summaryrefslogtreecommitdiff
path: root/src
diff options
context:
space:
mode:
Diffstat (limited to 'src')
-rw-r--r--src/audio_core/cubeb_sink.cpp6
-rw-r--r--src/audio_core/cubeb_sink.h2
-rw-r--r--src/common/CMakeLists.txt1
-rw-r--r--src/core/core_cpu.cpp9
-rw-r--r--src/core/core_cpu.h8
-rw-r--r--src/core/cpu_core_manager.cpp3
-rw-r--r--src/core/hle/kernel/scheduler.cpp14
-rw-r--r--src/core/hle/kernel/scheduler.h6
-rw-r--r--src/core/hle/service/audio/hwopus.cpp176
-rw-r--r--src/core/settings.cpp3
-rw-r--r--src/video_core/CMakeLists.txt1
-rw-r--r--src/video_core/engines/shader_bytecode.h25
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp37
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp41
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.h28
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp56
-rw-r--r--src/video_core/shader/decode.cpp1
-rw-r--r--src/video_core/shader/decode/memory.cpp493
-rw-r--r--src/video_core/shader/decode/texture.cpp534
-rw-r--r--src/video_core/shader/shader_ir.h5
-rw-r--r--src/yuzu_cmd/config.cpp11
21 files changed, 787 insertions, 673 deletions
diff --git a/src/audio_core/cubeb_sink.cpp b/src/audio_core/cubeb_sink.cpp
index 1da0b9f2a..7047ed9cf 100644
--- a/src/audio_core/cubeb_sink.cpp
+++ b/src/audio_core/cubeb_sink.cpp
@@ -12,7 +12,7 @@
#include "common/ring_buffer.h"
#include "core/settings.h"
-#ifdef _MSC_VER
+#ifdef _WIN32
#include <objbase.h>
#endif
@@ -113,7 +113,7 @@ private:
CubebSink::CubebSink(std::string_view target_device_name) {
// Cubeb requires COM to be initialized on the thread calling cubeb_init on Windows
-#ifdef _MSC_VER
+#ifdef _WIN32
com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
#endif
@@ -152,7 +152,7 @@ CubebSink::~CubebSink() {
cubeb_destroy(ctx);
-#ifdef _MSC_VER
+#ifdef _WIN32
if (SUCCEEDED(com_init_result)) {
CoUninitialize();
}
diff --git a/src/audio_core/cubeb_sink.h b/src/audio_core/cubeb_sink.h
index 511df7bb1..7ce850f47 100644
--- a/src/audio_core/cubeb_sink.h
+++ b/src/audio_core/cubeb_sink.h
@@ -26,7 +26,7 @@ private:
cubeb_devid output_device{};
std::vector<SinkStreamPtr> sink_streams;
-#ifdef _MSC_VER
+#ifdef _WIN32
u32 com_init_result = 0;
#endif
};
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index bdd885273..3d30f0e3e 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -47,6 +47,7 @@ add_custom_command(OUTPUT scm_rev.cpp
"${VIDEO_CORE}/shader/decode/integer_set.cpp"
"${VIDEO_CORE}/shader/decode/integer_set_predicate.cpp"
"${VIDEO_CORE}/shader/decode/memory.cpp"
+ "${VIDEO_CORE}/shader/decode/texture.cpp"
"${VIDEO_CORE}/shader/decode/other.cpp"
"${VIDEO_CORE}/shader/decode/predicate_set_predicate.cpp"
"${VIDEO_CORE}/shader/decode/predicate_set_register.cpp"
diff --git a/src/core/core_cpu.cpp b/src/core/core_cpu.cpp
index 54aa21a3a..1eefed6d0 100644
--- a/src/core/core_cpu.cpp
+++ b/src/core/core_cpu.cpp
@@ -11,6 +11,7 @@
#endif
#include "core/arm/exclusive_monitor.h"
#include "core/arm/unicorn/arm_unicorn.h"
+#include "core/core.h"
#include "core/core_cpu.h"
#include "core/core_timing.h"
#include "core/hle/kernel/scheduler.h"
@@ -49,9 +50,9 @@ bool CpuBarrier::Rendezvous() {
return false;
}
-Cpu::Cpu(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
- CpuBarrier& cpu_barrier, std::size_t core_index)
- : cpu_barrier{cpu_barrier}, core_timing{core_timing}, core_index{core_index} {
+Cpu::Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier,
+ std::size_t core_index)
+ : cpu_barrier{cpu_barrier}, core_timing{system.CoreTiming()}, core_index{core_index} {
if (Settings::values.use_cpu_jit) {
#ifdef ARCHITECTURE_x86_64
arm_interface = std::make_unique<ARM_Dynarmic>(core_timing, exclusive_monitor, core_index);
@@ -63,7 +64,7 @@ Cpu::Cpu(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
arm_interface = std::make_unique<ARM_Unicorn>(core_timing);
}
- scheduler = std::make_unique<Kernel::Scheduler>(*arm_interface);
+ scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface);
}
Cpu::~Cpu() = default;
diff --git a/src/core/core_cpu.h b/src/core/core_cpu.h
index e2204c6b0..7589beb8c 100644
--- a/src/core/core_cpu.h
+++ b/src/core/core_cpu.h
@@ -15,6 +15,10 @@ namespace Kernel {
class Scheduler;
}
+namespace Core {
+class System;
+}
+
namespace Core::Timing {
class CoreTiming;
}
@@ -45,8 +49,8 @@ private:
class Cpu {
public:
- Cpu(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
- CpuBarrier& cpu_barrier, std::size_t core_index);
+ Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier,
+ std::size_t core_index);
~Cpu();
void RunLoop(bool tight_loop = true);
diff --git a/src/core/cpu_core_manager.cpp b/src/core/cpu_core_manager.cpp
index 2ddb3610d..93bc5619c 100644
--- a/src/core/cpu_core_manager.cpp
+++ b/src/core/cpu_core_manager.cpp
@@ -27,8 +27,7 @@ void CpuCoreManager::Initialize(System& system) {
exclusive_monitor = Cpu::MakeExclusiveMonitor(cores.size());
for (std::size_t index = 0; index < cores.size(); ++index) {
- cores[index] =
- std::make_unique<Cpu>(system.CoreTiming(), *exclusive_monitor, *barrier, index);
+ cores[index] = std::make_unique<Cpu>(system, *exclusive_monitor, *barrier, index);
}
// Create threads for CPU cores 1-3, and build thread_to_cpu map
diff --git a/src/core/hle/kernel/scheduler.cpp b/src/core/hle/kernel/scheduler.cpp
index 44f30d070..5fccfd9f4 100644
--- a/src/core/hle/kernel/scheduler.cpp
+++ b/src/core/hle/kernel/scheduler.cpp
@@ -19,7 +19,8 @@ namespace Kernel {
std::mutex Scheduler::scheduler_mutex;
-Scheduler::Scheduler(Core::ARM_Interface& cpu_core) : cpu_core(cpu_core) {}
+Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core)
+ : cpu_core{cpu_core}, system{system} {}
Scheduler::~Scheduler() {
for (auto& thread : thread_list) {
@@ -61,7 +62,7 @@ Thread* Scheduler::PopNextReadyThread() {
void Scheduler::SwitchContext(Thread* new_thread) {
Thread* const previous_thread = GetCurrentThread();
- Process* const previous_process = Core::CurrentProcess();
+ Process* const previous_process = system.Kernel().CurrentProcess();
UpdateLastContextSwitchTime(previous_thread, previous_process);
@@ -94,8 +95,8 @@ void Scheduler::SwitchContext(Thread* new_thread) {
auto* const thread_owner_process = current_thread->GetOwnerProcess();
if (previous_process != thread_owner_process) {
- Core::System::GetInstance().Kernel().MakeCurrentProcess(thread_owner_process);
- SetCurrentPageTable(&Core::CurrentProcess()->VMManager().page_table);
+ system.Kernel().MakeCurrentProcess(thread_owner_process);
+ SetCurrentPageTable(&thread_owner_process->VMManager().page_table);
}
cpu_core.LoadContext(new_thread->GetContext());
@@ -111,7 +112,7 @@ void Scheduler::SwitchContext(Thread* new_thread) {
void Scheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) {
const u64 prev_switch_ticks = last_context_switch_time;
- const u64 most_recent_switch_ticks = Core::System::GetInstance().CoreTiming().GetTicks();
+ const u64 most_recent_switch_ticks = system.CoreTiming().GetTicks();
const u64 update_ticks = most_recent_switch_ticks - prev_switch_ticks;
if (thread != nullptr) {
@@ -223,8 +224,7 @@ void Scheduler::YieldWithLoadBalancing(Thread* thread) {
// Take the first non-nullptr one
for (unsigned cur_core = 0; cur_core < Core::NUM_CPU_CORES; ++cur_core) {
const auto res =
- Core::System::GetInstance().CpuCore(cur_core).Scheduler().GetNextSuggestedThread(
- core, priority);
+ system.CpuCore(cur_core).Scheduler().GetNextSuggestedThread(core, priority);
// If scheduler provides a suggested thread
if (res != nullptr) {
diff --git a/src/core/hle/kernel/scheduler.h b/src/core/hle/kernel/scheduler.h
index 97ced4dfc..1c5bf57d9 100644
--- a/src/core/hle/kernel/scheduler.h
+++ b/src/core/hle/kernel/scheduler.h
@@ -13,7 +13,8 @@
namespace Core {
class ARM_Interface;
-}
+class System;
+} // namespace Core
namespace Kernel {
@@ -21,7 +22,7 @@ class Process;
class Scheduler final {
public:
- explicit Scheduler(Core::ARM_Interface& cpu_core);
+ explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core);
~Scheduler();
/// Returns whether there are any threads that are ready to run.
@@ -162,6 +163,7 @@ private:
Core::ARM_Interface& cpu_core;
u64 last_context_switch_time = 0;
+ Core::System& system;
static std::mutex scheduler_mutex;
};
diff --git a/src/core/hle/service/audio/hwopus.cpp b/src/core/hle/service/audio/hwopus.cpp
index 11eba4a12..377e12cfa 100644
--- a/src/core/hle/service/audio/hwopus.cpp
+++ b/src/core/hle/service/audio/hwopus.cpp
@@ -9,43 +9,32 @@
#include <opus.h>
-#include "common/common_funcs.h"
+#include "common/assert.h"
#include "common/logging/log.h"
#include "core/hle/ipc_helpers.h"
#include "core/hle/kernel/hle_ipc.h"
#include "core/hle/service/audio/hwopus.h"
namespace Service::Audio {
-
+namespace {
struct OpusDeleter {
void operator()(void* ptr) const {
operator delete(ptr);
}
};
-class IHardwareOpusDecoderManager final : public ServiceFramework<IHardwareOpusDecoderManager> {
-public:
- IHardwareOpusDecoderManager(std::unique_ptr<OpusDecoder, OpusDeleter> decoder, u32 sample_rate,
- u32 channel_count)
- : ServiceFramework("IHardwareOpusDecoderManager"), decoder(std::move(decoder)),
- sample_rate(sample_rate), channel_count(channel_count) {
- // clang-format off
- static const FunctionInfo functions[] = {
- {0, &IHardwareOpusDecoderManager::DecodeInterleavedOld, "DecodeInterleavedOld"},
- {1, nullptr, "SetContext"},
- {2, nullptr, "DecodeInterleavedForMultiStreamOld"},
- {3, nullptr, "SetContextForMultiStream"},
- {4, &IHardwareOpusDecoderManager::DecodeInterleavedWithPerfOld, "DecodeInterleavedWithPerfOld"},
- {5, nullptr, "DecodeInterleavedForMultiStreamWithPerfOld"},
- {6, &IHardwareOpusDecoderManager::DecodeInterleaved, "DecodeInterleaved"},
- {7, nullptr, "DecodeInterleavedForMultiStream"},
- };
- // clang-format on
+using OpusDecoderPtr = std::unique_ptr<OpusDecoder, OpusDeleter>;
- RegisterHandlers(functions);
- }
+struct OpusPacketHeader {
+ // Packet size in bytes.
+ u32_be size;
+ // Indicates the final range of the codec's entropy coder.
+ u32_be final_range;
+};
+static_assert(sizeof(OpusPacketHeader) == 0x8, "OpusHeader is an invalid size");
-private:
+class OpusDecoderStateBase {
+public:
/// Describes extra behavior that may be asked of the decoding context.
enum class ExtraBehavior {
/// No extra behavior.
@@ -55,30 +44,36 @@ private:
ResetContext,
};
- void DecodeInterleavedOld(Kernel::HLERequestContext& ctx) {
- LOG_DEBUG(Audio, "called");
-
- DecodeInterleavedHelper(ctx, nullptr, ExtraBehavior::None);
- }
-
- void DecodeInterleavedWithPerfOld(Kernel::HLERequestContext& ctx) {
- LOG_DEBUG(Audio, "called");
+ enum class PerfTime {
+ Disabled,
+ Enabled,
+ };
- u64 performance = 0;
- DecodeInterleavedHelper(ctx, &performance, ExtraBehavior::None);
- }
+ virtual ~OpusDecoderStateBase() = default;
- void DecodeInterleaved(Kernel::HLERequestContext& ctx) {
- LOG_DEBUG(Audio, "called");
-
- IPC::RequestParser rp{ctx};
- const auto extra_behavior =
- rp.Pop<bool>() ? ExtraBehavior::ResetContext : ExtraBehavior::None;
+ // Decodes interleaved Opus packets. Optionally allows reporting time taken to
+ // perform the decoding, as well as any relevant extra behavior.
+ virtual void DecodeInterleaved(Kernel::HLERequestContext& ctx, PerfTime perf_time,
+ ExtraBehavior extra_behavior) = 0;
+};
- u64 performance = 0;
- DecodeInterleavedHelper(ctx, &performance, extra_behavior);
+// Represents the decoder state for a non-multistream decoder.
+class OpusDecoderState final : public OpusDecoderStateBase {
+public:
+ explicit OpusDecoderState(OpusDecoderPtr decoder, u32 sample_rate, u32 channel_count)
+ : decoder{std::move(decoder)}, sample_rate{sample_rate}, channel_count{channel_count} {}
+
+ void DecodeInterleaved(Kernel::HLERequestContext& ctx, PerfTime perf_time,
+ ExtraBehavior extra_behavior) override {
+ if (perf_time == PerfTime::Disabled) {
+ DecodeInterleavedHelper(ctx, nullptr, extra_behavior);
+ } else {
+ u64 performance = 0;
+ DecodeInterleavedHelper(ctx, &performance, extra_behavior);
+ }
}
+private:
void DecodeInterleavedHelper(Kernel::HLERequestContext& ctx, u64* performance,
ExtraBehavior extra_behavior) {
u32 consumed = 0;
@@ -89,8 +84,7 @@ private:
ResetDecoderContext();
}
- if (!Decoder_DecodeInterleaved(consumed, sample_count, ctx.ReadBuffer(), samples,
- performance)) {
+ if (!DecodeOpusData(consumed, sample_count, ctx.ReadBuffer(), samples, performance)) {
LOG_ERROR(Audio, "Failed to decode opus data");
IPC::ResponseBuilder rb{ctx, 2};
// TODO(ogniK): Use correct error code
@@ -109,27 +103,27 @@ private:
ctx.WriteBuffer(samples.data(), samples.size() * sizeof(s16));
}
- bool Decoder_DecodeInterleaved(u32& consumed, u32& sample_count, const std::vector<u8>& input,
- std::vector<opus_int16>& output, u64* out_performance_time) {
+ bool DecodeOpusData(u32& consumed, u32& sample_count, const std::vector<u8>& input,
+ std::vector<opus_int16>& output, u64* out_performance_time) const {
const auto start_time = std::chrono::high_resolution_clock::now();
const std::size_t raw_output_sz = output.size() * sizeof(opus_int16);
- if (sizeof(OpusHeader) > input.size()) {
+ if (sizeof(OpusPacketHeader) > input.size()) {
LOG_ERROR(Audio, "Input is smaller than the header size, header_sz={}, input_sz={}",
- sizeof(OpusHeader), input.size());
+ sizeof(OpusPacketHeader), input.size());
return false;
}
- OpusHeader hdr{};
- std::memcpy(&hdr, input.data(), sizeof(OpusHeader));
- if (sizeof(OpusHeader) + static_cast<u32>(hdr.sz) > input.size()) {
+ OpusPacketHeader hdr{};
+ std::memcpy(&hdr, input.data(), sizeof(OpusPacketHeader));
+ if (sizeof(OpusPacketHeader) + static_cast<u32>(hdr.size) > input.size()) {
LOG_ERROR(Audio, "Input does not fit in the opus header size. data_sz={}, input_sz={}",
- sizeof(OpusHeader) + static_cast<u32>(hdr.sz), input.size());
+ sizeof(OpusPacketHeader) + static_cast<u32>(hdr.size), input.size());
return false;
}
- const auto frame = input.data() + sizeof(OpusHeader);
+ const auto frame = input.data() + sizeof(OpusPacketHeader);
const auto decoded_sample_count = opus_packet_get_nb_samples(
- frame, static_cast<opus_int32>(input.size() - sizeof(OpusHeader)),
+ frame, static_cast<opus_int32>(input.size() - sizeof(OpusPacketHeader)),
static_cast<opus_int32>(sample_rate));
if (decoded_sample_count * channel_count * sizeof(u16) > raw_output_sz) {
LOG_ERROR(
@@ -141,18 +135,18 @@ private:
const int frame_size = (static_cast<int>(raw_output_sz / sizeof(s16) / channel_count));
const auto out_sample_count =
- opus_decode(decoder.get(), frame, hdr.sz, output.data(), frame_size, 0);
+ opus_decode(decoder.get(), frame, hdr.size, output.data(), frame_size, 0);
if (out_sample_count < 0) {
LOG_ERROR(Audio,
"Incorrect sample count received from opus_decode, "
"output_sample_count={}, frame_size={}, data_sz_from_hdr={}",
- out_sample_count, frame_size, static_cast<u32>(hdr.sz));
+ out_sample_count, frame_size, static_cast<u32>(hdr.size));
return false;
}
const auto end_time = std::chrono::high_resolution_clock::now() - start_time;
sample_count = out_sample_count;
- consumed = static_cast<u32>(sizeof(OpusHeader) + hdr.sz);
+ consumed = static_cast<u32>(sizeof(OpusPacketHeader) + hdr.size);
if (out_performance_time != nullptr) {
*out_performance_time =
std::chrono::duration_cast<std::chrono::milliseconds>(end_time).count();
@@ -167,21 +161,66 @@ private:
opus_decoder_ctl(decoder.get(), OPUS_RESET_STATE);
}
- struct OpusHeader {
- u32_be sz; // Needs to be BE for some odd reason
- INSERT_PADDING_WORDS(1);
- };
- static_assert(sizeof(OpusHeader) == 0x8, "OpusHeader is an invalid size");
-
- std::unique_ptr<OpusDecoder, OpusDeleter> decoder;
+ OpusDecoderPtr decoder;
u32 sample_rate;
u32 channel_count;
};
-static std::size_t WorkerBufferSize(u32 channel_count) {
+class IHardwareOpusDecoderManager final : public ServiceFramework<IHardwareOpusDecoderManager> {
+public:
+ explicit IHardwareOpusDecoderManager(std::unique_ptr<OpusDecoderStateBase> decoder_state)
+ : ServiceFramework("IHardwareOpusDecoderManager"), decoder_state{std::move(decoder_state)} {
+ // clang-format off
+ static const FunctionInfo functions[] = {
+ {0, &IHardwareOpusDecoderManager::DecodeInterleavedOld, "DecodeInterleavedOld"},
+ {1, nullptr, "SetContext"},
+ {2, nullptr, "DecodeInterleavedForMultiStreamOld"},
+ {3, nullptr, "SetContextForMultiStream"},
+ {4, &IHardwareOpusDecoderManager::DecodeInterleavedWithPerfOld, "DecodeInterleavedWithPerfOld"},
+ {5, nullptr, "DecodeInterleavedForMultiStreamWithPerfOld"},
+ {6, &IHardwareOpusDecoderManager::DecodeInterleaved, "DecodeInterleaved"},
+ {7, nullptr, "DecodeInterleavedForMultiStream"},
+ };
+ // clang-format on
+
+ RegisterHandlers(functions);
+ }
+
+private:
+ void DecodeInterleavedOld(Kernel::HLERequestContext& ctx) {
+ LOG_DEBUG(Audio, "called");
+
+ decoder_state->DecodeInterleaved(ctx, OpusDecoderStateBase::PerfTime::Disabled,
+ OpusDecoderStateBase::ExtraBehavior::None);
+ }
+
+ void DecodeInterleavedWithPerfOld(Kernel::HLERequestContext& ctx) {
+ LOG_DEBUG(Audio, "called");
+
+ decoder_state->DecodeInterleaved(ctx, OpusDecoderStateBase::PerfTime::Enabled,
+ OpusDecoderStateBase::ExtraBehavior::None);
+ }
+
+ void DecodeInterleaved(Kernel::HLERequestContext& ctx) {
+ LOG_DEBUG(Audio, "called");
+
+ IPC::RequestParser rp{ctx};
+ const auto extra_behavior = rp.Pop<bool>()
+ ? OpusDecoderStateBase::ExtraBehavior::ResetContext
+ : OpusDecoderStateBase::ExtraBehavior::None;
+
+ decoder_state->DecodeInterleaved(ctx, OpusDecoderStateBase::PerfTime::Enabled,
+ extra_behavior);
+ }
+
+ std::unique_ptr<OpusDecoderStateBase> decoder_state;
+};
+
+std::size_t WorkerBufferSize(u32 channel_count) {
ASSERT_MSG(channel_count == 1 || channel_count == 2, "Invalid channel count");
return opus_decoder_get_size(static_cast<int>(channel_count));
}
+} // Anonymous namespace
void HwOpus::GetWorkBufferSize(Kernel::HLERequestContext& ctx) {
IPC::RequestParser rp{ctx};
@@ -220,8 +259,7 @@ void HwOpus::OpenOpusDecoder(Kernel::HLERequestContext& ctx) {
const std::size_t worker_sz = WorkerBufferSize(channel_count);
ASSERT_MSG(buffer_sz >= worker_sz, "Worker buffer too large");
- std::unique_ptr<OpusDecoder, OpusDeleter> decoder{
- static_cast<OpusDecoder*>(operator new(worker_sz))};
+ OpusDecoderPtr decoder{static_cast<OpusDecoder*>(operator new(worker_sz))};
if (const int err = opus_decoder_init(decoder.get(), sample_rate, channel_count)) {
LOG_ERROR(Audio, "Failed to init opus decoder with error={}", err);
IPC::ResponseBuilder rb{ctx, 2};
@@ -232,8 +270,8 @@ void HwOpus::OpenOpusDecoder(Kernel::HLERequestContext& ctx) {
IPC::ResponseBuilder rb{ctx, 2, 0, 1};
rb.Push(RESULT_SUCCESS);
- rb.PushIpcInterface<IHardwareOpusDecoderManager>(std::move(decoder), sample_rate,
- channel_count);
+ rb.PushIpcInterface<IHardwareOpusDecoderManager>(
+ std::make_unique<OpusDecoderState>(std::move(decoder), sample_rate, channel_count));
}
HwOpus::HwOpus() : ServiceFramework("hwopus") {
diff --git a/src/core/settings.cpp b/src/core/settings.cpp
index 2e232e1e7..6dd3139cc 100644
--- a/src/core/settings.cpp
+++ b/src/core/settings.cpp
@@ -91,7 +91,10 @@ void LogSettings() {
LogSetting("Renderer_UseResolutionFactor", Settings::values.resolution_factor);
LogSetting("Renderer_UseFrameLimit", Settings::values.use_frame_limit);
LogSetting("Renderer_FrameLimit", Settings::values.frame_limit);
+ LogSetting("Renderer_UseDiskShaderCache", Settings::values.use_disk_shader_cache);
LogSetting("Renderer_UseAccurateGpuEmulation", Settings::values.use_accurate_gpu_emulation);
+ LogSetting("Renderer_UseAsynchronousGpuEmulation",
+ Settings::values.use_asynchronous_gpu_emulation);
LogSetting("Audio_OutputEngine", Settings::values.sink_id);
LogSetting("Audio_EnableAudioStretching", Settings::values.enable_audio_stretching);
LogSetting("Audio_OutputDevice", Settings::values.audio_device_id);
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index dac992d44..0c3038c52 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -80,6 +80,7 @@ add_library(video_core STATIC
shader/decode/hfma2.cpp
shader/decode/conversion.cpp
shader/decode/memory.cpp
+ shader/decode/texture.cpp
shader/decode/float_set_predicate.cpp
shader/decode/integer_set_predicate.cpp
shader/decode/half_set_predicate.cpp
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index c7eb15b6a..7f613370b 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -324,11 +324,11 @@ enum class TextureQueryType : u64 {
enum class TextureProcessMode : u64 {
None = 0,
- LZ = 1, // Unknown, appears to be the same as none.
+ LZ = 1, // Load LOD of zero.
LB = 2, // Load Bias.
- LL = 3, // Load LOD (LevelOfDetail)
- LBA = 6, // Load Bias. The A is unknown, does not appear to differ with LB
- LLA = 7 // Load LOD. The A is unknown, does not appear to differ with LL
+ LL = 3, // Load LOD.
+ LBA = 6, // Load Bias. The A is unknown, does not appear to differ with LB.
+ LLA = 7 // Load LOD. The A is unknown, does not appear to differ with LL.
};
enum class TextureMiscMode : u64 {
@@ -1445,6 +1445,7 @@ public:
Flow,
Synch,
Memory,
+ Texture,
FloatSet,
FloatSetPredicate,
IntegerSet,
@@ -1575,14 +1576,14 @@ private:
INST("1110111101010---", Id::ST_L, Type::Memory, "ST_L"),
INST("1110111011010---", Id::LDG, Type::Memory, "LDG"),
INST("1110111011011---", Id::STG, Type::Memory, "STG"),
- INST("110000----111---", Id::TEX, Type::Memory, "TEX"),
- INST("1101111101001---", Id::TXQ, Type::Memory, "TXQ"),
- INST("1101-00---------", Id::TEXS, Type::Memory, "TEXS"),
- INST("1101101---------", Id::TLDS, Type::Memory, "TLDS"),
- INST("110010----111---", Id::TLD4, Type::Memory, "TLD4"),
- INST("1101111100------", Id::TLD4S, Type::Memory, "TLD4S"),
- INST("110111110110----", Id::TMML_B, Type::Memory, "TMML_B"),
- INST("1101111101011---", Id::TMML, Type::Memory, "TMML"),
+ INST("110000----111---", Id::TEX, Type::Texture, "TEX"),
+ INST("1101111101001---", Id::TXQ, Type::Texture, "TXQ"),
+ INST("1101-00---------", Id::TEXS, Type::Texture, "TEXS"),
+ INST("1101101---------", Id::TLDS, Type::Texture, "TLDS"),
+ INST("110010----111---", Id::TLD4, Type::Texture, "TLD4"),
+ INST("1101111100------", Id::TLD4S, Type::Texture, "TLD4S"),
+ INST("110111110110----", Id::TMML_B, Type::Texture, "TMML_B"),
+ INST("1101111101011---", Id::TMML, Type::Texture, "TMML"),
INST("111000110000----", Id::EXIT, Type::Trivial, "EXIT"),
INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"),
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index e33848bc1..824863561 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -118,7 +118,7 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::Syst
glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &uniform_buffer_alignment);
- LOG_CRITICAL(Render_OpenGL, "Sync fixed function OpenGL state here!");
+ LOG_DEBUG(Render_OpenGL, "Sync fixed function OpenGL state here");
CheckExtensions();
}
@@ -177,7 +177,7 @@ GLuint RasterizerOpenGL::SetupVertexFormat() {
continue;
const auto& buffer = regs.vertex_array[attrib.buffer];
- LOG_TRACE(HW_GPU,
+ LOG_TRACE(Render_OpenGL,
"vertex attrib {}, count={}, size={}, type={}, offset={}, normalize={}",
index, attrib.ComponentCount(), attrib.SizeString(), attrib.TypeString(),
attrib.offset.Value(), attrib.IsNormalized());
@@ -343,9 +343,8 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
shader_program_manager->UseProgrammableFragmentShader(program_handle);
break;
default:
- LOG_CRITICAL(HW_GPU, "Unimplemented shader index={}, enable={}, offset=0x{:08X}", index,
- shader_config.enable.Value(), shader_config.offset);
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unimplemented shader index={}, enable={}, offset=0x{:08X}", index,
+ shader_config.enable.Value(), shader_config.offset);
}
const auto stage_enum = static_cast<Maxwell::ShaderStage>(stage);
@@ -793,7 +792,10 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
VideoCore::Surface::PixelFormatFromGPUPixelFormat(config.pixel_format)};
ASSERT_MSG(params.width == config.width, "Framebuffer width is different");
ASSERT_MSG(params.height == config.height, "Framebuffer height is different");
- ASSERT_MSG(params.pixel_format == pixel_format, "Framebuffer pixel_format is different");
+
+ if (params.pixel_format != pixel_format) {
+ LOG_WARNING(Render_OpenGL, "Framebuffer pixel_format is different");
+ }
screen_info.display_texture = surface->Texture().handle;
@@ -922,8 +924,8 @@ void RasterizerOpenGL::SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::Shader
size = buffer.size;
if (size > MaxConstbufferSize) {
- LOG_CRITICAL(HW_GPU, "indirect constbuffer size {} exceeds maximum {}", size,
- MaxConstbufferSize);
+ LOG_WARNING(Render_OpenGL, "Indirect constbuffer size {} exceeds maximum {}", size,
+ MaxConstbufferSize);
size = MaxConstbufferSize;
}
} else {
@@ -983,10 +985,9 @@ void RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, const Shader& s
texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc);
- Surface surface = res_cache.GetTextureSurface(texture, entry);
- if (surface != nullptr) {
+ if (Surface surface = res_cache.GetTextureSurface(texture, entry); surface) {
state.texture_units[current_bindpoint].texture =
- entry.IsArray() ? surface->TextureLayer().handle : surface->Texture().handle;
+ surface->Texture(entry.IsArray()).handle;
surface->UpdateSwizzle(texture.tic.x_source, texture.tic.y_source, texture.tic.z_source,
texture.tic.w_source);
} else {
@@ -1218,11 +1219,7 @@ void RasterizerOpenGL::SyncScissorTest(OpenGLState& current_state) {
void RasterizerOpenGL::SyncTransformFeedback() {
const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-
- if (regs.tfb_enabled != 0) {
- LOG_CRITICAL(Render_OpenGL, "Transform feedbacks are not implemented");
- UNREACHABLE();
- }
+ UNIMPLEMENTED_IF_MSG(regs.tfb_enabled != 0, "Transform feedbacks are not implemented");
}
void RasterizerOpenGL::SyncPointState() {
@@ -1242,12 +1239,8 @@ void RasterizerOpenGL::SyncPolygonOffset() {
void RasterizerOpenGL::CheckAlphaTests() {
const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-
- if (regs.alpha_test_enabled != 0 && regs.rt_control.count > 1) {
- LOG_CRITICAL(Render_OpenGL, "Alpha Testing is enabled with Multiple Render Targets, "
- "this behavior is undefined.");
- UNREACHABLE();
- }
+ UNIMPLEMENTED_IF_MSG(regs.alpha_test_enabled != 0 && regs.rt_control.count > 1,
+ "Alpha Testing is enabled with more than one rendertarget");
}
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 876698b37..e9eb6e921 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -400,6 +400,27 @@ static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType
return format;
}
+/// Returns the discrepant array target
+constexpr GLenum GetArrayDiscrepantTarget(SurfaceTarget target) {
+ switch (target) {
+ case SurfaceTarget::Texture1D:
+ return GL_TEXTURE_1D_ARRAY;
+ case SurfaceTarget::Texture2D:
+ return GL_TEXTURE_2D_ARRAY;
+ case SurfaceTarget::Texture3D:
+ return GL_NONE;
+ case SurfaceTarget::Texture1DArray:
+ return GL_TEXTURE_1D;
+ case SurfaceTarget::Texture2DArray:
+ return GL_TEXTURE_2D;
+ case SurfaceTarget::TextureCubemap:
+ return GL_TEXTURE_CUBE_MAP_ARRAY;
+ case SurfaceTarget::TextureCubeArray:
+ return GL_TEXTURE_CUBE_MAP;
+ }
+ return GL_NONE;
+}
+
Common::Rectangle<u32> SurfaceParams::GetRect(u32 mip_level) const {
u32 actual_height{std::max(1U, unaligned_height >> mip_level)};
if (IsPixelFormatASTC(pixel_format)) {
@@ -795,20 +816,22 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle,
glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
}
-void CachedSurface::EnsureTextureView() {
- if (texture_view.handle != 0)
+void CachedSurface::EnsureTextureDiscrepantView() {
+ if (discrepant_view.handle != 0)
return;
- const GLenum target{TargetLayer()};
+ const GLenum target{GetArrayDiscrepantTarget(params.target)};
+ ASSERT(target != GL_NONE);
+
const GLuint num_layers{target == GL_TEXTURE_CUBE_MAP_ARRAY ? 6u : 1u};
constexpr GLuint min_layer = 0;
constexpr GLuint min_level = 0;
- glGenTextures(1, &texture_view.handle);
- glTextureView(texture_view.handle, target, texture.handle, gl_internal_format, min_level,
+ glGenTextures(1, &discrepant_view.handle);
+ glTextureView(discrepant_view.handle, target, texture.handle, gl_internal_format, min_level,
params.max_mip_level, min_layer, num_layers);
- ApplyTextureDefaults(texture_view.handle, params.max_mip_level);
- glTextureParameteriv(texture_view.handle, GL_TEXTURE_SWIZZLE_RGBA,
+ ApplyTextureDefaults(discrepant_view.handle, params.max_mip_level);
+ glTextureParameteriv(discrepant_view.handle, GL_TEXTURE_SWIZZLE_RGBA,
reinterpret_cast<const GLint*>(swizzle.data()));
}
@@ -834,8 +857,8 @@ void CachedSurface::UpdateSwizzle(Tegra::Texture::SwizzleSource swizzle_x,
swizzle = {new_x, new_y, new_z, new_w};
const auto swizzle_data = reinterpret_cast<const GLint*>(swizzle.data());
glTextureParameteriv(texture.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
- if (texture_view.handle != 0) {
- glTextureParameteriv(texture_view.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
+ if (discrepant_view.handle != 0) {
+ glTextureParameteriv(discrepant_view.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
}
}
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index 797bbdc9c..9cf6f50be 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -367,31 +367,19 @@ public:
return texture;
}
- const OGLTexture& TextureLayer() {
- if (params.is_array) {
- return Texture();
+ const OGLTexture& Texture(bool as_array) {
+ if (params.is_array == as_array) {
+ return texture;
+ } else {
+ EnsureTextureDiscrepantView();
+ return discrepant_view;
}
- EnsureTextureView();
- return texture_view;
}
GLenum Target() const {
return gl_target;
}
- GLenum TargetLayer() const {
- using VideoCore::Surface::SurfaceTarget;
- switch (params.target) {
- case SurfaceTarget::Texture1D:
- return GL_TEXTURE_1D_ARRAY;
- case SurfaceTarget::Texture2D:
- return GL_TEXTURE_2D_ARRAY;
- case SurfaceTarget::TextureCubemap:
- return GL_TEXTURE_CUBE_MAP_ARRAY;
- }
- return Target();
- }
-
const SurfaceParams& GetSurfaceParams() const {
return params;
}
@@ -431,10 +419,10 @@ public:
private:
void UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle, GLuint draw_fb_handle);
- void EnsureTextureView();
+ void EnsureTextureDiscrepantView();
OGLTexture texture;
- OGLTexture texture_view;
+ OGLTexture discrepant_view;
std::vector<std::vector<u8>> gl_buffer;
SurfaceParams params{};
GLenum gl_target{};
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 72ff6ac6a..11d1169f0 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -5,7 +5,9 @@
#include <array>
#include <string>
#include <string_view>
+#include <utility>
#include <variant>
+#include <vector>
#include <fmt/format.h>
@@ -717,7 +719,7 @@ private:
}
std::string GenerateTexture(Operation operation, const std::string& func,
- bool is_extra_int = false) {
+ const std::vector<std::pair<Type, Node>>& extras) {
constexpr std::array<const char*, 4> coord_constructors = {"float", "vec2", "vec3", "vec4"};
const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
@@ -738,36 +740,47 @@ private:
expr += Visit(operation[i]);
const std::size_t next = i + 1;
- if (next < count || has_array || has_shadow)
+ if (next < count)
expr += ", ";
}
if (has_array) {
- expr += "float(ftoi(" + Visit(meta->array) + "))";
+ expr += ", float(ftoi(" + Visit(meta->array) + "))";
}
if (has_shadow) {
- if (has_array)
- expr += ", ";
- expr += Visit(meta->depth_compare);
+ expr += ", " + Visit(meta->depth_compare);
}
expr += ')';
- for (const Node extra : meta->extras) {
+ for (const auto& extra_pair : extras) {
+ const auto [type, operand] = extra_pair;
+ if (operand == nullptr) {
+ continue;
+ }
expr += ", ";
- if (is_extra_int) {
- if (const auto immediate = std::get_if<ImmediateNode>(extra)) {
+
+ switch (type) {
+ case Type::Int:
+ if (const auto immediate = std::get_if<ImmediateNode>(operand)) {
// Inline the string as an immediate integer in GLSL (some extra arguments are
// required to be constant)
expr += std::to_string(static_cast<s32>(immediate->GetValue()));
} else {
- expr += "ftoi(" + Visit(extra) + ')';
+ expr += "ftoi(" + Visit(operand) + ')';
}
- } else {
- expr += Visit(extra);
+ break;
+ case Type::Float:
+ expr += Visit(operand);
+ break;
+ default: {
+ const auto type_int = static_cast<u32>(type);
+ UNIMPLEMENTED_MSG("Unimplemented extra type={}", type_int);
+ expr += '0';
+ break;
+ }
}
}
- expr += ')';
- return expr;
+ return expr + ')';
}
std::string Assign(Operation operation) {
@@ -1146,7 +1159,7 @@ private:
const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
ASSERT(meta);
- std::string expr = GenerateTexture(operation, "texture");
+ std::string expr = GenerateTexture(operation, "texture", {{Type::Float, meta->bias}});
if (meta->sampler.IsShadow()) {
expr = "vec4(" + expr + ')';
}
@@ -1157,7 +1170,7 @@ private:
const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
ASSERT(meta);
- std::string expr = GenerateTexture(operation, "textureLod");
+ std::string expr = GenerateTexture(operation, "textureLod", {{Type::Float, meta->lod}});
if (meta->sampler.IsShadow()) {
expr = "vec4(" + expr + ')';
}
@@ -1168,7 +1181,8 @@ private:
const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
ASSERT(meta);
- return GenerateTexture(operation, "textureGather", !meta->sampler.IsShadow()) +
+ const auto type = meta->sampler.IsShadow() ? Type::Float : Type::Int;
+ return GenerateTexture(operation, "textureGather", {{type, meta->component}}) +
GetSwizzle(meta->element);
}
@@ -1197,8 +1211,8 @@ private:
ASSERT(meta);
if (meta->element < 2) {
- return "itof(int((" + GenerateTexture(operation, "textureQueryLod") + " * vec2(256))" +
- GetSwizzle(meta->element) + "))";
+ return "itof(int((" + GenerateTexture(operation, "textureQueryLod", {}) +
+ " * vec2(256))" + GetSwizzle(meta->element) + "))";
}
return "0";
}
@@ -1224,9 +1238,9 @@ private:
else if (next < count)
expr += ", ";
}
- for (std::size_t i = 0; i < meta->extras.size(); ++i) {
+ if (meta->lod) {
expr += ", ";
- expr += CastOperand(Visit(meta->extras.at(i)), Type::Int);
+ expr += CastOperand(Visit(meta->lod), Type::Int);
}
expr += ')';
diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp
index 740ac3118..e4c438792 100644
--- a/src/video_core/shader/decode.cpp
+++ b/src/video_core/shader/decode.cpp
@@ -165,6 +165,7 @@ u32 ShaderIR::DecodeInstr(NodeBlock& bb, u32 pc) {
{OpCode::Type::Hfma2, &ShaderIR::DecodeHfma2},
{OpCode::Type::Conversion, &ShaderIR::DecodeConversion},
{OpCode::Type::Memory, &ShaderIR::DecodeMemory},
+ {OpCode::Type::Texture, &ShaderIR::DecodeTexture},
{OpCode::Type::FloatSetPredicate, &ShaderIR::DecodeFloatSetPredicate},
{OpCode::Type::IntegerSetPredicate, &ShaderIR::DecodeIntegerSetPredicate},
{OpCode::Type::HalfSetPredicate, &ShaderIR::DecodeHalfSetPredicate},
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index 38f01ca50..ea3c71eed 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/src/video_core/shader/decode/memory.cpp
@@ -17,24 +17,6 @@ using Tegra::Shader::Attribute;
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
using Tegra::Shader::Register;
-using Tegra::Shader::TextureMiscMode;
-using Tegra::Shader::TextureProcessMode;
-using Tegra::Shader::TextureType;
-
-static std::size_t GetCoordCount(TextureType texture_type) {
- switch (texture_type) {
- case TextureType::Texture1D:
- return 1;
- case TextureType::Texture2D:
- return 2;
- case TextureType::Texture3D:
- case TextureType::TextureCube:
- return 3;
- default:
- UNIMPLEMENTED_MSG("Unhandled texture type: {}", static_cast<u32>(texture_type));
- return 0;
- }
-}
u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
@@ -247,194 +229,6 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
}
break;
}
- case OpCode::Id::TEX: {
- UNIMPLEMENTED_IF_MSG(instr.tex.UsesMiscMode(TextureMiscMode::AOFFI),
- "AOFFI is not implemented");
-
- if (instr.tex.UsesMiscMode(TextureMiscMode::NODEP)) {
- LOG_WARNING(HW_GPU, "TEX.NODEP implementation is incomplete");
- }
-
- const TextureType texture_type{instr.tex.texture_type};
- const bool is_array = instr.tex.array != 0;
- const bool depth_compare = instr.tex.UsesMiscMode(TextureMiscMode::DC);
- const auto process_mode = instr.tex.GetTextureProcessMode();
- WriteTexInstructionFloat(
- bb, instr, GetTexCode(instr, texture_type, process_mode, depth_compare, is_array));
- break;
- }
- case OpCode::Id::TEXS: {
- const TextureType texture_type{instr.texs.GetTextureType()};
- const bool is_array{instr.texs.IsArrayTexture()};
- const bool depth_compare = instr.texs.UsesMiscMode(TextureMiscMode::DC);
- const auto process_mode = instr.texs.GetTextureProcessMode();
-
- if (instr.texs.UsesMiscMode(TextureMiscMode::NODEP)) {
- LOG_WARNING(HW_GPU, "TEXS.NODEP implementation is incomplete");
- }
-
- const Node4 components =
- GetTexsCode(instr, texture_type, process_mode, depth_compare, is_array);
-
- if (instr.texs.fp32_flag) {
- WriteTexsInstructionFloat(bb, instr, components);
- } else {
- WriteTexsInstructionHalfFloat(bb, instr, components);
- }
- break;
- }
- case OpCode::Id::TLD4: {
- ASSERT(instr.tld4.array == 0);
- UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::AOFFI),
- "AOFFI is not implemented");
- UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::NDV),
- "NDV is not implemented");
- UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::PTP),
- "PTP is not implemented");
-
- if (instr.tld4.UsesMiscMode(TextureMiscMode::NODEP)) {
- LOG_WARNING(HW_GPU, "TLD4.NODEP implementation is incomplete");
- }
-
- const auto texture_type = instr.tld4.texture_type.Value();
- const bool depth_compare = instr.tld4.UsesMiscMode(TextureMiscMode::DC);
- const bool is_array = instr.tld4.array != 0;
- WriteTexInstructionFloat(bb, instr,
- GetTld4Code(instr, texture_type, depth_compare, is_array));
- break;
- }
- case OpCode::Id::TLD4S: {
- UNIMPLEMENTED_IF_MSG(instr.tld4s.UsesMiscMode(TextureMiscMode::AOFFI),
- "AOFFI is not implemented");
- if (instr.tld4s.UsesMiscMode(TextureMiscMode::NODEP)) {
- LOG_WARNING(HW_GPU, "TLD4S.NODEP implementation is incomplete");
- }
-
- const bool depth_compare = instr.tld4s.UsesMiscMode(TextureMiscMode::DC);
- const Node op_a = GetRegister(instr.gpr8);
- const Node op_b = GetRegister(instr.gpr20);
-
- // TODO(Subv): Figure out how the sampler type is encoded in the TLD4S instruction.
- std::vector<Node> coords;
- if (depth_compare) {
- // Note: TLD4S coordinate encoding works just like TEXS's
- const Node op_y = GetRegister(instr.gpr8.Value() + 1);
- coords.push_back(op_a);
- coords.push_back(op_y);
- coords.push_back(op_b);
- } else {
- coords.push_back(op_a);
- coords.push_back(op_b);
- }
- std::vector<Node> extras;
- extras.push_back(Immediate(static_cast<u32>(instr.tld4s.component)));
-
- const auto& sampler =
- GetSampler(instr.sampler, TextureType::Texture2D, false, depth_compare);
-
- Node4 values;
- for (u32 element = 0; element < values.size(); ++element) {
- auto coords_copy = coords;
- MetaTexture meta{sampler, {}, {}, extras, element};
- values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
- }
-
- WriteTexsInstructionFloat(bb, instr, values);
- break;
- }
- case OpCode::Id::TXQ: {
- if (instr.txq.UsesMiscMode(TextureMiscMode::NODEP)) {
- LOG_WARNING(HW_GPU, "TXQ.NODEP implementation is incomplete");
- }
-
- // TODO: The new commits on the texture refactor, change the way samplers work.
- // Sadly, not all texture instructions specify the type of texture their sampler
- // uses. This must be fixed at a later instance.
- const auto& sampler =
- GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false);
-
- u32 indexer = 0;
- switch (instr.txq.query_type) {
- case Tegra::Shader::TextureQueryType::Dimension: {
- for (u32 element = 0; element < 4; ++element) {
- if (!instr.txq.IsComponentEnabled(element)) {
- continue;
- }
- MetaTexture meta{sampler, {}, {}, {}, element};
- const Node value =
- Operation(OperationCode::TextureQueryDimensions, meta, GetRegister(instr.gpr8));
- SetTemporal(bb, indexer++, value);
- }
- for (u32 i = 0; i < indexer; ++i) {
- SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
- }
- break;
- }
- default:
- UNIMPLEMENTED_MSG("Unhandled texture query type: {}",
- static_cast<u32>(instr.txq.query_type.Value()));
- }
- break;
- }
- case OpCode::Id::TMML: {
- UNIMPLEMENTED_IF_MSG(instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
- "NDV is not implemented");
-
- if (instr.tmml.UsesMiscMode(TextureMiscMode::NODEP)) {
- LOG_WARNING(HW_GPU, "TMML.NODEP implementation is incomplete");
- }
-
- auto texture_type = instr.tmml.texture_type.Value();
- const bool is_array = instr.tmml.array != 0;
- const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
-
- std::vector<Node> coords;
-
- // TODO: Add coordinates for different samplers once other texture types are implemented.
- switch (texture_type) {
- case TextureType::Texture1D:
- coords.push_back(GetRegister(instr.gpr8));
- break;
- case TextureType::Texture2D:
- coords.push_back(GetRegister(instr.gpr8.Value() + 0));
- coords.push_back(GetRegister(instr.gpr8.Value() + 1));
- break;
- default:
- UNIMPLEMENTED_MSG("Unhandled texture type {}", static_cast<u32>(texture_type));
-
- // Fallback to interpreting as a 2D texture for now
- coords.push_back(GetRegister(instr.gpr8.Value() + 0));
- coords.push_back(GetRegister(instr.gpr8.Value() + 1));
- texture_type = TextureType::Texture2D;
- }
-
- for (u32 element = 0; element < 2; ++element) {
- auto params = coords;
- MetaTexture meta{sampler, {}, {}, {}, element};
- const Node value = Operation(OperationCode::TextureQueryLod, meta, std::move(params));
- SetTemporal(bb, element, value);
- }
- for (u32 element = 0; element < 2; ++element) {
- SetRegister(bb, instr.gpr0.Value() + element, GetTemporal(element));
- }
-
- break;
- }
- case OpCode::Id::TLDS: {
- const Tegra::Shader::TextureType texture_type{instr.tlds.GetTextureType()};
- const bool is_array{instr.tlds.IsArrayTexture()};
-
- UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::AOFFI),
- "AOFFI is not implemented");
- UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::MZ), "MZ is not implemented");
-
- if (instr.tlds.UsesMiscMode(TextureMiscMode::NODEP)) {
- LOG_WARNING(HW_GPU, "TLDS.NODEP implementation is incomplete");
- }
-
- WriteTexsInstructionFloat(bb, instr, GetTldsCode(instr, texture_type, is_array));
- break;
- }
default:
UNIMPLEMENTED_MSG("Unhandled memory instruction: {}", opcode->get().GetName());
}
@@ -442,291 +236,4 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
return pc;
}
-const Sampler& ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, TextureType type,
- bool is_array, bool is_shadow) {
- const auto offset = static_cast<std::size_t>(sampler.index.Value());
-
- // If this sampler has already been used, return the existing mapping.
- const auto itr =
- std::find_if(used_samplers.begin(), used_samplers.end(),
- [&](const Sampler& entry) { return entry.GetOffset() == offset; });
- if (itr != used_samplers.end()) {
- ASSERT(itr->GetType() == type && itr->IsArray() == is_array &&
- itr->IsShadow() == is_shadow);
- return *itr;
- }
-
- // Otherwise create a new mapping for this sampler
- const std::size_t next_index = used_samplers.size();
- const Sampler entry{offset, next_index, type, is_array, is_shadow};
- return *used_samplers.emplace(entry).first;
-}
-
-void ShaderIR::WriteTexInstructionFloat(NodeBlock& bb, Instruction instr, const Node4& components) {
- u32 dest_elem = 0;
- for (u32 elem = 0; elem < 4; ++elem) {
- if (!instr.tex.IsComponentEnabled(elem)) {
- // Skip disabled components
- continue;
- }
- SetTemporal(bb, dest_elem++, components[elem]);
- }
- // After writing values in temporals, move them to the real registers
- for (u32 i = 0; i < dest_elem; ++i) {
- SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
- }
-}
-
-void ShaderIR::WriteTexsInstructionFloat(NodeBlock& bb, Instruction instr,
- const Node4& components) {
- // TEXS has two destination registers and a swizzle. The first two elements in the swizzle
- // go into gpr0+0 and gpr0+1, and the rest goes into gpr28+0 and gpr28+1
-
- u32 dest_elem = 0;
- for (u32 component = 0; component < 4; ++component) {
- if (!instr.texs.IsComponentEnabled(component))
- continue;
- SetTemporal(bb, dest_elem++, components[component]);
- }
-
- for (u32 i = 0; i < dest_elem; ++i) {
- if (i < 2) {
- // Write the first two swizzle components to gpr0 and gpr0+1
- SetRegister(bb, instr.gpr0.Value() + i % 2, GetTemporal(i));
- } else {
- ASSERT(instr.texs.HasTwoDestinations());
- // Write the rest of the swizzle components to gpr28 and gpr28+1
- SetRegister(bb, instr.gpr28.Value() + i % 2, GetTemporal(i));
- }
- }
-}
-
-void ShaderIR::WriteTexsInstructionHalfFloat(NodeBlock& bb, Instruction instr,
- const Node4& components) {
- // TEXS.F16 destionation registers are packed in two registers in pairs (just like any half
- // float instruction).
-
- Node4 values;
- u32 dest_elem = 0;
- for (u32 component = 0; component < 4; ++component) {
- if (!instr.texs.IsComponentEnabled(component))
- continue;
- values[dest_elem++] = components[component];
- }
- if (dest_elem == 0)
- return;
-
- std::generate(values.begin() + dest_elem, values.end(), [&]() { return Immediate(0); });
-
- const Node first_value = Operation(OperationCode::HPack2, values[0], values[1]);
- if (dest_elem <= 2) {
- SetRegister(bb, instr.gpr0, first_value);
- return;
- }
-
- SetTemporal(bb, 0, first_value);
- SetTemporal(bb, 1, Operation(OperationCode::HPack2, values[2], values[3]));
-
- SetRegister(bb, instr.gpr0, GetTemporal(0));
- SetRegister(bb, instr.gpr28, GetTemporal(1));
-}
-
-Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
- TextureProcessMode process_mode, std::vector<Node> coords,
- Node array, Node depth_compare, u32 bias_offset) {
- const bool is_array = array;
- const bool is_shadow = depth_compare;
-
- UNIMPLEMENTED_IF_MSG((texture_type == TextureType::Texture3D && (is_array || is_shadow)) ||
- (texture_type == TextureType::TextureCube && is_array && is_shadow),
- "This method is not supported.");
-
- const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, is_shadow);
-
- const bool lod_needed = process_mode == TextureProcessMode::LZ ||
- process_mode == TextureProcessMode::LL ||
- process_mode == TextureProcessMode::LLA;
-
- // LOD selection (either via bias or explicit textureLod) not supported in GL for
- // sampler2DArrayShadow and samplerCubeArrayShadow.
- const bool gl_lod_supported =
- !((texture_type == Tegra::Shader::TextureType::Texture2D && is_array && is_shadow) ||
- (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && is_shadow));
-
- const OperationCode read_method =
- lod_needed && gl_lod_supported ? OperationCode::TextureLod : OperationCode::Texture;
-
- UNIMPLEMENTED_IF(process_mode != TextureProcessMode::None && !gl_lod_supported);
-
- std::vector<Node> extras;
- if (process_mode != TextureProcessMode::None && gl_lod_supported) {
- if (process_mode == TextureProcessMode::LZ) {
- extras.push_back(Immediate(0.0f));
- } else {
- // If present, lod or bias are always stored in the register indexed by the gpr20
- // field with an offset depending on the usage of the other registers
- extras.push_back(GetRegister(instr.gpr20.Value() + bias_offset));
- }
- }
-
- Node4 values;
- for (u32 element = 0; element < values.size(); ++element) {
- auto copy_coords = coords;
- MetaTexture meta{sampler, array, depth_compare, extras, element};
- values[element] = Operation(read_method, meta, std::move(copy_coords));
- }
-
- return values;
-}
-
-Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
- TextureProcessMode process_mode, bool depth_compare, bool is_array) {
- const bool lod_bias_enabled =
- (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
-
- const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
- texture_type, depth_compare, is_array, lod_bias_enabled, 4, 5);
- // If enabled arrays index is always stored in the gpr8 field
- const u64 array_register = instr.gpr8.Value();
- // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
- const u64 coord_register = array_register + (is_array ? 1 : 0);
-
- std::vector<Node> coords;
- for (std::size_t i = 0; i < coord_count; ++i) {
- coords.push_back(GetRegister(coord_register + i));
- }
- // 1D.DC in OpenGL the 2nd component is ignored.
- if (depth_compare && !is_array && texture_type == TextureType::Texture1D) {
- coords.push_back(Immediate(0.0f));
- }
-
- const Node array = is_array ? GetRegister(array_register) : nullptr;
-
- Node dc{};
- if (depth_compare) {
- // Depth is always stored in the register signaled by gpr20 or in the next register if lod
- // or bias are used
- const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
- dc = GetRegister(depth_register);
- }
-
- return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0);
-}
-
-Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
- TextureProcessMode process_mode, bool depth_compare, bool is_array) {
- const bool lod_bias_enabled =
- (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
-
- const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
- texture_type, depth_compare, is_array, lod_bias_enabled, 4, 4);
- // If enabled arrays index is always stored in the gpr8 field
- const u64 array_register = instr.gpr8.Value();
- // First coordinate index is stored in gpr8 field or (gpr8 + 1) when arrays are used
- const u64 coord_register = array_register + (is_array ? 1 : 0);
- const u64 last_coord_register =
- (is_array || !(lod_bias_enabled || depth_compare) || (coord_count > 2))
- ? static_cast<u64>(instr.gpr20.Value())
- : coord_register + 1;
- const u32 bias_offset = coord_count > 2 ? 1 : 0;
-
- std::vector<Node> coords;
- for (std::size_t i = 0; i < coord_count; ++i) {
- const bool last = (i == (coord_count - 1)) && (coord_count > 1);
- coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
- }
-
- const Node array = is_array ? GetRegister(array_register) : nullptr;
-
- Node dc{};
- if (depth_compare) {
- // Depth is always stored in the register signaled by gpr20 or in the next register if lod
- // or bias are used
- const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
- dc = GetRegister(depth_register);
- }
-
- return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset);
-}
-
-Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool depth_compare,
- bool is_array) {
- const std::size_t coord_count = GetCoordCount(texture_type);
- const std::size_t total_coord_count = coord_count + (is_array ? 1 : 0);
- const std::size_t total_reg_count = total_coord_count + (depth_compare ? 1 : 0);
-
- // If enabled arrays index is always stored in the gpr8 field
- const u64 array_register = instr.gpr8.Value();
- // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
- const u64 coord_register = array_register + (is_array ? 1 : 0);
-
- std::vector<Node> coords;
- for (size_t i = 0; i < coord_count; ++i)
- coords.push_back(GetRegister(coord_register + i));
-
- const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
-
- Node4 values;
- for (u32 element = 0; element < values.size(); ++element) {
- auto coords_copy = coords;
- MetaTexture meta{sampler, GetRegister(array_register), {}, {}, element};
- values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
- }
-
- return values;
-}
-
-Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is_array) {
- const std::size_t type_coord_count = GetCoordCount(texture_type);
- const bool lod_enabled = instr.tlds.GetTextureProcessMode() == TextureProcessMode::LL;
-
- // If enabled arrays index is always stored in the gpr8 field
- const u64 array_register = instr.gpr8.Value();
- // if is array gpr20 is used
- const u64 coord_register = is_array ? instr.gpr20.Value() : instr.gpr8.Value();
-
- const u64 last_coord_register =
- ((type_coord_count > 2) || (type_coord_count == 2 && !lod_enabled)) && !is_array
- ? static_cast<u64>(instr.gpr20.Value())
- : coord_register + 1;
-
- std::vector<Node> coords;
- for (std::size_t i = 0; i < type_coord_count; ++i) {
- const bool last = (i == (type_coord_count - 1)) && (type_coord_count > 1);
- coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
- }
-
- const Node array = is_array ? GetRegister(array_register) : nullptr;
- // When lod is used always is in gpr20
- const Node lod = lod_enabled ? GetRegister(instr.gpr20) : Immediate(0);
-
- const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
-
- Node4 values;
- for (u32 element = 0; element < values.size(); ++element) {
- auto coords_copy = coords;
- MetaTexture meta{sampler, array, {}, {lod}, element};
- values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy));
- }
- return values;
-}
-
-std::tuple<std::size_t, std::size_t> ShaderIR::ValidateAndGetCoordinateElement(
- TextureType texture_type, bool depth_compare, bool is_array, bool lod_bias_enabled,
- std::size_t max_coords, std::size_t max_inputs) {
- const std::size_t coord_count = GetCoordCount(texture_type);
-
- std::size_t total_coord_count = coord_count + (is_array ? 1 : 0) + (depth_compare ? 1 : 0);
- const std::size_t total_reg_count = total_coord_count + (lod_bias_enabled ? 1 : 0);
- if (total_coord_count > max_coords || total_reg_count > max_inputs) {
- UNIMPLEMENTED_MSG("Unsupported Texture operation");
- total_coord_count = std::min(total_coord_count, max_coords);
- }
- // 1D.DC OpenGL is using a vec3 but 2nd component is ignored later.
- total_coord_count +=
- (depth_compare && !is_array && texture_type == TextureType::Texture1D) ? 1 : 0;
-
- return {coord_count, total_coord_count};
-}
-
} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
new file mode 100644
index 000000000..a99ae19bf
--- /dev/null
+++ b/src/video_core/shader/decode/texture.cpp
@@ -0,0 +1,534 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <vector>
+#include <fmt/format.h>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Register;
+using Tegra::Shader::TextureMiscMode;
+using Tegra::Shader::TextureProcessMode;
+using Tegra::Shader::TextureType;
+
+static std::size_t GetCoordCount(TextureType texture_type) {
+ switch (texture_type) {
+ case TextureType::Texture1D:
+ return 1;
+ case TextureType::Texture2D:
+ return 2;
+ case TextureType::Texture3D:
+ case TextureType::TextureCube:
+ return 3;
+ default:
+ UNIMPLEMENTED_MSG("Unhandled texture type: {}", static_cast<u32>(texture_type));
+ return 0;
+ }
+}
+
+u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
+ const Instruction instr = {program_code[pc]};
+ const auto opcode = OpCode::Decode(instr);
+
+ switch (opcode->get().GetId()) {
+ case OpCode::Id::TEX: {
+ UNIMPLEMENTED_IF_MSG(instr.tex.UsesMiscMode(TextureMiscMode::AOFFI),
+ "AOFFI is not implemented");
+
+ if (instr.tex.UsesMiscMode(TextureMiscMode::NODEP)) {
+ LOG_WARNING(HW_GPU, "TEX.NODEP implementation is incomplete");
+ }
+
+ const TextureType texture_type{instr.tex.texture_type};
+ const bool is_array = instr.tex.array != 0;
+ const bool depth_compare = instr.tex.UsesMiscMode(TextureMiscMode::DC);
+ const auto process_mode = instr.tex.GetTextureProcessMode();
+ WriteTexInstructionFloat(
+ bb, instr, GetTexCode(instr, texture_type, process_mode, depth_compare, is_array));
+ break;
+ }
+ case OpCode::Id::TEXS: {
+ const TextureType texture_type{instr.texs.GetTextureType()};
+ const bool is_array{instr.texs.IsArrayTexture()};
+ const bool depth_compare = instr.texs.UsesMiscMode(TextureMiscMode::DC);
+ const auto process_mode = instr.texs.GetTextureProcessMode();
+
+ if (instr.texs.UsesMiscMode(TextureMiscMode::NODEP)) {
+ LOG_WARNING(HW_GPU, "TEXS.NODEP implementation is incomplete");
+ }
+
+ const Node4 components =
+ GetTexsCode(instr, texture_type, process_mode, depth_compare, is_array);
+
+ if (instr.texs.fp32_flag) {
+ WriteTexsInstructionFloat(bb, instr, components);
+ } else {
+ WriteTexsInstructionHalfFloat(bb, instr, components);
+ }
+ break;
+ }
+ case OpCode::Id::TLD4: {
+ ASSERT(instr.tld4.array == 0);
+ UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::AOFFI),
+ "AOFFI is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::NDV),
+ "NDV is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::PTP),
+ "PTP is not implemented");
+
+ if (instr.tld4.UsesMiscMode(TextureMiscMode::NODEP)) {
+ LOG_WARNING(HW_GPU, "TLD4.NODEP implementation is incomplete");
+ }
+
+ const auto texture_type = instr.tld4.texture_type.Value();
+ const bool depth_compare = instr.tld4.UsesMiscMode(TextureMiscMode::DC);
+ const bool is_array = instr.tld4.array != 0;
+ WriteTexInstructionFloat(bb, instr,
+ GetTld4Code(instr, texture_type, depth_compare, is_array));
+ break;
+ }
+ case OpCode::Id::TLD4S: {
+ UNIMPLEMENTED_IF_MSG(instr.tld4s.UsesMiscMode(TextureMiscMode::AOFFI),
+ "AOFFI is not implemented");
+ if (instr.tld4s.UsesMiscMode(TextureMiscMode::NODEP)) {
+ LOG_WARNING(HW_GPU, "TLD4S.NODEP implementation is incomplete");
+ }
+
+ const bool depth_compare = instr.tld4s.UsesMiscMode(TextureMiscMode::DC);
+ const Node op_a = GetRegister(instr.gpr8);
+ const Node op_b = GetRegister(instr.gpr20);
+
+ // TODO(Subv): Figure out how the sampler type is encoded in the TLD4S instruction.
+ std::vector<Node> coords;
+ if (depth_compare) {
+ // Note: TLD4S coordinate encoding works just like TEXS's
+ const Node op_y = GetRegister(instr.gpr8.Value() + 1);
+ coords.push_back(op_a);
+ coords.push_back(op_y);
+ coords.push_back(op_b);
+ } else {
+ coords.push_back(op_a);
+ coords.push_back(op_b);
+ }
+ const Node component = Immediate(static_cast<u32>(instr.tld4s.component));
+
+ const auto& sampler =
+ GetSampler(instr.sampler, TextureType::Texture2D, false, depth_compare);
+
+ Node4 values;
+ for (u32 element = 0; element < values.size(); ++element) {
+ auto coords_copy = coords;
+ MetaTexture meta{sampler, {}, {}, {}, {}, component, element};
+ values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
+ }
+
+ WriteTexsInstructionFloat(bb, instr, values);
+ break;
+ }
+ case OpCode::Id::TXQ: {
+ if (instr.txq.UsesMiscMode(TextureMiscMode::NODEP)) {
+ LOG_WARNING(HW_GPU, "TXQ.NODEP implementation is incomplete");
+ }
+
+ // TODO: The new commits on the texture refactor, change the way samplers work.
+ // Sadly, not all texture instructions specify the type of texture their sampler
+ // uses. This must be fixed at a later instance.
+ const auto& sampler =
+ GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false);
+
+ u32 indexer = 0;
+ switch (instr.txq.query_type) {
+ case Tegra::Shader::TextureQueryType::Dimension: {
+ for (u32 element = 0; element < 4; ++element) {
+ if (!instr.txq.IsComponentEnabled(element)) {
+ continue;
+ }
+ MetaTexture meta{sampler, {}, {}, {}, {}, {}, element};
+ const Node value =
+ Operation(OperationCode::TextureQueryDimensions, meta, GetRegister(instr.gpr8));
+ SetTemporal(bb, indexer++, value);
+ }
+ for (u32 i = 0; i < indexer; ++i) {
+ SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
+ }
+ break;
+ }
+ default:
+ UNIMPLEMENTED_MSG("Unhandled texture query type: {}",
+ static_cast<u32>(instr.txq.query_type.Value()));
+ }
+ break;
+ }
+ case OpCode::Id::TMML: {
+ UNIMPLEMENTED_IF_MSG(instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
+ "NDV is not implemented");
+
+ if (instr.tmml.UsesMiscMode(TextureMiscMode::NODEP)) {
+ LOG_WARNING(HW_GPU, "TMML.NODEP implementation is incomplete");
+ }
+
+ auto texture_type = instr.tmml.texture_type.Value();
+ const bool is_array = instr.tmml.array != 0;
+ const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
+
+ std::vector<Node> coords;
+
+ // TODO: Add coordinates for different samplers once other texture types are implemented.
+ switch (texture_type) {
+ case TextureType::Texture1D:
+ coords.push_back(GetRegister(instr.gpr8));
+ break;
+ case TextureType::Texture2D:
+ coords.push_back(GetRegister(instr.gpr8.Value() + 0));
+ coords.push_back(GetRegister(instr.gpr8.Value() + 1));
+ break;
+ default:
+ UNIMPLEMENTED_MSG("Unhandled texture type {}", static_cast<u32>(texture_type));
+
+ // Fallback to interpreting as a 2D texture for now
+ coords.push_back(GetRegister(instr.gpr8.Value() + 0));
+ coords.push_back(GetRegister(instr.gpr8.Value() + 1));
+ texture_type = TextureType::Texture2D;
+ }
+
+ for (u32 element = 0; element < 2; ++element) {
+ auto params = coords;
+ MetaTexture meta{sampler, {}, {}, {}, {}, {}, element};
+ const Node value = Operation(OperationCode::TextureQueryLod, meta, std::move(params));
+ SetTemporal(bb, element, value);
+ }
+ for (u32 element = 0; element < 2; ++element) {
+ SetRegister(bb, instr.gpr0.Value() + element, GetTemporal(element));
+ }
+
+ break;
+ }
+ case OpCode::Id::TLDS: {
+ const Tegra::Shader::TextureType texture_type{instr.tlds.GetTextureType()};
+ const bool is_array{instr.tlds.IsArrayTexture()};
+
+ UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::AOFFI),
+ "AOFFI is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::MZ), "MZ is not implemented");
+
+ if (instr.tlds.UsesMiscMode(TextureMiscMode::NODEP)) {
+ LOG_WARNING(HW_GPU, "TLDS.NODEP implementation is incomplete");
+ }
+
+ WriteTexsInstructionFloat(bb, instr, GetTldsCode(instr, texture_type, is_array));
+ break;
+ }
+ default:
+ UNIMPLEMENTED_MSG("Unhandled memory instruction: {}", opcode->get().GetName());
+ }
+
+ return pc;
+}
+
+const Sampler& ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, TextureType type,
+ bool is_array, bool is_shadow) {
+ const auto offset = static_cast<std::size_t>(sampler.index.Value());
+
+ // If this sampler has already been used, return the existing mapping.
+ const auto itr =
+ std::find_if(used_samplers.begin(), used_samplers.end(),
+ [&](const Sampler& entry) { return entry.GetOffset() == offset; });
+ if (itr != used_samplers.end()) {
+ ASSERT(itr->GetType() == type && itr->IsArray() == is_array &&
+ itr->IsShadow() == is_shadow);
+ return *itr;
+ }
+
+ // Otherwise create a new mapping for this sampler
+ const std::size_t next_index = used_samplers.size();
+ const Sampler entry{offset, next_index, type, is_array, is_shadow};
+ return *used_samplers.emplace(entry).first;
+}
+
+void ShaderIR::WriteTexInstructionFloat(NodeBlock& bb, Instruction instr, const Node4& components) {
+ u32 dest_elem = 0;
+ for (u32 elem = 0; elem < 4; ++elem) {
+ if (!instr.tex.IsComponentEnabled(elem)) {
+ // Skip disabled components
+ continue;
+ }
+ SetTemporal(bb, dest_elem++, components[elem]);
+ }
+ // After writing values in temporals, move them to the real registers
+ for (u32 i = 0; i < dest_elem; ++i) {
+ SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
+ }
+}
+
+void ShaderIR::WriteTexsInstructionFloat(NodeBlock& bb, Instruction instr,
+ const Node4& components) {
+ // TEXS has two destination registers and a swizzle. The first two elements in the swizzle
+ // go into gpr0+0 and gpr0+1, and the rest goes into gpr28+0 and gpr28+1
+
+ u32 dest_elem = 0;
+ for (u32 component = 0; component < 4; ++component) {
+ if (!instr.texs.IsComponentEnabled(component))
+ continue;
+ SetTemporal(bb, dest_elem++, components[component]);
+ }
+
+ for (u32 i = 0; i < dest_elem; ++i) {
+ if (i < 2) {
+ // Write the first two swizzle components to gpr0 and gpr0+1
+ SetRegister(bb, instr.gpr0.Value() + i % 2, GetTemporal(i));
+ } else {
+ ASSERT(instr.texs.HasTwoDestinations());
+ // Write the rest of the swizzle components to gpr28 and gpr28+1
+ SetRegister(bb, instr.gpr28.Value() + i % 2, GetTemporal(i));
+ }
+ }
+}
+
+void ShaderIR::WriteTexsInstructionHalfFloat(NodeBlock& bb, Instruction instr,
+ const Node4& components) {
+ // TEXS.F16 destionation registers are packed in two registers in pairs (just like any half
+ // float instruction).
+
+ Node4 values;
+ u32 dest_elem = 0;
+ for (u32 component = 0; component < 4; ++component) {
+ if (!instr.texs.IsComponentEnabled(component))
+ continue;
+ values[dest_elem++] = components[component];
+ }
+ if (dest_elem == 0)
+ return;
+
+ std::generate(values.begin() + dest_elem, values.end(), [&]() { return Immediate(0); });
+
+ const Node first_value = Operation(OperationCode::HPack2, values[0], values[1]);
+ if (dest_elem <= 2) {
+ SetRegister(bb, instr.gpr0, first_value);
+ return;
+ }
+
+ SetTemporal(bb, 0, first_value);
+ SetTemporal(bb, 1, Operation(OperationCode::HPack2, values[2], values[3]));
+
+ SetRegister(bb, instr.gpr0, GetTemporal(0));
+ SetRegister(bb, instr.gpr28, GetTemporal(1));
+}
+
+Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
+ TextureProcessMode process_mode, std::vector<Node> coords,
+ Node array, Node depth_compare, u32 bias_offset) {
+ const bool is_array = array;
+ const bool is_shadow = depth_compare;
+
+ UNIMPLEMENTED_IF_MSG((texture_type == TextureType::Texture3D && (is_array || is_shadow)) ||
+ (texture_type == TextureType::TextureCube && is_array && is_shadow),
+ "This method is not supported.");
+
+ const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, is_shadow);
+
+ const bool lod_needed = process_mode == TextureProcessMode::LZ ||
+ process_mode == TextureProcessMode::LL ||
+ process_mode == TextureProcessMode::LLA;
+
+ // LOD selection (either via bias or explicit textureLod) not supported in GL for
+ // sampler2DArrayShadow and samplerCubeArrayShadow.
+ const bool gl_lod_supported =
+ !((texture_type == Tegra::Shader::TextureType::Texture2D && is_array && is_shadow) ||
+ (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && is_shadow));
+
+ const OperationCode read_method =
+ (lod_needed && gl_lod_supported) ? OperationCode::TextureLod : OperationCode::Texture;
+
+ UNIMPLEMENTED_IF(process_mode != TextureProcessMode::None && !gl_lod_supported);
+
+ Node bias = {};
+ Node lod = {};
+ if (process_mode != TextureProcessMode::None && gl_lod_supported) {
+ switch (process_mode) {
+ case TextureProcessMode::LZ:
+ lod = Immediate(0.0f);
+ break;
+ case TextureProcessMode::LB:
+ // If present, lod or bias are always stored in the register indexed by the gpr20
+ // field with an offset depending on the usage of the other registers
+ bias = GetRegister(instr.gpr20.Value() + bias_offset);
+ break;
+ case TextureProcessMode::LL:
+ lod = GetRegister(instr.gpr20.Value() + bias_offset);
+ break;
+ default:
+ UNIMPLEMENTED_MSG("Unimplemented process mode={}", static_cast<u32>(process_mode));
+ break;
+ }
+ }
+
+ Node4 values;
+ for (u32 element = 0; element < values.size(); ++element) {
+ auto copy_coords = coords;
+ MetaTexture meta{sampler, array, depth_compare, bias, lod, {}, element};
+ values[element] = Operation(read_method, meta, std::move(copy_coords));
+ }
+
+ return values;
+}
+
+Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
+ TextureProcessMode process_mode, bool depth_compare, bool is_array) {
+ const bool lod_bias_enabled =
+ (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
+
+ const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
+ texture_type, depth_compare, is_array, lod_bias_enabled, 4, 5);
+ // If enabled arrays index is always stored in the gpr8 field
+ const u64 array_register = instr.gpr8.Value();
+ // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
+ const u64 coord_register = array_register + (is_array ? 1 : 0);
+
+ std::vector<Node> coords;
+ for (std::size_t i = 0; i < coord_count; ++i) {
+ coords.push_back(GetRegister(coord_register + i));
+ }
+ // 1D.DC in OpenGL the 2nd component is ignored.
+ if (depth_compare && !is_array && texture_type == TextureType::Texture1D) {
+ coords.push_back(Immediate(0.0f));
+ }
+
+ const Node array = is_array ? GetRegister(array_register) : nullptr;
+
+ Node dc{};
+ if (depth_compare) {
+ // Depth is always stored in the register signaled by gpr20 or in the next register if lod
+ // or bias are used
+ const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
+ dc = GetRegister(depth_register);
+ }
+
+ return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0);
+}
+
+Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
+ TextureProcessMode process_mode, bool depth_compare, bool is_array) {
+ const bool lod_bias_enabled =
+ (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
+
+ const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
+ texture_type, depth_compare, is_array, lod_bias_enabled, 4, 4);
+ // If enabled arrays index is always stored in the gpr8 field
+ const u64 array_register = instr.gpr8.Value();
+ // First coordinate index is stored in gpr8 field or (gpr8 + 1) when arrays are used
+ const u64 coord_register = array_register + (is_array ? 1 : 0);
+ const u64 last_coord_register =
+ (is_array || !(lod_bias_enabled || depth_compare) || (coord_count > 2))
+ ? static_cast<u64>(instr.gpr20.Value())
+ : coord_register + 1;
+ const u32 bias_offset = coord_count > 2 ? 1 : 0;
+
+ std::vector<Node> coords;
+ for (std::size_t i = 0; i < coord_count; ++i) {
+ const bool last = (i == (coord_count - 1)) && (coord_count > 1);
+ coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
+ }
+
+ const Node array = is_array ? GetRegister(array_register) : nullptr;
+
+ Node dc{};
+ if (depth_compare) {
+ // Depth is always stored in the register signaled by gpr20 or in the next register if lod
+ // or bias are used
+ const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
+ dc = GetRegister(depth_register);
+ }
+
+ return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset);
+}
+
+Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool depth_compare,
+ bool is_array) {
+ const std::size_t coord_count = GetCoordCount(texture_type);
+ const std::size_t total_coord_count = coord_count + (is_array ? 1 : 0);
+ const std::size_t total_reg_count = total_coord_count + (depth_compare ? 1 : 0);
+
+ // If enabled arrays index is always stored in the gpr8 field
+ const u64 array_register = instr.gpr8.Value();
+ // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
+ const u64 coord_register = array_register + (is_array ? 1 : 0);
+
+ std::vector<Node> coords;
+ for (size_t i = 0; i < coord_count; ++i)
+ coords.push_back(GetRegister(coord_register + i));
+
+ const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
+
+ Node4 values;
+ for (u32 element = 0; element < values.size(); ++element) {
+ auto coords_copy = coords;
+ MetaTexture meta{sampler, GetRegister(array_register), {}, {}, {}, {}, element};
+ values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
+ }
+
+ return values;
+}
+
+Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is_array) {
+ const std::size_t type_coord_count = GetCoordCount(texture_type);
+ const bool lod_enabled = instr.tlds.GetTextureProcessMode() == TextureProcessMode::LL;
+
+ // If enabled arrays index is always stored in the gpr8 field
+ const u64 array_register = instr.gpr8.Value();
+ // if is array gpr20 is used
+ const u64 coord_register = is_array ? instr.gpr20.Value() : instr.gpr8.Value();
+
+ const u64 last_coord_register =
+ ((type_coord_count > 2) || (type_coord_count == 2 && !lod_enabled)) && !is_array
+ ? static_cast<u64>(instr.gpr20.Value())
+ : coord_register + 1;
+
+ std::vector<Node> coords;
+ for (std::size_t i = 0; i < type_coord_count; ++i) {
+ const bool last = (i == (type_coord_count - 1)) && (type_coord_count > 1);
+ coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
+ }
+
+ const Node array = is_array ? GetRegister(array_register) : nullptr;
+ // When lod is used always is in gpr20
+ const Node lod = lod_enabled ? GetRegister(instr.gpr20) : Immediate(0);
+
+ const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
+
+ Node4 values;
+ for (u32 element = 0; element < values.size(); ++element) {
+ auto coords_copy = coords;
+ MetaTexture meta{sampler, array, {}, {}, lod, {}, element};
+ values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy));
+ }
+ return values;
+}
+
+std::tuple<std::size_t, std::size_t> ShaderIR::ValidateAndGetCoordinateElement(
+ TextureType texture_type, bool depth_compare, bool is_array, bool lod_bias_enabled,
+ std::size_t max_coords, std::size_t max_inputs) {
+ const std::size_t coord_count = GetCoordCount(texture_type);
+
+ std::size_t total_coord_count = coord_count + (is_array ? 1 : 0) + (depth_compare ? 1 : 0);
+ const std::size_t total_reg_count = total_coord_count + (lod_bias_enabled ? 1 : 0);
+ if (total_coord_count > max_coords || total_reg_count > max_inputs) {
+ UNIMPLEMENTED_MSG("Unsupported Texture operation");
+ total_coord_count = std::min(total_coord_count, max_coords);
+ }
+ // 1D.DC OpenGL is using a vec3 but 2nd component is ignored later.
+ total_coord_count +=
+ (depth_compare && !is_array && texture_type == TextureType::Texture1D) ? 1 : 0;
+
+ return {coord_count, total_coord_count};
+}
+
+} // namespace VideoCommon::Shader \ No newline at end of file
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 52c7f2c4e..5bc3a3900 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -290,7 +290,9 @@ struct MetaTexture {
const Sampler& sampler;
Node array{};
Node depth_compare{};
- std::vector<Node> extras;
+ Node bias{};
+ Node lod{};
+ Node component{};
u32 element{};
};
@@ -614,6 +616,7 @@ private:
u32 DecodeHfma2(NodeBlock& bb, u32 pc);
u32 DecodeConversion(NodeBlock& bb, u32 pc);
u32 DecodeMemory(NodeBlock& bb, u32 pc);
+ u32 DecodeTexture(NodeBlock& bb, u32 pc);
u32 DecodeFloatSetPredicate(NodeBlock& bb, u32 pc);
u32 DecodeIntegerSetPredicate(NodeBlock& bb, u32 pc);
u32 DecodeHalfSetPredicate(NodeBlock& bb, u32 pc);
diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp
index ca880dc65..32e78049c 100644
--- a/src/yuzu_cmd/config.cpp
+++ b/src/yuzu_cmd/config.cpp
@@ -346,7 +346,7 @@ void Config::ReadValues() {
// Renderer
Settings::values.resolution_factor =
- (float)sdl2_config->GetReal("Renderer", "resolution_factor", 1.0);
+ static_cast<float>(sdl2_config->GetReal("Renderer", "resolution_factor", 1.0));
Settings::values.use_frame_limit = sdl2_config->GetBoolean("Renderer", "use_frame_limit", true);
Settings::values.frame_limit =
static_cast<u16>(sdl2_config->GetInteger("Renderer", "frame_limit", 100));
@@ -357,16 +357,17 @@ void Config::ReadValues() {
Settings::values.use_asynchronous_gpu_emulation =
sdl2_config->GetBoolean("Renderer", "use_asynchronous_gpu_emulation", false);
- Settings::values.bg_red = (float)sdl2_config->GetReal("Renderer", "bg_red", 0.0);
- Settings::values.bg_green = (float)sdl2_config->GetReal("Renderer", "bg_green", 0.0);
- Settings::values.bg_blue = (float)sdl2_config->GetReal("Renderer", "bg_blue", 0.0);
+ Settings::values.bg_red = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_red", 0.0));
+ Settings::values.bg_green =
+ static_cast<float>(sdl2_config->GetReal("Renderer", "bg_green", 0.0));
+ Settings::values.bg_blue = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_blue", 0.0));
// Audio
Settings::values.sink_id = sdl2_config->Get("Audio", "output_engine", "auto");
Settings::values.enable_audio_stretching =
sdl2_config->GetBoolean("Audio", "enable_audio_stretching", true);
Settings::values.audio_device_id = sdl2_config->Get("Audio", "output_device", "auto");
- Settings::values.volume = sdl2_config->GetReal("Audio", "volume", 1);
+ Settings::values.volume = static_cast<float>(sdl2_config->GetReal("Audio", "volume", 1));
Settings::values.language_index = sdl2_config->GetInteger("System", "language_index", 1);
generated by cgit v1.2.3 (git 2.25.1) at 2025-05-01 11:43:49 +0000