
-rw-r--r--src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp11
-rw-r--r--src/core/hle/service/nvdrv/devices/nvdisp_disp0.h5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp15
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp152
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_ctrl.h15
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp7
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp44
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_gpu.h41
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_nvdec.h5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_vic.cpp5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvhost_vic.h5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvmap.cpp5
-rw-r--r--src/core/hle/service/nvdrv/devices/nvmap.h5
-rw-r--r--src/core/hle/service/nvdrv/interface.cpp48
-rw-r--r--src/core/hle/service/nvdrv/interface.h4
-rw-r--r--src/core/hle/service/nvdrv/nvdata.h48
-rw-r--r--src/core/hle/service/nvdrv/nvdrv.cpp59
-rw-r--r--src/core/hle/service/nvdrv/nvdrv.h88
-rw-r--r--src/core/hle/service/nvflinger/buffer_queue.cpp23
-rw-r--r--src/core/hle/service/nvflinger/buffer_queue.h11
-rw-r--r--src/core/hle/service/nvflinger/nvflinger.cpp23
-rw-r--r--src/core/hle/service/nvflinger/nvflinger.h4
-rw-r--r--src/core/hle/service/service.cpp4
-rw-r--r--src/core/hle/service/vi/vi.cpp48
-rw-r--r--src/core/loader/elf.cpp2
-rw-r--r--src/core/loader/kip.cpp2
-rw-r--r--src/core/loader/nro.cpp2
-rw-r--r--src/core/loader/nso.cpp2
-rw-r--r--src/video_core/CMakeLists.txt5
-rw-r--r--src/video_core/buffer_cache.h299
-rw-r--r--src/video_core/buffer_cache/buffer_block.h76
-rw-r--r--src/video_core/buffer_cache/buffer_cache.h447
-rw-r--r--src/video_core/buffer_cache/map_interval.h89
-rw-r--r--src/video_core/dma_pusher.cpp1
-rw-r--r--src/video_core/engines/fermi_2d.cpp3
-rw-r--r--src/video_core/engines/fermi_2d.h3
-rw-r--r--src/video_core/engines/kepler_memory.cpp2
-rw-r--r--src/video_core/engines/kepler_memory.h1
-rw-r--r--src/video_core/engines/maxwell_3d.cpp15
-rw-r--r--src/video_core/engines/maxwell_dma.cpp8
-rw-r--r--src/video_core/engines/maxwell_dma.h9
-rw-r--r--src/video_core/engines/shader_bytecode.h25
-rw-r--r--src/video_core/gpu.cpp76
-rw-r--r--src/video_core/gpu.h49
-rw-r--r--src/video_core/gpu_asynch.cpp9
-rw-r--r--src/video_core/gpu_asynch.h3
-rw-r--r--src/video_core/gpu_synch.cpp2
-rw-r--r--src/video_core/gpu_synch.h4
-rw-r--r--src/video_core/gpu_thread.cpp27
-rw-r--r--src/video_core/gpu_thread.h32
-rw-r--r--src/video_core/rasterizer_interface.h3
-rw-r--r--src/video_core/renderer_opengl/gl_buffer_cache.cpp52
-rw-r--r--src/video_core/renderer_opengl/gl_buffer_cache.h39
-rw-r--r--src/video_core/renderer_opengl/gl_device.cpp3
-rw-r--r--src/video_core/renderer_opengl/gl_device.h5
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp8
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.h1
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.cpp14
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.h1
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp67
-rw-r--r--src/video_core/renderer_opengl/gl_texture_cache.cpp5
-rw-r--r--src/video_core/renderer_opengl/gl_texture_cache.h2
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_decompiler.cpp43
-rw-r--r--src/video_core/shader/control_flow.cpp47
-rw-r--r--src/video_core/shader/control_flow.h30
-rw-r--r--src/video_core/shader/decode.cpp5
-rw-r--r--src/video_core/shader/decode/arithmetic.cpp13
-rw-r--r--src/video_core/shader/decode/arithmetic_half_immediate.cpp4
-rw-r--r--src/video_core/shader/decode/conversion.cpp30
-rw-r--r--src/video_core/shader/decode/ffma.cpp10
-rw-r--r--src/video_core/shader/decode/float_set.cpp1
-rw-r--r--src/video_core/shader/decode/float_set_predicate.cpp1
-rw-r--r--src/video_core/shader/decode/half_set_predicate.cpp10
-rw-r--r--src/video_core/shader/decode/hfma2.cpp4
-rw-r--r--src/video_core/shader/decode/integer_set.cpp1
-rw-r--r--src/video_core/shader/decode/integer_set_predicate.cpp1
-rw-r--r--src/video_core/shader/decode/other.cpp13
-rw-r--r--src/video_core/shader/decode/predicate_set_register.cpp1
-rw-r--r--src/video_core/shader/decode/warp.cpp55
-rw-r--r--src/video_core/shader/node.h30
-rw-r--r--src/video_core/shader/shader_ir.cpp5
-rw-r--r--src/video_core/shader/shader_ir.h4
-rw-r--r--src/video_core/shader/track.cpp4
-rw-r--r--src/video_core/texture_cache/surface_base.cpp5
-rw-r--r--src/video_core/texture_cache/surface_params.h1
-rw-r--r--src/video_core/texture_cache/texture_cache.h2
-rw-r--r--src/video_core/textures/texture.h2
-rw-r--r--src/yuzu/CMakeLists.txt55
-rw-r--r--src/yuzu/configuration/config.cpp2
-rw-r--r--src/yuzu/configuration/configure_debug.cpp4
-rw-r--r--src/yuzu/configuration/configure_gamelist.cpp2
-rw-r--r--src/yuzu/configuration/configure_general.cpp2
-rw-r--r--src/yuzu/configuration/configure_input.cpp12
-rw-r--r--src/yuzu/configuration/configure_input_player.cpp10
-rw-r--r--src/yuzu/configuration/configure_input_simple.cpp4
-rw-r--r--src/yuzu/configuration/configure_mouse_advanced.cpp6
-rw-r--r--src/yuzu/configuration/configure_per_general.cpp2
-rw-r--r--src/yuzu/configuration/configure_profile_manager.cpp8
-rw-r--r--src/yuzu/configuration/configure_touchscreen_advanced.cpp2
-rw-r--r--src/yuzu/configuration/configure_web.cpp2
-rw-r--r--src/yuzu/debugger/console.cpp2
-rw-r--r--src/yuzu/discord_impl.cpp2
-rw-r--r--src/yuzu/game_list.cpp2
-rw-r--r--src/yuzu/game_list_p.h2
-rw-r--r--src/yuzu/game_list_worker.cpp2
-rw-r--r--src/yuzu/hotkeys.cpp2
-rw-r--r--src/yuzu/main.cpp17
-rw-r--r--src/yuzu/uisettings.cpp (renamed from src/yuzu/ui_settings.cpp)2
-rw-r--r--src/yuzu/uisettings.h (renamed from src/yuzu/ui_settings.h)0
-rw-r--r--src/yuzu_tester/yuzu.cpp3
145 files changed, 2203 insertions, 917 deletions
diff --git a/.ci/templates/build-single.yml b/.ci/templates/build-single.yml
index c411e25d1..357731eb9 100644
--- a/.ci/templates/build-single.yml
+++ b/.ci/templates/build-single.yml
@@ -1,17 +1,19 @@
parameters:
artifactSource: 'true'
+ cache: 'false'
steps:
- task: DockerInstaller@0
displayName: 'Prepare Environment'
inputs:
dockerVersion: '17.09.0-ce'
-- task: CacheBeta@0
- displayName: 'Cache Build System'
- inputs:
- key: yuzu-v1-$(BuildName)-$(BuildSuffix)-$(CacheSuffix)
- path: $(System.DefaultWorkingDirectory)/ccache
- cacheHitVar: CACHE_RESTORED
+- ${{ if eq(parameters.cache, 'true') }}:
+ - task: CacheBeta@0
+ displayName: 'Cache Build System'
+ inputs:
+ key: yuzu-v1-$(BuildName)-$(BuildSuffix)-$(CacheSuffix)
+ path: $(System.DefaultWorkingDirectory)/ccache
+ cacheHitVar: CACHE_RESTORED
- script: chmod a+x ./.ci/scripts/$(ScriptFolder)/exec.sh && ./.ci/scripts/$(ScriptFolder)/exec.sh
displayName: 'Build'
- script: chmod a+x ./.ci/scripts/$(ScriptFolder)/upload.sh && RELEASE_NAME=$(BuildName) ./.ci/scripts/$(ScriptFolder)/upload.sh
diff --git a/.ci/templates/build-standard.yml b/.ci/templates/build-standard.yml
index 6cd209dbf..aa180894e 100644
--- a/.ci/templates/build-standard.yml
+++ b/.ci/templates/build-standard.yml
@@ -19,4 +19,5 @@ jobs:
needSubmodules: 'true'
- template: ./build-single.yml
parameters:
- artifactSource: 'false' \ No newline at end of file
+ artifactSource: 'false'
+ cache: $(parameters.cache) \ No newline at end of file
diff --git a/.ci/templates/build-testing.yml b/.ci/templates/build-testing.yml
index cb7736205..a307addfd 100644
--- a/.ci/templates/build-testing.yml
+++ b/.ci/templates/build-testing.yml
@@ -10,7 +10,7 @@ jobs:
BuildSuffix: 'windows-testing'
ScriptFolder: 'windows'
steps:
- - script: pip install requests urllib3
+ - script: sudo apt upgrade python3-pip && pip install requests urllib3
displayName: 'Prepare Environment'
- task: PythonScript@0
condition: eq(variables['Build.Reason'], 'PullRequest')
@@ -30,3 +30,4 @@ jobs:
- template: ./build-single.yml
parameters:
artifactSource: 'false'
+ cache: 'false'
diff --git a/.ci/yuzu-mainline.yml b/.ci/yuzu-mainline.yml
index 164bcb165..2930a8564 100644
--- a/.ci/yuzu-mainline.yml
+++ b/.ci/yuzu-mainline.yml
@@ -21,3 +21,5 @@ stages:
dependsOn: format
jobs:
- template: ./templates/build-standard.yml
+ parameters:
+ cache: 'true'
diff --git a/.ci/yuzu-verify.yml b/.ci/yuzu-verify.yml
index d01c1feed..5492e696a 100644
--- a/.ci/yuzu-verify.yml
+++ b/.ci/yuzu-verify.yml
@@ -15,4 +15,6 @@ stages:
dependsOn: format
jobs:
- template: ./templates/build-standard.yml
- - template: ./templates/build-testing.yml \ No newline at end of file
+ parameters:
+ cache: 'false'
+ - template: ./templates/build-testing.yml
diff --git a/CMakeModules/GenerateSCMRev.cmake b/CMakeModules/GenerateSCMRev.cmake
index abdc74428..a1ace89cb 100644
--- a/CMakeModules/GenerateSCMRev.cmake
+++ b/CMakeModules/GenerateSCMRev.cmake
@@ -81,6 +81,7 @@ set(HASH_FILES
"${VIDEO_CORE}/shader/decode/register_set_predicate.cpp"
"${VIDEO_CORE}/shader/decode/shift.cpp"
"${VIDEO_CORE}/shader/decode/video.cpp"
+ "${VIDEO_CORE}/shader/decode/warp.cpp"
"${VIDEO_CORE}/shader/decode/xmad.cpp"
"${VIDEO_CORE}/shader/control_flow.cpp"
"${VIDEO_CORE}/shader/control_flow.h"
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index 2b4266f29..01abdb3bb 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -55,6 +55,7 @@ add_custom_command(OUTPUT scm_rev.cpp
"${VIDEO_CORE}/shader/decode/register_set_predicate.cpp"
"${VIDEO_CORE}/shader/decode/shift.cpp"
"${VIDEO_CORE}/shader/decode/video.cpp"
+ "${VIDEO_CORE}/shader/decode/warp.cpp"
"${VIDEO_CORE}/shader/decode/xmad.cpp"
"${VIDEO_CORE}/shader/control_flow.cpp"
"${VIDEO_CORE}/shader/control_flow.h"
diff --git a/src/common/alignment.h b/src/common/alignment.h
index 617b14d9b..88d5d3a65 100644
--- a/src/common/alignment.h
+++ b/src/common/alignment.h
@@ -3,6 +3,7 @@
#pragma once
#include <cstddef>
+#include <memory>
#include <type_traits>
namespace Common {
@@ -37,4 +38,63 @@ constexpr bool IsWordAligned(T value) {
return (value & 0b11) == 0;
}
+template <typename T, std::size_t Align = 16>
+class AlignmentAllocator {
+public:
+ using value_type = T;
+ using size_type = std::size_t;
+ using difference_type = std::ptrdiff_t;
+
+ using pointer = T*;
+ using const_pointer = const T*;
+
+ using reference = T&;
+ using const_reference = const T&;
+
+public:
+ pointer address(reference r) noexcept {
+ return std::addressof(r);
+ }
+
+ const_pointer address(const_reference r) const noexcept {
+ return std::addressof(r);
+ }
+
+ pointer allocate(size_type n) {
+ return static_cast<pointer>(::operator new (n, std::align_val_t{Align}));
+ }
+
+ void deallocate(pointer p, size_type) {
+ ::operator delete (p, std::align_val_t{Align});
+ }
+
+ void construct(pointer p, const value_type& wert) {
+ new (p) value_type(wert);
+ }
+
+ void destroy(pointer p) {
+ p->~value_type();
+ }
+
+ size_type max_size() const noexcept {
+ return size_type(-1) / sizeof(value_type);
+ }
+
+ template <typename T2>
+ struct rebind {
+ using other = AlignmentAllocator<T2, Align>;
+ };
+
+ bool operator!=(const AlignmentAllocator<T, Align>& other) const noexcept {
+ return !(*this == other);
+ }
+
+ // Returns true if and only if storage allocated from *this
+ // can be deallocated from other, and vice versa.
+ // Always returns true for stateless allocators.
+ bool operator==(const AlignmentAllocator<T, Align>& other) const noexcept {
+ return true;
+ }
+};
+
} // namespace Common
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index f4325f0f8..5462decee 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -111,6 +111,8 @@ add_library(core STATIC
frontend/scope_acquire_window_context.h
gdbstub/gdbstub.cpp
gdbstub/gdbstub.h
+ hardware_interrupt_manager.cpp
+ hardware_interrupt_manager.h
hle/ipc.h
hle/ipc_helpers.h
hle/kernel/address_arbiter.cpp
@@ -372,6 +374,7 @@ add_library(core STATIC
hle/service/nvdrv/devices/nvmap.h
hle/service/nvdrv/interface.cpp
hle/service/nvdrv/interface.h
+ hle/service/nvdrv/nvdata.h
hle/service/nvdrv/nvdrv.cpp
hle/service/nvdrv/nvdrv.h
hle/service/nvdrv/nvmemp.cpp
diff --git a/src/core/core.cpp b/src/core/core.cpp
index 4aceee785..20d64f3b0 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -19,6 +19,7 @@
#include "core/file_sys/vfs_concat.h"
#include "core/file_sys/vfs_real.h"
#include "core/gdbstub/gdbstub.h"
+#include "core/hardware_interrupt_manager.h"
#include "core/hle/kernel/client_port.h"
#include "core/hle/kernel/kernel.h"
#include "core/hle/kernel/process.h"
@@ -151,7 +152,7 @@ struct System::Impl {
if (!renderer->Init()) {
return ResultStatus::ErrorVideoCore;
}
-
+ interrupt_manager = std::make_unique<Core::Hardware::InterruptManager>(system);
gpu_core = VideoCore::CreateGPU(system);
is_powered_on = true;
@@ -298,6 +299,7 @@ struct System::Impl {
std::unique_ptr<VideoCore::RendererBase> renderer;
std::unique_ptr<Tegra::GPU> gpu_core;
std::shared_ptr<Tegra::DebugContext> debug_context;
+ std::unique_ptr<Core::Hardware::InterruptManager> interrupt_manager;
CpuCoreManager cpu_core_manager;
bool is_powered_on = false;
@@ -444,6 +446,14 @@ const Tegra::GPU& System::GPU() const {
return *impl->gpu_core;
}
+Core::Hardware::InterruptManager& System::InterruptManager() {
+ return *impl->interrupt_manager;
+}
+
+const Core::Hardware::InterruptManager& System::InterruptManager() const {
+ return *impl->interrupt_manager;
+}
+
VideoCore::RendererBase& System::Renderer() {
return *impl->renderer;
}
diff --git a/src/core/core.h b/src/core/core.h
index 8ebb385ac..0138d93b0 100644
--- a/src/core/core.h
+++ b/src/core/core.h
@@ -70,6 +70,10 @@ namespace Core::Timing {
class CoreTiming;
}
+namespace Core::Hardware {
+class InterruptManager;
+}
+
namespace Core {
class ARM_Interface;
@@ -234,6 +238,12 @@ public:
/// Provides a constant reference to the core timing instance.
const Timing::CoreTiming& CoreTiming() const;
+ /// Provides a reference to the interrupt manager instance.
+ Core::Hardware::InterruptManager& InterruptManager();
+
+ /// Provides a constant reference to the interrupt manager instance.
+ const Core::Hardware::InterruptManager& InterruptManager() const;
+
/// Provides a reference to the kernel instance.
Kernel::KernelCore& Kernel();
diff --git a/src/core/hardware_interrupt_manager.cpp b/src/core/hardware_interrupt_manager.cpp
new file mode 100644
index 000000000..c2115db2d
--- /dev/null
+++ b/src/core/hardware_interrupt_manager.cpp
@@ -0,0 +1,30 @@
+// Copyright 2019 Yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/hardware_interrupt_manager.h"
+#include "core/hle/service/nvdrv/interface.h"
+#include "core/hle/service/sm/sm.h"
+
+namespace Core::Hardware {
+
+InterruptManager::InterruptManager(Core::System& system_in) : system(system_in) {
+ gpu_interrupt_event =
+ system.CoreTiming().RegisterEvent("GPUInterrupt", [this](u64 message, s64) {
+ auto nvdrv = system.ServiceManager().GetService<Service::Nvidia::NVDRV>("nvdrv");
+ const u32 syncpt = static_cast<u32>(message >> 32);
+ const u32 value = static_cast<u32>(message);
+ nvdrv->SignalGPUInterruptSyncpt(syncpt, value);
+ });
+}
+
+InterruptManager::~InterruptManager() = default;
+
+void InterruptManager::GPUInterruptSyncpt(const u32 syncpoint_id, const u32 value) {
+ const u64 msg = (static_cast<u64>(syncpoint_id) << 32ULL) | value;
+ system.CoreTiming().ScheduleEvent(10, gpu_interrupt_event, msg);
+}
+
+} // namespace Core::Hardware
diff --git a/src/core/hardware_interrupt_manager.h b/src/core/hardware_interrupt_manager.h
new file mode 100644
index 000000000..494db883a
--- /dev/null
+++ b/src/core/hardware_interrupt_manager.h
@@ -0,0 +1,31 @@
+// Copyright 2019 Yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Core {
+class System;
+}
+
+namespace Core::Timing {
+struct EventType;
+}
+
+namespace Core::Hardware {
+
+class InterruptManager {
+public:
+ explicit InterruptManager(Core::System& system);
+ ~InterruptManager();
+
+ void GPUInterruptSyncpt(u32 syncpoint_id, u32 value);
+
+private:
+ Core::System& system;
+ Core::Timing::EventType* gpu_interrupt_event{};
+};
+
+} // namespace Core::Hardware
diff --git a/src/core/hle/kernel/code_set.h b/src/core/hle/kernel/code_set.h
index 879957dcb..d8ad54030 100644
--- a/src/core/hle/kernel/code_set.h
+++ b/src/core/hle/kernel/code_set.h
@@ -8,6 +8,7 @@
#include <vector>
#include "common/common_types.h"
+#include "core/hle/kernel/physical_memory.h"
namespace Kernel {
@@ -77,7 +78,7 @@ struct CodeSet final {
}
/// The overall data that backs this code set.
- std::vector<u8> memory;
+ Kernel::PhysicalMemory memory;
/// The segments that comprise this code set.
std::array<Segment, 3> segments;
diff --git a/src/core/hle/kernel/physical_memory.h b/src/core/hle/kernel/physical_memory.h
new file mode 100644
index 000000000..090565310
--- /dev/null
+++ b/src/core/hle/kernel/physical_memory.h
@@ -0,0 +1,19 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/alignment.h"
+
+namespace Kernel {
+
+// This encapsulation serves 2 purposes:
+// - First, to encapsulate host physical memory under a single type and set an
+// standard for managing it.
+// - Second to ensure all host backing memory used is aligned to 256 bytes due
+// to strict alignment restrictions on GPU memory.
+
+using PhysicalMemory = std::vector<u8, Common::AlignmentAllocator<u8, 256>>;
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/process.cpp b/src/core/hle/kernel/process.cpp
index 92169a97b..e80a12ac3 100644
--- a/src/core/hle/kernel/process.cpp
+++ b/src/core/hle/kernel/process.cpp
@@ -247,7 +247,7 @@ VAddr Process::CreateTLSRegion() {
ASSERT(region_address.Succeeded());
const auto map_result = vm_manager.MapMemoryBlock(
- *region_address, std::make_shared<std::vector<u8>>(Memory::PAGE_SIZE), 0,
+ *region_address, std::make_shared<PhysicalMemory>(Memory::PAGE_SIZE), 0,
Memory::PAGE_SIZE, MemoryState::ThreadLocal);
ASSERT(map_result.Succeeded());
@@ -277,7 +277,7 @@ void Process::FreeTLSRegion(VAddr tls_address) {
}
void Process::LoadModule(CodeSet module_, VAddr base_addr) {
- const auto memory = std::make_shared<std::vector<u8>>(std::move(module_.memory));
+ const auto memory = std::make_shared<PhysicalMemory>(std::move(module_.memory));
const auto MapSegment = [&](const CodeSet::Segment& segment, VMAPermission permissions,
MemoryState memory_state) {
@@ -327,7 +327,7 @@ void Process::AllocateMainThreadStack(u64 stack_size) {
// Allocate and map the main thread stack
const VAddr mapping_address = vm_manager.GetTLSIORegionEndAddress() - main_thread_stack_size;
vm_manager
- .MapMemoryBlock(mapping_address, std::make_shared<std::vector<u8>>(main_thread_stack_size),
+ .MapMemoryBlock(mapping_address, std::make_shared<PhysicalMemory>(main_thread_stack_size),
0, main_thread_stack_size, MemoryState::Stack)
.Unwrap();
}
diff --git a/src/core/hle/kernel/shared_memory.cpp b/src/core/hle/kernel/shared_memory.cpp
index f15c5ee36..a815c4eea 100644
--- a/src/core/hle/kernel/shared_memory.cpp
+++ b/src/core/hle/kernel/shared_memory.cpp
@@ -28,7 +28,7 @@ SharedPtr<SharedMemory> SharedMemory::Create(KernelCore& kernel, Process* owner_
shared_memory->other_permissions = other_permissions;
if (address == 0) {
- shared_memory->backing_block = std::make_shared<std::vector<u8>>(size);
+ shared_memory->backing_block = std::make_shared<Kernel::PhysicalMemory>(size);
shared_memory->backing_block_offset = 0;
// Refresh the address mappings for the current process.
@@ -59,8 +59,8 @@ SharedPtr<SharedMemory> SharedMemory::Create(KernelCore& kernel, Process* owner_
}
SharedPtr<SharedMemory> SharedMemory::CreateForApplet(
- KernelCore& kernel, std::shared_ptr<std::vector<u8>> heap_block, std::size_t offset, u64 size,
- MemoryPermission permissions, MemoryPermission other_permissions, std::string name) {
+ KernelCore& kernel, std::shared_ptr<Kernel::PhysicalMemory> heap_block, std::size_t offset,
+ u64 size, MemoryPermission permissions, MemoryPermission other_permissions, std::string name) {
SharedPtr<SharedMemory> shared_memory(new SharedMemory(kernel));
shared_memory->owner_process = nullptr;
diff --git a/src/core/hle/kernel/shared_memory.h b/src/core/hle/kernel/shared_memory.h
index c2b6155e1..01ca6dcd2 100644
--- a/src/core/hle/kernel/shared_memory.h
+++ b/src/core/hle/kernel/shared_memory.h
@@ -10,6 +10,7 @@
#include "common/common_types.h"
#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/physical_memory.h"
#include "core/hle/kernel/process.h"
#include "core/hle/result.h"
@@ -62,12 +63,10 @@ public:
* block.
* @param name Optional object name, used for debugging purposes.
*/
- static SharedPtr<SharedMemory> CreateForApplet(KernelCore& kernel,
- std::shared_ptr<std::vector<u8>> heap_block,
- std::size_t offset, u64 size,
- MemoryPermission permissions,
- MemoryPermission other_permissions,
- std::string name = "Unknown Applet");
+ static SharedPtr<SharedMemory> CreateForApplet(
+ KernelCore& kernel, std::shared_ptr<Kernel::PhysicalMemory> heap_block, std::size_t offset,
+ u64 size, MemoryPermission permissions, MemoryPermission other_permissions,
+ std::string name = "Unknown Applet");
std::string GetTypeName() const override {
return "SharedMemory";
@@ -135,7 +134,7 @@ private:
~SharedMemory() override;
/// Backing memory for this shared memory block.
- std::shared_ptr<std::vector<u8>> backing_block;
+ std::shared_ptr<PhysicalMemory> backing_block;
/// Offset into the backing block for this shared memory.
std::size_t backing_block_offset = 0;
/// Size of the memory block. Page-aligned.
diff --git a/src/core/hle/kernel/transfer_memory.cpp b/src/core/hle/kernel/transfer_memory.cpp
index 26c4e5e67..1113c815e 100644
--- a/src/core/hle/kernel/transfer_memory.cpp
+++ b/src/core/hle/kernel/transfer_memory.cpp
@@ -47,7 +47,7 @@ ResultCode TransferMemory::MapMemory(VAddr address, u64 size, MemoryPermission p
return ERR_INVALID_STATE;
}
- backing_block = std::make_shared<std::vector<u8>>(size);
+ backing_block = std::make_shared<PhysicalMemory>(size);
const auto map_state = owner_permissions == MemoryPermission::None
? MemoryState::TransferMemoryIsolated
diff --git a/src/core/hle/kernel/transfer_memory.h b/src/core/hle/kernel/transfer_memory.h
index a140b1e2b..6be9dc094 100644
--- a/src/core/hle/kernel/transfer_memory.h
+++ b/src/core/hle/kernel/transfer_memory.h
@@ -8,6 +8,7 @@
#include <vector>
#include "core/hle/kernel/object.h"
+#include "core/hle/kernel/physical_memory.h"
union ResultCode;
@@ -82,7 +83,7 @@ private:
~TransferMemory() override;
/// Memory block backing this instance.
- std::shared_ptr<std::vector<u8>> backing_block;
+ std::shared_ptr<PhysicalMemory> backing_block;
/// The base address for the memory managed by this instance.
VAddr base_address = 0;
diff --git a/src/core/hle/kernel/vm_manager.cpp b/src/core/hle/kernel/vm_manager.cpp
index 4f45fb03b..40cea1e7c 100644
--- a/src/core/hle/kernel/vm_manager.cpp
+++ b/src/core/hle/kernel/vm_manager.cpp
@@ -5,6 +5,7 @@
#include <algorithm>
#include <iterator>
#include <utility>
+#include "common/alignment.h"
#include "common/assert.h"
#include "common/logging/log.h"
#include "common/memory_hook.h"
@@ -103,7 +104,7 @@ bool VMManager::IsValidHandle(VMAHandle handle) const {
}
ResultVal<VMManager::VMAHandle> VMManager::MapMemoryBlock(VAddr target,
- std::shared_ptr<std::vector<u8>> block,
+ std::shared_ptr<PhysicalMemory> block,
std::size_t offset, u64 size,
MemoryState state, VMAPermission perm) {
ASSERT(block != nullptr);
@@ -260,7 +261,7 @@ ResultVal<VAddr> VMManager::SetHeapSize(u64 size) {
if (heap_memory == nullptr) {
// Initialize heap
- heap_memory = std::make_shared<std::vector<u8>>(size);
+ heap_memory = std::make_shared<PhysicalMemory>(size);
heap_end = heap_region_base + size;
} else {
UnmapRange(heap_region_base, GetCurrentHeapSize());
@@ -341,7 +342,7 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
const auto map_size = std::min(end_addr - cur_addr, vma_end - cur_addr);
if (vma.state == MemoryState::Unmapped) {
const auto map_res =
- MapMemoryBlock(cur_addr, std::make_shared<std::vector<u8>>(map_size, 0), 0,
+ MapMemoryBlock(cur_addr, std::make_shared<PhysicalMemory>(map_size, 0), 0,
map_size, MemoryState::Heap, VMAPermission::ReadWrite);
result = map_res.Code();
if (result.IsError()) {
@@ -442,7 +443,7 @@ ResultCode VMManager::UnmapPhysicalMemory(VAddr target, u64 size) {
if (result.IsError()) {
for (const auto [map_address, map_size] : unmapped_regions) {
const auto remap_res =
- MapMemoryBlock(map_address, std::make_shared<std::vector<u8>>(map_size, 0), 0,
+ MapMemoryBlock(map_address, std::make_shared<PhysicalMemory>(map_size, 0), 0,
map_size, MemoryState::Heap, VMAPermission::None);
ASSERT_MSG(remap_res.Succeeded(), "UnmapPhysicalMemory re-map on error");
}
@@ -593,7 +594,7 @@ ResultCode VMManager::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size, Mem
ASSERT_MSG(vma_offset + size <= vma->second.size,
"Shared memory exceeds bounds of mapped block");
- const std::shared_ptr<std::vector<u8>>& backing_block = vma->second.backing_block;
+ const std::shared_ptr<PhysicalMemory>& backing_block = vma->second.backing_block;
const std::size_t backing_block_offset = vma->second.offset + vma_offset;
CASCADE_RESULT(auto new_vma,
@@ -606,7 +607,7 @@ ResultCode VMManager::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size, Mem
return RESULT_SUCCESS;
}
-void VMManager::RefreshMemoryBlockMappings(const std::vector<u8>* block) {
+void VMManager::RefreshMemoryBlockMappings(const PhysicalMemory* block) {
// If this ever proves to have a noticeable performance impact, allow users of the function to
// specify a specific range of addresses to limit the scan to.
for (const auto& p : vma_map) {
@@ -764,7 +765,7 @@ void VMManager::MergeAdjacentVMA(VirtualMemoryArea& left, const VirtualMemoryAre
right.backing_block->begin() + right.offset + right.size);
} else {
// Slow case: make a new memory block for left and right.
- auto new_memory = std::make_shared<std::vector<u8>>();
+ auto new_memory = std::make_shared<PhysicalMemory>();
new_memory->insert(new_memory->end(), left.backing_block->begin() + left.offset,
left.backing_block->begin() + left.offset + left.size);
new_memory->insert(new_memory->end(), right.backing_block->begin() + right.offset,
diff --git a/src/core/hle/kernel/vm_manager.h b/src/core/hle/kernel/vm_manager.h
index 0aecb7499..b18cde619 100644
--- a/src/core/hle/kernel/vm_manager.h
+++ b/src/core/hle/kernel/vm_manager.h
@@ -11,6 +11,7 @@
#include "common/common_types.h"
#include "common/memory_hook.h"
#include "common/page_table.h"
+#include "core/hle/kernel/physical_memory.h"
#include "core/hle/result.h"
#include "core/memory.h"
@@ -290,7 +291,7 @@ struct VirtualMemoryArea {
// Settings for type = AllocatedMemoryBlock
/// Memory block backing this VMA.
- std::shared_ptr<std::vector<u8>> backing_block = nullptr;
+ std::shared_ptr<PhysicalMemory> backing_block = nullptr;
/// Offset into the backing_memory the mapping starts from.
std::size_t offset = 0;
@@ -348,7 +349,7 @@ public:
* @param size Size of the mapping.
* @param state MemoryState tag to attach to the VMA.
*/
- ResultVal<VMAHandle> MapMemoryBlock(VAddr target, std::shared_ptr<std::vector<u8>> block,
+ ResultVal<VMAHandle> MapMemoryBlock(VAddr target, std::shared_ptr<PhysicalMemory> block,
std::size_t offset, u64 size, MemoryState state,
VMAPermission perm = VMAPermission::ReadWrite);
@@ -547,7 +548,7 @@ public:
* Scans all VMAs and updates the page table range of any that use the given vector as backing
* memory. This should be called after any operation that causes reallocation of the vector.
*/
- void RefreshMemoryBlockMappings(const std::vector<u8>* block);
+ void RefreshMemoryBlockMappings(const PhysicalMemory* block);
/// Dumps the address space layout to the log, for debugging
void LogLayout() const;
@@ -777,7 +778,7 @@ private:
// the entire virtual address space extents that bound the allocations, including any holes.
// This makes deallocation and reallocation of holes fast and keeps process memory contiguous
// in the emulator address space, allowing Memory::GetPointer to be reasonably safe.
- std::shared_ptr<std::vector<u8>> heap_memory;
+ std::shared_ptr<PhysicalMemory> heap_memory;
// The end of the currently allocated heap. This is not an inclusive
// end of the range. This is essentially 'base_address + current_size'.
diff --git a/src/core/hle/service/audio/audio.cpp b/src/core/hle/service/audio/audio.cpp
index 128df7db5..1781bec83 100644
--- a/src/core/hle/service/audio/audio.cpp
+++ b/src/core/hle/service/audio/audio.cpp
@@ -19,16 +19,16 @@
namespace Service::Audio {
-void InstallInterfaces(SM::ServiceManager& service_manager) {
+void InstallInterfaces(SM::ServiceManager& service_manager, Core::System& system) {
std::make_shared<AudCtl>()->InstallAsService(service_manager);
std::make_shared<AudOutA>()->InstallAsService(service_manager);
- std::make_shared<AudOutU>()->InstallAsService(service_manager);
+ std::make_shared<AudOutU>(system)->InstallAsService(service_manager);
std::make_shared<AudInA>()->InstallAsService(service_manager);
std::make_shared<AudInU>()->InstallAsService(service_manager);
std::make_shared<AudRecA>()->InstallAsService(service_manager);
std::make_shared<AudRecU>()->InstallAsService(service_manager);
std::make_shared<AudRenA>()->InstallAsService(service_manager);
- std::make_shared<AudRenU>()->InstallAsService(service_manager);
+ std::make_shared<AudRenU>(system)->InstallAsService(service_manager);
std::make_shared<CodecCtl>()->InstallAsService(service_manager);
std::make_shared<HwOpus>()->InstallAsService(service_manager);
diff --git a/src/core/hle/service/audio/audio.h b/src/core/hle/service/audio/audio.h
index f5bd3bf5f..b6d13912e 100644
--- a/src/core/hle/service/audio/audio.h
+++ b/src/core/hle/service/audio/audio.h
@@ -4,6 +4,10 @@
#pragma once
+namespace Core {
+class System;
+}
+
namespace Service::SM {
class ServiceManager;
}
@@ -11,6 +15,6 @@ class ServiceManager;
namespace Service::Audio {
/// Registers all Audio services with the specified service manager.
-void InstallInterfaces(SM::ServiceManager& service_manager);
+void InstallInterfaces(SM::ServiceManager& service_manager, Core::System& system);
} // namespace Service::Audio
diff --git a/src/core/hle/service/audio/audout_u.cpp b/src/core/hle/service/audio/audout_u.cpp
index 7db6eb08d..fb84a8f13 100644
--- a/src/core/hle/service/audio/audout_u.cpp
+++ b/src/core/hle/service/audio/audout_u.cpp
@@ -40,8 +40,8 @@ enum class AudioState : u32 {
class IAudioOut final : public ServiceFramework<IAudioOut> {
public:
- IAudioOut(AudoutParams audio_params, AudioCore::AudioOut& audio_core, std::string&& device_name,
- std::string&& unique_name)
+ IAudioOut(Core::System& system, AudoutParams audio_params, AudioCore::AudioOut& audio_core,
+ std::string&& device_name, std::string&& unique_name)
: ServiceFramework("IAudioOut"), audio_core(audio_core),
device_name(std::move(device_name)), audio_params(audio_params) {
// clang-format off
@@ -65,7 +65,6 @@ public:
RegisterHandlers(functions);
// This is the event handle used to check if the audio buffer was released
- auto& system = Core::System::GetInstance();
buffer_event = Kernel::WritableEvent::CreateEventPair(
system.Kernel(), Kernel::ResetType::Manual, "IAudioOutBufferReleased");
@@ -212,6 +211,22 @@ private:
Kernel::EventPair buffer_event;
};
+AudOutU::AudOutU(Core::System& system_) : ServiceFramework("audout:u"), system{system_} {
+ // clang-format off
+ static const FunctionInfo functions[] = {
+ {0, &AudOutU::ListAudioOutsImpl, "ListAudioOuts"},
+ {1, &AudOutU::OpenAudioOutImpl, "OpenAudioOut"},
+ {2, &AudOutU::ListAudioOutsImpl, "ListAudioOutsAuto"},
+ {3, &AudOutU::OpenAudioOutImpl, "OpenAudioOutAuto"},
+ };
+ // clang-format on
+
+ RegisterHandlers(functions);
+ audio_core = std::make_unique<AudioCore::AudioOut>();
+}
+
+AudOutU::~AudOutU() = default;
+
void AudOutU::ListAudioOutsImpl(Kernel::HLERequestContext& ctx) {
LOG_DEBUG(Service_Audio, "called");
@@ -248,7 +263,7 @@ void AudOutU::OpenAudioOutImpl(Kernel::HLERequestContext& ctx) {
std::string unique_name{fmt::format("{}-{}", device_name, audio_out_interfaces.size())};
auto audio_out_interface = std::make_shared<IAudioOut>(
- params, *audio_core, std::move(device_name), std::move(unique_name));
+ system, params, *audio_core, std::move(device_name), std::move(unique_name));
IPC::ResponseBuilder rb{ctx, 6, 0, 1};
rb.Push(RESULT_SUCCESS);
@@ -256,20 +271,9 @@ void AudOutU::OpenAudioOutImpl(Kernel::HLERequestContext& ctx) {
rb.Push<u32>(params.channel_count);
rb.Push<u32>(static_cast<u32>(AudioCore::Codec::PcmFormat::Int16));
rb.Push<u32>(static_cast<u32>(AudioState::Stopped));
- rb.PushIpcInterface<Audio::IAudioOut>(audio_out_interface);
+ rb.PushIpcInterface<IAudioOut>(audio_out_interface);
audio_out_interfaces.push_back(std::move(audio_out_interface));
}
-AudOutU::AudOutU() : ServiceFramework("audout:u") {
- static const FunctionInfo functions[] = {{0, &AudOutU::ListAudioOutsImpl, "ListAudioOuts"},
- {1, &AudOutU::OpenAudioOutImpl, "OpenAudioOut"},
- {2, &AudOutU::ListAudioOutsImpl, "ListAudioOutsAuto"},
- {3, &AudOutU::OpenAudioOutImpl, "OpenAudioOutAuto"}};
- RegisterHandlers(functions);
- audio_core = std::make_unique<AudioCore::AudioOut>();
-}
-
-AudOutU::~AudOutU() = default;
-
} // namespace Service::Audio
diff --git a/src/core/hle/service/audio/audout_u.h b/src/core/hle/service/audio/audout_u.h
index aed4c43b2..c9f532ccd 100644
--- a/src/core/hle/service/audio/audout_u.h
+++ b/src/core/hle/service/audio/audout_u.h
@@ -11,6 +11,10 @@ namespace AudioCore {
class AudioOut;
}
+namespace Core {
+class System;
+}
+
namespace Kernel {
class HLERequestContext;
}
@@ -21,15 +25,17 @@ class IAudioOut;
class AudOutU final : public ServiceFramework<AudOutU> {
public:
- AudOutU();
+ explicit AudOutU(Core::System& system_);
~AudOutU() override;
private:
+ void ListAudioOutsImpl(Kernel::HLERequestContext& ctx);
+ void OpenAudioOutImpl(Kernel::HLERequestContext& ctx);
+
std::vector<std::shared_ptr<IAudioOut>> audio_out_interfaces;
std::unique_ptr<AudioCore::AudioOut> audio_core;
- void ListAudioOutsImpl(Kernel::HLERequestContext& ctx);
- void OpenAudioOutImpl(Kernel::HLERequestContext& ctx);
+ Core::System& system;
};
} // namespace Service::Audio
diff --git a/src/core/hle/service/audio/audren_u.cpp b/src/core/hle/service/audio/audren_u.cpp
index 679299f68..5b0b7f17e 100644
--- a/src/core/hle/service/audio/audren_u.cpp
+++ b/src/core/hle/service/audio/audren_u.cpp
@@ -5,6 +5,7 @@
#include <algorithm>
#include <array>
#include <memory>
+#include <string_view>
#include "audio_core/audio_renderer.h"
#include "common/alignment.h"
@@ -25,7 +26,7 @@ namespace Service::Audio {
class IAudioRenderer final : public ServiceFramework<IAudioRenderer> {
public:
- explicit IAudioRenderer(AudioCore::AudioRendererParameter audren_params,
+ explicit IAudioRenderer(Core::System& system, AudioCore::AudioRendererParameter audren_params,
const std::size_t instance_number)
: ServiceFramework("IAudioRenderer") {
// clang-format off
@@ -46,7 +47,6 @@ public:
// clang-format on
RegisterHandlers(functions);
- auto& system = Core::System::GetInstance();
system_event = Kernel::WritableEvent::CreateEventPair(
system.Kernel(), Kernel::ResetType::Manual, "IAudioRenderer:SystemEvent");
renderer = std::make_unique<AudioCore::AudioRenderer>(
@@ -160,7 +160,8 @@ private:
class IAudioDevice final : public ServiceFramework<IAudioDevice> {
public:
- IAudioDevice() : ServiceFramework("IAudioDevice") {
+ explicit IAudioDevice(Core::System& system, u32_le revision_num)
+ : ServiceFramework("IAudioDevice"), revision{revision_num} {
static const FunctionInfo functions[] = {
{0, &IAudioDevice::ListAudioDeviceName, "ListAudioDeviceName"},
{1, &IAudioDevice::SetAudioDeviceOutputVolume, "SetAudioDeviceOutputVolume"},
@@ -178,7 +179,7 @@ public:
};
RegisterHandlers(functions);
- auto& kernel = Core::System::GetInstance().Kernel();
+ auto& kernel = system.Kernel();
buffer_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Automatic,
"IAudioOutBufferReleasedEvent");
@@ -189,15 +190,47 @@ public:
}
private:
+ using AudioDeviceName = std::array<char, 256>;
+ static constexpr std::array<std::string_view, 4> audio_device_names{{
+ "AudioStereoJackOutput",
+ "AudioBuiltInSpeakerOutput",
+ "AudioTvOutput",
+ "AudioUsbDeviceOutput",
+ }};
+ enum class DeviceType {
+ AHUBHeadphones,
+ AHUBSpeakers,
+ HDA,
+ USBOutput,
+ };
+
void ListAudioDeviceName(Kernel::HLERequestContext& ctx) {
- LOG_WARNING(Service_Audio, "(STUBBED) called");
+ LOG_DEBUG(Service_Audio, "called");
- constexpr std::array<char, 15> audio_interface{{"AudioInterface"}};
- ctx.WriteBuffer(audio_interface);
+ const bool usb_output_supported =
+ IsFeatureSupported(AudioFeatures::AudioUSBDeviceOutput, revision);
+ const std::size_t count = ctx.GetWriteBufferSize() / sizeof(AudioDeviceName);
+
+ std::vector<AudioDeviceName> name_buffer;
+ name_buffer.reserve(audio_device_names.size());
+
+ for (std::size_t i = 0; i < count && i < audio_device_names.size(); i++) {
+ const auto type = static_cast<DeviceType>(i);
+
+ if (!usb_output_supported && type == DeviceType::USBOutput) {
+ continue;
+ }
+
+ const auto& device_name = audio_device_names[i];
+ auto& entry = name_buffer.emplace_back();
+ device_name.copy(entry.data(), device_name.size());
+ }
+
+ ctx.WriteBuffer(name_buffer);
IPC::ResponseBuilder rb{ctx, 3};
rb.Push(RESULT_SUCCESS);
- rb.Push<u32>(1);
+ rb.Push(static_cast<u32>(name_buffer.size()));
}
void SetAudioDeviceOutputVolume(Kernel::HLERequestContext& ctx) {
@@ -216,12 +249,16 @@ private:
void GetActiveAudioDeviceName(Kernel::HLERequestContext& ctx) {
LOG_WARNING(Service_Audio, "(STUBBED) called");
- constexpr std::array<char, 12> audio_interface{{"AudioDevice"}};
- ctx.WriteBuffer(audio_interface);
+ // Currently set to always be TV audio output.
+ const auto& device_name = audio_device_names[2];
- IPC::ResponseBuilder rb{ctx, 3};
+ AudioDeviceName out_device_name{};
+ device_name.copy(out_device_name.data(), device_name.size());
+
+ ctx.WriteBuffer(out_device_name);
+
+ IPC::ResponseBuilder rb{ctx, 2};
rb.Push(RESULT_SUCCESS);
- rb.Push<u32>(1);
}
void QueryAudioDeviceSystemEvent(Kernel::HLERequestContext& ctx) {
@@ -250,12 +287,13 @@ private:
rb.PushCopyObjects(audio_output_device_switch_event.readable);
}
+ u32_le revision = 0;
Kernel::EventPair buffer_event;
Kernel::EventPair audio_output_device_switch_event;
}; // namespace Audio
-AudRenU::AudRenU() : ServiceFramework("audren:u") {
+AudRenU::AudRenU(Core::System& system_) : ServiceFramework("audren:u"), system{system_} {
// clang-format off
static const FunctionInfo functions[] = {
{0, &AudRenU::OpenAudioRenderer, "OpenAudioRenderer"},
@@ -328,7 +366,7 @@ void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
};
// Calculates the portion of the size related to the mix data (and the sorting thereof).
- const auto calculate_mix_info_size = [this](const AudioCore::AudioRendererParameter& params) {
+ const auto calculate_mix_info_size = [](const AudioCore::AudioRendererParameter& params) {
// The size of the mixing info data structure.
constexpr u64 mix_info_size = 0x940;
@@ -400,7 +438,7 @@ void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
// Calculates the part of the size related to the splitter context.
const auto calculate_splitter_context_size =
- [this](const AudioCore::AudioRendererParameter& params) -> u64 {
+ [](const AudioCore::AudioRendererParameter& params) -> u64 {
if (!IsFeatureSupported(AudioFeatures::Splitter, params.revision)) {
return 0;
}
@@ -447,7 +485,7 @@ void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
};
// Calculates the part of the size related to performance statistics.
- const auto calculate_perf_size = [this](const AudioCore::AudioRendererParameter& params) {
+ const auto calculate_perf_size = [](const AudioCore::AudioRendererParameter& params) {
// Extra size value appended to the end of the calculation.
constexpr u64 appended = 128;
@@ -474,78 +512,76 @@ void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
};
// Calculates the part of the size that relates to the audio command buffer.
- const auto calculate_command_buffer_size =
- [this](const AudioCore::AudioRendererParameter& params) {
- constexpr u64 alignment = (buffer_alignment_size - 1) * 2;
+ const auto calculate_command_buffer_size = [](const AudioCore::AudioRendererParameter& params) {
+ constexpr u64 alignment = (buffer_alignment_size - 1) * 2;
- if (!IsFeatureSupported(AudioFeatures::VariadicCommandBuffer, params.revision)) {
- constexpr u64 command_buffer_size = 0x18000;
+ if (!IsFeatureSupported(AudioFeatures::VariadicCommandBuffer, params.revision)) {
+ constexpr u64 command_buffer_size = 0x18000;
- return command_buffer_size + alignment;
- }
+ return command_buffer_size + alignment;
+ }
- // When the variadic command buffer is supported, this means
- // the command generator for the audio renderer can issue commands
- // that are (as one would expect), variable in size. So what we need to do
- // is determine the maximum possible size for a few command data structures
- // then multiply them by the amount of present commands indicated by the given
- // respective audio parameters.
+ // When the variadic command buffer is supported, this means
+ // the command generator for the audio renderer can issue commands
+ // that are (as one would expect), variable in size. So what we need to do
+ // is determine the maximum possible size for a few command data structures
+ // then multiply them by the amount of present commands indicated by the given
+ // respective audio parameters.
- constexpr u64 max_biquad_filters = 2;
- constexpr u64 max_mix_buffers = 24;
+ constexpr u64 max_biquad_filters = 2;
+ constexpr u64 max_mix_buffers = 24;
- constexpr u64 biquad_filter_command_size = 0x2C;
+ constexpr u64 biquad_filter_command_size = 0x2C;
- constexpr u64 depop_mix_command_size = 0x24;
- constexpr u64 depop_setup_command_size = 0x50;
+ constexpr u64 depop_mix_command_size = 0x24;
+ constexpr u64 depop_setup_command_size = 0x50;
- constexpr u64 effect_command_max_size = 0x540;
+ constexpr u64 effect_command_max_size = 0x540;
- constexpr u64 mix_command_size = 0x1C;
- constexpr u64 mix_ramp_command_size = 0x24;
- constexpr u64 mix_ramp_grouped_command_size = 0x13C;
+ constexpr u64 mix_command_size = 0x1C;
+ constexpr u64 mix_ramp_command_size = 0x24;
+ constexpr u64 mix_ramp_grouped_command_size = 0x13C;
- constexpr u64 perf_command_size = 0x28;
+ constexpr u64 perf_command_size = 0x28;
- constexpr u64 sink_command_size = 0x130;
+ constexpr u64 sink_command_size = 0x130;
- constexpr u64 submix_command_max_size =
- depop_mix_command_size + (mix_command_size * max_mix_buffers) * max_mix_buffers;
+ constexpr u64 submix_command_max_size =
+ depop_mix_command_size + (mix_command_size * max_mix_buffers) * max_mix_buffers;
- constexpr u64 volume_command_size = 0x1C;
- constexpr u64 volume_ramp_command_size = 0x20;
+ constexpr u64 volume_command_size = 0x1C;
+ constexpr u64 volume_ramp_command_size = 0x20;
- constexpr u64 voice_biquad_filter_command_size =
- biquad_filter_command_size * max_biquad_filters;
- constexpr u64 voice_data_command_size = 0x9C;
- const u64 voice_command_max_size =
- (params.splitter_count * depop_setup_command_size) +
- (voice_data_command_size + voice_biquad_filter_command_size +
- volume_ramp_command_size + mix_ramp_grouped_command_size);
+ constexpr u64 voice_biquad_filter_command_size =
+ biquad_filter_command_size * max_biquad_filters;
+ constexpr u64 voice_data_command_size = 0x9C;
+ const u64 voice_command_max_size =
+ (params.splitter_count * depop_setup_command_size) +
+ (voice_data_command_size + voice_biquad_filter_command_size + volume_ramp_command_size +
+ mix_ramp_grouped_command_size);
- // Now calculate the individual elements that comprise the size and add them together.
- const u64 effect_commands_size = params.effect_count * effect_command_max_size;
+ // Now calculate the individual elements that comprise the size and add them together.
+ const u64 effect_commands_size = params.effect_count * effect_command_max_size;
- const u64 final_mix_commands_size =
- depop_mix_command_size + volume_command_size * max_mix_buffers;
+ const u64 final_mix_commands_size =
+ depop_mix_command_size + volume_command_size * max_mix_buffers;
- const u64 perf_commands_size =
- perf_command_size *
- (CalculateNumPerformanceEntries(params) + max_perf_detail_entries);
+ const u64 perf_commands_size =
+ perf_command_size * (CalculateNumPerformanceEntries(params) + max_perf_detail_entries);
- const u64 sink_commands_size = params.sink_count * sink_command_size;
+ const u64 sink_commands_size = params.sink_count * sink_command_size;
- const u64 splitter_commands_size =
- params.num_splitter_send_channels * max_mix_buffers * mix_ramp_command_size;
+ const u64 splitter_commands_size =
+ params.num_splitter_send_channels * max_mix_buffers * mix_ramp_command_size;
- const u64 submix_commands_size = params.submix_count * submix_command_max_size;
+ const u64 submix_commands_size = params.submix_count * submix_command_max_size;
- const u64 voice_commands_size = params.voice_count * voice_command_max_size;
+ const u64 voice_commands_size = params.voice_count * voice_command_max_size;
- return effect_commands_size + final_mix_commands_size + perf_commands_size +
- sink_commands_size + splitter_commands_size + submix_commands_size +
- voice_commands_size + alignment;
- };
+ return effect_commands_size + final_mix_commands_size + perf_commands_size +
+ sink_commands_size + splitter_commands_size + submix_commands_size +
+ voice_commands_size + alignment;
+ };
IPC::RequestParser rp{ctx};
const auto params = rp.PopRaw<AudioCore::AudioRendererParameter>();
@@ -578,12 +614,16 @@ void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
}
void AudRenU::GetAudioDeviceService(Kernel::HLERequestContext& ctx) {
- LOG_DEBUG(Service_Audio, "called");
+ IPC::RequestParser rp{ctx};
+ const u64 aruid = rp.Pop<u64>();
- IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+ LOG_DEBUG(Service_Audio, "called. aruid={:016X}", aruid);
+ // Revisionless variant of GetAudioDeviceServiceWithRevisionInfo that
+ // always assumes the initial release revision (REV1).
+ IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(RESULT_SUCCESS);
- rb.PushIpcInterface<Audio::IAudioDevice>();
+ rb.PushIpcInterface<IAudioDevice>(system, Common::MakeMagic('R', 'E', 'V', '1'));
}
void AudRenU::OpenAudioRendererAuto(Kernel::HLERequestContext& ctx) {
@@ -593,13 +633,19 @@ void AudRenU::OpenAudioRendererAuto(Kernel::HLERequestContext& ctx) {
}
void AudRenU::GetAudioDeviceServiceWithRevisionInfo(Kernel::HLERequestContext& ctx) {
- LOG_WARNING(Service_Audio, "(STUBBED) called");
+ struct Parameters {
+ u32 revision;
+ u64 aruid;
+ };
- IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+ IPC::RequestParser rp{ctx};
+ const auto [revision, aruid] = rp.PopRaw<Parameters>();
+
+ LOG_DEBUG(Service_Audio, "called. revision={:08X}, aruid={:016X}", revision, aruid);
+ IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(RESULT_SUCCESS);
- rb.PushIpcInterface<Audio::IAudioDevice>(); // TODO(ogniK): Figure out what is different
- // based on the current revision
+ rb.PushIpcInterface<IAudioDevice>(system, revision);
}
void AudRenU::OpenAudioRendererImpl(Kernel::HLERequestContext& ctx) {
@@ -608,14 +654,16 @@ void AudRenU::OpenAudioRendererImpl(Kernel::HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(RESULT_SUCCESS);
- rb.PushIpcInterface<IAudioRenderer>(params, audren_instance_count++);
+ rb.PushIpcInterface<IAudioRenderer>(system, params, audren_instance_count++);
}
-bool AudRenU::IsFeatureSupported(AudioFeatures feature, u32_le revision) const {
+bool IsFeatureSupported(AudioFeatures feature, u32_le revision) {
// Byte swap
const u32_be version_num = revision - Common::MakeMagic('R', 'E', 'V', '0');
switch (feature) {
+ case AudioFeatures::AudioUSBDeviceOutput:
+ return version_num >= 4U;
case AudioFeatures::Splitter:
return version_num >= 2U;
case AudioFeatures::PerformanceMetricsVersion2:
diff --git a/src/core/hle/service/audio/audren_u.h b/src/core/hle/service/audio/audren_u.h
index 49f2733cf..4e0ccc792 100644
--- a/src/core/hle/service/audio/audren_u.h
+++ b/src/core/hle/service/audio/audren_u.h
@@ -6,6 +6,10 @@
#include "core/hle/service/service.h"
+namespace Core {
+class System;
+}
+
namespace Kernel {
class HLERequestContext;
}
@@ -14,7 +18,7 @@ namespace Service::Audio {
class AudRenU final : public ServiceFramework<AudRenU> {
public:
- explicit AudRenU();
+ explicit AudRenU(Core::System& system_);
~AudRenU() override;
private:
@@ -26,14 +30,19 @@ private:
void OpenAudioRendererImpl(Kernel::HLERequestContext& ctx);
- enum class AudioFeatures : u32 {
- Splitter,
- PerformanceMetricsVersion2,
- VariadicCommandBuffer,
- };
-
- bool IsFeatureSupported(AudioFeatures feature, u32_le revision) const;
std::size_t audren_instance_count = 0;
+ Core::System& system;
};
+// Describes a particular audio feature that may be supported in a particular revision.
+enum class AudioFeatures : u32 {
+ AudioUSBDeviceOutput,
+ Splitter,
+ PerformanceMetricsVersion2,
+ VariadicCommandBuffer,
+};
+
+// Tests if a particular audio feature is supported with a given audio revision.
+bool IsFeatureSupported(AudioFeatures feature, u32_le revision);
+
} // namespace Service::Audio
diff --git a/src/core/hle/service/ns/pl_u.cpp b/src/core/hle/service/ns/pl_u.cpp
index ad176f89d..2a522136d 100644
--- a/src/core/hle/service/ns/pl_u.cpp
+++ b/src/core/hle/service/ns/pl_u.cpp
@@ -77,7 +77,7 @@ enum class LoadState : u32 {
Done = 1,
};
-static void DecryptSharedFont(const std::vector<u32>& input, std::vector<u8>& output,
+static void DecryptSharedFont(const std::vector<u32>& input, Kernel::PhysicalMemory& output,
std::size_t& offset) {
ASSERT_MSG(offset + (input.size() * sizeof(u32)) < SHARED_FONT_MEM_SIZE,
"Shared fonts exceeds 17mb!");
@@ -94,7 +94,7 @@ static void DecryptSharedFont(const std::vector<u32>& input, std::vector<u8>& ou
offset += transformed_font.size() * sizeof(u32);
}
-static void EncryptSharedFont(const std::vector<u8>& input, std::vector<u8>& output,
+static void EncryptSharedFont(const std::vector<u8>& input, Kernel::PhysicalMemory& output,
std::size_t& offset) {
ASSERT_MSG(offset + input.size() + 8 < SHARED_FONT_MEM_SIZE, "Shared fonts exceeds 17mb!");
const u32 KEY = EXPECTED_MAGIC ^ EXPECTED_RESULT;
@@ -121,7 +121,7 @@ struct PL_U::Impl {
return shared_font_regions.at(index);
}
- void BuildSharedFontsRawRegions(const std::vector<u8>& input) {
+ void BuildSharedFontsRawRegions(const Kernel::PhysicalMemory& input) {
// As we can derive the xor key we can just populate the offsets
// based on the shared memory dump
unsigned cur_offset = 0;
@@ -144,7 +144,7 @@ struct PL_U::Impl {
Kernel::SharedPtr<Kernel::SharedMemory> shared_font_mem;
/// Backing memory for the shared font data
- std::shared_ptr<std::vector<u8>> shared_font;
+ std::shared_ptr<Kernel::PhysicalMemory> shared_font;
// Automatically populated based on shared_fonts dump or system archives.
std::vector<FontRegion> shared_font_regions;
@@ -166,7 +166,7 @@ PL_U::PL_U() : ServiceFramework("pl:u"), impl{std::make_unique<Impl>()} {
// Rebuild shared fonts from data ncas
if (nand->HasEntry(static_cast<u64>(FontArchives::Standard),
FileSys::ContentRecordType::Data)) {
- impl->shared_font = std::make_shared<std::vector<u8>>(SHARED_FONT_MEM_SIZE);
+ impl->shared_font = std::make_shared<Kernel::PhysicalMemory>(SHARED_FONT_MEM_SIZE);
for (auto font : SHARED_FONTS) {
const auto nca =
nand->GetEntry(static_cast<u64>(font.first), FileSys::ContentRecordType::Data);
@@ -207,7 +207,7 @@ PL_U::PL_U() : ServiceFramework("pl:u"), impl{std::make_unique<Impl>()} {
}
} else {
- impl->shared_font = std::make_shared<std::vector<u8>>(
+ impl->shared_font = std::make_shared<Kernel::PhysicalMemory>(
SHARED_FONT_MEM_SIZE); // Shared memory needs to always be allocated and a fixed size
const std::string user_path = FileUtil::GetUserPath(FileUtil::UserPath::SysDataDir);
diff --git a/src/core/hle/service/nvdrv/devices/nvdevice.h b/src/core/hle/service/nvdrv/devices/nvdevice.h
index 4f6042b00..5b8248433 100644
--- a/src/core/hle/service/nvdrv/devices/nvdevice.h
+++ b/src/core/hle/service/nvdrv/devices/nvdevice.h
@@ -8,6 +8,11 @@
#include "common/bit_field.h"
#include "common/common_types.h"
#include "common/swap.h"
+#include "core/hle/service/nvdrv/nvdata.h"
+
+namespace Core {
+class System;
+}
namespace Service::Nvidia::Devices {
@@ -15,7 +20,7 @@ namespace Service::Nvidia::Devices {
/// implement the ioctl interface.
class nvdevice {
public:
- nvdevice() = default;
+ explicit nvdevice(Core::System& system) : system{system} {};
virtual ~nvdevice() = default;
union Ioctl {
u32_le raw;
@@ -33,7 +38,11 @@ public:
* @param output A buffer where the output data will be written to.
* @returns The result code of the ioctl.
*/
- virtual u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) = 0;
+ virtual u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) = 0;
+
+protected:
+ Core::System& system;
};
} // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp
index 20c7c39aa..76494f0b7 100644
--- a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp
@@ -13,10 +13,12 @@
namespace Service::Nvidia::Devices {
-nvdisp_disp0::nvdisp_disp0(std::shared_ptr<nvmap> nvmap_dev) : nvmap_dev(std::move(nvmap_dev)) {}
+nvdisp_disp0::nvdisp_disp0(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
+ : nvdevice(system), nvmap_dev(std::move(nvmap_dev)) {}
nvdisp_disp0 ::~nvdisp_disp0() = default;
-u32 nvdisp_disp0::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvdisp_disp0::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
UNIMPLEMENTED_MSG("Unimplemented ioctl");
return 0;
}
@@ -34,9 +36,8 @@ void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u3
addr, offset, width, height, stride, static_cast<PixelFormat>(format),
transform, crop_rect};
- auto& instance = Core::System::GetInstance();
- instance.GetPerfStats().EndGameFrame();
- instance.GPU().SwapBuffers(framebuffer);
+ system.GetPerfStats().EndGameFrame();
+ system.GPU().SwapBuffers(framebuffer);
}
} // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h
index 12f3ef825..e79e490ff 100644
--- a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h
+++ b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h
@@ -17,10 +17,11 @@ class nvmap;
class nvdisp_disp0 final : public nvdevice {
public:
- explicit nvdisp_disp0(std::shared_ptr<nvmap> nvmap_dev);
+ explicit nvdisp_disp0(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
~nvdisp_disp0() override;
- u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+ u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) override;
/// Performs a screen flip, drawing the buffer pointed to by the handle.
void flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height, u32 stride,
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
index af62d33d2..24ab3f2e9 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
@@ -22,10 +22,12 @@ enum {
};
}
-nvhost_as_gpu::nvhost_as_gpu(std::shared_ptr<nvmap> nvmap_dev) : nvmap_dev(std::move(nvmap_dev)) {}
+nvhost_as_gpu::nvhost_as_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
+ : nvdevice(system), nvmap_dev(std::move(nvmap_dev)) {}
nvhost_as_gpu::~nvhost_as_gpu() = default;
-u32 nvhost_as_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_as_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
LOG_DEBUG(Service_NVDRV, "called, command=0x{:08X}, input_size=0x{:X}, output_size=0x{:X}",
command.raw, input.size(), output.size());
@@ -65,7 +67,7 @@ u32 nvhost_as_gpu::AllocateSpace(const std::vector<u8>& input, std::vector<u8>&
LOG_DEBUG(Service_NVDRV, "called, pages={:X}, page_size={:X}, flags={:X}", params.pages,
params.page_size, params.flags);
- auto& gpu = Core::System::GetInstance().GPU();
+ auto& gpu = system.GPU();
const u64 size{static_cast<u64>(params.pages) * static_cast<u64>(params.page_size)};
if (params.flags & 1) {
params.offset = gpu.MemoryManager().AllocateSpace(params.offset, size, 1);
@@ -85,7 +87,7 @@ u32 nvhost_as_gpu::Remap(const std::vector<u8>& input, std::vector<u8>& output)
std::vector<IoctlRemapEntry> entries(num_entries);
std::memcpy(entries.data(), input.data(), input.size());
- auto& gpu = Core::System::GetInstance().GPU();
+ auto& gpu = system.GPU();
for (const auto& entry : entries) {
LOG_WARNING(Service_NVDRV, "remap entry, offset=0x{:X} handle=0x{:X} pages=0x{:X}",
entry.offset, entry.nvmap_handle, entry.pages);
@@ -136,7 +138,7 @@ u32 nvhost_as_gpu::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& ou
// case to prevent unexpected behavior.
ASSERT(object->id == params.nvmap_handle);
- auto& gpu = Core::System::GetInstance().GPU();
+ auto& gpu = system.GPU();
if (params.flags & 1) {
params.offset = gpu.MemoryManager().MapBufferEx(object->addr, params.offset, object->size);
@@ -173,8 +175,7 @@ u32 nvhost_as_gpu::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& ou
return 0;
}
- params.offset = Core::System::GetInstance().GPU().MemoryManager().UnmapBuffer(params.offset,
- itr->second.size);
+ params.offset = system.GPU().MemoryManager().UnmapBuffer(params.offset, itr->second.size);
buffer_mappings.erase(itr->second.offset);
std::memcpy(output.data(), &params, output.size());
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h
index eb14b1da8..30ca5f4c3 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h
@@ -17,10 +17,11 @@ class nvmap;
class nvhost_as_gpu final : public nvdevice {
public:
- explicit nvhost_as_gpu(std::shared_ptr<nvmap> nvmap_dev);
+ explicit nvhost_as_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
~nvhost_as_gpu() override;
- u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+ u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) override;
private:
enum class IoctlCommand : u32_le {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp
index b39fb9ef9..9a66a5f88 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp
@@ -7,14 +7,20 @@
#include "common/assert.h"
#include "common/logging/log.h"
+#include "core/core.h"
+#include "core/hle/kernel/readable_event.h"
+#include "core/hle/kernel/writable_event.h"
#include "core/hle/service/nvdrv/devices/nvhost_ctrl.h"
+#include "video_core/gpu.h"
namespace Service::Nvidia::Devices {
-nvhost_ctrl::nvhost_ctrl() = default;
+nvhost_ctrl::nvhost_ctrl(Core::System& system, EventInterface& events_interface)
+ : nvdevice(system), events_interface{events_interface} {}
nvhost_ctrl::~nvhost_ctrl() = default;
-u32 nvhost_ctrl::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_ctrl::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
LOG_DEBUG(Service_NVDRV, "called, command=0x{:08X}, input_size=0x{:X}, output_size=0x{:X}",
command.raw, input.size(), output.size());
@@ -22,11 +28,15 @@ u32 nvhost_ctrl::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<
case IoctlCommand::IocGetConfigCommand:
return NvOsGetConfigU32(input, output);
case IoctlCommand::IocCtrlEventWaitCommand:
- return IocCtrlEventWait(input, output, false);
+ return IocCtrlEventWait(input, output, false, ctrl);
case IoctlCommand::IocCtrlEventWaitAsyncCommand:
- return IocCtrlEventWait(input, output, true);
+ return IocCtrlEventWait(input, output, true, ctrl);
case IoctlCommand::IocCtrlEventRegisterCommand:
return IocCtrlEventRegister(input, output);
+ case IoctlCommand::IocCtrlEventUnregisterCommand:
+ return IocCtrlEventUnregister(input, output);
+ case IoctlCommand::IocCtrlEventSignalCommand:
+ return IocCtrlEventSignal(input, output);
}
UNIMPLEMENTED_MSG("Unimplemented ioctl");
return 0;
@@ -41,23 +51,137 @@ u32 nvhost_ctrl::NvOsGetConfigU32(const std::vector<u8>& input, std::vector<u8>&
}
u32 nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output,
- bool is_async) {
+ bool is_async, IoctlCtrl& ctrl) {
IocCtrlEventWaitParams params{};
std::memcpy(&params, input.data(), sizeof(params));
- LOG_WARNING(Service_NVDRV,
- "(STUBBED) called, syncpt_id={}, threshold={}, timeout={}, is_async={}",
- params.syncpt_id, params.threshold, params.timeout, is_async);
+ LOG_DEBUG(Service_NVDRV, "syncpt_id={}, threshold={}, timeout={}, is_async={}",
+ params.syncpt_id, params.threshold, params.timeout, is_async);
- // TODO(Subv): Implement actual syncpt waiting.
- params.value = 0;
+ if (params.syncpt_id >= MaxSyncPoints) {
+ return NvResult::BadParameter;
+ }
+
+ auto& gpu = system.GPU();
+ // This is mostly to take into account unimplemented features. As synced
+ // gpu is always synced.
+ if (!gpu.IsAsync()) {
+ return NvResult::Success;
+ }
+ auto lock = gpu.LockSync();
+ const u32 current_syncpoint_value = gpu.GetSyncpointValue(params.syncpt_id);
+ const s32 diff = current_syncpoint_value - params.threshold;
+ if (diff >= 0) {
+ params.value = current_syncpoint_value;
+ std::memcpy(output.data(), &params, sizeof(params));
+ return NvResult::Success;
+ }
+ const u32 target_value = current_syncpoint_value - diff;
+
+ if (!is_async) {
+ params.value = 0;
+ }
+
+ if (params.timeout == 0) {
+ std::memcpy(output.data(), &params, sizeof(params));
+ return NvResult::Timeout;
+ }
+
+ u32 event_id;
+ if (is_async) {
+ event_id = params.value & 0x00FF;
+ if (event_id >= MaxNvEvents) {
+ std::memcpy(output.data(), &params, sizeof(params));
+ return NvResult::BadParameter;
+ }
+ } else {
+ if (ctrl.fresh_call) {
+ const auto result = events_interface.GetFreeEvent();
+ if (result) {
+ event_id = *result;
+ } else {
+ LOG_CRITICAL(Service_NVDRV, "No Free Events available!");
+ event_id = params.value & 0x00FF;
+ }
+ } else {
+ event_id = ctrl.event_id;
+ }
+ }
+
+ EventState status = events_interface.status[event_id];
+ if (event_id < MaxNvEvents || status == EventState::Free || status == EventState::Registered) {
+ events_interface.SetEventStatus(event_id, EventState::Waiting);
+ events_interface.assigned_syncpt[event_id] = params.syncpt_id;
+ events_interface.assigned_value[event_id] = target_value;
+ if (is_async) {
+ params.value = params.syncpt_id << 4;
+ } else {
+ params.value = ((params.syncpt_id & 0xfff) << 16) | 0x10000000;
+ }
+ params.value |= event_id;
+ events_interface.events[event_id].writable->Clear();
+ gpu.RegisterSyncptInterrupt(params.syncpt_id, target_value);
+ if (!is_async && ctrl.fresh_call) {
+ ctrl.must_delay = true;
+ ctrl.timeout = params.timeout;
+ ctrl.event_id = event_id;
+ return NvResult::Timeout;
+ }
+ std::memcpy(output.data(), &params, sizeof(params));
+ return NvResult::Timeout;
+ }
std::memcpy(output.data(), &params, sizeof(params));
- return 0;
+ return NvResult::BadParameter;
}
u32 nvhost_ctrl::IocCtrlEventRegister(const std::vector<u8>& input, std::vector<u8>& output) {
- LOG_WARNING(Service_NVDRV, "(STUBBED) called");
- // TODO(bunnei): Implement this.
- return 0;
+ IocCtrlEventRegisterParams params{};
+ std::memcpy(&params, input.data(), sizeof(params));
+ const u32 event_id = params.user_event_id & 0x00FF;
+ LOG_DEBUG(Service_NVDRV, " called, user_event_id: {:X}", event_id);
+ if (event_id >= MaxNvEvents) {
+ return NvResult::BadParameter;
+ }
+ if (events_interface.registered[event_id]) {
+ return NvResult::BadParameter;
+ }
+ events_interface.RegisterEvent(event_id);
+ return NvResult::Success;
+}
+
+u32 nvhost_ctrl::IocCtrlEventUnregister(const std::vector<u8>& input, std::vector<u8>& output) {
+ IocCtrlEventUnregisterParams params{};
+ std::memcpy(&params, input.data(), sizeof(params));
+ const u32 event_id = params.user_event_id & 0x00FF;
+ LOG_DEBUG(Service_NVDRV, " called, user_event_id: {:X}", event_id);
+ if (event_id >= MaxNvEvents) {
+ return NvResult::BadParameter;
+ }
+ if (!events_interface.registered[event_id]) {
+ return NvResult::BadParameter;
+ }
+ events_interface.UnregisterEvent(event_id);
+ return NvResult::Success;
+}
+
+u32 nvhost_ctrl::IocCtrlEventSignal(const std::vector<u8>& input, std::vector<u8>& output) {
+ IocCtrlEventSignalParams params{};
+ std::memcpy(&params, input.data(), sizeof(params));
+ // TODO(Blinkhawk): This is normally called when an NvEvents timeout on WaitSynchronization
+ // It is believed from RE to cancel the GPU Event. However, better research is required
+ u32 event_id = params.user_event_id & 0x00FF;
+ LOG_WARNING(Service_NVDRV, "(STUBBED) called, user_event_id: {:X}", event_id);
+ if (event_id >= MaxNvEvents) {
+ return NvResult::BadParameter;
+ }
+ if (events_interface.status[event_id] == EventState::Waiting) {
+ auto& gpu = system.GPU();
+ if (gpu.CancelSyncptInterrupt(events_interface.assigned_syncpt[event_id],
+ events_interface.assigned_value[event_id])) {
+ events_interface.LiberateEvent(event_id);
+ events_interface.events[event_id].writable->Signal();
+ }
+ }
+ return NvResult::Success;
}
} // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h
index 6d0de2212..14e6e7e57 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h
@@ -8,15 +8,17 @@
#include <vector>
#include "common/common_types.h"
#include "core/hle/service/nvdrv/devices/nvdevice.h"
+#include "core/hle/service/nvdrv/nvdrv.h"
namespace Service::Nvidia::Devices {
class nvhost_ctrl final : public nvdevice {
public:
- nvhost_ctrl();
+ explicit nvhost_ctrl(Core::System& system, EventInterface& events_interface);
~nvhost_ctrl() override;
- u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+ u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) override;
private:
enum class IoctlCommand : u32_le {
@@ -132,9 +134,16 @@ private:
u32 NvOsGetConfigU32(const std::vector<u8>& input, std::vector<u8>& output);
- u32 IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output, bool is_async);
+ u32 IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output, bool is_async,
+ IoctlCtrl& ctrl);
u32 IocCtrlEventRegister(const std::vector<u8>& input, std::vector<u8>& output);
+
+ u32 IocCtrlEventUnregister(const std::vector<u8>& input, std::vector<u8>& output);
+
+ u32 IocCtrlEventSignal(const std::vector<u8>& input, std::vector<u8>& output);
+
+ EventInterface& events_interface;
};
} // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
index 0e28755bd..988effd90 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
@@ -12,10 +12,11 @@
namespace Service::Nvidia::Devices {
-nvhost_ctrl_gpu::nvhost_ctrl_gpu() = default;
+nvhost_ctrl_gpu::nvhost_ctrl_gpu(Core::System& system) : nvdevice(system) {}
nvhost_ctrl_gpu::~nvhost_ctrl_gpu() = default;
-u32 nvhost_ctrl_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_ctrl_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
LOG_DEBUG(Service_NVDRV, "called, command=0x{:08X}, input_size=0x{:X}, output_size=0x{:X}",
command.raw, input.size(), output.size());
@@ -185,7 +186,7 @@ u32 nvhost_ctrl_gpu::GetGpuTime(const std::vector<u8>& input, std::vector<u8>& o
IoctlGetGpuTime params{};
std::memcpy(&params, input.data(), input.size());
- const auto ns = Core::Timing::CyclesToNs(Core::System::GetInstance().CoreTiming().GetTicks());
+ const auto ns = Core::Timing::CyclesToNs(system.CoreTiming().GetTicks());
params.gpu_time = static_cast<u64_le>(ns.count());
std::memcpy(output.data(), &params, output.size());
return 0;
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h
index 240435eea..2b035ae3f 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h
@@ -13,10 +13,11 @@ namespace Service::Nvidia::Devices {
class nvhost_ctrl_gpu final : public nvdevice {
public:
- nvhost_ctrl_gpu();
+ explicit nvhost_ctrl_gpu(Core::System& system);
~nvhost_ctrl_gpu() override;
- u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+ u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) override;
private:
enum class IoctlCommand : u32_le {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
index 8ce7bc7a5..241dac881 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
@@ -13,10 +13,12 @@
namespace Service::Nvidia::Devices {
-nvhost_gpu::nvhost_gpu(std::shared_ptr<nvmap> nvmap_dev) : nvmap_dev(std::move(nvmap_dev)) {}
+nvhost_gpu::nvhost_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
+ : nvdevice(system), nvmap_dev(std::move(nvmap_dev)) {}
nvhost_gpu::~nvhost_gpu() = default;
-u32 nvhost_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_gpu::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
LOG_DEBUG(Service_NVDRV, "called, command=0x{:08X}, input_size=0x{:X}, output_size=0x{:X}",
command.raw, input.size(), output.size());
@@ -119,8 +121,10 @@ u32 nvhost_gpu::AllocGPFIFOEx2(const std::vector<u8>& input, std::vector<u8>& ou
params.num_entries, params.flags, params.unk0, params.unk1, params.unk2,
params.unk3);
- params.fence_out.id = 0;
- params.fence_out.value = 0;
+ auto& gpu = system.GPU();
+ params.fence_out.id = assigned_syncpoints;
+ params.fence_out.value = gpu.GetSyncpointValue(assigned_syncpoints);
+ assigned_syncpoints++;
std::memcpy(output.data(), &params, output.size());
return 0;
}
@@ -143,7 +147,7 @@ u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& outp
IoctlSubmitGpfifo params{};
std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
LOG_WARNING(Service_NVDRV, "(STUBBED) called, gpfifo={:X}, num_entries={:X}, flags={:X}",
- params.address, params.num_entries, params.flags);
+ params.address, params.num_entries, params.flags.raw);
ASSERT_MSG(input.size() == sizeof(IoctlSubmitGpfifo) +
params.num_entries * sizeof(Tegra::CommandListHeader),
@@ -153,10 +157,18 @@ u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& outp
std::memcpy(entries.data(), &input[sizeof(IoctlSubmitGpfifo)],
params.num_entries * sizeof(Tegra::CommandListHeader));
- Core::System::GetInstance().GPU().PushGPUEntries(std::move(entries));
+ UNIMPLEMENTED_IF(params.flags.add_wait.Value() != 0);
+ UNIMPLEMENTED_IF(params.flags.add_increment.Value() != 0);
+
+ auto& gpu = system.GPU();
+ u32 current_syncpoint_value = gpu.GetSyncpointValue(params.fence_out.id);
+ if (params.flags.increment.Value()) {
+ params.fence_out.value += current_syncpoint_value;
+ } else {
+ params.fence_out.value = current_syncpoint_value;
+ }
+ gpu.PushGPUEntries(std::move(entries));
- params.fence_out.id = 0;
- params.fence_out.value = 0;
std::memcpy(output.data(), &params, sizeof(IoctlSubmitGpfifo));
return 0;
}
@@ -168,16 +180,24 @@ u32 nvhost_gpu::KickoffPB(const std::vector<u8>& input, std::vector<u8>& output)
IoctlSubmitGpfifo params{};
std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
LOG_WARNING(Service_NVDRV, "(STUBBED) called, gpfifo={:X}, num_entries={:X}, flags={:X}",
- params.address, params.num_entries, params.flags);
+ params.address, params.num_entries, params.flags.raw);
Tegra::CommandList entries(params.num_entries);
Memory::ReadBlock(params.address, entries.data(),
params.num_entries * sizeof(Tegra::CommandListHeader));
- Core::System::GetInstance().GPU().PushGPUEntries(std::move(entries));
+ UNIMPLEMENTED_IF(params.flags.add_wait.Value() != 0);
+ UNIMPLEMENTED_IF(params.flags.add_increment.Value() != 0);
+
+ auto& gpu = system.GPU();
+ u32 current_syncpoint_value = gpu.GetSyncpointValue(params.fence_out.id);
+ if (params.flags.increment.Value()) {
+ params.fence_out.value += current_syncpoint_value;
+ } else {
+ params.fence_out.value = current_syncpoint_value;
+ }
+ gpu.PushGPUEntries(std::move(entries));
- params.fence_out.id = 0;
- params.fence_out.value = 0;
std::memcpy(output.data(), &params, output.size());
return 0;
}
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
index 62beb5c0c..d2e8fbae9 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
@@ -10,6 +10,7 @@
#include "common/common_types.h"
#include "common/swap.h"
#include "core/hle/service/nvdrv/devices/nvdevice.h"
+#include "core/hle/service/nvdrv/nvdata.h"
namespace Service::Nvidia::Devices {
@@ -20,10 +21,11 @@ constexpr u32 NVGPU_IOCTL_CHANNEL_KICKOFF_PB(0x1b);
class nvhost_gpu final : public nvdevice {
public:
- explicit nvhost_gpu(std::shared_ptr<nvmap> nvmap_dev);
+ explicit nvhost_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
~nvhost_gpu() override;
- u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+ u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) override;
private:
enum class IoctlCommand : u32_le {
@@ -113,11 +115,7 @@ private:
static_assert(sizeof(IoctlGetErrorNotification) == 16,
"IoctlGetErrorNotification is incorrect size");
- struct IoctlFence {
- u32_le id;
- u32_le value;
- };
- static_assert(sizeof(IoctlFence) == 8, "IoctlFence is incorrect size");
+ static_assert(sizeof(Fence) == 8, "Fence is incorrect size");
struct IoctlAllocGpfifoEx {
u32_le num_entries;
@@ -132,13 +130,13 @@ private:
static_assert(sizeof(IoctlAllocGpfifoEx) == 32, "IoctlAllocGpfifoEx is incorrect size");
struct IoctlAllocGpfifoEx2 {
- u32_le num_entries; // in
- u32_le flags; // in
- u32_le unk0; // in (1 works)
- IoctlFence fence_out; // out
- u32_le unk1; // in
- u32_le unk2; // in
- u32_le unk3; // in
+ u32_le num_entries; // in
+ u32_le flags; // in
+ u32_le unk0; // in (1 works)
+ Fence fence_out; // out
+ u32_le unk1; // in
+ u32_le unk2; // in
+ u32_le unk3; // in
};
static_assert(sizeof(IoctlAllocGpfifoEx2) == 32, "IoctlAllocGpfifoEx2 is incorrect size");
@@ -153,10 +151,16 @@ private:
struct IoctlSubmitGpfifo {
u64_le address; // pointer to gpfifo entry structs
u32_le num_entries; // number of fence objects being submitted
- u32_le flags;
- IoctlFence fence_out; // returned new fence object for others to wait on
- };
- static_assert(sizeof(IoctlSubmitGpfifo) == 16 + sizeof(IoctlFence),
+ union {
+ u32_le raw;
+ BitField<0, 1, u32_le> add_wait; // append a wait sync_point to the list
+ BitField<1, 1, u32_le> add_increment; // append an increment to the list
+ BitField<2, 1, u32_le> new_hw_format; // Mostly ignored
+ BitField<8, 1, u32_le> increment; // increment the returned fence
+ } flags;
+ Fence fence_out; // returned new fence object for others to wait on
+ };
+ static_assert(sizeof(IoctlSubmitGpfifo) == 16 + sizeof(Fence),
"IoctlSubmitGpfifo is incorrect size");
struct IoctlGetWaitbase {
@@ -184,6 +188,7 @@ private:
u32 ChannelSetTimeout(const std::vector<u8>& input, std::vector<u8>& output);
std::shared_ptr<nvmap> nvmap_dev;
+ u32 assigned_syncpoints{};
};
} // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
index f5e8ea7c3..f572ad30f 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
@@ -10,10 +10,11 @@
namespace Service::Nvidia::Devices {
-nvhost_nvdec::nvhost_nvdec() = default;
+nvhost_nvdec::nvhost_nvdec(Core::System& system) : nvdevice(system) {}
nvhost_nvdec::~nvhost_nvdec() = default;
-u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_nvdec::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
LOG_DEBUG(Service_NVDRV, "called, command=0x{:08X}, input_size=0x{:X}, output_size=0x{:X}",
command.raw, input.size(), output.size());
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
index 0e7b284f8..2710f0511 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
@@ -13,10 +13,11 @@ namespace Service::Nvidia::Devices {
class nvhost_nvdec final : public nvdevice {
public:
- nvhost_nvdec();
+ explicit nvhost_nvdec(Core::System& system);
~nvhost_nvdec() override;
- u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+ u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) override;
private:
enum class IoctlCommand : u32_le {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp
index 3e0951ab0..38282956f 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp
@@ -10,10 +10,11 @@
namespace Service::Nvidia::Devices {
-nvhost_nvjpg::nvhost_nvjpg() = default;
+nvhost_nvjpg::nvhost_nvjpg(Core::System& system) : nvdevice(system) {}
nvhost_nvjpg::~nvhost_nvjpg() = default;
-u32 nvhost_nvjpg::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_nvjpg::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
LOG_DEBUG(Service_NVDRV, "called, command=0x{:08X}, input_size=0x{:X}, output_size=0x{:X}",
command.raw, input.size(), output.size());
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h b/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h
index 89fd5e95e..379766693 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h
@@ -13,10 +13,11 @@ namespace Service::Nvidia::Devices {
class nvhost_nvjpg final : public nvdevice {
public:
- nvhost_nvjpg();
+ explicit nvhost_nvjpg(Core::System& system);
~nvhost_nvjpg() override;
- u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+ u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) override;
private:
enum class IoctlCommand : u32_le {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
index d544f0f31..70e8091db 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
@@ -10,10 +10,11 @@
namespace Service::Nvidia::Devices {
-nvhost_vic::nvhost_vic() = default;
+nvhost_vic::nvhost_vic(Core::System& system) : nvdevice(system) {}
nvhost_vic::~nvhost_vic() = default;
-u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_vic::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
LOG_DEBUG(Service_NVDRV, "called, command=0x{:08X}, input_size=0x{:X}, output_size=0x{:X}",
command.raw, input.size(), output.size());
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_vic.h b/src/core/hle/service/nvdrv/devices/nvhost_vic.h
index fc24c3f9c..7d111977e 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_vic.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_vic.h
@@ -13,10 +13,11 @@ namespace Service::Nvidia::Devices {
class nvhost_vic final : public nvdevice {
public:
- nvhost_vic();
+ explicit nvhost_vic(Core::System& system);
~nvhost_vic() override;
- u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+ u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) override;
private:
enum class IoctlCommand : u32_le {
diff --git a/src/core/hle/service/nvdrv/devices/nvmap.cpp b/src/core/hle/service/nvdrv/devices/nvmap.cpp
index 1ec796fc6..223b496b7 100644
--- a/src/core/hle/service/nvdrv/devices/nvmap.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvmap.cpp
@@ -18,7 +18,7 @@ enum {
};
}
-nvmap::nvmap() = default;
+nvmap::nvmap(Core::System& system) : nvdevice(system) {}
nvmap::~nvmap() = default;
VAddr nvmap::GetObjectAddress(u32 handle) const {
@@ -28,7 +28,8 @@ VAddr nvmap::GetObjectAddress(u32 handle) const {
return object->addr;
}
-u32 nvmap::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvmap::ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
switch (static_cast<IoctlCommand>(command.raw)) {
case IoctlCommand::Create:
return IocCreate(input, output);
diff --git a/src/core/hle/service/nvdrv/devices/nvmap.h b/src/core/hle/service/nvdrv/devices/nvmap.h
index 396230c19..bf4a101c2 100644
--- a/src/core/hle/service/nvdrv/devices/nvmap.h
+++ b/src/core/hle/service/nvdrv/devices/nvmap.h
@@ -16,13 +16,14 @@ namespace Service::Nvidia::Devices {
class nvmap final : public nvdevice {
public:
- nvmap();
+ explicit nvmap(Core::System& system);
~nvmap() override;
/// Returns the allocated address of an nvmap object given its handle.
VAddr GetObjectAddress(u32 handle) const;
- u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
+ u32 ioctl(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) override;
/// Represents an nvmap object.
struct Object {
diff --git a/src/core/hle/service/nvdrv/interface.cpp b/src/core/hle/service/nvdrv/interface.cpp
index b60fc748b..d5be64ed2 100644
--- a/src/core/hle/service/nvdrv/interface.cpp
+++ b/src/core/hle/service/nvdrv/interface.cpp
@@ -8,12 +8,18 @@
#include "core/hle/ipc_helpers.h"
#include "core/hle/kernel/kernel.h"
#include "core/hle/kernel/readable_event.h"
+#include "core/hle/kernel/thread.h"
#include "core/hle/kernel/writable_event.h"
#include "core/hle/service/nvdrv/interface.h"
+#include "core/hle/service/nvdrv/nvdata.h"
#include "core/hle/service/nvdrv/nvdrv.h"
namespace Service::Nvidia {
+void NVDRV::SignalGPUInterruptSyncpt(const u32 syncpoint_id, const u32 value) {
+ nvdrv->SignalSyncpt(syncpoint_id, value);
+}
+
void NVDRV::Open(Kernel::HLERequestContext& ctx) {
LOG_DEBUG(Service_NVDRV, "called");
@@ -36,11 +42,31 @@ void NVDRV::Ioctl(Kernel::HLERequestContext& ctx) {
std::vector<u8> output(ctx.GetWriteBufferSize());
+ IoctlCtrl ctrl{};
+
+ u32 result = nvdrv->Ioctl(fd, command, ctx.ReadBuffer(), output, ctrl);
+
+ if (ctrl.must_delay) {
+ ctrl.fresh_call = false;
+ ctx.SleepClientThread(
+ "NVServices::DelayedResponse", ctrl.timeout,
+ [=](Kernel::SharedPtr<Kernel::Thread> thread, Kernel::HLERequestContext& ctx,
+ Kernel::ThreadWakeupReason reason) {
+ IoctlCtrl ctrl2{ctrl};
+ std::vector<u8> output2 = output;
+ u32 result = nvdrv->Ioctl(fd, command, ctx.ReadBuffer(), output2, ctrl2);
+ ctx.WriteBuffer(output2);
+ IPC::ResponseBuilder rb{ctx, 3};
+ rb.Push(RESULT_SUCCESS);
+ rb.Push(result);
+ },
+ nvdrv->GetEventWriteable(ctrl.event_id));
+ } else {
+ ctx.WriteBuffer(output);
+ }
IPC::ResponseBuilder rb{ctx, 3};
rb.Push(RESULT_SUCCESS);
- rb.Push(nvdrv->Ioctl(fd, command, ctx.ReadBuffer(), output));
-
- ctx.WriteBuffer(output);
+ rb.Push(result);
}
void NVDRV::Close(Kernel::HLERequestContext& ctx) {
@@ -66,13 +92,19 @@ void NVDRV::Initialize(Kernel::HLERequestContext& ctx) {
void NVDRV::QueryEvent(Kernel::HLERequestContext& ctx) {
IPC::RequestParser rp{ctx};
u32 fd = rp.Pop<u32>();
- u32 event_id = rp.Pop<u32>();
+ // TODO(Blinkhawk): Figure the meaning of the flag at bit 16
+ u32 event_id = rp.Pop<u32>() & 0x000000FF;
LOG_WARNING(Service_NVDRV, "(STUBBED) called, fd={:X}, event_id={:X}", fd, event_id);
IPC::ResponseBuilder rb{ctx, 3, 1};
rb.Push(RESULT_SUCCESS);
- rb.PushCopyObjects(query_event.readable);
- rb.Push<u32>(0);
+ if (event_id < MaxNvEvents) {
+ rb.PushCopyObjects(nvdrv->GetEvent(event_id));
+ rb.Push<u32>(NvResult::Success);
+ } else {
+ rb.Push<u32>(0);
+ rb.Push<u32>(NvResult::BadParameter);
+ }
}
void NVDRV::SetClientPID(Kernel::HLERequestContext& ctx) {
@@ -127,10 +159,6 @@ NVDRV::NVDRV(std::shared_ptr<Module> nvdrv, const char* name)
{13, &NVDRV::FinishInitialize, "FinishInitialize"},
};
RegisterHandlers(functions);
-
- auto& kernel = Core::System::GetInstance().Kernel();
- query_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Automatic,
- "NVDRV::query_event");
}
NVDRV::~NVDRV() = default;
diff --git a/src/core/hle/service/nvdrv/interface.h b/src/core/hle/service/nvdrv/interface.h
index 5b4889910..10a0ecd52 100644
--- a/src/core/hle/service/nvdrv/interface.h
+++ b/src/core/hle/service/nvdrv/interface.h
@@ -19,6 +19,8 @@ public:
NVDRV(std::shared_ptr<Module> nvdrv, const char* name);
~NVDRV() override;
+ void SignalGPUInterruptSyncpt(const u32 syncpoint_id, const u32 value);
+
private:
void Open(Kernel::HLERequestContext& ctx);
void Ioctl(Kernel::HLERequestContext& ctx);
@@ -33,8 +35,6 @@ private:
std::shared_ptr<Module> nvdrv;
u64 pid{};
-
- Kernel::EventPair query_event;
};
} // namespace Service::Nvidia
diff --git a/src/core/hle/service/nvdrv/nvdata.h b/src/core/hle/service/nvdrv/nvdata.h
new file mode 100644
index 000000000..ac03cbc23
--- /dev/null
+++ b/src/core/hle/service/nvdrv/nvdata.h
@@ -0,0 +1,48 @@
+#pragma once
+
+#include <array>
+#include "common/common_types.h"
+
+namespace Service::Nvidia {
+
+constexpr u32 MaxSyncPoints = 192;
+constexpr u32 MaxNvEvents = 64;
+
+struct Fence {
+ s32 id;
+ u32 value;
+};
+
+static_assert(sizeof(Fence) == 8, "Fence has wrong size");
+
+struct MultiFence {
+ u32 num_fences;
+ std::array<Fence, 4> fences;
+};
+
+enum NvResult : u32 {
+ Success = 0,
+ BadParameter = 4,
+ Timeout = 5,
+ ResourceError = 15,
+};
+
+enum class EventState {
+ Free = 0,
+ Registered = 1,
+ Waiting = 2,
+ Busy = 3,
+};
+
+struct IoctlCtrl {
+ // First call done to the servioce for services that call itself again after a call.
+ bool fresh_call{true};
+ // Tells the Ioctl Wrapper that it must delay the IPC response and send the thread to sleep
+ bool must_delay{};
+ // Timeout for the delay
+ s64 timeout{};
+ // NV Event Id
+ s32 event_id{-1};
+};
+
+} // namespace Service::Nvidia
diff --git a/src/core/hle/service/nvdrv/nvdrv.cpp b/src/core/hle/service/nvdrv/nvdrv.cpp
index 6e4b8f2c6..2011a226a 100644
--- a/src/core/hle/service/nvdrv/nvdrv.cpp
+++ b/src/core/hle/service/nvdrv/nvdrv.cpp
@@ -4,7 +4,10 @@
#include <utility>
+#include <fmt/format.h>
#include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/readable_event.h"
+#include "core/hle/kernel/writable_event.h"
#include "core/hle/service/nvdrv/devices/nvdevice.h"
#include "core/hle/service/nvdrv/devices/nvdisp_disp0.h"
#include "core/hle/service/nvdrv/devices/nvhost_as_gpu.h"
@@ -22,8 +25,9 @@
namespace Service::Nvidia {
-void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger& nvflinger) {
- auto module_ = std::make_shared<Module>();
+void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger& nvflinger,
+ Core::System& system) {
+ auto module_ = std::make_shared<Module>(system);
std::make_shared<NVDRV>(module_, "nvdrv")->InstallAsService(service_manager);
std::make_shared<NVDRV>(module_, "nvdrv:a")->InstallAsService(service_manager);
std::make_shared<NVDRV>(module_, "nvdrv:s")->InstallAsService(service_manager);
@@ -32,17 +36,25 @@ void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger
nvflinger.SetNVDrvInstance(module_);
}
-Module::Module() {
- auto nvmap_dev = std::make_shared<Devices::nvmap>();
- devices["/dev/nvhost-as-gpu"] = std::make_shared<Devices::nvhost_as_gpu>(nvmap_dev);
- devices["/dev/nvhost-gpu"] = std::make_shared<Devices::nvhost_gpu>(nvmap_dev);
- devices["/dev/nvhost-ctrl-gpu"] = std::make_shared<Devices::nvhost_ctrl_gpu>();
+Module::Module(Core::System& system) {
+ auto& kernel = system.Kernel();
+ for (u32 i = 0; i < MaxNvEvents; i++) {
+ std::string event_label = fmt::format("NVDRV::NvEvent_{}", i);
+ events_interface.events[i] = Kernel::WritableEvent::CreateEventPair(
+ kernel, Kernel::ResetType::Automatic, event_label);
+ events_interface.status[i] = EventState::Free;
+ events_interface.registered[i] = false;
+ }
+ auto nvmap_dev = std::make_shared<Devices::nvmap>(system);
+ devices["/dev/nvhost-as-gpu"] = std::make_shared<Devices::nvhost_as_gpu>(system, nvmap_dev);
+ devices["/dev/nvhost-gpu"] = std::make_shared<Devices::nvhost_gpu>(system, nvmap_dev);
+ devices["/dev/nvhost-ctrl-gpu"] = std::make_shared<Devices::nvhost_ctrl_gpu>(system);
devices["/dev/nvmap"] = nvmap_dev;
- devices["/dev/nvdisp_disp0"] = std::make_shared<Devices::nvdisp_disp0>(nvmap_dev);
- devices["/dev/nvhost-ctrl"] = std::make_shared<Devices::nvhost_ctrl>();
- devices["/dev/nvhost-nvdec"] = std::make_shared<Devices::nvhost_nvdec>();
- devices["/dev/nvhost-nvjpg"] = std::make_shared<Devices::nvhost_nvjpg>();
- devices["/dev/nvhost-vic"] = std::make_shared<Devices::nvhost_vic>();
+ devices["/dev/nvdisp_disp0"] = std::make_shared<Devices::nvdisp_disp0>(system, nvmap_dev);
+ devices["/dev/nvhost-ctrl"] = std::make_shared<Devices::nvhost_ctrl>(system, events_interface);
+ devices["/dev/nvhost-nvdec"] = std::make_shared<Devices::nvhost_nvdec>(system);
+ devices["/dev/nvhost-nvjpg"] = std::make_shared<Devices::nvhost_nvjpg>(system);
+ devices["/dev/nvhost-vic"] = std::make_shared<Devices::nvhost_vic>(system);
}
Module::~Module() = default;
@@ -59,12 +71,13 @@ u32 Module::Open(const std::string& device_name) {
return fd;
}
-u32 Module::Ioctl(u32 fd, u32 command, const std::vector<u8>& input, std::vector<u8>& output) {
+u32 Module::Ioctl(u32 fd, u32 command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl) {
auto itr = open_files.find(fd);
ASSERT_MSG(itr != open_files.end(), "Tried to talk to an invalid device");
auto& device = itr->second;
- return device->ioctl({command}, input, output);
+ return device->ioctl({command}, input, output, ctrl);
}
ResultCode Module::Close(u32 fd) {
@@ -77,4 +90,22 @@ ResultCode Module::Close(u32 fd) {
return RESULT_SUCCESS;
}
+void Module::SignalSyncpt(const u32 syncpoint_id, const u32 value) {
+ for (u32 i = 0; i < MaxNvEvents; i++) {
+ if (events_interface.assigned_syncpt[i] == syncpoint_id &&
+ events_interface.assigned_value[i] == value) {
+ events_interface.LiberateEvent(i);
+ events_interface.events[i].writable->Signal();
+ }
+ }
+}
+
+Kernel::SharedPtr<Kernel::ReadableEvent> Module::GetEvent(const u32 event_id) const {
+ return events_interface.events[event_id].readable;
+}
+
+Kernel::SharedPtr<Kernel::WritableEvent> Module::GetEventWriteable(const u32 event_id) const {
+ return events_interface.events[event_id].writable;
+}
+
} // namespace Service::Nvidia
diff --git a/src/core/hle/service/nvdrv/nvdrv.h b/src/core/hle/service/nvdrv/nvdrv.h
index 53564f696..a339ab672 100644
--- a/src/core/hle/service/nvdrv/nvdrv.h
+++ b/src/core/hle/service/nvdrv/nvdrv.h
@@ -8,8 +8,14 @@
#include <unordered_map>
#include <vector>
#include "common/common_types.h"
+#include "core/hle/kernel/writable_event.h"
+#include "core/hle/service/nvdrv/nvdata.h"
#include "core/hle/service/service.h"
+namespace Core {
+class System;
+}
+
namespace Service::NVFlinger {
class NVFlinger;
}
@@ -20,16 +26,72 @@ namespace Devices {
class nvdevice;
}
-struct IoctlFence {
- u32 id;
- u32 value;
+struct EventInterface {
+ // Mask representing currently busy events
+ u64 events_mask{};
+ // Each kernel event associated to an NV event
+ std::array<Kernel::EventPair, MaxNvEvents> events;
+ // The status of the current NVEvent
+ std::array<EventState, MaxNvEvents> status{};
+ // Tells if an NVEvent is registered or not
+ std::array<bool, MaxNvEvents> registered{};
+ // When an NVEvent is waiting on GPU interrupt, this is the sync_point
+ // associated with it.
+ std::array<u32, MaxNvEvents> assigned_syncpt{};
+ // This is the value of the GPU interrupt for which the NVEvent is waiting
+ // for.
+ std::array<u32, MaxNvEvents> assigned_value{};
+ // Constant to denote an unasigned syncpoint.
+ static constexpr u32 unassigned_syncpt = 0xFFFFFFFF;
+ std::optional<u32> GetFreeEvent() const {
+ u64 mask = events_mask;
+ for (u32 i = 0; i < MaxNvEvents; i++) {
+ const bool is_free = (mask & 0x1) == 0;
+ if (is_free) {
+ if (status[i] == EventState::Registered || status[i] == EventState::Free) {
+ return {i};
+ }
+ }
+ mask = mask >> 1;
+ }
+ return {};
+ }
+ void SetEventStatus(const u32 event_id, EventState new_status) {
+ EventState old_status = status[event_id];
+ if (old_status == new_status) {
+ return;
+ }
+ status[event_id] = new_status;
+ if (new_status == EventState::Registered) {
+ registered[event_id] = true;
+ }
+ if (new_status == EventState::Waiting || new_status == EventState::Busy) {
+ events_mask |= (1ULL << event_id);
+ }
+ }
+ void RegisterEvent(const u32 event_id) {
+ registered[event_id] = true;
+ if (status[event_id] == EventState::Free) {
+ status[event_id] = EventState::Registered;
+ }
+ }
+ void UnregisterEvent(const u32 event_id) {
+ registered[event_id] = false;
+ if (status[event_id] == EventState::Registered) {
+ status[event_id] = EventState::Free;
+ }
+ }
+ void LiberateEvent(const u32 event_id) {
+ status[event_id] = registered[event_id] ? EventState::Registered : EventState::Free;
+ events_mask &= ~(1ULL << event_id);
+ assigned_syncpt[event_id] = unassigned_syncpt;
+ assigned_value[event_id] = 0;
+ }
};
-static_assert(sizeof(IoctlFence) == 8, "IoctlFence has wrong size");
-
class Module final {
public:
- Module();
+ Module(Core::System& system);
~Module();
/// Returns a pointer to one of the available devices, identified by its name.
@@ -44,10 +106,17 @@ public:
/// Opens a device node and returns a file descriptor to it.
u32 Open(const std::string& device_name);
/// Sends an ioctl command to the specified file descriptor.
- u32 Ioctl(u32 fd, u32 command, const std::vector<u8>& input, std::vector<u8>& output);
+ u32 Ioctl(u32 fd, u32 command, const std::vector<u8>& input, std::vector<u8>& output,
+ IoctlCtrl& ctrl);
/// Closes a device file descriptor and returns operation success.
ResultCode Close(u32 fd);
+ void SignalSyncpt(const u32 syncpoint_id, const u32 value);
+
+ Kernel::SharedPtr<Kernel::ReadableEvent> GetEvent(u32 event_id) const;
+
+ Kernel::SharedPtr<Kernel::WritableEvent> GetEventWriteable(u32 event_id) const;
+
private:
/// Id to use for the next open file descriptor.
u32 next_fd = 1;
@@ -57,9 +126,12 @@ private:
/// Mapping of device node names to their implementation.
std::unordered_map<std::string, std::shared_ptr<Devices::nvdevice>> devices;
+
+ EventInterface events_interface;
};
/// Registers all NVDRV services with the specified service manager.
-void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger& nvflinger);
+void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger& nvflinger,
+ Core::System& system);
} // namespace Service::Nvidia
diff --git a/src/core/hle/service/nvflinger/buffer_queue.cpp b/src/core/hle/service/nvflinger/buffer_queue.cpp
index 5731e815f..e1a07d3ee 100644
--- a/src/core/hle/service/nvflinger/buffer_queue.cpp
+++ b/src/core/hle/service/nvflinger/buffer_queue.cpp
@@ -34,7 +34,8 @@ void BufferQueue::SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer)
buffer_wait_event.writable->Signal();
}
-std::optional<u32> BufferQueue::DequeueBuffer(u32 width, u32 height) {
+std::optional<std::pair<u32, Service::Nvidia::MultiFence*>> BufferQueue::DequeueBuffer(u32 width,
+ u32 height) {
auto itr = std::find_if(queue.begin(), queue.end(), [&](const Buffer& buffer) {
// Only consider free buffers. Buffers become free once again after they've been Acquired
// and Released by the compositor, see the NVFlinger::Compose method.
@@ -51,7 +52,7 @@ std::optional<u32> BufferQueue::DequeueBuffer(u32 width, u32 height) {
}
itr->status = Buffer::Status::Dequeued;
- return itr->slot;
+ return {{itr->slot, &itr->multi_fence}};
}
const IGBPBuffer& BufferQueue::RequestBuffer(u32 slot) const {
@@ -63,7 +64,8 @@ const IGBPBuffer& BufferQueue::RequestBuffer(u32 slot) const {
}
void BufferQueue::QueueBuffer(u32 slot, BufferTransformFlags transform,
- const Common::Rectangle<int>& crop_rect) {
+ const Common::Rectangle<int>& crop_rect, u32 swap_interval,
+ Service::Nvidia::MultiFence& multi_fence) {
auto itr = std::find_if(queue.begin(), queue.end(),
[&](const Buffer& buffer) { return buffer.slot == slot; });
ASSERT(itr != queue.end());
@@ -71,12 +73,21 @@ void BufferQueue::QueueBuffer(u32 slot, BufferTransformFlags transform,
itr->status = Buffer::Status::Queued;
itr->transform = transform;
itr->crop_rect = crop_rect;
+ itr->swap_interval = swap_interval;
+ itr->multi_fence = multi_fence;
+ queue_sequence.push_back(slot);
}
std::optional<std::reference_wrapper<const BufferQueue::Buffer>> BufferQueue::AcquireBuffer() {
- auto itr = std::find_if(queue.begin(), queue.end(), [](const Buffer& buffer) {
- return buffer.status == Buffer::Status::Queued;
- });
+ auto itr = queue.end();
+ // Iterate to find a queued buffer matching the requested slot.
+ while (itr == queue.end() && !queue_sequence.empty()) {
+ u32 slot = queue_sequence.front();
+ itr = std::find_if(queue.begin(), queue.end(), [&slot](const Buffer& buffer) {
+ return buffer.status == Buffer::Status::Queued && buffer.slot == slot;
+ });
+ queue_sequence.pop_front();
+ }
if (itr == queue.end())
return {};
itr->status = Buffer::Status::Acquired;
diff --git a/src/core/hle/service/nvflinger/buffer_queue.h b/src/core/hle/service/nvflinger/buffer_queue.h
index e1ccb6171..356bedb81 100644
--- a/src/core/hle/service/nvflinger/buffer_queue.h
+++ b/src/core/hle/service/nvflinger/buffer_queue.h
@@ -4,6 +4,7 @@
#pragma once
+#include <list>
#include <optional>
#include <vector>
@@ -12,6 +13,7 @@
#include "common/swap.h"
#include "core/hle/kernel/object.h"
#include "core/hle/kernel/writable_event.h"
+#include "core/hle/service/nvdrv/nvdata.h"
namespace Service::NVFlinger {
@@ -68,13 +70,17 @@ public:
IGBPBuffer igbp_buffer;
BufferTransformFlags transform;
Common::Rectangle<int> crop_rect;
+ u32 swap_interval;
+ Service::Nvidia::MultiFence multi_fence;
};
void SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer);
- std::optional<u32> DequeueBuffer(u32 width, u32 height);
+ std::optional<std::pair<u32, Service::Nvidia::MultiFence*>> DequeueBuffer(u32 width,
+ u32 height);
const IGBPBuffer& RequestBuffer(u32 slot) const;
void QueueBuffer(u32 slot, BufferTransformFlags transform,
- const Common::Rectangle<int>& crop_rect);
+ const Common::Rectangle<int>& crop_rect, u32 swap_interval,
+ Service::Nvidia::MultiFence& multi_fence);
std::optional<std::reference_wrapper<const Buffer>> AcquireBuffer();
void ReleaseBuffer(u32 slot);
u32 Query(QueryType type);
@@ -92,6 +98,7 @@ private:
u64 layer_id;
std::vector<Buffer> queue;
+ std::list<u32> queue_sequence;
Kernel::EventPair buffer_wait_event;
};
diff --git a/src/core/hle/service/nvflinger/nvflinger.cpp b/src/core/hle/service/nvflinger/nvflinger.cpp
index 3c5c53e24..f9db79370 100644
--- a/src/core/hle/service/nvflinger/nvflinger.cpp
+++ b/src/core/hle/service/nvflinger/nvflinger.cpp
@@ -37,15 +37,14 @@ NVFlinger::NVFlinger(Core::Timing::CoreTiming& core_timing) : core_timing{core_t
displays.emplace_back(4, "Null");
// Schedule the screen composition events
- const auto ticks = Settings::values.force_30fps_mode ? frame_ticks_30fps : frame_ticks;
-
- composition_event = core_timing.RegisterEvent(
- "ScreenComposition", [this, ticks](u64 userdata, s64 cycles_late) {
- Compose();
- this->core_timing.ScheduleEvent(ticks - cycles_late, composition_event);
- });
-
- core_timing.ScheduleEvent(ticks, composition_event);
+ composition_event = core_timing.RegisterEvent("ScreenComposition", [this](u64 userdata,
+ s64 cycles_late) {
+ Compose();
+ const auto ticks = Settings::values.force_30fps_mode ? frame_ticks_30fps : GetNextTicks();
+ this->core_timing.ScheduleEvent(std::max<s64>(0LL, ticks - cycles_late), composition_event);
+ });
+
+ core_timing.ScheduleEvent(frame_ticks, composition_event);
}
NVFlinger::~NVFlinger() {
@@ -206,8 +205,14 @@ void NVFlinger::Compose() {
igbp_buffer.width, igbp_buffer.height, igbp_buffer.stride,
buffer->get().transform, buffer->get().crop_rect);
+ swap_interval = buffer->get().swap_interval;
buffer_queue.ReleaseBuffer(buffer->get().slot);
}
}
+s64 NVFlinger::GetNextTicks() const {
+ constexpr s64 max_hertz = 120LL;
+ return (Core::Timing::BASE_CLOCK_RATE * (1LL << swap_interval)) / max_hertz;
+}
+
} // namespace Service::NVFlinger
diff --git a/src/core/hle/service/nvflinger/nvflinger.h b/src/core/hle/service/nvflinger/nvflinger.h
index c0a83fffb..988be8726 100644
--- a/src/core/hle/service/nvflinger/nvflinger.h
+++ b/src/core/hle/service/nvflinger/nvflinger.h
@@ -74,6 +74,8 @@ public:
/// finished.
void Compose();
+ s64 GetNextTicks() const;
+
private:
/// Finds the display identified by the specified ID.
VI::Display* FindDisplay(u64 display_id);
@@ -98,6 +100,8 @@ private:
/// layers.
u32 next_buffer_queue_id = 1;
+ u32 swap_interval = 1;
+
/// Event that handles screen composition.
Core::Timing::EventType* composition_event;
diff --git a/src/core/hle/service/service.cpp b/src/core/hle/service/service.cpp
index 7eefd733f..3a0f8c3f6 100644
--- a/src/core/hle/service/service.cpp
+++ b/src/core/hle/service/service.cpp
@@ -206,7 +206,7 @@ void Init(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system) {
AM::InstallInterfaces(*sm, nv_flinger, system);
AOC::InstallInterfaces(*sm);
APM::InstallInterfaces(system);
- Audio::InstallInterfaces(*sm);
+ Audio::InstallInterfaces(*sm, system);
BCAT::InstallInterfaces(*sm);
BPC::InstallInterfaces(*sm);
BtDrv::InstallInterfaces(*sm);
@@ -236,7 +236,7 @@ void Init(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system) {
NIM::InstallInterfaces(*sm);
NPNS::InstallInterfaces(*sm);
NS::InstallInterfaces(*sm);
- Nvidia::InstallInterfaces(*sm, *nv_flinger);
+ Nvidia::InstallInterfaces(*sm, *nv_flinger, system);
PCIe::InstallInterfaces(*sm);
PCTL::InstallInterfaces(*sm);
PCV::InstallInterfaces(*sm);
diff --git a/src/core/hle/service/vi/vi.cpp b/src/core/hle/service/vi/vi.cpp
index f1fa6ccd1..199b30635 100644
--- a/src/core/hle/service/vi/vi.cpp
+++ b/src/core/hle/service/vi/vi.cpp
@@ -21,6 +21,7 @@
#include "core/hle/kernel/readable_event.h"
#include "core/hle/kernel/thread.h"
#include "core/hle/kernel/writable_event.h"
+#include "core/hle/service/nvdrv/nvdata.h"
#include "core/hle/service/nvdrv/nvdrv.h"
#include "core/hle/service/nvflinger/buffer_queue.h"
#include "core/hle/service/nvflinger/nvflinger.h"
@@ -328,32 +329,22 @@ public:
Data data;
};
-struct BufferProducerFence {
- u32 is_valid;
- std::array<Nvidia::IoctlFence, 4> fences;
-};
-static_assert(sizeof(BufferProducerFence) == 36, "BufferProducerFence has wrong size");
-
class IGBPDequeueBufferResponseParcel : public Parcel {
public:
- explicit IGBPDequeueBufferResponseParcel(u32 slot) : slot(slot) {}
+ explicit IGBPDequeueBufferResponseParcel(u32 slot, Service::Nvidia::MultiFence& multi_fence)
+ : slot(slot), multi_fence(multi_fence) {}
~IGBPDequeueBufferResponseParcel() override = default;
protected:
void SerializeData() override {
- // TODO(Subv): Find out how this Fence is used.
- BufferProducerFence fence = {};
- fence.is_valid = 1;
- for (auto& fence_ : fence.fences)
- fence_.id = -1;
-
Write(slot);
Write<u32_le>(1);
- WriteObject(fence);
+ WriteObject(multi_fence);
Write<u32_le>(0);
}
u32_le slot;
+ Service::Nvidia::MultiFence multi_fence;
};
class IGBPRequestBufferRequestParcel : public Parcel {
@@ -400,12 +391,6 @@ public:
data = Read<Data>();
}
- struct Fence {
- u32_le id;
- u32_le value;
- };
- static_assert(sizeof(Fence) == 8, "Fence has wrong size");
-
struct Data {
u32_le slot;
INSERT_PADDING_WORDS(3);
@@ -418,15 +403,15 @@ public:
s32_le scaling_mode;
NVFlinger::BufferQueue::BufferTransformFlags transform;
u32_le sticky_transform;
- INSERT_PADDING_WORDS(2);
- u32_le fence_is_valid;
- std::array<Fence, 2> fences;
+ INSERT_PADDING_WORDS(1);
+ u32_le swap_interval;
+ Service::Nvidia::MultiFence multi_fence;
Common::Rectangle<int> GetCropRect() const {
return {crop_left, crop_top, crop_right, crop_bottom};
}
};
- static_assert(sizeof(Data) == 80, "ParcelData has wrong size");
+ static_assert(sizeof(Data) == 96, "ParcelData has wrong size");
Data data;
};
@@ -547,11 +532,11 @@ private:
IGBPDequeueBufferRequestParcel request{ctx.ReadBuffer()};
const u32 width{request.data.width};
const u32 height{request.data.height};
- std::optional<u32> slot = buffer_queue.DequeueBuffer(width, height);
+ auto result = buffer_queue.DequeueBuffer(width, height);
- if (slot) {
+ if (result) {
// Buffer is available
- IGBPDequeueBufferResponseParcel response{*slot};
+ IGBPDequeueBufferResponseParcel response{result->first, *result->second};
ctx.WriteBuffer(response.Serialize());
} else {
// Wait the current thread until a buffer becomes available
@@ -561,10 +546,10 @@ private:
Kernel::ThreadWakeupReason reason) {
// Repeat TransactParcel DequeueBuffer when a buffer is available
auto& buffer_queue = nv_flinger->FindBufferQueue(id);
- std::optional<u32> slot = buffer_queue.DequeueBuffer(width, height);
- ASSERT_MSG(slot != std::nullopt, "Could not dequeue buffer.");
+ auto result = buffer_queue.DequeueBuffer(width, height);
+ ASSERT_MSG(result != std::nullopt, "Could not dequeue buffer.");
- IGBPDequeueBufferResponseParcel response{*slot};
+ IGBPDequeueBufferResponseParcel response{result->first, *result->second};
ctx.WriteBuffer(response.Serialize());
IPC::ResponseBuilder rb{ctx, 2};
rb.Push(RESULT_SUCCESS);
@@ -582,7 +567,8 @@ private:
IGBPQueueBufferRequestParcel request{ctx.ReadBuffer()};
buffer_queue.QueueBuffer(request.data.slot, request.data.transform,
- request.data.GetCropRect());
+ request.data.GetCropRect(), request.data.swap_interval,
+ request.data.multi_fence);
IGBPQueueBufferResponseParcel response{1280, 720};
ctx.WriteBuffer(response.Serialize());
diff --git a/src/core/loader/elf.cpp b/src/core/loader/elf.cpp
index 6d4b02375..f1795fdd6 100644
--- a/src/core/loader/elf.cpp
+++ b/src/core/loader/elf.cpp
@@ -295,7 +295,7 @@ Kernel::CodeSet ElfReader::LoadInto(VAddr vaddr) {
}
}
- std::vector<u8> program_image(total_image_size);
+ Kernel::PhysicalMemory program_image(total_image_size);
std::size_t current_image_position = 0;
Kernel::CodeSet codeset;
diff --git a/src/core/loader/kip.cpp b/src/core/loader/kip.cpp
index 70051c13a..474b55cb1 100644
--- a/src/core/loader/kip.cpp
+++ b/src/core/loader/kip.cpp
@@ -69,7 +69,7 @@ AppLoader::LoadResult AppLoader_KIP::Load(Kernel::Process& process) {
const VAddr base_address = process.VMManager().GetCodeRegionBaseAddress();
Kernel::CodeSet codeset;
- std::vector<u8> program_image;
+ Kernel::PhysicalMemory program_image;
const auto load_segment = [&program_image](Kernel::CodeSet::Segment& segment,
const std::vector<u8>& data, u32 offset) {
diff --git a/src/core/loader/nro.cpp b/src/core/loader/nro.cpp
index 6a0ca389b..e92e2e06e 100644
--- a/src/core/loader/nro.cpp
+++ b/src/core/loader/nro.cpp
@@ -143,7 +143,7 @@ static bool LoadNroImpl(Kernel::Process& process, const std::vector<u8>& data,
}
// Build program image
- std::vector<u8> program_image(PageAlignSize(nro_header.file_size));
+ Kernel::PhysicalMemory program_image(PageAlignSize(nro_header.file_size));
std::memcpy(program_image.data(), data.data(), program_image.size());
if (program_image.size() != PageAlignSize(nro_header.file_size)) {
return {};
diff --git a/src/core/loader/nso.cpp b/src/core/loader/nso.cpp
index 29311404a..70c90109f 100644
--- a/src/core/loader/nso.cpp
+++ b/src/core/loader/nso.cpp
@@ -89,7 +89,7 @@ std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::Process& process,
// Build program image
Kernel::CodeSet codeset;
- std::vector<u8> program_image;
+ Kernel::PhysicalMemory program_image;
for (std::size_t i = 0; i < nso_header.segments.size(); ++i) {
std::vector<u8> data =
file.ReadBytes(nso_header.segments_compressed_size[i], nso_header.segments[i].offset);
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 7c18c27b3..e2f85c5f1 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -1,5 +1,7 @@
add_library(video_core STATIC
- buffer_cache.h
+ buffer_cache/buffer_block.h
+ buffer_cache/buffer_cache.h
+ buffer_cache/map_interval.h
dma_pusher.cpp
dma_pusher.h
debug_utils/debug_utils.cpp
@@ -100,6 +102,7 @@ add_library(video_core STATIC
shader/decode/integer_set.cpp
shader/decode/half_set.cpp
shader/decode/video.cpp
+ shader/decode/warp.cpp
shader/decode/xmad.cpp
shader/decode/other.cpp
shader/control_flow.cpp
diff --git a/src/video_core/buffer_cache.h b/src/video_core/buffer_cache.h
deleted file mode 100644
index 6f868b8b4..000000000
--- a/src/video_core/buffer_cache.h
+++ /dev/null
@@ -1,299 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-#include <memory>
-#include <mutex>
-#include <unordered_map>
-#include <unordered_set>
-#include <utility>
-#include <vector>
-
-#include "common/alignment.h"
-#include "common/common_types.h"
-#include "core/core.h"
-#include "video_core/memory_manager.h"
-#include "video_core/rasterizer_cache.h"
-
-namespace VideoCore {
-class RasterizerInterface;
-}
-
-namespace VideoCommon {
-
-template <typename BufferStorageType>
-class CachedBuffer final : public RasterizerCacheObject {
-public:
- explicit CachedBuffer(VAddr cpu_addr, u8* host_ptr)
- : RasterizerCacheObject{host_ptr}, host_ptr{host_ptr}, cpu_addr{cpu_addr} {}
- ~CachedBuffer() override = default;
-
- VAddr GetCpuAddr() const override {
- return cpu_addr;
- }
-
- std::size_t GetSizeInBytes() const override {
- return size;
- }
-
- u8* GetWritableHostPtr() const {
- return host_ptr;
- }
-
- std::size_t GetSize() const {
- return size;
- }
-
- std::size_t GetCapacity() const {
- return capacity;
- }
-
- bool IsInternalized() const {
- return is_internal;
- }
-
- const BufferStorageType& GetBuffer() const {
- return buffer;
- }
-
- void SetSize(std::size_t new_size) {
- size = new_size;
- }
-
- void SetInternalState(bool is_internal_) {
- is_internal = is_internal_;
- }
-
- BufferStorageType ExchangeBuffer(BufferStorageType buffer_, std::size_t new_capacity) {
- capacity = new_capacity;
- std::swap(buffer, buffer_);
- return buffer_;
- }
-
-private:
- u8* host_ptr{};
- VAddr cpu_addr{};
- std::size_t size{};
- std::size_t capacity{};
- bool is_internal{};
- BufferStorageType buffer;
-};
-
-template <typename BufferStorageType, typename BufferType, typename StreamBuffer>
-class BufferCache : public RasterizerCache<std::shared_ptr<CachedBuffer<BufferStorageType>>> {
-public:
- using Buffer = std::shared_ptr<CachedBuffer<BufferStorageType>>;
- using BufferInfo = std::pair<const BufferType*, u64>;
-
- explicit BufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system,
- std::unique_ptr<StreamBuffer> stream_buffer)
- : RasterizerCache<Buffer>{rasterizer}, system{system},
- stream_buffer{std::move(stream_buffer)}, stream_buffer_handle{
- this->stream_buffer->GetHandle()} {}
- ~BufferCache() = default;
-
- void Unregister(const Buffer& entry) override {
- std::lock_guard lock{RasterizerCache<Buffer>::mutex};
- if (entry->IsInternalized()) {
- internalized_entries.erase(entry->GetCacheAddr());
- }
- ReserveBuffer(entry);
- RasterizerCache<Buffer>::Unregister(entry);
- }
-
- void TickFrame() {
- marked_for_destruction_index =
- (marked_for_destruction_index + 1) % marked_for_destruction_ring_buffer.size();
- MarkedForDestruction().clear();
- }
-
- BufferInfo UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::size_t alignment = 4,
- bool internalize = false, bool is_written = false) {
- std::lock_guard lock{RasterizerCache<Buffer>::mutex};
-
- auto& memory_manager = system.GPU().MemoryManager();
- const auto host_ptr = memory_manager.GetPointer(gpu_addr);
- if (!host_ptr) {
- return {GetEmptyBuffer(size), 0};
- }
- const auto cache_addr = ToCacheAddr(host_ptr);
-
- // Cache management is a big overhead, so only cache entries with a given size.
- // TODO: Figure out which size is the best for given games.
- constexpr std::size_t max_stream_size = 0x800;
- if (!internalize && size < max_stream_size &&
- internalized_entries.find(cache_addr) == internalized_entries.end()) {
- return StreamBufferUpload(host_ptr, size, alignment);
- }
-
- auto entry = RasterizerCache<Buffer>::TryGet(cache_addr);
- if (!entry) {
- return FixedBufferUpload(gpu_addr, host_ptr, size, internalize, is_written);
- }
-
- if (entry->GetSize() < size) {
- IncreaseBufferSize(entry, size);
- }
- if (is_written) {
- entry->MarkAsModified(true, *this);
- }
- return {ToHandle(entry->GetBuffer()), 0};
- }
-
- /// Uploads from a host memory. Returns the OpenGL buffer where it's located and its offset.
- BufferInfo UploadHostMemory(const void* raw_pointer, std::size_t size,
- std::size_t alignment = 4) {
- std::lock_guard lock{RasterizerCache<Buffer>::mutex};
- return StreamBufferUpload(raw_pointer, size, alignment);
- }
-
- void Map(std::size_t max_size) {
- std::tie(buffer_ptr, buffer_offset_base, invalidated) = stream_buffer->Map(max_size, 4);
- buffer_offset = buffer_offset_base;
- }
-
- /// Finishes the upload stream, returns true on bindings invalidation.
- bool Unmap() {
- stream_buffer->Unmap(buffer_offset - buffer_offset_base);
- return std::exchange(invalidated, false);
- }
-
- virtual const BufferType* GetEmptyBuffer(std::size_t size) = 0;
-
-protected:
- void FlushObjectInner(const Buffer& entry) override {
- DownloadBufferData(entry->GetBuffer(), 0, entry->GetSize(), entry->GetWritableHostPtr());
- }
-
- virtual BufferStorageType CreateBuffer(std::size_t size) = 0;
-
- virtual const BufferType* ToHandle(const BufferStorageType& storage) = 0;
-
- virtual void UploadBufferData(const BufferStorageType& buffer, std::size_t offset,
- std::size_t size, const u8* data) = 0;
-
- virtual void DownloadBufferData(const BufferStorageType& buffer, std::size_t offset,
- std::size_t size, u8* data) = 0;
-
- virtual void CopyBufferData(const BufferStorageType& src, const BufferStorageType& dst,
- std::size_t src_offset, std::size_t dst_offset,
- std::size_t size) = 0;
-
-private:
- BufferInfo StreamBufferUpload(const void* raw_pointer, std::size_t size,
- std::size_t alignment) {
- AlignBuffer(alignment);
- const std::size_t uploaded_offset = buffer_offset;
- std::memcpy(buffer_ptr, raw_pointer, size);
-
- buffer_ptr += size;
- buffer_offset += size;
- return {&stream_buffer_handle, uploaded_offset};
- }
-
- BufferInfo FixedBufferUpload(GPUVAddr gpu_addr, u8* host_ptr, std::size_t size,
- bool internalize, bool is_written) {
- auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
- const auto cpu_addr = memory_manager.GpuToCpuAddress(gpu_addr);
- ASSERT(cpu_addr);
-
- auto entry = GetUncachedBuffer(*cpu_addr, host_ptr);
- entry->SetSize(size);
- entry->SetInternalState(internalize);
- RasterizerCache<Buffer>::Register(entry);
-
- if (internalize) {
- internalized_entries.emplace(ToCacheAddr(host_ptr));
- }
- if (is_written) {
- entry->MarkAsModified(true, *this);
- }
-
- if (entry->GetCapacity() < size) {
- MarkedForDestruction().push_back(entry->ExchangeBuffer(CreateBuffer(size), size));
- }
-
- UploadBufferData(entry->GetBuffer(), 0, size, host_ptr);
- return {ToHandle(entry->GetBuffer()), 0};
- }
-
- void IncreaseBufferSize(Buffer& entry, std::size_t new_size) {
- const std::size_t old_size = entry->GetSize();
- if (entry->GetCapacity() < new_size) {
- const auto& old_buffer = entry->GetBuffer();
- auto new_buffer = CreateBuffer(new_size);
-
- // Copy bits from the old buffer to the new buffer.
- CopyBufferData(old_buffer, new_buffer, 0, 0, old_size);
- MarkedForDestruction().push_back(
- entry->ExchangeBuffer(std::move(new_buffer), new_size));
-
- // This buffer could have been used
- invalidated = true;
- }
- // Upload the new bits.
- const std::size_t size_diff = new_size - old_size;
- UploadBufferData(entry->GetBuffer(), old_size, size_diff, entry->GetHostPtr() + old_size);
-
- // Update entry's size in the object and in the cache.
- Unregister(entry);
-
- entry->SetSize(new_size);
- RasterizerCache<Buffer>::Register(entry);
- }
-
- Buffer GetUncachedBuffer(VAddr cpu_addr, u8* host_ptr) {
- if (auto entry = TryGetReservedBuffer(host_ptr)) {
- return entry;
- }
- return std::make_shared<CachedBuffer<BufferStorageType>>(cpu_addr, host_ptr);
- }
-
- Buffer TryGetReservedBuffer(u8* host_ptr) {
- const auto it = buffer_reserve.find(ToCacheAddr(host_ptr));
- if (it == buffer_reserve.end()) {
- return {};
- }
- auto& reserve = it->second;
- auto entry = reserve.back();
- reserve.pop_back();
- return entry;
- }
-
- void ReserveBuffer(Buffer entry) {
- buffer_reserve[entry->GetCacheAddr()].push_back(std::move(entry));
- }
-
- void AlignBuffer(std::size_t alignment) {
- // Align the offset, not the mapped pointer
- const std::size_t offset_aligned = Common::AlignUp(buffer_offset, alignment);
- buffer_ptr += offset_aligned - buffer_offset;
- buffer_offset = offset_aligned;
- }
-
- std::vector<BufferStorageType>& MarkedForDestruction() {
- return marked_for_destruction_ring_buffer[marked_for_destruction_index];
- }
-
- Core::System& system;
-
- std::unique_ptr<StreamBuffer> stream_buffer;
- BufferType stream_buffer_handle{};
-
- bool invalidated = false;
-
- u8* buffer_ptr = nullptr;
- u64 buffer_offset = 0;
- u64 buffer_offset_base = 0;
-
- std::size_t marked_for_destruction_index = 0;
- std::array<std::vector<BufferStorageType>, 4> marked_for_destruction_ring_buffer;
-
- std::unordered_set<CacheAddr> internalized_entries;
- std::unordered_map<CacheAddr, std::vector<Buffer>> buffer_reserve;
-};
-
-} // namespace VideoCommon
diff --git a/src/video_core/buffer_cache/buffer_block.h b/src/video_core/buffer_cache/buffer_block.h
new file mode 100644
index 000000000..4b9193182
--- /dev/null
+++ b/src/video_core/buffer_cache/buffer_block.h
@@ -0,0 +1,76 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <unordered_set>
+#include <utility>
+
+#include "common/alignment.h"
+#include "common/common_types.h"
+#include "video_core/gpu.h"
+
+namespace VideoCommon {
+
+class BufferBlock {
+public:
+ bool Overlaps(const CacheAddr start, const CacheAddr end) const {
+ return (cache_addr < end) && (cache_addr_end > start);
+ }
+
+ bool IsInside(const CacheAddr other_start, const CacheAddr other_end) const {
+ return cache_addr <= other_start && other_end <= cache_addr_end;
+ }
+
+ u8* GetWritableHostPtr() const {
+ return FromCacheAddr(cache_addr);
+ }
+
+ u8* GetWritableHostPtr(std::size_t offset) const {
+ return FromCacheAddr(cache_addr + offset);
+ }
+
+ std::size_t GetOffset(const CacheAddr in_addr) {
+ return static_cast<std::size_t>(in_addr - cache_addr);
+ }
+
+ CacheAddr GetCacheAddr() const {
+ return cache_addr;
+ }
+
+ CacheAddr GetCacheAddrEnd() const {
+ return cache_addr_end;
+ }
+
+ void SetCacheAddr(const CacheAddr new_addr) {
+ cache_addr = new_addr;
+ cache_addr_end = new_addr + size;
+ }
+
+ std::size_t GetSize() const {
+ return size;
+ }
+
+ void SetEpoch(u64 new_epoch) {
+ epoch = new_epoch;
+ }
+
+ u64 GetEpoch() {
+ return epoch;
+ }
+
+protected:
+ explicit BufferBlock(CacheAddr cache_addr, const std::size_t size) : size{size} {
+ SetCacheAddr(cache_addr);
+ }
+ ~BufferBlock() = default;
+
+private:
+ CacheAddr cache_addr{};
+ CacheAddr cache_addr_end{};
+ std::size_t size{};
+ u64 epoch{};
+};
+
+} // namespace VideoCommon
diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h
new file mode 100644
index 000000000..2442ddfd6
--- /dev/null
+++ b/src/video_core/buffer_cache/buffer_cache.h
@@ -0,0 +1,447 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+#include <mutex>
+#include <unordered_map>
+#include <unordered_set>
+#include <utility>
+#include <vector>
+
+#include "common/alignment.h"
+#include "common/common_types.h"
+#include "core/core.h"
+#include "video_core/buffer_cache/buffer_block.h"
+#include "video_core/buffer_cache/map_interval.h"
+#include "video_core/memory_manager.h"
+#include "video_core/rasterizer_interface.h"
+
+namespace VideoCommon {
+
+using MapInterval = std::shared_ptr<MapIntervalBase>;
+
+template <typename TBuffer, typename TBufferType, typename StreamBuffer>
+class BufferCache {
+public:
+ using BufferInfo = std::pair<const TBufferType*, u64>;
+
+ BufferInfo UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::size_t alignment = 4,
+ bool is_written = false) {
+ std::lock_guard lock{mutex};
+
+ auto& memory_manager = system.GPU().MemoryManager();
+ const auto host_ptr = memory_manager.GetPointer(gpu_addr);
+ if (!host_ptr) {
+ return {GetEmptyBuffer(size), 0};
+ }
+ const auto cache_addr = ToCacheAddr(host_ptr);
+
+ // Cache management is a big overhead, so only cache entries with a given size.
+ // TODO: Figure out which size is the best for given games.
+ constexpr std::size_t max_stream_size = 0x800;
+ if (size < max_stream_size) {
+ if (!is_written && !IsRegionWritten(cache_addr, cache_addr + size - 1)) {
+ return StreamBufferUpload(host_ptr, size, alignment);
+ }
+ }
+
+ auto block = GetBlock(cache_addr, size);
+ auto map = MapAddress(block, gpu_addr, cache_addr, size);
+ if (is_written) {
+ map->MarkAsModified(true, GetModifiedTicks());
+ if (!map->IsWritten()) {
+ map->MarkAsWritten(true);
+ MarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
+ }
+ } else {
+ if (map->IsWritten()) {
+ WriteBarrier();
+ }
+ }
+
+ const u64 offset = static_cast<u64>(block->GetOffset(cache_addr));
+
+ return {ToHandle(block), offset};
+ }
+
+ /// Uploads from a host memory. Returns the OpenGL buffer where it's located and its offset.
+ BufferInfo UploadHostMemory(const void* raw_pointer, std::size_t size,
+ std::size_t alignment = 4) {
+ std::lock_guard lock{mutex};
+ return StreamBufferUpload(raw_pointer, size, alignment);
+ }
+
+ void Map(std::size_t max_size) {
+ std::lock_guard lock{mutex};
+
+ std::tie(buffer_ptr, buffer_offset_base, invalidated) = stream_buffer->Map(max_size, 4);
+ buffer_offset = buffer_offset_base;
+ }
+
+ /// Finishes the upload stream, returns true on bindings invalidation.
+ bool Unmap() {
+ std::lock_guard lock{mutex};
+
+ stream_buffer->Unmap(buffer_offset - buffer_offset_base);
+ return std::exchange(invalidated, false);
+ }
+
+ void TickFrame() {
+ ++epoch;
+ while (!pending_destruction.empty()) {
+ if (pending_destruction.front()->GetEpoch() + 1 > epoch) {
+ break;
+ }
+ pending_destruction.pop_front();
+ }
+ }
+
+ /// Write any cached resources overlapping the specified region back to memory
+ void FlushRegion(CacheAddr addr, std::size_t size) {
+ std::lock_guard lock{mutex};
+
+ std::vector<MapInterval> objects = GetMapsInRange(addr, size);
+ std::sort(objects.begin(), objects.end(), [](const MapInterval& a, const MapInterval& b) {
+ return a->GetModificationTick() < b->GetModificationTick();
+ });
+ for (auto& object : objects) {
+ if (object->IsModified() && object->IsRegistered()) {
+ FlushMap(object);
+ }
+ }
+ }
+
+ /// Mark the specified region as being invalidated
+ void InvalidateRegion(CacheAddr addr, u64 size) {
+ std::lock_guard lock{mutex};
+
+ std::vector<MapInterval> objects = GetMapsInRange(addr, size);
+ for (auto& object : objects) {
+ if (object->IsRegistered()) {
+ Unregister(object);
+ }
+ }
+ }
+
+ virtual const TBufferType* GetEmptyBuffer(std::size_t size) = 0;
+
+protected:
+ explicit BufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system,
+ std::unique_ptr<StreamBuffer> stream_buffer)
+ : rasterizer{rasterizer}, system{system}, stream_buffer{std::move(stream_buffer)},
+ stream_buffer_handle{this->stream_buffer->GetHandle()} {}
+
+ ~BufferCache() = default;
+
+ virtual const TBufferType* ToHandle(const TBuffer& storage) = 0;
+
+ virtual void WriteBarrier() = 0;
+
+ virtual TBuffer CreateBlock(CacheAddr cache_addr, std::size_t size) = 0;
+
+ virtual void UploadBlockData(const TBuffer& buffer, std::size_t offset, std::size_t size,
+ const u8* data) = 0;
+
+ virtual void DownloadBlockData(const TBuffer& buffer, std::size_t offset, std::size_t size,
+ u8* data) = 0;
+
+ virtual void CopyBlock(const TBuffer& src, const TBuffer& dst, std::size_t src_offset,
+ std::size_t dst_offset, std::size_t size) = 0;
+
+ /// Register an object into the cache
+ void Register(const MapInterval& new_map, bool inherit_written = false) {
+ const CacheAddr cache_ptr = new_map->GetStart();
+ const std::optional<VAddr> cpu_addr =
+ system.GPU().MemoryManager().GpuToCpuAddress(new_map->GetGpuAddress());
+ if (!cache_ptr || !cpu_addr) {
+ LOG_CRITICAL(HW_GPU, "Failed to register buffer with unmapped gpu_address 0x{:016x}",
+ new_map->GetGpuAddress());
+ return;
+ }
+ const std::size_t size = new_map->GetEnd() - new_map->GetStart();
+ new_map->SetCpuAddress(*cpu_addr);
+ new_map->MarkAsRegistered(true);
+ const IntervalType interval{new_map->GetStart(), new_map->GetEnd()};
+ mapped_addresses.insert({interval, new_map});
+ rasterizer.UpdatePagesCachedCount(*cpu_addr, size, 1);
+ if (inherit_written) {
+ MarkRegionAsWritten(new_map->GetStart(), new_map->GetEnd() - 1);
+ new_map->MarkAsWritten(true);
+ }
+ }
+
+ /// Unregisters an object from the cache
+ void Unregister(MapInterval& map) {
+ const std::size_t size = map->GetEnd() - map->GetStart();
+ rasterizer.UpdatePagesCachedCount(map->GetCpuAddress(), size, -1);
+ map->MarkAsRegistered(false);
+ if (map->IsWritten()) {
+ UnmarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
+ }
+ const IntervalType delete_interval{map->GetStart(), map->GetEnd()};
+ mapped_addresses.erase(delete_interval);
+ }
+
+private:
+ MapInterval CreateMap(const CacheAddr start, const CacheAddr end, const GPUVAddr gpu_addr) {
+ return std::make_shared<MapIntervalBase>(start, end, gpu_addr);
+ }
+
+ MapInterval MapAddress(const TBuffer& block, const GPUVAddr gpu_addr,
+ const CacheAddr cache_addr, const std::size_t size) {
+
+ std::vector<MapInterval> overlaps = GetMapsInRange(cache_addr, size);
+ if (overlaps.empty()) {
+ const CacheAddr cache_addr_end = cache_addr + size;
+ MapInterval new_map = CreateMap(cache_addr, cache_addr_end, gpu_addr);
+ u8* host_ptr = FromCacheAddr(cache_addr);
+ UploadBlockData(block, block->GetOffset(cache_addr), size, host_ptr);
+ Register(new_map);
+ return new_map;
+ }
+
+ const CacheAddr cache_addr_end = cache_addr + size;
+ if (overlaps.size() == 1) {
+ MapInterval& current_map = overlaps[0];
+ if (current_map->IsInside(cache_addr, cache_addr_end)) {
+ return current_map;
+ }
+ }
+ CacheAddr new_start = cache_addr;
+ CacheAddr new_end = cache_addr_end;
+ bool write_inheritance = false;
+ bool modified_inheritance = false;
+ // Calculate new buffer parameters
+ for (auto& overlap : overlaps) {
+ new_start = std::min(overlap->GetStart(), new_start);
+ new_end = std::max(overlap->GetEnd(), new_end);
+ write_inheritance |= overlap->IsWritten();
+ modified_inheritance |= overlap->IsModified();
+ }
+ GPUVAddr new_gpu_addr = gpu_addr + new_start - cache_addr;
+ for (auto& overlap : overlaps) {
+ Unregister(overlap);
+ }
+ UpdateBlock(block, new_start, new_end, overlaps);
+ MapInterval new_map = CreateMap(new_start, new_end, new_gpu_addr);
+ if (modified_inheritance) {
+ new_map->MarkAsModified(true, GetModifiedTicks());
+ }
+ Register(new_map, write_inheritance);
+ return new_map;
+ }
+
+ void UpdateBlock(const TBuffer& block, CacheAddr start, CacheAddr end,
+ std::vector<MapInterval>& overlaps) {
+ const IntervalType base_interval{start, end};
+ IntervalSet interval_set{};
+ interval_set.add(base_interval);
+ for (auto& overlap : overlaps) {
+ const IntervalType subtract{overlap->GetStart(), overlap->GetEnd()};
+ interval_set.subtract(subtract);
+ }
+ for (auto& interval : interval_set) {
+ std::size_t size = interval.upper() - interval.lower();
+ if (size > 0) {
+ u8* host_ptr = FromCacheAddr(interval.lower());
+ UploadBlockData(block, block->GetOffset(interval.lower()), size, host_ptr);
+ }
+ }
+ }
+
+ std::vector<MapInterval> GetMapsInRange(CacheAddr addr, std::size_t size) {
+ if (size == 0) {
+ return {};
+ }
+
+ std::vector<MapInterval> objects{};
+ const IntervalType interval{addr, addr + size};
+ for (auto& pair : boost::make_iterator_range(mapped_addresses.equal_range(interval))) {
+ objects.push_back(pair.second);
+ }
+
+ return objects;
+ }
+
+ /// Returns a ticks counter used for tracking when cached objects were last modified
+ u64 GetModifiedTicks() {
+ return ++modified_ticks;
+ }
+
+ void FlushMap(MapInterval map) {
+ std::size_t size = map->GetEnd() - map->GetStart();
+ TBuffer block = blocks[map->GetStart() >> block_page_bits];
+ u8* host_ptr = FromCacheAddr(map->GetStart());
+ DownloadBlockData(block, block->GetOffset(map->GetStart()), size, host_ptr);
+ map->MarkAsModified(false, 0);
+ }
+
+ BufferInfo StreamBufferUpload(const void* raw_pointer, std::size_t size,
+ std::size_t alignment) {
+ AlignBuffer(alignment);
+ const std::size_t uploaded_offset = buffer_offset;
+ std::memcpy(buffer_ptr, raw_pointer, size);
+
+ buffer_ptr += size;
+ buffer_offset += size;
+ return {&stream_buffer_handle, uploaded_offset};
+ }
+
+ void AlignBuffer(std::size_t alignment) {
+ // Align the offset, not the mapped pointer
+ const std::size_t offset_aligned = Common::AlignUp(buffer_offset, alignment);
+ buffer_ptr += offset_aligned - buffer_offset;
+ buffer_offset = offset_aligned;
+ }
+
+ TBuffer EnlargeBlock(TBuffer buffer) {
+ const std::size_t old_size = buffer->GetSize();
+ const std::size_t new_size = old_size + block_page_size;
+ const CacheAddr cache_addr = buffer->GetCacheAddr();
+ TBuffer new_buffer = CreateBlock(cache_addr, new_size);
+ CopyBlock(buffer, new_buffer, 0, 0, old_size);
+ buffer->SetEpoch(epoch);
+ pending_destruction.push_back(buffer);
+ const CacheAddr cache_addr_end = cache_addr + new_size - 1;
+ u64 page_start = cache_addr >> block_page_bits;
+ const u64 page_end = cache_addr_end >> block_page_bits;
+ while (page_start <= page_end) {
+ blocks[page_start] = new_buffer;
+ ++page_start;
+ }
+ return new_buffer;
+ }
+
+ TBuffer MergeBlocks(TBuffer first, TBuffer second) {
+ const std::size_t size_1 = first->GetSize();
+ const std::size_t size_2 = second->GetSize();
+ const CacheAddr first_addr = first->GetCacheAddr();
+ const CacheAddr second_addr = second->GetCacheAddr();
+ const CacheAddr new_addr = std::min(first_addr, second_addr);
+ const std::size_t new_size = size_1 + size_2;
+ TBuffer new_buffer = CreateBlock(new_addr, new_size);
+ CopyBlock(first, new_buffer, 0, new_buffer->GetOffset(first_addr), size_1);
+ CopyBlock(second, new_buffer, 0, new_buffer->GetOffset(second_addr), size_2);
+ first->SetEpoch(epoch);
+ second->SetEpoch(epoch);
+ pending_destruction.push_back(first);
+ pending_destruction.push_back(second);
+ const CacheAddr cache_addr_end = new_addr + new_size - 1;
+ u64 page_start = new_addr >> block_page_bits;
+ const u64 page_end = cache_addr_end >> block_page_bits;
+ while (page_start <= page_end) {
+ blocks[page_start] = new_buffer;
+ ++page_start;
+ }
+ return new_buffer;
+ }
+
+ TBuffer GetBlock(const CacheAddr cache_addr, const std::size_t size) {
+ TBuffer found{};
+ const CacheAddr cache_addr_end = cache_addr + size - 1;
+ u64 page_start = cache_addr >> block_page_bits;
+ const u64 page_end = cache_addr_end >> block_page_bits;
+ while (page_start <= page_end) {
+ auto it = blocks.find(page_start);
+ if (it == blocks.end()) {
+ if (found) {
+ found = EnlargeBlock(found);
+ } else {
+ const CacheAddr start_addr = (page_start << block_page_bits);
+ found = CreateBlock(start_addr, block_page_size);
+ blocks[page_start] = found;
+ }
+ } else {
+ if (found) {
+ if (found == it->second) {
+ ++page_start;
+ continue;
+ }
+ found = MergeBlocks(found, it->second);
+ } else {
+ found = it->second;
+ }
+ }
+ ++page_start;
+ }
+ return found;
+ }
+
+ void MarkRegionAsWritten(const CacheAddr start, const CacheAddr end) {
+ u64 page_start = start >> write_page_bit;
+ const u64 page_end = end >> write_page_bit;
+ while (page_start <= page_end) {
+ auto it = written_pages.find(page_start);
+ if (it != written_pages.end()) {
+ it->second = it->second + 1;
+ } else {
+ written_pages[page_start] = 1;
+ }
+ page_start++;
+ }
+ }
+
+ void UnmarkRegionAsWritten(const CacheAddr start, const CacheAddr end) {
+ u64 page_start = start >> write_page_bit;
+ const u64 page_end = end >> write_page_bit;
+ while (page_start <= page_end) {
+ auto it = written_pages.find(page_start);
+ if (it != written_pages.end()) {
+ if (it->second > 1) {
+ it->second = it->second - 1;
+ } else {
+ written_pages.erase(it);
+ }
+ }
+ page_start++;
+ }
+ }
+
+ bool IsRegionWritten(const CacheAddr start, const CacheAddr end) const {
+ u64 page_start = start >> write_page_bit;
+ const u64 page_end = end >> write_page_bit;
+ while (page_start <= page_end) {
+ if (written_pages.count(page_start) > 0) {
+ return true;
+ }
+ page_start++;
+ }
+ return false;
+ }
+
+ VideoCore::RasterizerInterface& rasterizer;
+ Core::System& system;
+ std::unique_ptr<StreamBuffer> stream_buffer;
+
+ TBufferType stream_buffer_handle{};
+
+ bool invalidated = false;
+
+ u8* buffer_ptr = nullptr;
+ u64 buffer_offset = 0;
+ u64 buffer_offset_base = 0;
+
+ using IntervalSet = boost::icl::interval_set<CacheAddr>;
+ using IntervalCache = boost::icl::interval_map<CacheAddr, MapInterval>;
+ using IntervalType = typename IntervalCache::interval_type;
+ IntervalCache mapped_addresses{};
+
+ static constexpr u64 write_page_bit{11};
+ std::unordered_map<u64, u32> written_pages{};
+
+ static constexpr u64 block_page_bits{21};
+ static constexpr u64 block_page_size{1 << block_page_bits};
+ std::unordered_map<u64, TBuffer> blocks{};
+
+ std::list<TBuffer> pending_destruction{};
+ u64 epoch{};
+ u64 modified_ticks{};
+
+ std::recursive_mutex mutex;
+};
+
+} // namespace VideoCommon
diff --git a/src/video_core/buffer_cache/map_interval.h b/src/video_core/buffer_cache/map_interval.h
new file mode 100644
index 000000000..3a104d5cd
--- /dev/null
+++ b/src/video_core/buffer_cache/map_interval.h
@@ -0,0 +1,89 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "video_core/gpu.h"
+
+namespace VideoCommon {
+
+class MapIntervalBase {
+public:
+ MapIntervalBase(const CacheAddr start, const CacheAddr end, const GPUVAddr gpu_addr)
+ : start{start}, end{end}, gpu_addr{gpu_addr} {}
+
+ void SetCpuAddress(VAddr new_cpu_addr) {
+ cpu_addr = new_cpu_addr;
+ }
+
+ VAddr GetCpuAddress() const {
+ return cpu_addr;
+ }
+
+ GPUVAddr GetGpuAddress() const {
+ return gpu_addr;
+ }
+
+ bool IsInside(const CacheAddr other_start, const CacheAddr other_end) const {
+ return (start <= other_start && other_end <= end);
+ }
+
+ bool operator==(const MapIntervalBase& rhs) const {
+ return std::tie(start, end) == std::tie(rhs.start, rhs.end);
+ }
+
+ bool operator!=(const MapIntervalBase& rhs) const {
+ return !operator==(rhs);
+ }
+
+ void MarkAsRegistered(const bool registered) {
+ is_registered = registered;
+ }
+
+ bool IsRegistered() const {
+ return is_registered;
+ }
+
+ CacheAddr GetStart() const {
+ return start;
+ }
+
+ CacheAddr GetEnd() const {
+ return end;
+ }
+
+ void MarkAsModified(const bool is_modified_, const u64 tick) {
+ is_modified = is_modified_;
+ ticks = tick;
+ }
+
+ bool IsModified() const {
+ return is_modified;
+ }
+
+ u64 GetModificationTick() const {
+ return ticks;
+ }
+
+ void MarkAsWritten(const bool is_written_) {
+ is_written = is_written_;
+ }
+
+ bool IsWritten() const {
+ return is_written;
+ }
+
+private:
+ CacheAddr start;
+ CacheAddr end;
+ GPUVAddr gpu_addr;
+ VAddr cpu_addr{};
+ bool is_written{};
+ bool is_modified{};
+ bool is_registered{};
+ u64 ticks{};
+};
+
+} // namespace VideoCommon
diff --git a/src/video_core/dma_pusher.cpp b/src/video_core/dma_pusher.cpp
index bd036cbe8..0094fd715 100644
--- a/src/video_core/dma_pusher.cpp
+++ b/src/video_core/dma_pusher.cpp
@@ -31,6 +31,7 @@ void DmaPusher::DispatchCalls() {
break;
}
}
+ gpu.FlushCommands();
}
bool DmaPusher::Step() {
diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp
index 0ee228e28..98a8b5337 100644
--- a/src/video_core/engines/fermi_2d.cpp
+++ b/src/video_core/engines/fermi_2d.cpp
@@ -10,8 +10,7 @@
namespace Tegra::Engines {
-Fermi2D::Fermi2D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager)
- : rasterizer{rasterizer}, memory_manager{memory_manager} {}
+Fermi2D::Fermi2D(VideoCore::RasterizerInterface& rasterizer) : rasterizer{rasterizer} {}
void Fermi2D::CallMethod(const GPU::MethodCall& method_call) {
ASSERT_MSG(method_call.method < Regs::NUM_REGS,
diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h
index 05421d185..0901cf2fa 100644
--- a/src/video_core/engines/fermi_2d.h
+++ b/src/video_core/engines/fermi_2d.h
@@ -33,7 +33,7 @@ namespace Tegra::Engines {
class Fermi2D final {
public:
- explicit Fermi2D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager);
+ explicit Fermi2D(VideoCore::RasterizerInterface& rasterizer);
~Fermi2D() = default;
/// Write the value to the register identified by method.
@@ -145,7 +145,6 @@ public:
private:
VideoCore::RasterizerInterface& rasterizer;
- MemoryManager& memory_manager;
/// Performs the copy from the source surface to the destination surface as configured in the
/// registers.
diff --git a/src/video_core/engines/kepler_memory.cpp b/src/video_core/engines/kepler_memory.cpp
index 44279de00..fa4a7c5c1 100644
--- a/src/video_core/engines/kepler_memory.cpp
+++ b/src/video_core/engines/kepler_memory.cpp
@@ -15,7 +15,7 @@
namespace Tegra::Engines {
KeplerMemory::KeplerMemory(Core::System& system, MemoryManager& memory_manager)
- : system{system}, memory_manager{memory_manager}, upload_state{memory_manager, regs.upload} {}
+ : system{system}, upload_state{memory_manager, regs.upload} {}
KeplerMemory::~KeplerMemory() = default;
diff --git a/src/video_core/engines/kepler_memory.h b/src/video_core/engines/kepler_memory.h
index f3bc675a9..e0e25c321 100644
--- a/src/video_core/engines/kepler_memory.h
+++ b/src/video_core/engines/kepler_memory.h
@@ -65,7 +65,6 @@ public:
private:
Core::System& system;
- MemoryManager& memory_manager;
Upload::State upload_state;
};
diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index 74c46ec04..6a5a4f5c4 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -524,9 +524,10 @@ void Maxwell3D::ProcessQueryCondition() {
void Maxwell3D::ProcessSyncPoint() {
const u32 sync_point = regs.sync_info.sync_point.Value();
const u32 increment = regs.sync_info.increment.Value();
- const u32 cache_flush = regs.sync_info.unknown.Value();
- LOG_DEBUG(HW_GPU, "Syncpoint set {}, increment: {}, unk: {}", sync_point, increment,
- cache_flush);
+ [[maybe_unused]] const u32 cache_flush = regs.sync_info.unknown.Value();
+ if (increment) {
+ system.GPU().IncrementSyncPoint(sync_point);
+ }
}
void Maxwell3D::DrawArrays() {
@@ -625,10 +626,10 @@ Texture::TICEntry Maxwell3D::GetTICEntry(u32 tic_index) const {
Texture::TICEntry tic_entry;
memory_manager.ReadBlockUnsafe(tic_address_gpu, &tic_entry, sizeof(Texture::TICEntry));
- const auto r_type{tic_entry.r_type.Value()};
- const auto g_type{tic_entry.g_type.Value()};
- const auto b_type{tic_entry.b_type.Value()};
- const auto a_type{tic_entry.a_type.Value()};
+ [[maybe_unused]] const auto r_type{tic_entry.r_type.Value()};
+ [[maybe_unused]] const auto g_type{tic_entry.g_type.Value()};
+ [[maybe_unused]] const auto b_type{tic_entry.b_type.Value()};
+ [[maybe_unused]] const auto a_type{tic_entry.a_type.Value()};
// TODO(Subv): Different data types for separate components are not supported
DEBUG_ASSERT(r_type == g_type && r_type == b_type && r_type == a_type);
diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp
index a71d98e36..ad8453c5f 100644
--- a/src/video_core/engines/maxwell_dma.cpp
+++ b/src/video_core/engines/maxwell_dma.cpp
@@ -9,15 +9,13 @@
#include "video_core/engines/maxwell_3d.h"
#include "video_core/engines/maxwell_dma.h"
#include "video_core/memory_manager.h"
-#include "video_core/rasterizer_interface.h"
#include "video_core/renderer_base.h"
#include "video_core/textures/decoders.h"
namespace Tegra::Engines {
-MaxwellDMA::MaxwellDMA(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
- MemoryManager& memory_manager)
- : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager} {}
+MaxwellDMA::MaxwellDMA(Core::System& system, MemoryManager& memory_manager)
+ : system{system}, memory_manager{memory_manager} {}
void MaxwellDMA::CallMethod(const GPU::MethodCall& method_call) {
ASSERT_MSG(method_call.method < Regs::NUM_REGS,
@@ -39,7 +37,7 @@ void MaxwellDMA::CallMethod(const GPU::MethodCall& method_call) {
}
void MaxwellDMA::HandleCopy() {
- LOG_WARNING(HW_GPU, "Requested a DMA copy");
+ LOG_TRACE(HW_GPU, "Requested a DMA copy");
const GPUVAddr source = regs.src_address.Address();
const GPUVAddr dest = regs.dst_address.Address();
diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h
index 17b015ca7..93808a9bb 100644
--- a/src/video_core/engines/maxwell_dma.h
+++ b/src/video_core/engines/maxwell_dma.h
@@ -20,10 +20,6 @@ namespace Tegra {
class MemoryManager;
}
-namespace VideoCore {
-class RasterizerInterface;
-}
-
namespace Tegra::Engines {
/**
@@ -33,8 +29,7 @@ namespace Tegra::Engines {
class MaxwellDMA final {
public:
- explicit MaxwellDMA(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
- MemoryManager& memory_manager);
+ explicit MaxwellDMA(Core::System& system, MemoryManager& memory_manager);
~MaxwellDMA() = default;
/// Write the value to the register identified by method.
@@ -180,8 +175,6 @@ public:
private:
Core::System& system;
- VideoCore::RasterizerInterface& rasterizer;
-
MemoryManager& memory_manager;
std::vector<u8> read_buffer;
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 8520a0143..bc8c2a1c5 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -538,6 +538,12 @@ enum class PhysicalAttributeDirection : u64 {
Output = 1,
};
+enum class VoteOperation : u64 {
+ All = 0, // allThreadsNV
+ Any = 1, // anyThreadNV
+ Eq = 2, // allThreadsEqualNV
+};
+
union Instruction {
Instruction& operator=(const Instruction& instr) {
value = instr.value;
@@ -560,6 +566,18 @@ union Instruction {
BitField<48, 16, u64> opcode;
union {
+ BitField<8, 5, ConditionCode> cc;
+ BitField<13, 1, u64> trigger;
+ } nop;
+
+ union {
+ BitField<48, 2, VoteOperation> operation;
+ BitField<45, 3, u64> dest_pred;
+ BitField<39, 3, u64> value;
+ BitField<42, 1, u64> negate_value;
+ } vote;
+
+ union {
BitField<8, 8, Register> gpr;
BitField<20, 24, s64> offset;
} gmem;
@@ -1018,8 +1036,6 @@ union Instruction {
} f2i;
union {
- BitField<8, 2, Register::Size> src_size;
- BitField<10, 2, Register::Size> dst_size;
BitField<39, 4, u64> rounding;
// H0, H1 extract for F16 missing
BitField<41, 1, u64> selector; // Guessed as some games set it, TODO: reverse this value
@@ -1484,6 +1500,7 @@ public:
SYNC,
BRK,
DEPBAR,
+ VOTE,
BFE_C,
BFE_R,
BFE_IMM,
@@ -1516,6 +1533,7 @@ public:
TMML, // Texture Mip Map Level
SUST, // Surface Store
EXIT,
+ NOP,
IPA,
OUT_R, // Emit vertex/primitive
ISBERD,
@@ -1645,6 +1663,7 @@ public:
Hfma2,
Flow,
Synch,
+ Warp,
Memory,
Texture,
Image,
@@ -1771,6 +1790,7 @@ private:
INST("111000110100---", Id::BRK, Type::Flow, "BRK"),
INST("111000110000----", Id::EXIT, Type::Flow, "EXIT"),
INST("1111000011110---", Id::DEPBAR, Type::Synch, "DEPBAR"),
+ INST("0101000011011---", Id::VOTE, Type::Warp, "VOTE"),
INST("1110111111011---", Id::LD_A, Type::Memory, "LD_A"),
INST("1110111101001---", Id::LD_S, Type::Memory, "LD_S"),
INST("1110111101000---", Id::LD_L, Type::Memory, "LD_L"),
@@ -1795,6 +1815,7 @@ private:
INST("110111110110----", Id::TMML_B, Type::Texture, "TMML_B"),
INST("1101111101011---", Id::TMML, Type::Texture, "TMML"),
INST("11101011001-----", Id::SUST, Type::Image, "SUST"),
+ INST("0101000010110---", Id::NOP, Type::Trivial, "NOP"),
INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"),
INST("1110111111010---", Id::ISBERD, Type::Trivial, "ISBERD"),
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index e25754e37..8d9db45f5 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -29,14 +29,15 @@ u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
UNREACHABLE();
}
-GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer) : renderer{renderer} {
+GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer, bool is_async)
+ : system{system}, renderer{renderer}, is_async{is_async} {
auto& rasterizer{renderer.Rasterizer()};
memory_manager = std::make_unique<Tegra::MemoryManager>(system, rasterizer);
dma_pusher = std::make_unique<Tegra::DmaPusher>(*this);
maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager);
- fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer, *memory_manager);
+ fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer);
kepler_compute = std::make_unique<Engines::KeplerCompute>(system, rasterizer, *memory_manager);
- maxwell_dma = std::make_unique<Engines::MaxwellDMA>(system, rasterizer, *memory_manager);
+ maxwell_dma = std::make_unique<Engines::MaxwellDMA>(system, *memory_manager);
kepler_memory = std::make_unique<Engines::KeplerMemory>(system, *memory_manager);
}
@@ -74,6 +75,55 @@ const DmaPusher& GPU::DmaPusher() const {
return *dma_pusher;
}
+void GPU::IncrementSyncPoint(const u32 syncpoint_id) {
+ syncpoints[syncpoint_id]++;
+ std::lock_guard lock{sync_mutex};
+ if (!syncpt_interrupts[syncpoint_id].empty()) {
+ u32 value = syncpoints[syncpoint_id].load();
+ auto it = syncpt_interrupts[syncpoint_id].begin();
+ while (it != syncpt_interrupts[syncpoint_id].end()) {
+ if (value >= *it) {
+ TriggerCpuInterrupt(syncpoint_id, *it);
+ it = syncpt_interrupts[syncpoint_id].erase(it);
+ continue;
+ }
+ it++;
+ }
+ }
+}
+
+u32 GPU::GetSyncpointValue(const u32 syncpoint_id) const {
+ return syncpoints[syncpoint_id].load();
+}
+
+void GPU::RegisterSyncptInterrupt(const u32 syncpoint_id, const u32 value) {
+ auto& interrupt = syncpt_interrupts[syncpoint_id];
+ bool contains = std::any_of(interrupt.begin(), interrupt.end(),
+ [value](u32 in_value) { return in_value == value; });
+ if (contains) {
+ return;
+ }
+ syncpt_interrupts[syncpoint_id].emplace_back(value);
+}
+
+bool GPU::CancelSyncptInterrupt(const u32 syncpoint_id, const u32 value) {
+ std::lock_guard lock{sync_mutex};
+ auto& interrupt = syncpt_interrupts[syncpoint_id];
+ const auto iter =
+ std::find_if(interrupt.begin(), interrupt.end(),
+ [value](u32 interrupt_value) { return value == interrupt_value; });
+
+ if (iter == interrupt.end()) {
+ return false;
+ }
+ interrupt.erase(iter);
+ return true;
+}
+
+void GPU::FlushCommands() {
+ renderer.Rasterizer().FlushCommands();
+}
+
u32 RenderTargetBytesPerPixel(RenderTargetFormat format) {
ASSERT(format != RenderTargetFormat::NONE);
@@ -151,12 +201,12 @@ enum class BufferMethods {
NotifyIntr = 0x8,
WrcacheFlush = 0x9,
Unk28 = 0xA,
- Unk2c = 0xB,
+ UnkCacheFlush = 0xB,
RefCnt = 0x14,
SemaphoreAcquire = 0x1A,
SemaphoreRelease = 0x1B,
- Unk70 = 0x1C,
- Unk74 = 0x1D,
+ FenceValue = 0x1C,
+ FenceAction = 0x1D,
Unk78 = 0x1E,
Unk7c = 0x1F,
Yield = 0x20,
@@ -202,6 +252,10 @@ void GPU::CallPullerMethod(const MethodCall& method_call) {
case BufferMethods::SemaphoreAddressLow:
case BufferMethods::SemaphoreSequence:
case BufferMethods::RefCnt:
+ case BufferMethods::UnkCacheFlush:
+ case BufferMethods::WrcacheFlush:
+ case BufferMethods::FenceValue:
+ case BufferMethods::FenceAction:
break;
case BufferMethods::SemaphoreTrigger: {
ProcessSemaphoreTriggerMethod();
@@ -212,21 +266,11 @@ void GPU::CallPullerMethod(const MethodCall& method_call) {
LOG_ERROR(HW_GPU, "Special puller engine method NotifyIntr not implemented");
break;
}
- case BufferMethods::WrcacheFlush: {
- // TODO(Kmather73): Research and implement this method.
- LOG_ERROR(HW_GPU, "Special puller engine method WrcacheFlush not implemented");
- break;
- }
case BufferMethods::Unk28: {
// TODO(Kmather73): Research and implement this method.
LOG_ERROR(HW_GPU, "Special puller engine method Unk28 not implemented");
break;
}
- case BufferMethods::Unk2c: {
- // TODO(Kmather73): Research and implement this method.
- LOG_ERROR(HW_GPU, "Special puller engine method Unk2c not implemented");
- break;
- }
case BufferMethods::SemaphoreAcquire: {
ProcessSemaphoreAcquire();
break;
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 0ace0ff4f..544340ecd 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -5,8 +5,12 @@
#pragma once
#include <array>
+#include <atomic>
+#include <list>
#include <memory>
+#include <mutex>
#include "common/common_types.h"
+#include "core/hle/service/nvdrv/nvdata.h"
#include "core/hle/service/nvflinger/buffer_queue.h"
#include "video_core/dma_pusher.h"
@@ -15,6 +19,10 @@ inline CacheAddr ToCacheAddr(const void* host_ptr) {
return reinterpret_cast<CacheAddr>(host_ptr);
}
+inline u8* FromCacheAddr(CacheAddr cache_addr) {
+ return reinterpret_cast<u8*>(cache_addr);
+}
+
namespace Core {
class System;
}
@@ -127,7 +135,7 @@ class MemoryManager;
class GPU {
public:
- explicit GPU(Core::System& system, VideoCore::RendererBase& renderer);
+ explicit GPU(Core::System& system, VideoCore::RendererBase& renderer, bool is_async);
virtual ~GPU();
@@ -149,6 +157,8 @@ public:
/// Calls a GPU method.
void CallMethod(const MethodCall& method_call);
+ void FlushCommands();
+
/// Returns a reference to the Maxwell3D GPU engine.
Engines::Maxwell3D& Maxwell3D();
@@ -170,6 +180,22 @@ public:
/// Returns a reference to the GPU DMA pusher.
Tegra::DmaPusher& DmaPusher();
+ void IncrementSyncPoint(u32 syncpoint_id);
+
+ u32 GetSyncpointValue(u32 syncpoint_id) const;
+
+ void RegisterSyncptInterrupt(u32 syncpoint_id, u32 value);
+
+ bool CancelSyncptInterrupt(u32 syncpoint_id, u32 value);
+
+ std::unique_lock<std::mutex> LockSync() {
+ return std::unique_lock{sync_mutex};
+ }
+
+ bool IsAsync() const {
+ return is_async;
+ }
+
/// Returns a const reference to the GPU DMA pusher.
const Tegra::DmaPusher& DmaPusher() const;
@@ -200,7 +226,12 @@ public:
u32 semaphore_acquire;
u32 semaphore_release;
- INSERT_PADDING_WORDS(0xE4);
+ u32 fence_value;
+ union {
+ BitField<4, 4, u32> operation;
+ BitField<8, 8, u32> id;
+ } fence_action;
+ INSERT_PADDING_WORDS(0xE2);
// Puller state
u32 acquire_mode;
@@ -234,6 +265,9 @@ public:
/// Notify rasterizer that any caches of the specified region should be flushed and invalidated
virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0;
+protected:
+ virtual void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const = 0;
+
private:
void ProcessBindMethod(const MethodCall& method_call);
void ProcessSemaphoreTriggerMethod();
@@ -251,6 +285,7 @@ private:
protected:
std::unique_ptr<Tegra::DmaPusher> dma_pusher;
+ Core::System& system;
VideoCore::RendererBase& renderer;
private:
@@ -268,6 +303,14 @@ private:
std::unique_ptr<Engines::MaxwellDMA> maxwell_dma;
/// Inline memory engine
std::unique_ptr<Engines::KeplerMemory> kepler_memory;
+
+ std::array<std::atomic<u32>, Service::Nvidia::MaxSyncPoints> syncpoints{};
+
+ std::array<std::list<u32>, Service::Nvidia::MaxSyncPoints> syncpt_interrupts;
+
+ std::mutex sync_mutex;
+
+ const bool is_async;
};
#define ASSERT_REG_POSITION(field_name, position) \
@@ -280,6 +323,8 @@ ASSERT_REG_POSITION(semaphore_trigger, 0x7);
ASSERT_REG_POSITION(reference_count, 0x14);
ASSERT_REG_POSITION(semaphore_acquire, 0x1A);
ASSERT_REG_POSITION(semaphore_release, 0x1B);
+ASSERT_REG_POSITION(fence_value, 0x1C);
+ASSERT_REG_POSITION(fence_action, 0x1D);
ASSERT_REG_POSITION(acquire_mode, 0x100);
ASSERT_REG_POSITION(acquire_source, 0x101);
diff --git a/src/video_core/gpu_asynch.cpp b/src/video_core/gpu_asynch.cpp
index d4e2553a9..ea67be831 100644
--- a/src/video_core/gpu_asynch.cpp
+++ b/src/video_core/gpu_asynch.cpp
@@ -2,6 +2,8 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
+#include "core/core.h"
+#include "core/hardware_interrupt_manager.h"
#include "video_core/gpu_asynch.h"
#include "video_core/gpu_thread.h"
#include "video_core/renderer_base.h"
@@ -9,7 +11,7 @@
namespace VideoCommon {
GPUAsynch::GPUAsynch(Core::System& system, VideoCore::RendererBase& renderer)
- : GPU(system, renderer), gpu_thread{system} {}
+ : GPU(system, renderer, true), gpu_thread{system} {}
GPUAsynch::~GPUAsynch() = default;
@@ -38,4 +40,9 @@ void GPUAsynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
gpu_thread.FlushAndInvalidateRegion(addr, size);
}
+void GPUAsynch::TriggerCpuInterrupt(const u32 syncpoint_id, const u32 value) const {
+ auto& interrupt_manager = system.InterruptManager();
+ interrupt_manager.GPUInterruptSyncpt(syncpoint_id, value);
+}
+
} // namespace VideoCommon
diff --git a/src/video_core/gpu_asynch.h b/src/video_core/gpu_asynch.h
index 30be74cba..36377d677 100644
--- a/src/video_core/gpu_asynch.h
+++ b/src/video_core/gpu_asynch.h
@@ -27,6 +27,9 @@ public:
void InvalidateRegion(CacheAddr addr, u64 size) override;
void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+protected:
+ void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const override;
+
private:
GPUThread::ThreadManager gpu_thread;
};
diff --git a/src/video_core/gpu_synch.cpp b/src/video_core/gpu_synch.cpp
index 45e43b1dc..d4ead9c47 100644
--- a/src/video_core/gpu_synch.cpp
+++ b/src/video_core/gpu_synch.cpp
@@ -8,7 +8,7 @@
namespace VideoCommon {
GPUSynch::GPUSynch(Core::System& system, VideoCore::RendererBase& renderer)
- : GPU(system, renderer) {}
+ : GPU(system, renderer, false) {}
GPUSynch::~GPUSynch() = default;
diff --git a/src/video_core/gpu_synch.h b/src/video_core/gpu_synch.h
index 3031fcf72..07bcc47f1 100644
--- a/src/video_core/gpu_synch.h
+++ b/src/video_core/gpu_synch.h
@@ -25,6 +25,10 @@ public:
void FlushRegion(CacheAddr addr, u64 size) override;
void InvalidateRegion(CacheAddr addr, u64 size) override;
void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+
+protected:
+ void TriggerCpuInterrupt([[maybe_unused]] u32 syncpoint_id,
+ [[maybe_unused]] u32 value) const override {}
};
} // namespace VideoCommon
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index 3f0939ec9..b441e92b0 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -21,7 +21,8 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
MicroProfileOnThreadCreate("GpuThread");
// Wait for first GPU command before acquiring the window context
- state.WaitForCommands();
+ while (state.queue.Empty())
+ ;
// If emulation was stopped during disk shader loading, abort before trying to acquire context
if (!state.is_running) {
@@ -32,7 +33,6 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
CommandDataContainer next;
while (state.is_running) {
- state.WaitForCommands();
while (!state.queue.Empty()) {
state.queue.Pop(next);
if (const auto submit_list = std::get_if<SubmitListCommand>(&next.data)) {
@@ -49,8 +49,7 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
} else {
UNREACHABLE();
}
- state.signaled_fence = next.fence;
- state.TrySynchronize();
+ state.signaled_fence.store(next.fence);
}
}
}
@@ -89,12 +88,7 @@ void ThreadManager::FlushRegion(CacheAddr addr, u64 size) {
}
void ThreadManager::InvalidateRegion(CacheAddr addr, u64 size) {
- if (state.queue.Empty()) {
- // It's quicker to invalidate a single region on the CPU if the queue is already empty
- system.Renderer().Rasterizer().InvalidateRegion(addr, size);
- } else {
- PushCommand(InvalidateRegionCommand(addr, size));
- }
+ system.Renderer().Rasterizer().InvalidateRegion(addr, size);
}
void ThreadManager::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
@@ -105,22 +99,13 @@ void ThreadManager::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
u64 ThreadManager::PushCommand(CommandData&& command_data) {
const u64 fence{++state.last_fence};
state.queue.Push(CommandDataContainer(std::move(command_data), fence));
- state.SignalCommands();
return fence;
}
MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192));
void SynchState::WaitForSynchronization(u64 fence) {
- if (signaled_fence >= fence) {
- return;
- }
-
- // Wait for the GPU to be idle (all commands to be executed)
- {
- MICROPROFILE_SCOPE(GPU_wait);
- std::unique_lock lock{synchronization_mutex};
- synchronization_condition.wait(lock, [this, fence] { return signaled_fence >= fence; });
- }
+ while (signaled_fence.load() < fence)
+ ;
}
} // namespace VideoCommon::GPUThread
diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h
index 05a168a72..1d9d0c39e 100644
--- a/src/video_core/gpu_thread.h
+++ b/src/video_core/gpu_thread.h
@@ -88,41 +88,9 @@ struct CommandDataContainer {
/// Struct used to synchronize the GPU thread
struct SynchState final {
std::atomic_bool is_running{true};
- std::atomic_int queued_frame_count{};
- std::mutex synchronization_mutex;
- std::mutex commands_mutex;
- std::condition_variable commands_condition;
- std::condition_variable synchronization_condition;
-
- /// Returns true if the gap in GPU commands is small enough that we can consider the CPU and GPU
- /// synchronized. This is entirely empirical.
- bool IsSynchronized() const {
- constexpr std::size_t max_queue_gap{5};
- return queue.Size() <= max_queue_gap;
- }
-
- void TrySynchronize() {
- if (IsSynchronized()) {
- std::lock_guard lock{synchronization_mutex};
- synchronization_condition.notify_one();
- }
- }
void WaitForSynchronization(u64 fence);
- void SignalCommands() {
- if (queue.Empty()) {
- return;
- }
-
- commands_condition.notify_one();
- }
-
- void WaitForCommands() {
- std::unique_lock lock{commands_mutex};
- commands_condition.wait(lock, [this] { return !queue.Empty(); });
- }
-
using CommandQueue = Common::SPSCQueue<CommandDataContainer>;
CommandQueue queue;
u64 last_fence{};
diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h
index 9881df0d5..6b3f2d50a 100644
--- a/src/video_core/rasterizer_interface.h
+++ b/src/video_core/rasterizer_interface.h
@@ -50,6 +50,9 @@ public:
/// and invalidated
virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0;
+ /// Notify the rasterizer to send all written commands to the host GPU.
+ virtual void FlushCommands() = 0;
+
/// Notify rasterizer that a frame is about to finish
virtual void TickFrame() = 0;
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
index 2a9b523f5..f8a807c84 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
@@ -7,28 +7,41 @@
#include <glad/glad.h>
#include "common/assert.h"
+#include "common/microprofile.h"
+#include "video_core/rasterizer_interface.h"
#include "video_core/renderer_opengl/gl_buffer_cache.h"
#include "video_core/renderer_opengl/gl_rasterizer.h"
#include "video_core/renderer_opengl/gl_resource_manager.h"
namespace OpenGL {
+MICROPROFILE_DEFINE(OpenGL_Buffer_Download, "OpenGL", "Buffer Download", MP_RGB(192, 192, 128));
+
+CachedBufferBlock::CachedBufferBlock(CacheAddr cache_addr, const std::size_t size)
+ : VideoCommon::BufferBlock{cache_addr, size} {
+ gl_buffer.Create();
+ glNamedBufferData(gl_buffer.handle, static_cast<GLsizeiptr>(size), nullptr, GL_DYNAMIC_DRAW);
+}
+
+CachedBufferBlock::~CachedBufferBlock() = default;
+
OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, Core::System& system,
std::size_t stream_size)
- : VideoCommon::BufferCache<OGLBuffer, GLuint, OGLStreamBuffer>{
+ : VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuffer>{
rasterizer, system, std::make_unique<OGLStreamBuffer>(stream_size, true)} {}
OGLBufferCache::~OGLBufferCache() = default;
-OGLBuffer OGLBufferCache::CreateBuffer(std::size_t size) {
- OGLBuffer buffer;
- buffer.Create();
- glNamedBufferData(buffer.handle, static_cast<GLsizeiptr>(size), nullptr, GL_DYNAMIC_DRAW);
- return buffer;
+Buffer OGLBufferCache::CreateBlock(CacheAddr cache_addr, std::size_t size) {
+ return std::make_shared<CachedBufferBlock>(cache_addr, size);
+}
+
+void OGLBufferCache::WriteBarrier() {
+ glMemoryBarrier(GL_ALL_BARRIER_BITS);
}
-const GLuint* OGLBufferCache::ToHandle(const OGLBuffer& buffer) {
- return &buffer.handle;
+const GLuint* OGLBufferCache::ToHandle(const Buffer& buffer) {
+ return buffer->GetHandle();
}
const GLuint* OGLBufferCache::GetEmptyBuffer(std::size_t) {
@@ -36,23 +49,24 @@ const GLuint* OGLBufferCache::GetEmptyBuffer(std::size_t) {
return &null_buffer;
}
-void OGLBufferCache::UploadBufferData(const OGLBuffer& buffer, std::size_t offset, std::size_t size,
- const u8* data) {
- glNamedBufferSubData(buffer.handle, static_cast<GLintptr>(offset),
+void OGLBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
+ const u8* data) {
+ glNamedBufferSubData(*buffer->GetHandle(), static_cast<GLintptr>(offset),
static_cast<GLsizeiptr>(size), data);
}
-void OGLBufferCache::DownloadBufferData(const OGLBuffer& buffer, std::size_t offset,
- std::size_t size, u8* data) {
- glGetNamedBufferSubData(buffer.handle, static_cast<GLintptr>(offset),
+void OGLBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
+ u8* data) {
+ MICROPROFILE_SCOPE(OpenGL_Buffer_Download);
+ glGetNamedBufferSubData(*buffer->GetHandle(), static_cast<GLintptr>(offset),
static_cast<GLsizeiptr>(size), data);
}
-void OGLBufferCache::CopyBufferData(const OGLBuffer& src, const OGLBuffer& dst,
- std::size_t src_offset, std::size_t dst_offset,
- std::size_t size) {
- glCopyNamedBufferSubData(src.handle, dst.handle, static_cast<GLintptr>(src_offset),
- static_cast<GLintptr>(dst_offset), static_cast<GLsizeiptr>(size));
+void OGLBufferCache::CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
+ std::size_t dst_offset, std::size_t size) {
+ glCopyNamedBufferSubData(*src->GetHandle(), *dst->GetHandle(),
+ static_cast<GLintptr>(src_offset), static_cast<GLintptr>(dst_offset),
+ static_cast<GLsizeiptr>(size));
}
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h
index 8c8ac4038..022e7bfa9 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.h
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.h
@@ -7,7 +7,7 @@
#include <memory>
#include "common/common_types.h"
-#include "video_core/buffer_cache.h"
+#include "video_core/buffer_cache/buffer_cache.h"
#include "video_core/rasterizer_cache.h"
#include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/renderer_opengl/gl_stream_buffer.h"
@@ -21,7 +21,24 @@ namespace OpenGL {
class OGLStreamBuffer;
class RasterizerOpenGL;
-class OGLBufferCache final : public VideoCommon::BufferCache<OGLBuffer, GLuint, OGLStreamBuffer> {
+class CachedBufferBlock;
+
+using Buffer = std::shared_ptr<CachedBufferBlock>;
+
+class CachedBufferBlock : public VideoCommon::BufferBlock {
+public:
+ explicit CachedBufferBlock(CacheAddr cache_addr, const std::size_t size);
+ ~CachedBufferBlock();
+
+ const GLuint* GetHandle() const {
+ return &gl_buffer.handle;
+ }
+
+private:
+ OGLBuffer gl_buffer{};
+};
+
+class OGLBufferCache final : public VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuffer> {
public:
explicit OGLBufferCache(RasterizerOpenGL& rasterizer, Core::System& system,
std::size_t stream_size);
@@ -30,18 +47,20 @@ public:
const GLuint* GetEmptyBuffer(std::size_t) override;
protected:
- OGLBuffer CreateBuffer(std::size_t size) override;
+ Buffer CreateBlock(CacheAddr cache_addr, std::size_t size) override;
+
+ void WriteBarrier() override;
- const GLuint* ToHandle(const OGLBuffer& buffer) override;
+ const GLuint* ToHandle(const Buffer& buffer) override;
- void UploadBufferData(const OGLBuffer& buffer, std::size_t offset, std::size_t size,
- const u8* data) override;
+ void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
+ const u8* data) override;
- void DownloadBufferData(const OGLBuffer& buffer, std::size_t offset, std::size_t size,
- u8* data) override;
+ void DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
+ u8* data) override;
- void CopyBufferData(const OGLBuffer& src, const OGLBuffer& dst, std::size_t src_offset,
- std::size_t dst_offset, std::size_t size) override;
+ void CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
+ std::size_t dst_offset, std::size_t size) override;
};
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index 85424a4c9..03d434b28 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -27,6 +27,8 @@ Device::Device() {
shader_storage_alignment = GetInteger<std::size_t>(GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT);
max_vertex_attributes = GetInteger<u32>(GL_MAX_VERTEX_ATTRIBS);
max_varyings = GetInteger<u32>(GL_MAX_VARYING_VECTORS);
+ has_warp_intrinsics = GLAD_GL_NV_gpu_shader5 && GLAD_GL_NV_shader_thread_group &&
+ GLAD_GL_NV_shader_thread_shuffle;
has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
has_variable_aoffi = TestVariableAoffi();
has_component_indexing_bug = TestComponentIndexingBug();
@@ -36,6 +38,7 @@ Device::Device(std::nullptr_t) {
uniform_buffer_alignment = 0;
max_vertex_attributes = 16;
max_varyings = 15;
+ has_warp_intrinsics = true;
has_vertex_viewport_layer = true;
has_variable_aoffi = true;
has_component_indexing_bug = false;
diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h
index dc883722d..3ef7c6dd8 100644
--- a/src/video_core/renderer_opengl/gl_device.h
+++ b/src/video_core/renderer_opengl/gl_device.h
@@ -30,6 +30,10 @@ public:
return max_varyings;
}
+ bool HasWarpIntrinsics() const {
+ return has_warp_intrinsics;
+ }
+
bool HasVertexViewportLayer() const {
return has_vertex_viewport_layer;
}
@@ -50,6 +54,7 @@ private:
std::size_t shader_storage_alignment{};
u32 max_vertex_attributes{};
u32 max_varyings{};
+ bool has_warp_intrinsics{};
bool has_vertex_viewport_layer{};
bool has_variable_aoffi{};
bool has_component_indexing_bug{};
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index c28ae795c..bb09ecd52 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -708,8 +708,6 @@ void RasterizerOpenGL::DrawArrays() {
return;
}
- const auto& regs = gpu.regs;
-
SyncColorMask();
SyncFragmentColorClampState();
SyncMultiSampleState();
@@ -863,6 +861,10 @@ void RasterizerOpenGL::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
InvalidateRegion(addr, size);
}
+void RasterizerOpenGL::FlushCommands() {
+ glFlush();
+}
+
void RasterizerOpenGL::TickFrame() {
buffer_cache.TickFrame();
}
@@ -976,7 +978,7 @@ void RasterizerOpenGL::SetupGlobalMemory(const GLShader::GlobalMemoryEntry& entr
GPUVAddr gpu_addr, std::size_t size) {
const auto alignment{device.GetShaderStorageBufferAlignment()};
const auto [ssbo, buffer_offset] =
- buffer_cache.UploadMemory(gpu_addr, size, alignment, true, entry.IsWritten());
+ buffer_cache.UploadMemory(gpu_addr, size, alignment, entry.IsWritten());
bind_ssbo_pushbuffer.Push(ssbo, buffer_offset, static_cast<GLsizeiptr>(size));
}
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 8b123c48d..9d20a4fbf 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -63,6 +63,7 @@ public:
void FlushRegion(CacheAddr addr, u64 size) override;
void InvalidateRegion(CacheAddr addr, u64 size) override;
void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+ void FlushCommands() override;
void TickFrame() override;
bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
const Tegra::Engines::Fermi2D::Regs::Surface& dst,
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 1c90facc3..cf6a5cddf 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -212,7 +212,9 @@ CachedProgram SpecializeShader(const std::string& code, const GLShader::ShaderEn
const auto texture_buffer_usage{variant.texture_buffer_usage};
std::string source = "#version 430 core\n"
- "#extension GL_ARB_separate_shader_objects : enable\n";
+ "#extension GL_ARB_separate_shader_objects : enable\n"
+ "#extension GL_NV_gpu_shader5 : enable\n"
+ "#extension GL_NV_shader_thread_group : enable\n";
if (entries.shader_viewport_layer_array) {
source += "#extension GL_ARB_shader_viewport_layer_array : enable\n";
}
@@ -247,20 +249,24 @@ CachedProgram SpecializeShader(const std::string& code, const GLShader::ShaderEn
if (!texture_buffer_usage.test(i)) {
continue;
}
- source += fmt::format("#define SAMPLER_{}_IS_BUFFER", i);
+ source += fmt::format("#define SAMPLER_{}_IS_BUFFER\n", i);
+ }
+ if (texture_buffer_usage.any()) {
+ source += '\n';
}
if (program_type == ProgramType::Geometry) {
const auto [glsl_topology, debug_name, max_vertices] =
GetPrimitiveDescription(primitive_mode);
- source += "layout (" + std::string(glsl_topology) + ") in;\n";
+ source += "layout (" + std::string(glsl_topology) + ") in;\n\n";
source += "#define MAX_VERTEX_INPUT " + std::to_string(max_vertices) + '\n';
}
if (program_type == ProgramType::Compute) {
source += "layout (local_size_variable) in;\n";
}
+ source += '\n';
source += code;
OGLShader shader;
@@ -289,7 +295,7 @@ std::set<GLenum> GetSupportedFormats() {
CachedShader::CachedShader(const ShaderParameters& params, ProgramType program_type,
GLShader::ProgramResult result)
- : RasterizerCacheObject{params.host_ptr}, host_ptr{params.host_ptr}, cpu_addr{params.cpu_addr},
+ : RasterizerCacheObject{params.host_ptr}, cpu_addr{params.cpu_addr},
unique_identifier{params.unique_identifier}, program_type{program_type},
disk_cache{params.disk_cache}, precompiled_programs{params.precompiled_programs},
entries{result.second}, code{std::move(result.first)}, shader_length{entries.shader_length} {}
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index a3106a0ff..2c8faf855 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -106,7 +106,6 @@ private:
ShaderDiskCacheUsage GetUsage(const ProgramVariant& variant) const;
- u8* host_ptr{};
VAddr cpu_addr{};
u64 unique_identifier{};
ProgramType program_type{};
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index ffe26b241..359d58cbe 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -565,7 +565,7 @@ private:
case Tegra::Shader::ImageType::Texture1D:
return "image1D";
case Tegra::Shader::ImageType::TextureBuffer:
- return "bufferImage";
+ return "imageBuffer";
case Tegra::Shader::ImageType::Texture1DArray:
return "image1DArray";
case Tegra::Shader::ImageType::Texture2D:
@@ -1136,6 +1136,16 @@ private:
Type::Float);
}
+ std::string FCastHalf0(Operation operation) {
+ const std::string op_a = VisitOperand(operation, 0, Type::HalfFloat);
+ return fmt::format("({})[0]", op_a);
+ }
+
+ std::string FCastHalf1(Operation operation) {
+ const std::string op_a = VisitOperand(operation, 0, Type::HalfFloat);
+ return fmt::format("({})[1]", op_a);
+ }
+
template <Type type>
std::string Min(Operation operation) {
return GenerateBinaryCall(operation, "min", type, type, type);
@@ -1292,6 +1302,11 @@ private:
return ApplyPrecise(operation, BitwiseCastResult(clamped, Type::HalfFloat));
}
+ std::string HCastFloat(Operation operation) {
+ const std::string op_a = VisitOperand(operation, 0, Type::Float);
+ return fmt::format("fromHalf2(vec2({}, 0.0f))", op_a);
+ }
+
std::string HUnpack(Operation operation) {
const std::string operand{VisitOperand(operation, 0, Type::HalfFloat)};
const auto value = [&]() -> std::string {
@@ -1720,6 +1735,48 @@ private:
return "utof(gl_WorkGroupID"s + GetSwizzle(element) + ')';
}
+ std::string BallotThread(Operation operation) {
+ const std::string value = VisitOperand(operation, 0, Type::Bool);
+ if (!device.HasWarpIntrinsics()) {
+ LOG_ERROR(Render_OpenGL,
+ "Nvidia warp intrinsics are not available and its required by a shader");
+ // Stub on non-Nvidia devices by simulating all threads voting the same as the active
+ // one.
+ return fmt::format("utof({} ? 0xFFFFFFFFU : 0U)", value);
+ }
+ return fmt::format("utof(ballotThreadNV({}))", value);
+ }
+
+ std::string Vote(Operation operation, const char* func) {
+ const std::string value = VisitOperand(operation, 0, Type::Bool);
+ if (!device.HasWarpIntrinsics()) {
+ LOG_ERROR(Render_OpenGL,
+ "Nvidia vote intrinsics are not available and its required by a shader");
+ // Stub with a warp size of one.
+ return value;
+ }
+ return fmt::format("{}({})", func, value);
+ }
+
+ std::string VoteAll(Operation operation) {
+ return Vote(operation, "allThreadsNV");
+ }
+
+ std::string VoteAny(Operation operation) {
+ return Vote(operation, "anyThreadNV");
+ }
+
+ std::string VoteEqual(Operation operation) {
+ if (!device.HasWarpIntrinsics()) {
+ LOG_ERROR(Render_OpenGL,
+ "Nvidia vote intrinsics are not available and its required by a shader");
+ // We must return true here since a stub for a theoretical warp size of 1 will always
+ // return an equal result for all its votes.
+ return "true";
+ }
+ return Vote(operation, "allThreadsEqualNV");
+ }
+
static constexpr std::array operation_decompilers = {
&GLSLDecompiler::Assign,
@@ -1732,6 +1789,8 @@ private:
&GLSLDecompiler::Negate<Type::Float>,
&GLSLDecompiler::Absolute<Type::Float>,
&GLSLDecompiler::FClamp,
+ &GLSLDecompiler::FCastHalf0,
+ &GLSLDecompiler::FCastHalf1,
&GLSLDecompiler::Min<Type::Float>,
&GLSLDecompiler::Max<Type::Float>,
&GLSLDecompiler::FCos,
@@ -1792,6 +1851,7 @@ private:
&GLSLDecompiler::Absolute<Type::HalfFloat>,
&GLSLDecompiler::HNegate,
&GLSLDecompiler::HClamp,
+ &GLSLDecompiler::HCastFloat,
&GLSLDecompiler::HUnpack,
&GLSLDecompiler::HMergeF32,
&GLSLDecompiler::HMergeH0,
@@ -1867,6 +1927,11 @@ private:
&GLSLDecompiler::WorkGroupId<0>,
&GLSLDecompiler::WorkGroupId<1>,
&GLSLDecompiler::WorkGroupId<2>,
+
+ &GLSLDecompiler::BallotThread,
+ &GLSLDecompiler::VoteAll,
+ &GLSLDecompiler::VoteAny,
+ &GLSLDecompiler::VoteEqual,
};
static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp
index 8fcd39a69..4f135fe03 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp
@@ -137,7 +137,6 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format
const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) {
ASSERT(static_cast<std::size_t>(pixel_format) < tex_format_tuples.size());
const auto& format{tex_format_tuples[static_cast<std::size_t>(pixel_format)]};
- ASSERT(component_type == format.component_type);
return format;
}
@@ -185,6 +184,9 @@ GLint GetSwizzleSource(SwizzleSource source) {
}
void ApplyTextureDefaults(const SurfaceParams& params, GLuint texture) {
+ if (params.IsBuffer()) {
+ return;
+ }
glTextureParameteri(texture, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTextureParameteri(texture, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTextureParameteri(texture, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
@@ -209,6 +211,7 @@ OGLTexture CreateTexture(const SurfaceParams& params, GLenum target, GLenum inte
glNamedBufferStorage(texture_buffer.handle, params.width * params.GetBytesPerPixel(),
nullptr, GL_DYNAMIC_STORAGE_BIT);
glTextureBuffer(texture.handle, internal_format, texture_buffer.handle);
+ break;
case SurfaceTarget::Texture2D:
case SurfaceTarget::TextureCubemap:
glTextureStorage2D(texture.handle, params.emulated_levels, internal_format, params.width,
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h
index ff6ab6988..21324488a 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.h
+++ b/src/video_core/renderer_opengl/gl_texture_cache.h
@@ -51,7 +51,7 @@ public:
}
protected:
- void DecorateSurfaceName();
+ void DecorateSurfaceName() override;
View CreateView(const ViewParams& view_key) override;
View CreateViewInner(const ViewParams& view_key, bool is_proxy);
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index d267712c9..a35b45c9c 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -735,6 +735,16 @@ private:
return {};
}
+ Id FCastHalf0(Operation operation) {
+ UNIMPLEMENTED();
+ return {};
+ }
+
+ Id FCastHalf1(Operation operation) {
+ UNIMPLEMENTED();
+ return {};
+ }
+
Id HNegate(Operation operation) {
UNIMPLEMENTED();
return {};
@@ -745,6 +755,11 @@ private:
return {};
}
+ Id HCastFloat(Operation operation) {
+ UNIMPLEMENTED();
+ return {};
+ }
+
Id HUnpack(Operation operation) {
UNIMPLEMENTED();
return {};
@@ -1057,6 +1072,26 @@ private:
return {};
}
+ Id BallotThread(Operation) {
+ UNIMPLEMENTED();
+ return {};
+ }
+
+ Id VoteAll(Operation) {
+ UNIMPLEMENTED();
+ return {};
+ }
+
+ Id VoteAny(Operation) {
+ UNIMPLEMENTED();
+ return {};
+ }
+
+ Id VoteEqual(Operation) {
+ UNIMPLEMENTED();
+ return {};
+ }
+
Id DeclareBuiltIn(spv::BuiltIn builtin, spv::StorageClass storage, Id type,
const std::string& name) {
const Id id = OpVariable(type, storage);
@@ -1210,6 +1245,8 @@ private:
&SPIRVDecompiler::Unary<&Module::OpFNegate, Type::Float>,
&SPIRVDecompiler::Unary<&Module::OpFAbs, Type::Float>,
&SPIRVDecompiler::Ternary<&Module::OpFClamp, Type::Float>,
+ &SPIRVDecompiler::FCastHalf0,
+ &SPIRVDecompiler::FCastHalf1,
&SPIRVDecompiler::Binary<&Module::OpFMin, Type::Float>,
&SPIRVDecompiler::Binary<&Module::OpFMax, Type::Float>,
&SPIRVDecompiler::Unary<&Module::OpCos, Type::Float>,
@@ -1270,6 +1307,7 @@ private:
&SPIRVDecompiler::Unary<&Module::OpFAbs, Type::HalfFloat>,
&SPIRVDecompiler::HNegate,
&SPIRVDecompiler::HClamp,
+ &SPIRVDecompiler::HCastFloat,
&SPIRVDecompiler::HUnpack,
&SPIRVDecompiler::HMergeF32,
&SPIRVDecompiler::HMergeH0,
@@ -1346,6 +1384,11 @@ private:
&SPIRVDecompiler::WorkGroupId<0>,
&SPIRVDecompiler::WorkGroupId<1>,
&SPIRVDecompiler::WorkGroupId<2>,
+
+ &SPIRVDecompiler::BallotThread,
+ &SPIRVDecompiler::VoteAll,
+ &SPIRVDecompiler::VoteAny,
+ &SPIRVDecompiler::VoteEqual,
};
static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
diff --git a/src/video_core/shader/control_flow.cpp b/src/video_core/shader/control_flow.cpp
index fdcc970ff..ec3a76690 100644
--- a/src/video_core/shader/control_flow.cpp
+++ b/src/video_core/shader/control_flow.cpp
@@ -15,7 +15,7 @@
#include "video_core/shader/shader_ir.h"
namespace VideoCommon::Shader {
-
+namespace {
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
@@ -29,8 +29,7 @@ struct Query {
struct BlockStack {
BlockStack() = default;
- BlockStack(const BlockStack& b) = default;
- BlockStack(const Query& q) : ssy_stack{q.ssy_stack}, pbk_stack{q.pbk_stack} {}
+ explicit BlockStack(const Query& q) : ssy_stack{q.ssy_stack}, pbk_stack{q.pbk_stack} {}
std::stack<u32> ssy_stack{};
std::stack<u32> pbk_stack{};
};
@@ -58,7 +57,7 @@ struct BlockInfo {
struct CFGRebuildState {
explicit CFGRebuildState(const ProgramCode& program_code, const std::size_t program_size,
const u32 start)
- : program_code{program_code}, program_size{program_size}, start{start} {}
+ : start{start}, program_code{program_code}, program_size{program_size} {}
u32 start{};
std::vector<BlockInfo> block_info{};
@@ -85,7 +84,7 @@ std::pair<BlockCollision, u32> TryGetBlock(CFGRebuildState& state, u32 address)
return {BlockCollision::Inside, index};
}
}
- return {BlockCollision::None, -1};
+ return {BlockCollision::None, 0xFFFFFFFF};
}
struct ParseInfo {
@@ -365,27 +364,29 @@ bool TryQuery(CFGRebuildState& state) {
const auto gather_end = labels.upper_bound(block.end);
while (gather_start != gather_end) {
cc.push(gather_start->second);
- gather_start++;
+ ++gather_start;
}
};
if (state.queries.empty()) {
return false;
}
+
Query& q = state.queries.front();
const u32 block_index = state.registered[q.address];
BlockInfo& block = state.block_info[block_index];
- // If the block is visted, check if the stacks match, else gather the ssy/pbk
+ // If the block is visited, check if the stacks match, else gather the ssy/pbk
// labels into the current stack and look if the branch at the end of the block
// consumes a label. Schedule new queries accordingly
if (block.visited) {
BlockStack& stack = state.stacks[q.address];
- const bool all_okay = (stack.ssy_stack.size() == 0 || q.ssy_stack == stack.ssy_stack) &&
- (stack.pbk_stack.size() == 0 || q.pbk_stack == stack.pbk_stack);
+ const bool all_okay = (stack.ssy_stack.empty() || q.ssy_stack == stack.ssy_stack) &&
+ (stack.pbk_stack.empty() || q.pbk_stack == stack.pbk_stack);
state.queries.pop_front();
return all_okay;
}
block.visited = true;
- state.stacks[q.address] = BlockStack{q};
+ state.stacks.insert_or_assign(q.address, BlockStack{q});
+
Query q2(q);
state.queries.pop_front();
gather_labels(q2.ssy_stack, state.ssy_labels, block);
@@ -394,6 +395,7 @@ bool TryQuery(CFGRebuildState& state) {
q2.address = block.end + 1;
state.queries.push_back(q2);
}
+
Query conditional_query{q2};
if (block.branch.is_sync) {
if (block.branch.address == unassigned_branch) {
@@ -408,13 +410,15 @@ bool TryQuery(CFGRebuildState& state) {
conditional_query.pbk_stack.pop();
}
conditional_query.address = block.branch.address;
- state.queries.push_back(conditional_query);
+ state.queries.push_back(std::move(conditional_query));
return true;
}
+} // Anonymous namespace
-std::optional<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u32 program_size,
- u32 start_address) {
+std::optional<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code,
+ std::size_t program_size, u32 start_address) {
CFGRebuildState state{program_code, program_size, start_address};
+
// Inspect Code and generate blocks
state.labels.clear();
state.labels.emplace(start_address);
@@ -424,10 +428,9 @@ std::optional<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u
return {};
}
}
+
// Decompile Stacks
- Query start_query{};
- start_query.address = state.start;
- state.queries.push_back(start_query);
+ state.queries.push_back(Query{state.start, {}, {}});
bool decompiled = true;
while (!state.queries.empty()) {
if (!TryQuery(state)) {
@@ -435,14 +438,15 @@ std::optional<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u
break;
}
}
+
// Sort and organize results
std::sort(state.block_info.begin(), state.block_info.end(),
- [](const BlockInfo& a, const BlockInfo& b) -> bool { return a.start < b.start; });
+ [](const BlockInfo& a, const BlockInfo& b) { return a.start < b.start; });
ShaderCharacteristics result_out{};
result_out.decompilable = decompiled;
result_out.start = start_address;
result_out.end = start_address;
- for (auto& block : state.block_info) {
+ for (const auto& block : state.block_info) {
ShaderBlock new_block{};
new_block.start = block.start;
new_block.end = block.end;
@@ -457,8 +461,9 @@ std::optional<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u
}
if (result_out.decompilable) {
result_out.labels = std::move(state.labels);
- return {result_out};
+ return {std::move(result_out)};
}
+
// If it's not decompilable, merge the unlabelled blocks together
auto back = result_out.blocks.begin();
auto next = std::next(back);
@@ -469,8 +474,8 @@ std::optional<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u
continue;
}
back = next;
- next++;
+ ++next;
}
- return {result_out};
+ return {std::move(result_out)};
}
} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/control_flow.h b/src/video_core/shader/control_flow.h
index 5e8ea3271..b0a5e4f8c 100644
--- a/src/video_core/shader/control_flow.h
+++ b/src/video_core/shader/control_flow.h
@@ -4,7 +4,6 @@
#pragma once
-#include <cstring>
#include <list>
#include <optional>
#include <unordered_set>
@@ -26,27 +25,44 @@ struct Condition {
bool IsUnconditional() const {
return predicate == Pred::UnusedIndex && cc == ConditionCode::T;
}
+
bool operator==(const Condition& other) const {
return std::tie(predicate, cc) == std::tie(other.predicate, other.cc);
}
+
+ bool operator!=(const Condition& other) const {
+ return !operator==(other);
+ }
};
struct ShaderBlock {
- u32 start{};
- u32 end{};
- bool ignore_branch{};
struct Branch {
Condition cond{};
bool kills{};
s32 address{};
+
bool operator==(const Branch& b) const {
return std::tie(cond, kills, address) == std::tie(b.cond, b.kills, b.address);
}
- } branch{};
+
+ bool operator!=(const Branch& b) const {
+ return !operator==(b);
+ }
+ };
+
+ u32 start{};
+ u32 end{};
+ bool ignore_branch{};
+ Branch branch{};
+
bool operator==(const ShaderBlock& sb) const {
return std::tie(start, end, ignore_branch, branch) ==
std::tie(sb.start, sb.end, sb.ignore_branch, sb.branch);
}
+
+ bool operator!=(const ShaderBlock& sb) const {
+ return !operator==(sb);
+ }
};
struct ShaderCharacteristics {
@@ -57,7 +73,7 @@ struct ShaderCharacteristics {
std::unordered_set<u32> labels{};
};
-std::optional<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u32 program_size,
- u32 start_address);
+std::optional<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code,
+ std::size_t program_size, u32 start_address);
} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp
index afffd157f..47a9fd961 100644
--- a/src/video_core/shader/decode.cpp
+++ b/src/video_core/shader/decode.cpp
@@ -47,14 +47,14 @@ void ShaderIR::Decode() {
if (shader_info.decompilable) {
disable_flow_stack = true;
const auto insert_block = [this](NodeBlock& nodes, u32 label) {
- if (label == exit_branch) {
+ if (label == static_cast<u32>(exit_branch)) {
return;
}
basic_blocks.insert({label, nodes});
};
const auto& blocks = shader_info.blocks;
NodeBlock current_block;
- u32 current_label = exit_branch;
+ u32 current_label = static_cast<u32>(exit_branch);
for (auto& block : blocks) {
if (shader_info.labels.count(block.start) != 0) {
insert_block(current_block, current_label);
@@ -176,6 +176,7 @@ u32 ShaderIR::DecodeInstr(NodeBlock& bb, u32 pc) {
{OpCode::Type::Ffma, &ShaderIR::DecodeFfma},
{OpCode::Type::Hfma2, &ShaderIR::DecodeHfma2},
{OpCode::Type::Conversion, &ShaderIR::DecodeConversion},
+ {OpCode::Type::Warp, &ShaderIR::DecodeWarp},
{OpCode::Type::Memory, &ShaderIR::DecodeMemory},
{OpCode::Type::Texture, &ShaderIR::DecodeTexture},
{OpCode::Type::Image, &ShaderIR::DecodeImage},
diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp
index 87d8fecaa..1473c282a 100644
--- a/src/video_core/shader/decode/arithmetic.cpp
+++ b/src/video_core/shader/decode/arithmetic.cpp
@@ -42,11 +42,14 @@ u32 ShaderIR::DecodeArithmetic(NodeBlock& bb, u32 pc) {
case OpCode::Id::FMUL_R:
case OpCode::Id::FMUL_IMM: {
// FMUL does not have 'abs' bits and only the second operand has a 'neg' bit.
- UNIMPLEMENTED_IF_MSG(instr.fmul.tab5cb8_2 != 0, "FMUL tab5cb8_2({}) is not implemented",
- instr.fmul.tab5cb8_2.Value());
- UNIMPLEMENTED_IF_MSG(
- instr.fmul.tab5c68_0 != 1, "FMUL tab5cb8_0({}) is not implemented",
- instr.fmul.tab5c68_0.Value()); // SMO typical sends 1 here which seems to be the default
+ if (instr.fmul.tab5cb8_2 != 0) {
+ LOG_WARNING(HW_GPU, "FMUL tab5cb8_2({}) is not implemented",
+ instr.fmul.tab5cb8_2.Value());
+ }
+ if (instr.fmul.tab5c68_0 != 1) {
+ LOG_WARNING(HW_GPU, "FMUL tab5cb8_0({}) is not implemented",
+ instr.fmul.tab5c68_0.Value());
+ }
op_b = GetOperandAbsNegFloat(op_b, false, instr.fmul.negate_b);
diff --git a/src/video_core/shader/decode/arithmetic_half_immediate.cpp b/src/video_core/shader/decode/arithmetic_half_immediate.cpp
index 7bcf38f23..6466fc011 100644
--- a/src/video_core/shader/decode/arithmetic_half_immediate.cpp
+++ b/src/video_core/shader/decode/arithmetic_half_immediate.cpp
@@ -23,7 +23,9 @@ u32 ShaderIR::DecodeArithmeticHalfImmediate(NodeBlock& bb, u32 pc) {
LOG_WARNING(HW_GPU, "{} FTZ not implemented", opcode->get().GetName());
}
} else {
- UNIMPLEMENTED_IF(instr.alu_half_imm.precision != Tegra::Shader::HalfPrecision::None);
+ if (instr.alu_half_imm.precision != Tegra::Shader::HalfPrecision::None) {
+ LOG_WARNING(HW_GPU, "{} FTZ not implemented", opcode->get().GetName());
+ }
}
Node op_a = UnpackHalfFloat(GetRegister(instr.gpr8), instr.alu_half_imm.type_a);
diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp
index 4221f0c58..8973fbefa 100644
--- a/src/video_core/shader/decode/conversion.cpp
+++ b/src/video_core/shader/decode/conversion.cpp
@@ -57,7 +57,7 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
case OpCode::Id::I2F_R:
case OpCode::Id::I2F_C:
case OpCode::Id::I2F_IMM: {
- UNIMPLEMENTED_IF(instr.conversion.dst_size != Register::Size::Word);
+ UNIMPLEMENTED_IF(instr.conversion.dst_size == Register::Size::Long);
UNIMPLEMENTED_IF(instr.conversion.selector);
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
"Condition codes generation in I2F is not implemented");
@@ -82,14 +82,19 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
value = GetOperandAbsNegFloat(value, false, instr.conversion.negate_a);
SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+
+ if (instr.conversion.dst_size == Register::Size::Short) {
+ value = Operation(OperationCode::HCastFloat, PRECISE, value);
+ }
+
SetRegister(bb, instr.gpr0, value);
break;
}
case OpCode::Id::F2F_R:
case OpCode::Id::F2F_C:
case OpCode::Id::F2F_IMM: {
- UNIMPLEMENTED_IF(instr.conversion.f2f.dst_size != Register::Size::Word);
- UNIMPLEMENTED_IF(instr.conversion.f2f.src_size != Register::Size::Word);
+ UNIMPLEMENTED_IF(instr.conversion.dst_size == Register::Size::Long);
+ UNIMPLEMENTED_IF(instr.conversion.src_size == Register::Size::Long);
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
"Condition codes generation in F2F is not implemented");
@@ -107,6 +112,11 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
}
}();
+ if (instr.conversion.src_size == Register::Size::Short) {
+ // TODO: figure where extract is sey in the encoding
+ value = Operation(OperationCode::FCastHalf0, PRECISE, value);
+ }
+
value = GetOperandAbsNegFloat(value, instr.conversion.abs_a, instr.conversion.negate_a);
value = [&]() {
@@ -124,19 +134,24 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
default:
UNIMPLEMENTED_MSG("Unimplemented F2F rounding mode {}",
static_cast<u32>(instr.conversion.f2f.rounding.Value()));
- return Immediate(0);
+ return value;
}
}();
value = GetSaturatedFloat(value, instr.alu.saturate_d);
SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+
+ if (instr.conversion.dst_size == Register::Size::Short) {
+ value = Operation(OperationCode::HCastFloat, PRECISE, value);
+ }
+
SetRegister(bb, instr.gpr0, value);
break;
}
case OpCode::Id::F2I_R:
case OpCode::Id::F2I_C:
case OpCode::Id::F2I_IMM: {
- UNIMPLEMENTED_IF(instr.conversion.src_size != Register::Size::Word);
+ UNIMPLEMENTED_IF(instr.conversion.src_size == Register::Size::Long);
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
"Condition codes generation in F2I is not implemented");
Node value = [&]() {
@@ -153,6 +168,11 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
}
}();
+ if (instr.conversion.src_size == Register::Size::Short) {
+ // TODO: figure where extract is sey in the encoding
+ value = Operation(OperationCode::FCastHalf0, PRECISE, value);
+ }
+
value = GetOperandAbsNegFloat(value, instr.conversion.abs_a, instr.conversion.negate_a);
value = [&]() {
diff --git a/src/video_core/shader/decode/ffma.cpp b/src/video_core/shader/decode/ffma.cpp
index 29be25ca3..ca2f39e8d 100644
--- a/src/video_core/shader/decode/ffma.cpp
+++ b/src/video_core/shader/decode/ffma.cpp
@@ -18,10 +18,12 @@ u32 ShaderIR::DecodeFfma(NodeBlock& bb, u32 pc) {
const auto opcode = OpCode::Decode(instr);
UNIMPLEMENTED_IF_MSG(instr.ffma.cc != 0, "FFMA cc not implemented");
- UNIMPLEMENTED_IF_MSG(instr.ffma.tab5980_0 != 1, "FFMA tab5980_0({}) not implemented",
- instr.ffma.tab5980_0.Value()); // Seems to be 1 by default based on SMO
- UNIMPLEMENTED_IF_MSG(instr.ffma.tab5980_1 != 0, "FFMA tab5980_1({}) not implemented",
- instr.ffma.tab5980_1.Value());
+ if (instr.ffma.tab5980_0 != 1) {
+ LOG_WARNING(HW_GPU, "FFMA tab5980_0({}) not implemented", instr.ffma.tab5980_0.Value());
+ }
+ if (instr.ffma.tab5980_1 != 0) {
+ LOG_WARNING(HW_GPU, "FFMA tab5980_1({}) not implemented", instr.ffma.tab5980_1.Value());
+ }
const Node op_a = GetRegister(instr.gpr8);
diff --git a/src/video_core/shader/decode/float_set.cpp b/src/video_core/shader/decode/float_set.cpp
index f5013e44a..5614e8a0d 100644
--- a/src/video_core/shader/decode/float_set.cpp
+++ b/src/video_core/shader/decode/float_set.cpp
@@ -15,7 +15,6 @@ using Tegra::Shader::OpCode;
u32 ShaderIR::DecodeFloatSet(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
- const auto opcode = OpCode::Decode(instr);
const Node op_a = GetOperandAbsNegFloat(GetRegister(instr.gpr8), instr.fset.abs_a != 0,
instr.fset.neg_a != 0);
diff --git a/src/video_core/shader/decode/float_set_predicate.cpp b/src/video_core/shader/decode/float_set_predicate.cpp
index 2323052b0..34854fcca 100644
--- a/src/video_core/shader/decode/float_set_predicate.cpp
+++ b/src/video_core/shader/decode/float_set_predicate.cpp
@@ -16,7 +16,6 @@ using Tegra::Shader::Pred;
u32 ShaderIR::DecodeFloatSetPredicate(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
- const auto opcode = OpCode::Decode(instr);
const Node op_a = GetOperandAbsNegFloat(GetRegister(instr.gpr8), instr.fsetp.abs_a != 0,
instr.fsetp.neg_a != 0);
diff --git a/src/video_core/shader/decode/half_set_predicate.cpp b/src/video_core/shader/decode/half_set_predicate.cpp
index ad180d6df..afea33e5f 100644
--- a/src/video_core/shader/decode/half_set_predicate.cpp
+++ b/src/video_core/shader/decode/half_set_predicate.cpp
@@ -18,7 +18,7 @@ u32 ShaderIR::DecodeHalfSetPredicate(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
const auto opcode = OpCode::Decode(instr);
- UNIMPLEMENTED_IF(instr.hsetp2.ftz != 0);
+ DEBUG_ASSERT(instr.hsetp2.ftz == 0);
Node op_a = UnpackHalfFloat(GetRegister(instr.gpr8), instr.hsetp2.type_a);
op_a = GetOperandAbsNegHalf(op_a, instr.hsetp2.abs_a, instr.hsetp2.negate_a);
@@ -30,7 +30,7 @@ u32 ShaderIR::DecodeHalfSetPredicate(NodeBlock& bb, u32 pc) {
case OpCode::Id::HSETP2_C:
cond = instr.hsetp2.cbuf_and_imm.cond;
h_and = instr.hsetp2.cbuf_and_imm.h_and;
- op_b = GetOperandAbsNegHalf(GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset),
+ op_b = GetOperandAbsNegHalf(GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset()),
instr.hsetp2.cbuf.abs_b, instr.hsetp2.cbuf.negate_b);
break;
case OpCode::Id::HSETP2_IMM:
@@ -52,15 +52,15 @@ u32 ShaderIR::DecodeHalfSetPredicate(NodeBlock& bb, u32 pc) {
}
const OperationCode combiner = GetPredicateCombiner(instr.hsetp2.op);
- const Node pred39 = GetPredicate(instr.hsetp2.pred39, instr.hsetp2.neg_pred);
+ const Node combined_pred = GetPredicate(instr.hsetp2.pred3, instr.hsetp2.neg_pred);
const auto Write = [&](u64 dest, Node src) {
- SetPredicate(bb, dest, Operation(combiner, std::move(src), pred39));
+ SetPredicate(bb, dest, Operation(combiner, std::move(src), combined_pred));
};
const Node comparison = GetPredicateComparisonHalf(cond, op_a, op_b);
const u64 first = instr.hsetp2.pred0;
- const u64 second = instr.hsetp2.pred3;
+ const u64 second = instr.hsetp2.pred39;
if (h_and) {
const Node joined = Operation(OperationCode::LogicalAnd2, comparison);
Write(first, joined);
diff --git a/src/video_core/shader/decode/hfma2.cpp b/src/video_core/shader/decode/hfma2.cpp
index c3bcf1ae9..5b44cb79c 100644
--- a/src/video_core/shader/decode/hfma2.cpp
+++ b/src/video_core/shader/decode/hfma2.cpp
@@ -22,9 +22,9 @@ u32 ShaderIR::DecodeHfma2(NodeBlock& bb, u32 pc) {
const auto opcode = OpCode::Decode(instr);
if (opcode->get().GetId() == OpCode::Id::HFMA2_RR) {
- UNIMPLEMENTED_IF(instr.hfma2.rr.precision != HalfPrecision::None);
+ DEBUG_ASSERT(instr.hfma2.rr.precision == HalfPrecision::None);
} else {
- UNIMPLEMENTED_IF(instr.hfma2.precision != HalfPrecision::None);
+ DEBUG_ASSERT(instr.hfma2.precision == HalfPrecision::None);
}
constexpr auto identity = HalfType::H0_H1;
diff --git a/src/video_core/shader/decode/integer_set.cpp b/src/video_core/shader/decode/integer_set.cpp
index 46e3d5905..59809bcd8 100644
--- a/src/video_core/shader/decode/integer_set.cpp
+++ b/src/video_core/shader/decode/integer_set.cpp
@@ -14,7 +14,6 @@ using Tegra::Shader::OpCode;
u32 ShaderIR::DecodeIntegerSet(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
- const auto opcode = OpCode::Decode(instr);
const Node op_a = GetRegister(instr.gpr8);
const Node op_b = [&]() {
diff --git a/src/video_core/shader/decode/integer_set_predicate.cpp b/src/video_core/shader/decode/integer_set_predicate.cpp
index dd20775d7..25e48fef8 100644
--- a/src/video_core/shader/decode/integer_set_predicate.cpp
+++ b/src/video_core/shader/decode/integer_set_predicate.cpp
@@ -16,7 +16,6 @@ using Tegra::Shader::Pred;
u32 ShaderIR::DecodeIntegerSetPredicate(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
- const auto opcode = OpCode::Decode(instr);
const Node op_a = GetRegister(instr.gpr8);
diff --git a/src/video_core/shader/decode/other.cpp b/src/video_core/shader/decode/other.cpp
index c0f64d7a0..d46e0f823 100644
--- a/src/video_core/shader/decode/other.cpp
+++ b/src/video_core/shader/decode/other.cpp
@@ -22,6 +22,12 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
const auto opcode = OpCode::Decode(instr);
switch (opcode->get().GetId()) {
+ case OpCode::Id::NOP: {
+ UNIMPLEMENTED_IF(instr.nop.cc != Tegra::Shader::ConditionCode::T);
+ UNIMPLEMENTED_IF(instr.nop.trigger != 0);
+ // With the previous preconditions, this instruction is a no-operation.
+ break;
+ }
case OpCode::Id::EXIT: {
const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T, "EXIT condition code used: {}",
@@ -68,6 +74,13 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
case SystemVariable::InvocationInfo:
LOG_WARNING(HW_GPU, "MOV_SYS instruction with InvocationInfo is incomplete");
return Immediate(0u);
+ case SystemVariable::Tid: {
+ Node value = Immediate(0);
+ value = BitfieldInsert(value, Operation(OperationCode::LocalInvocationIdX), 0, 9);
+ value = BitfieldInsert(value, Operation(OperationCode::LocalInvocationIdY), 16, 9);
+ value = BitfieldInsert(value, Operation(OperationCode::LocalInvocationIdZ), 26, 5);
+ return value;
+ }
case SystemVariable::TidX:
return Operation(OperationCode::LocalInvocationIdX);
case SystemVariable::TidY:
diff --git a/src/video_core/shader/decode/predicate_set_register.cpp b/src/video_core/shader/decode/predicate_set_register.cpp
index febbfeb50..84dbc50fe 100644
--- a/src/video_core/shader/decode/predicate_set_register.cpp
+++ b/src/video_core/shader/decode/predicate_set_register.cpp
@@ -15,7 +15,6 @@ using Tegra::Shader::OpCode;
u32 ShaderIR::DecodePredicateSetRegister(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
- const auto opcode = OpCode::Decode(instr);
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
"Condition codes generation in PSET is not implemented");
diff --git a/src/video_core/shader/decode/warp.cpp b/src/video_core/shader/decode/warp.cpp
new file mode 100644
index 000000000..04ca74f46
--- /dev/null
+++ b/src/video_core/shader/decode/warp.cpp
@@ -0,0 +1,55 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/node_helper.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Pred;
+using Tegra::Shader::VoteOperation;
+
+namespace {
+OperationCode GetOperationCode(VoteOperation vote_op) {
+ switch (vote_op) {
+ case VoteOperation::All:
+ return OperationCode::VoteAll;
+ case VoteOperation::Any:
+ return OperationCode::VoteAny;
+ case VoteOperation::Eq:
+ return OperationCode::VoteEqual;
+ default:
+ UNREACHABLE_MSG("Invalid vote operation={}", static_cast<u64>(vote_op));
+ return OperationCode::VoteAll;
+ }
+}
+} // Anonymous namespace
+
+u32 ShaderIR::DecodeWarp(NodeBlock& bb, u32 pc) {
+ const Instruction instr = {program_code[pc]};
+ const auto opcode = OpCode::Decode(instr);
+
+ switch (opcode->get().GetId()) {
+ case OpCode::Id::VOTE: {
+ const Node value = GetPredicate(instr.vote.value, instr.vote.negate_value != 0);
+ const Node active = Operation(OperationCode::BallotThread, value);
+ const Node vote = Operation(GetOperationCode(instr.vote.operation), value);
+ SetRegister(bb, instr.gpr0, active);
+ SetPredicate(bb, instr.vote.dest_pred, vote);
+ break;
+ }
+ default:
+ UNIMPLEMENTED_MSG("Unhandled warp instruction: {}", opcode->get().GetName());
+ break;
+ }
+
+ return pc;
+}
+
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h
index 715184d67..5db9313c4 100644
--- a/src/video_core/shader/node.h
+++ b/src/video_core/shader/node.h
@@ -30,6 +30,8 @@ enum class OperationCode {
FNegate, /// (MetaArithmetic, float a) -> float
FAbsolute, /// (MetaArithmetic, float a) -> float
FClamp, /// (MetaArithmetic, float value, float min, float max) -> float
+ FCastHalf0, /// (MetaArithmetic, f16vec2 a) -> float
+ FCastHalf1, /// (MetaArithmetic, f16vec2 a) -> float
FMin, /// (MetaArithmetic, float a, float b) -> float
FMax, /// (MetaArithmetic, float a, float b) -> float
FCos, /// (MetaArithmetic, float a) -> float
@@ -83,17 +85,18 @@ enum class OperationCode {
UBitfieldExtract, /// (MetaArithmetic, uint value, int offset, int offset) -> uint
UBitCount, /// (MetaArithmetic, uint) -> uint
- HAdd, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2
- HMul, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2
- HFma, /// (MetaArithmetic, f16vec2 a, f16vec2 b, f16vec2 c) -> f16vec2
- HAbsolute, /// (f16vec2 a) -> f16vec2
- HNegate, /// (f16vec2 a, bool first, bool second) -> f16vec2
- HClamp, /// (f16vec2 src, float min, float max) -> f16vec2
- HUnpack, /// (Tegra::Shader::HalfType, T value) -> f16vec2
- HMergeF32, /// (f16vec2 src) -> float
- HMergeH0, /// (f16vec2 dest, f16vec2 src) -> f16vec2
- HMergeH1, /// (f16vec2 dest, f16vec2 src) -> f16vec2
- HPack2, /// (float a, float b) -> f16vec2
+ HAdd, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2
+ HMul, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2
+ HFma, /// (MetaArithmetic, f16vec2 a, f16vec2 b, f16vec2 c) -> f16vec2
+ HAbsolute, /// (f16vec2 a) -> f16vec2
+ HNegate, /// (f16vec2 a, bool first, bool second) -> f16vec2
+ HClamp, /// (f16vec2 src, float min, float max) -> f16vec2
+ HCastFloat, /// (MetaArithmetic, float a) -> f16vec2
+ HUnpack, /// (Tegra::Shader::HalfType, T value) -> f16vec2
+ HMergeF32, /// (f16vec2 src) -> float
+ HMergeH0, /// (f16vec2 dest, f16vec2 src) -> f16vec2
+ HMergeH1, /// (f16vec2 dest, f16vec2 src) -> f16vec2
+ HPack2, /// (float a, float b) -> f16vec2
LogicalAssign, /// (bool& dst, bool src) -> void
LogicalAnd, /// (bool a, bool b) -> bool
@@ -165,6 +168,11 @@ enum class OperationCode {
WorkGroupIdY, /// () -> uint
WorkGroupIdZ, /// () -> uint
+ BallotThread, /// (bool) -> uint
+ VoteAll, /// (bool) -> bool
+ VoteAny, /// (bool) -> bool
+ VoteEqual, /// (bool) -> bool
+
Amount,
};
diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp
index 5e91fe129..1e5c7f660 100644
--- a/src/video_core/shader/shader_ir.cpp
+++ b/src/video_core/shader/shader_ir.cpp
@@ -405,4 +405,9 @@ Node ShaderIR::BitfieldExtract(Node value, u32 offset, u32 bits) {
Immediate(offset), Immediate(bits));
}
+Node ShaderIR::BitfieldInsert(Node base, Node insert, u32 offset, u32 bits) {
+ return Operation(OperationCode::UBitfieldInsert, NO_PRECISE, base, insert, Immediate(offset),
+ Immediate(bits));
+}
+
} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 59a083d90..bcc9b79b6 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -167,6 +167,7 @@ private:
u32 DecodeFfma(NodeBlock& bb, u32 pc);
u32 DecodeHfma2(NodeBlock& bb, u32 pc);
u32 DecodeConversion(NodeBlock& bb, u32 pc);
+ u32 DecodeWarp(NodeBlock& bb, u32 pc);
u32 DecodeMemory(NodeBlock& bb, u32 pc);
u32 DecodeTexture(NodeBlock& bb, u32 pc);
u32 DecodeImage(NodeBlock& bb, u32 pc);
@@ -279,6 +280,9 @@ private:
/// Extracts a sequence of bits from a node
Node BitfieldExtract(Node value, u32 offset, u32 bits);
+ /// Inserts a sequence of bits from a node
+ Node BitfieldInsert(Node base, Node insert, u32 offset, u32 bits);
+
void WriteTexInstructionFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
const Node4& components);
diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp
index a53e02253..55f5949e4 100644
--- a/src/video_core/shader/track.cpp
+++ b/src/video_core/shader/track.cpp
@@ -59,8 +59,8 @@ std::tuple<Node, u32, u32> ShaderIR::TrackCbuf(Node tracked, const NodeBlock& co
return TrackCbuf(source, code, new_cursor);
}
if (const auto operation = std::get_if<OperationNode>(&*tracked)) {
- for (std::size_t i = 0; i < operation->GetOperandsCount(); ++i) {
- if (auto found = TrackCbuf((*operation)[i], code, cursor); std::get<0>(found)) {
+ for (std::size_t i = operation->GetOperandsCount(); i > 0; --i) {
+ if (auto found = TrackCbuf((*operation)[i - 1], code, cursor); std::get<0>(found)) {
// Cbuf found in operand.
return found;
}
diff --git a/src/video_core/texture_cache/surface_base.cpp b/src/video_core/texture_cache/surface_base.cpp
index 6af9044ca..683c49207 100644
--- a/src/video_core/texture_cache/surface_base.cpp
+++ b/src/video_core/texture_cache/surface_base.cpp
@@ -24,9 +24,8 @@ StagingCache::StagingCache() = default;
StagingCache::~StagingCache() = default;
SurfaceBaseImpl::SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params)
- : params{params}, mipmap_sizes(params.num_levels),
- mipmap_offsets(params.num_levels), gpu_addr{gpu_addr}, host_memory_size{
- params.GetHostSizeInBytes()} {
+ : params{params}, host_memory_size{params.GetHostSizeInBytes()}, gpu_addr{gpu_addr},
+ mipmap_sizes(params.num_levels), mipmap_offsets(params.num_levels) {
std::size_t offset = 0;
for (u32 level = 0; level < params.num_levels; ++level) {
const std::size_t mipmap_size{params.GetGuestMipmapSize(level)};
diff --git a/src/video_core/texture_cache/surface_params.h b/src/video_core/texture_cache/surface_params.h
index 358d6757c..e7ef66ee2 100644
--- a/src/video_core/texture_cache/surface_params.h
+++ b/src/video_core/texture_cache/surface_params.h
@@ -58,7 +58,6 @@ public:
std::size_t GetHostSizeInBytes() const {
std::size_t host_size_in_bytes;
if (GetCompressionType() == SurfaceCompression::Converted) {
- constexpr std::size_t rgb8_bpp = 4ULL;
// ASTC is uncompressed in software, in emulated as RGBA8
host_size_in_bytes = 0;
for (u32 level = 0; level < num_levels; ++level) {
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index a3a3770a7..2ec0203d1 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -308,8 +308,6 @@ protected:
if (!guard_render_targets && surface->IsRenderTarget()) {
ManageRenderTargetUnregister(surface);
}
- const GPUVAddr gpu_addr = surface->GetGpuAddr();
- const CacheAddr cache_ptr = surface->GetCacheAddr();
const std::size_t size = surface->GetSizeInBytes();
const VAddr cpu_addr = surface->GetCpuAddr();
rasterizer.UpdatePagesCachedCount(cpu_addr, size, -1);
diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h
index e3be018b9..e36bc2c04 100644
--- a/src/video_core/textures/texture.h
+++ b/src/video_core/textures/texture.h
@@ -213,7 +213,7 @@ struct TICEntry {
if (header_version != TICHeaderVersion::OneDBuffer) {
return width_minus_1 + 1;
}
- return (buffer_high_width_minus_one << 16) | buffer_low_width_minus_one;
+ return ((buffer_high_width_minus_one << 16) | buffer_low_width_minus_one) + 1;
}
u32 Height() const {
diff --git a/src/yuzu/CMakeLists.txt b/src/yuzu/CMakeLists.txt
index 3dc0e47d0..f051e17b4 100644
--- a/src/yuzu/CMakeLists.txt
+++ b/src/yuzu/CMakeLists.txt
@@ -1,5 +1,6 @@
set(CMAKE_AUTOMOC ON)
set(CMAKE_AUTORCC ON)
+set(CMAKE_AUTOUIC ON)
set(CMAKE_INCLUDE_CURRENT_DIR ON)
set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${PROJECT_SOURCE_DIR}/CMakeModules)
@@ -7,6 +8,7 @@ add_executable(yuzu
Info.plist
about_dialog.cpp
about_dialog.h
+ aboutdialog.ui
applets/error.cpp
applets/error.h
applets/profile_select.cpp
@@ -17,42 +19,59 @@ add_executable(yuzu
applets/web_browser.h
bootmanager.cpp
bootmanager.h
+ compatdb.ui
compatibility_list.cpp
compatibility_list.h
configuration/config.cpp
configuration/config.h
+ configuration/configure.ui
configuration/configure_audio.cpp
configuration/configure_audio.h
+ configuration/configure_audio.ui
configuration/configure_debug.cpp
configuration/configure_debug.h
+ configuration/configure_debug.ui
configuration/configure_dialog.cpp
configuration/configure_dialog.h
configuration/configure_gamelist.cpp
configuration/configure_gamelist.h
+ configuration/configure_gamelist.ui
configuration/configure_general.cpp
configuration/configure_general.h
+ configuration/configure_general.ui
configuration/configure_graphics.cpp
configuration/configure_graphics.h
+ configuration/configure_graphics.ui
configuration/configure_hotkeys.cpp
configuration/configure_hotkeys.h
+ configuration/configure_hotkeys.ui
configuration/configure_input.cpp
configuration/configure_input.h
+ configuration/configure_input.ui
configuration/configure_input_player.cpp
configuration/configure_input_player.h
+ configuration/configure_input_player.ui
configuration/configure_input_simple.cpp
configuration/configure_input_simple.h
+ configuration/configure_input_simple.ui
configuration/configure_mouse_advanced.cpp
configuration/configure_mouse_advanced.h
+ configuration/configure_mouse_advanced.ui
+ configuration/configure_per_general.cpp
+ configuration/configure_per_general.h
+ configuration/configure_per_general.ui
configuration/configure_profile_manager.cpp
configuration/configure_profile_manager.h
+ configuration/configure_profile_manager.ui
configuration/configure_system.cpp
configuration/configure_system.h
- configuration/configure_per_general.cpp
- configuration/configure_per_general.h
+ configuration/configure_system.ui
configuration/configure_touchscreen_advanced.cpp
configuration/configure_touchscreen_advanced.h
+ configuration/configure_touchscreen_advanced.ui
configuration/configure_web.cpp
configuration/configure_web.h
+ configuration/configure_web.ui
debugger/graphics/graphics_breakpoint_observer.cpp
debugger/graphics/graphics_breakpoint_observer.h
debugger/graphics/graphics_breakpoints.cpp
@@ -72,12 +91,14 @@ add_executable(yuzu
game_list_worker.h
loading_screen.cpp
loading_screen.h
+ loading_screen.ui
hotkeys.cpp
hotkeys.h
main.cpp
main.h
- ui_settings.cpp
- ui_settings.h
+ main.ui
+ uisettings.cpp
+ uisettings.h
util/limitable_input_dialog.cpp
util/limitable_input_dialog.h
util/sequence_dialog/sequence_dialog.cpp
@@ -89,44 +110,18 @@ add_executable(yuzu
yuzu.rc
)
-set(UIS
- aboutdialog.ui
- configuration/configure.ui
- configuration/configure_audio.ui
- configuration/configure_debug.ui
- configuration/configure_gamelist.ui
- configuration/configure_general.ui
- configuration/configure_graphics.ui
- configuration/configure_hotkeys.ui
- configuration/configure_input.ui
- configuration/configure_input_player.ui
- configuration/configure_input_simple.ui
- configuration/configure_mouse_advanced.ui
- configuration/configure_per_general.ui
- configuration/configure_profile_manager.ui
- configuration/configure_system.ui
- configuration/configure_touchscreen_advanced.ui
- configuration/configure_web.ui
- compatdb.ui
- loading_screen.ui
- main.ui
-)
-
file(GLOB COMPAT_LIST
${PROJECT_BINARY_DIR}/dist/compatibility_list/compatibility_list.qrc
${PROJECT_BINARY_DIR}/dist/compatibility_list/compatibility_list.json)
file(GLOB_RECURSE ICONS ${PROJECT_SOURCE_DIR}/dist/icons/*)
file(GLOB_RECURSE THEMES ${PROJECT_SOURCE_DIR}/dist/qt_themes/*)
-qt5_wrap_ui(UI_HDRS ${UIS})
target_sources(yuzu
PRIVATE
${COMPAT_LIST}
${ICONS}
${THEMES}
- ${UI_HDRS}
- ${UIS}
)
if (APPLE)
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index b7f3fdf75..5d0fb3f9f 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -11,7 +11,7 @@
#include "core/hle/service/hid/controllers/npad.h"
#include "input_common/main.h"
#include "yuzu/configuration/config.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
Config::Config() {
// TODO: Don't hardcode the path; let the frontend decide where to put the config files.
diff --git a/src/yuzu/configuration/configure_debug.cpp b/src/yuzu/configuration/configure_debug.cpp
index 9a13bb797..5b7e03056 100644
--- a/src/yuzu/configuration/configure_debug.cpp
+++ b/src/yuzu/configuration/configure_debug.cpp
@@ -12,13 +12,13 @@
#include "ui_configure_debug.h"
#include "yuzu/configuration/configure_debug.h"
#include "yuzu/debugger/console.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
ConfigureDebug::ConfigureDebug(QWidget* parent) : QWidget(parent), ui(new Ui::ConfigureDebug) {
ui->setupUi(this);
SetConfiguration();
- connect(ui->open_log_button, &QPushButton::pressed, []() {
+ connect(ui->open_log_button, &QPushButton::clicked, []() {
QString path = QString::fromStdString(FileUtil::GetUserPath(FileUtil::UserPath::LogDir));
QDesktopServices::openUrl(QUrl::fromLocalFile(path));
});
diff --git a/src/yuzu/configuration/configure_gamelist.cpp b/src/yuzu/configuration/configure_gamelist.cpp
index d1724ba89..daedbc33e 100644
--- a/src/yuzu/configuration/configure_gamelist.cpp
+++ b/src/yuzu/configuration/configure_gamelist.cpp
@@ -9,7 +9,7 @@
#include "core/settings.h"
#include "ui_configure_gamelist.h"
#include "yuzu/configuration/configure_gamelist.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
namespace {
constexpr std::array default_icon_sizes{
diff --git a/src/yuzu/configuration/configure_general.cpp b/src/yuzu/configuration/configure_general.cpp
index 7a6e921cd..75fcbfea3 100644
--- a/src/yuzu/configuration/configure_general.cpp
+++ b/src/yuzu/configuration/configure_general.cpp
@@ -6,7 +6,7 @@
#include "core/settings.h"
#include "ui_configure_general.h"
#include "yuzu/configuration/configure_general.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
ConfigureGeneral::ConfigureGeneral(QWidget* parent)
: QWidget(parent), ui(new Ui::ConfigureGeneral) {
diff --git a/src/yuzu/configuration/configure_input.cpp b/src/yuzu/configuration/configure_input.cpp
index 4dd775aab..7613197f2 100644
--- a/src/yuzu/configuration/configure_input.cpp
+++ b/src/yuzu/configuration/configure_input.cpp
@@ -79,7 +79,7 @@ ConfigureInput::ConfigureInput(QWidget* parent)
LoadConfiguration();
UpdateUIEnabled();
- connect(ui->restore_defaults_button, &QPushButton::pressed, this,
+ connect(ui->restore_defaults_button, &QPushButton::clicked, this,
&ConfigureInput::RestoreDefaults);
for (auto* enabled : players_controller) {
@@ -96,20 +96,20 @@ ConfigureInput::ConfigureInput(QWidget* parent)
&ConfigureInput::UpdateUIEnabled);
for (std::size_t i = 0; i < players_configure.size(); ++i) {
- connect(players_configure[i], &QPushButton::pressed, this,
+ connect(players_configure[i], &QPushButton::clicked, this,
[this, i] { CallConfigureDialog<ConfigureInputPlayer>(*this, i, false); });
}
- connect(ui->handheld_configure, &QPushButton::pressed, this,
+ connect(ui->handheld_configure, &QPushButton::clicked, this,
[this] { CallConfigureDialog<ConfigureInputPlayer>(*this, 8, false); });
- connect(ui->debug_configure, &QPushButton::pressed, this,
+ connect(ui->debug_configure, &QPushButton::clicked, this,
[this] { CallConfigureDialog<ConfigureInputPlayer>(*this, 9, true); });
- connect(ui->mouse_advanced, &QPushButton::pressed, this,
+ connect(ui->mouse_advanced, &QPushButton::clicked, this,
[this] { CallConfigureDialog<ConfigureMouseAdvanced>(*this); });
- connect(ui->touchscreen_advanced, &QPushButton::pressed, this,
+ connect(ui->touchscreen_advanced, &QPushButton::clicked, this,
[this] { CallConfigureDialog<ConfigureTouchscreenAdvanced>(*this); });
}
diff --git a/src/yuzu/configuration/configure_input_player.cpp b/src/yuzu/configuration/configure_input_player.cpp
index 916baccc1..7b70f307c 100644
--- a/src/yuzu/configuration/configure_input_player.cpp
+++ b/src/yuzu/configuration/configure_input_player.cpp
@@ -244,7 +244,7 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i
}
button->setContextMenuPolicy(Qt::CustomContextMenu);
- connect(button, &QPushButton::released, [=] {
+ connect(button, &QPushButton::clicked, [=] {
HandleClick(
button_map[button_id],
[=](const Common::ParamPackage& params) { buttons_param[button_id] = params; },
@@ -273,7 +273,7 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i
}
analog_button->setContextMenuPolicy(Qt::CustomContextMenu);
- connect(analog_button, &QPushButton::released, [=]() {
+ connect(analog_button, &QPushButton::clicked, [=]() {
HandleClick(analog_map_buttons[analog_id][sub_button_id],
[=](const Common::ParamPackage& params) {
SetAnalogButton(params, analogs_param[analog_id],
@@ -300,7 +300,7 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i
menu_location));
});
}
- connect(analog_map_stick[analog_id], &QPushButton::released, [=] {
+ connect(analog_map_stick[analog_id], &QPushButton::clicked, [=] {
QMessageBox::information(this, tr("Information"),
tr("After pressing OK, first move your joystick horizontally, "
"and then vertically."));
@@ -311,8 +311,8 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i
});
}
- connect(ui->buttonClearAll, &QPushButton::released, [this] { ClearAll(); });
- connect(ui->buttonRestoreDefaults, &QPushButton::released, [this] { RestoreDefaults(); });
+ connect(ui->buttonClearAll, &QPushButton::clicked, [this] { ClearAll(); });
+ connect(ui->buttonRestoreDefaults, &QPushButton::clicked, [this] { RestoreDefaults(); });
timeout_timer->setSingleShot(true);
connect(timeout_timer.get(), &QTimer::timeout, [this] { SetPollingResult({}, true); });
diff --git a/src/yuzu/configuration/configure_input_simple.cpp b/src/yuzu/configuration/configure_input_simple.cpp
index 864803ea3..ab3a11d30 100644
--- a/src/yuzu/configuration/configure_input_simple.cpp
+++ b/src/yuzu/configuration/configure_input_simple.cpp
@@ -9,7 +9,7 @@
#include "yuzu/configuration/configure_input.h"
#include "yuzu/configuration/configure_input_player.h"
#include "yuzu/configuration/configure_input_simple.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
namespace {
@@ -101,7 +101,7 @@ ConfigureInputSimple::ConfigureInputSimple(QWidget* parent)
connect(ui->profile_combobox, QOverload<int>::of(&QComboBox::currentIndexChanged), this,
&ConfigureInputSimple::OnSelectProfile);
- connect(ui->profile_configure, &QPushButton::pressed, this, &ConfigureInputSimple::OnConfigure);
+ connect(ui->profile_configure, &QPushButton::clicked, this, &ConfigureInputSimple::OnConfigure);
LoadConfiguration();
}
diff --git a/src/yuzu/configuration/configure_mouse_advanced.cpp b/src/yuzu/configuration/configure_mouse_advanced.cpp
index b7305e653..0a4abe34f 100644
--- a/src/yuzu/configuration/configure_mouse_advanced.cpp
+++ b/src/yuzu/configuration/configure_mouse_advanced.cpp
@@ -83,7 +83,7 @@ ConfigureMouseAdvanced::ConfigureMouseAdvanced(QWidget* parent)
}
button->setContextMenuPolicy(Qt::CustomContextMenu);
- connect(button, &QPushButton::released, [=] {
+ connect(button, &QPushButton::clicked, [=] {
HandleClick(
button_map[button_id],
[=](const Common::ParamPackage& params) { buttons_param[button_id] = params; },
@@ -104,8 +104,8 @@ ConfigureMouseAdvanced::ConfigureMouseAdvanced(QWidget* parent)
});
}
- connect(ui->buttonClearAll, &QPushButton::released, [this] { ClearAll(); });
- connect(ui->buttonRestoreDefaults, &QPushButton::released, [this] { RestoreDefaults(); });
+ connect(ui->buttonClearAll, &QPushButton::clicked, [this] { ClearAll(); });
+ connect(ui->buttonRestoreDefaults, &QPushButton::clicked, [this] { RestoreDefaults(); });
timeout_timer->setSingleShot(true);
connect(timeout_timer.get(), &QTimer::timeout, [this] { SetPollingResult({}, true); });
diff --git a/src/yuzu/configuration/configure_per_general.cpp b/src/yuzu/configuration/configure_per_general.cpp
index 90336e235..d7f259f12 100644
--- a/src/yuzu/configuration/configure_per_general.cpp
+++ b/src/yuzu/configuration/configure_per_general.cpp
@@ -23,7 +23,7 @@
#include "yuzu/configuration/config.h"
#include "yuzu/configuration/configure_input.h"
#include "yuzu/configuration/configure_per_general.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
#include "yuzu/util/util.h"
ConfigurePerGameGeneral::ConfigurePerGameGeneral(QWidget* parent, u64 title_id)
diff --git a/src/yuzu/configuration/configure_profile_manager.cpp b/src/yuzu/configuration/configure_profile_manager.cpp
index c90f4cdd8..f53423440 100644
--- a/src/yuzu/configuration/configure_profile_manager.cpp
+++ b/src/yuzu/configuration/configure_profile_manager.cpp
@@ -108,10 +108,10 @@ ConfigureProfileManager ::ConfigureProfileManager(QWidget* parent)
connect(tree_view, &QTreeView::clicked, this, &ConfigureProfileManager::SelectUser);
- connect(ui->pm_add, &QPushButton::pressed, this, &ConfigureProfileManager::AddUser);
- connect(ui->pm_rename, &QPushButton::pressed, this, &ConfigureProfileManager::RenameUser);
- connect(ui->pm_remove, &QPushButton::pressed, this, &ConfigureProfileManager::DeleteUser);
- connect(ui->pm_set_image, &QPushButton::pressed, this, &ConfigureProfileManager::SetUserImage);
+ connect(ui->pm_add, &QPushButton::clicked, this, &ConfigureProfileManager::AddUser);
+ connect(ui->pm_rename, &QPushButton::clicked, this, &ConfigureProfileManager::RenameUser);
+ connect(ui->pm_remove, &QPushButton::clicked, this, &ConfigureProfileManager::DeleteUser);
+ connect(ui->pm_set_image, &QPushButton::clicked, this, &ConfigureProfileManager::SetUserImage);
scene = new QGraphicsScene;
ui->current_user_icon->setScene(scene);
diff --git a/src/yuzu/configuration/configure_touchscreen_advanced.cpp b/src/yuzu/configuration/configure_touchscreen_advanced.cpp
index 8ced28c75..7d7cc00b7 100644
--- a/src/yuzu/configuration/configure_touchscreen_advanced.cpp
+++ b/src/yuzu/configuration/configure_touchscreen_advanced.cpp
@@ -11,7 +11,7 @@ ConfigureTouchscreenAdvanced::ConfigureTouchscreenAdvanced(QWidget* parent)
: QDialog(parent), ui(std::make_unique<Ui::ConfigureTouchscreenAdvanced>()) {
ui->setupUi(this);
- connect(ui->restore_defaults_button, &QPushButton::pressed, this,
+ connect(ui->restore_defaults_button, &QPushButton::clicked, this,
&ConfigureTouchscreenAdvanced::RestoreDefaults);
LoadConfiguration();
diff --git a/src/yuzu/configuration/configure_web.cpp b/src/yuzu/configuration/configure_web.cpp
index 5a70ef168..336b062b3 100644
--- a/src/yuzu/configuration/configure_web.cpp
+++ b/src/yuzu/configuration/configure_web.cpp
@@ -9,7 +9,7 @@
#include "core/telemetry_session.h"
#include "ui_configure_web.h"
#include "yuzu/configuration/configure_web.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
ConfigureWeb::ConfigureWeb(QWidget* parent)
: QWidget(parent), ui(std::make_unique<Ui::ConfigureWeb>()) {
diff --git a/src/yuzu/debugger/console.cpp b/src/yuzu/debugger/console.cpp
index 320898f6a..207ff4d58 100644
--- a/src/yuzu/debugger/console.cpp
+++ b/src/yuzu/debugger/console.cpp
@@ -10,7 +10,7 @@
#include "common/logging/backend.h"
#include "yuzu/debugger/console.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
namespace Debugger {
void ToggleConsole() {
diff --git a/src/yuzu/discord_impl.cpp b/src/yuzu/discord_impl.cpp
index 9d87a41eb..ea0079353 100644
--- a/src/yuzu/discord_impl.cpp
+++ b/src/yuzu/discord_impl.cpp
@@ -9,7 +9,7 @@
#include "core/core.h"
#include "core/loader/loader.h"
#include "yuzu/discord_impl.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
namespace DiscordRPC {
diff --git a/src/yuzu/game_list.cpp b/src/yuzu/game_list.cpp
index 1885587af..d18b96519 100644
--- a/src/yuzu/game_list.cpp
+++ b/src/yuzu/game_list.cpp
@@ -23,7 +23,7 @@
#include "yuzu/game_list_p.h"
#include "yuzu/game_list_worker.h"
#include "yuzu/main.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
GameListSearchField::KeyReleaseEater::KeyReleaseEater(GameList* gamelist) : gamelist{gamelist} {}
diff --git a/src/yuzu/game_list_p.h b/src/yuzu/game_list_p.h
index 0b458ef48..ece534dd6 100644
--- a/src/yuzu/game_list_p.h
+++ b/src/yuzu/game_list_p.h
@@ -19,7 +19,7 @@
#include "common/common_types.h"
#include "common/logging/log.h"
#include "common/string_util.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
#include "yuzu/util/util.h"
/**
diff --git a/src/yuzu/game_list_worker.cpp b/src/yuzu/game_list_worker.cpp
index 4f30e9147..77f358630 100644
--- a/src/yuzu/game_list_worker.cpp
+++ b/src/yuzu/game_list_worker.cpp
@@ -29,7 +29,7 @@
#include "yuzu/game_list.h"
#include "yuzu/game_list_p.h"
#include "yuzu/game_list_worker.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
namespace {
diff --git a/src/yuzu/hotkeys.cpp b/src/yuzu/hotkeys.cpp
index 4582e7f21..d4e97fa16 100644
--- a/src/yuzu/hotkeys.cpp
+++ b/src/yuzu/hotkeys.cpp
@@ -7,7 +7,7 @@
#include <QTreeWidgetItem>
#include <QtGlobal>
#include "yuzu/hotkeys.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
HotkeyRegistry::HotkeyRegistry() = default;
HotkeyRegistry::~HotkeyRegistry() = default;
diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp
index ae21f4753..a7c656fdb 100644
--- a/src/yuzu/main.cpp
+++ b/src/yuzu/main.cpp
@@ -100,7 +100,7 @@ static FileSys::VirtualFile VfsDirectoryCreateFileWrapper(const FileSys::Virtual
#include "yuzu/hotkeys.h"
#include "yuzu/loading_screen.h"
#include "yuzu/main.h"
-#include "yuzu/ui_settings.h"
+#include "yuzu/uisettings.h"
#ifdef USE_DISCORD_PRESENCE
#include "yuzu/discord_impl.h"
@@ -1843,13 +1843,14 @@ void GMainWindow::OnCoreError(Core::System::ResultStatus result, std::string det
"data, or other bugs.");
switch (result) {
case Core::System::ResultStatus::ErrorSystemFiles: {
- QString message = tr("yuzu was unable to locate a Switch system archive");
- if (!details.empty()) {
- message.append(tr(": %1. ").arg(QString::fromStdString(details)));
+ QString message;
+ if (details.empty()) {
+ message =
+ tr("yuzu was unable to locate a Switch system archive. %1").arg(common_message);
} else {
- message.append(tr(". "));
+ message = tr("yuzu was unable to locate a Switch system archive: %1. %2")
+ .arg(QString::fromStdString(details), common_message);
}
- message.append(common_message);
answer = QMessageBox::question(this, tr("System Archive Not Found"), message,
QMessageBox::Yes | QMessageBox::No, QMessageBox::No);
@@ -1858,8 +1859,8 @@ void GMainWindow::OnCoreError(Core::System::ResultStatus result, std::string det
}
case Core::System::ResultStatus::ErrorSharedFont: {
- QString message = tr("yuzu was unable to locate the Switch shared fonts. ");
- message.append(common_message);
+ const QString message =
+ tr("yuzu was unable to locate the Switch shared fonts. %1").arg(common_message);
answer = QMessageBox::question(this, tr("Shared Fonts Not Found"), message,
QMessageBox::Yes | QMessageBox::No, QMessageBox::No);
status_message = tr("Shared Font Missing");
diff --git a/src/yuzu/ui_settings.cpp b/src/yuzu/uisettings.cpp
index 4bdc302e0..7f7d247a3 100644
--- a/src/yuzu/ui_settings.cpp
+++ b/src/yuzu/uisettings.cpp
@@ -2,7 +2,7 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
-#include "ui_settings.h"
+#include "yuzu/uisettings.h"
namespace UISettings {
diff --git a/src/yuzu/ui_settings.h b/src/yuzu/uisettings.h
index a62cd6911..a62cd6911 100644
--- a/src/yuzu/ui_settings.h
+++ b/src/yuzu/uisettings.h
diff --git a/src/yuzu_tester/yuzu.cpp b/src/yuzu_tester/yuzu.cpp
index b589c3de3..0ee97aa54 100644
--- a/src/yuzu_tester/yuzu.cpp
+++ b/src/yuzu_tester/yuzu.cpp
@@ -92,7 +92,6 @@ int main(int argc, char** argv) {
int option_index = 0;
- char* endarg;
#ifdef _WIN32
int argc_w;
auto argv_w = CommandLineToArgvW(GetCommandLineW(), &argc_w);
@@ -226,7 +225,7 @@ int main(int argc, char** argv) {
switch (load_result) {
case Core::System::ResultStatus::ErrorGetLoader:
- LOG_CRITICAL(Frontend, "Failed to obtain loader for %s!", filepath.c_str());
+ LOG_CRITICAL(Frontend, "Failed to obtain loader for {}!", filepath);
return -1;
case Core::System::ResultStatus::ErrorLoader:
LOG_CRITICAL(Frontend, "Failed to load ROM!");