VERY experimental Nintendo Switch 2 emulator
C++
video_core: make gpu context aware of rendering program
Liame15b1021a4da
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cppindex bf12d69a5d..efc9cca1cc 100644--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp@@ -5,6 +5,7 @@#include "common/assert.h"#include "common/logging/log.h"#include "core/core.h"+#include "core/hle/kernel/k_process.h"#include "core/hle/service/nvdrv/core/container.h"#include "core/hle/service/nvdrv/core/nvmap.h"#include "core/hle/service/nvdrv/core/syncpoint_manager.h"@@ -75,7 +76,7 @@ NvResult nvhost_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpucase 0xd:return WrapFixed(this, &nvhost_gpu::SetChannelPriority, input, output);case 0x1a:- return WrapFixed(this, &nvhost_gpu::AllocGPFIFOEx2, input, output);+ return WrapFixed(this, &nvhost_gpu::AllocGPFIFOEx2, input, output, fd);case 0x1b:return WrapFixedVariable(this, &nvhost_gpu::SubmitGPFIFOBase1, input, output, true);case 0x1d:@@ -120,8 +121,13 @@ NvResult nvhost_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> inpureturn NvResult::NotImplemented;}-void nvhost_gpu::OnOpen(NvCore::SessionId session_id, DeviceFD fd) {}-void nvhost_gpu::OnClose(DeviceFD fd) {}+void nvhost_gpu::OnOpen(NvCore::SessionId session_id, DeviceFD fd) {+ sessions[fd] = session_id;+}++void nvhost_gpu::OnClose(DeviceFD fd) {+ sessions.erase(fd);+}NvResult nvhost_gpu::SetNVMAPfd(IoctlSetNvmapFD& params) {LOG_DEBUG(Service_NVDRV, "called, fd={}", params.nvmap_fd);@@ -161,7 +167,7 @@ NvResult nvhost_gpu::SetChannelPriority(IoctlChannelSetPriority& params) {return NvResult::Success;}-NvResult nvhost_gpu::AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params) {+NvResult nvhost_gpu::AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params, DeviceFD fd) {LOG_WARNING(Service_NVDRV,"(STUBBED) called, num_entries={:X}, flags={:X}, unk0={:X}, ""unk1={:X}, unk2={:X}, unk3={:X}",@@ -173,7 +179,12 @@ NvResult nvhost_gpu::AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params) {return NvResult::AlreadyAllocated;}- system.GPU().InitChannel(*channel_state);+ u64 program_id{};+ if (auto* const session = core.GetSession(sessions[fd]); session != nullptr) {+ program_id = session->process->GetProgramId();+ }++ system.GPU().InitChannel(*channel_state, program_id);params.fence_out = syncpoint_manager.GetSyncpointFence(channel_syncpoint);diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_gpu.hindex e34a978db0..e0aeef9536 100644--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h@@ -192,7 +192,7 @@ private:NvResult ZCullBind(IoctlZCullBind& params);NvResult SetErrorNotifier(IoctlSetErrorNotifier& params);NvResult SetChannelPriority(IoctlChannelSetPriority& params);- NvResult AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params);+ NvResult AllocGPFIFOEx2(IoctlAllocGpfifoEx2& params, DeviceFD fd);NvResult AllocateObjectContext(IoctlAllocObjCtx& params);NvResult SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, Tegra::CommandList&& entries);@@ -210,6 +210,7 @@ private:NvCore::SyncpointManager& syncpoint_manager;NvCore::NvMap& nvmap;std::shared_ptr<Tegra::Control::ChannelState> channel_state;+ std::unordered_map<DeviceFD, NvCore::SessionId> sessions;u32 channel_syncpoint;std::mutex channel_mutex;diff --git a/src/video_core/control/channel_state.cpp b/src/video_core/control/channel_state.cppindex 832025d753..2539997d53 100644--- a/src/video_core/control/channel_state.cpp+++ b/src/video_core/control/channel_state.cpp@@ -16,8 +16,9 @@ namespace Tegra::Control {ChannelState::ChannelState(s32 bind_id_) : bind_id{bind_id_}, initialized{} {}-void ChannelState::Init(Core::System& system, GPU& gpu) {+void ChannelState::Init(Core::System& system, GPU& gpu, u64 program_id_) {ASSERT(memory_manager);+ program_id = program_id_;dma_pusher = std::make_unique<Tegra::DmaPusher>(system, gpu, *memory_manager, *this);maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, *memory_manager);fermi_2d = std::make_unique<Engines::Fermi2D>(*memory_manager);diff --git a/src/video_core/control/channel_state.h b/src/video_core/control/channel_state.hindex 3a7b9872c1..b385f4939f 100644--- a/src/video_core/control/channel_state.h+++ b/src/video_core/control/channel_state.h@@ -40,11 +40,12 @@ struct ChannelState {ChannelState(ChannelState&& other) noexcept = default;ChannelState& operator=(ChannelState&& other) noexcept = default;- void Init(Core::System& system, GPU& gpu);+ void Init(Core::System& system, GPU& gpu, u64 program_id);void BindRasterizer(VideoCore::RasterizerInterface* rasterizer);s32 bind_id = -1;+ u64 program_id = 0;/// 3D enginestd::unique_ptr<Engines::Maxwell3D> maxwell_3d;/// 2D enginediff --git a/src/video_core/control/channel_state_cache.cpp b/src/video_core/control/channel_state_cache.cppindex 4ebeb6356b..f8c6a762d2 100644--- a/src/video_core/control/channel_state_cache.cpp+++ b/src/video_core/control/channel_state_cache.cpp@@ -7,7 +7,7 @@ namespace VideoCommon {ChannelInfo::ChannelInfo(Tegra::Control::ChannelState& channel_state): maxwell3d{*channel_state.maxwell_3d}, kepler_compute{*channel_state.kepler_compute},- gpu_memory{*channel_state.memory_manager} {}+ gpu_memory{*channel_state.memory_manager}, program_id{channel_state.program_id} {}template class VideoCommon::ChannelSetupCaches<VideoCommon::ChannelInfo>;diff --git a/src/video_core/control/channel_state_cache.h b/src/video_core/control/channel_state_cache.hindex 1dbfda2992..7480d60d13 100644--- a/src/video_core/control/channel_state_cache.h+++ b/src/video_core/control/channel_state_cache.h@@ -39,6 +39,7 @@ public:Tegra::Engines::Maxwell3D& maxwell3d;Tegra::Engines::KeplerCompute& kepler_compute;Tegra::MemoryManager& gpu_memory;+ u64 program_id;};template <class P>@@ -77,9 +78,10 @@ protected:P* channel_state;size_t current_channel_id{UNSET_CHANNEL};size_t current_address_space{};- Tegra::Engines::Maxwell3D* maxwell3d;- Tegra::Engines::KeplerCompute* kepler_compute;- Tegra::MemoryManager* gpu_memory;+ Tegra::Engines::Maxwell3D* maxwell3d{};+ Tegra::Engines::KeplerCompute* kepler_compute{};+ Tegra::MemoryManager* gpu_memory{};+ u64 program_id{};std::deque<P> channel_storage;std::deque<size_t> free_channel_ids;diff --git a/src/video_core/control/channel_state_cache.inc b/src/video_core/control/channel_state_cache.incindex 31f792ddda..d882d8222f 100644--- a/src/video_core/control/channel_state_cache.inc+++ b/src/video_core/control/channel_state_cache.inc@@ -58,6 +58,7 @@ void ChannelSetupCaches<P>::BindToChannel(s32 id) {maxwell3d = &channel_state->maxwell3d;kepler_compute = &channel_state->kepler_compute;gpu_memory = &channel_state->gpu_memory;+ program_id = channel_state->program_id;current_address_space = gpu_memory->GetID();}@@ -76,6 +77,7 @@ void ChannelSetupCaches<P>::EraseChannel(s32 id) {maxwell3d = nullptr;kepler_compute = nullptr;gpu_memory = nullptr;+ program_id = 0;} else if (current_channel_id != UNSET_CHANNEL) {channel_state = &channel_storage[current_channel_id];}diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cppindex 8e663f2a8f..6d0b323395 100644--- a/src/video_core/gpu.cpp+++ b/src/video_core/gpu.cpp@@ -67,8 +67,8 @@ struct GPU::Impl {return CreateChannel(new_channel_id++);}- void InitChannel(Control::ChannelState& to_init) {- to_init.Init(system, gpu);+ void InitChannel(Control::ChannelState& to_init, u64 program_id) {+ to_init.Init(system, gpu, program_id);to_init.BindRasterizer(rasterizer);rasterizer->InitializeChannel(to_init);}@@ -412,8 +412,8 @@ std::shared_ptr<Control::ChannelState> GPU::AllocateChannel() {return impl->AllocateChannel();}-void GPU::InitChannel(Control::ChannelState& to_init) {- impl->InitChannel(to_init);+void GPU::InitChannel(Control::ChannelState& to_init, u64 program_id) {+ impl->InitChannel(to_init, program_id);}void GPU::BindChannel(s32 channel_id) {diff --git a/src/video_core/gpu.h b/src/video_core/gpu.hindex ad535512c9..50014e51f6 100644--- a/src/video_core/gpu.h+++ b/src/video_core/gpu.h@@ -149,7 +149,7 @@ public:std::shared_ptr<Control::ChannelState> AllocateChannel();- void InitChannel(Control::ChannelState& to_init);+ void InitChannel(Control::ChannelState& to_init, u64 program_id);void BindChannel(s32 channel_id);diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cppindex 74f9f099e9..6cfec2e6f2 100644--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp@@ -1061,7 +1061,9 @@ void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) {if (device.IsExtDepthBiasControlSupported()) {return true;}- if (!Settings::values.renderer_amdvlk_depth_bias_workaround) {+ // Only activate this in Super Smash Brothers Ultimate+ // Affects AMD cards using AMDVLK+ if (program_id != 0x1006A800016E000ULL) {return false;}// the base formulas can be obtained from here: