early-access version 1860

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 1859.
+This is the source code for early-access 1860.
 
 ## Legal Notice
 

src/audio_core/audio_renderer.cpp

@@ -29,10 +29,9 @@ namespace {
                                        (static_cast<float>(r_channel) * r_mix_amount)));
 }
 
-[[nodiscard]] static constexpr std::tuple<s16, s16> Mix6To2(s16 fl_channel, s16 fr_channel,
+[[maybe_unused, nodiscard]] static constexpr std::tuple<s16, s16> Mix6To2(
-                                                            s16 fc_channel,
+    s16 fl_channel, s16 fr_channel, s16 fc_channel, [[maybe_unused]] s16 lf_channel, s16 bl_channel,
-                                                            [[maybe_unused]] s16 lf_channel,
+    s16 br_channel) {
-                                                            s16 bl_channel, s16 br_channel) {
     // Front channels are mixed 36.94%, Center channels are mixed to be 26.12% & the back channels
     // are mixed to be 36.94%
 
@@ -57,11 +56,11 @@ namespace {
     const std::array<float_le, 4>& coeff) {
     const auto left =
         static_cast<float>(fl_channel) * coeff[0] + static_cast<float>(fc_channel) * coeff[1] +
-        static_cast<float>(lf_channel) * coeff[2] + static_cast<float>(bl_channel) * coeff[0];
+        static_cast<float>(lf_channel) * coeff[2] + static_cast<float>(bl_channel) * coeff[3];
 
     const auto right =
         static_cast<float>(fr_channel) * coeff[0] + static_cast<float>(fc_channel) * coeff[1] +
-        static_cast<float>(lf_channel) * coeff[2] + static_cast<float>(br_channel) * coeff[0];
+        static_cast<float>(lf_channel) * coeff[2] + static_cast<float>(br_channel) * coeff[3];
 
     return {ClampToS16(static_cast<s32>(left)), ClampToS16(static_cast<s32>(right))};
 }
@@ -241,7 +240,7 @@ void AudioRenderer::QueueMixedBuffer(Buffer::Tag tag) {
         const auto channel_count = buffer_offsets.size();
         const auto& final_mix = mix_context.GetFinalMixInfo();
         const auto& in_params = final_mix.GetInParams();
-        std::vector<s32*> mix_buffers(channel_count);
+        std::vector<std::span<s32>> mix_buffers(channel_count);
         for (std::size_t i = 0; i < channel_count; i++) {
             mix_buffers[i] =
                 command_generator.GetMixBuffer(in_params.buffer_offset + buffer_offsets[i]);
@@ -294,18 +293,11 @@ void AudioRenderer::QueueMixedBuffer(Buffer::Tag tag) {
                     buffer[i * stream_channel_count + 0] = Mix2To1(fl_sample, fr_sample);
                 } else if (stream_channel_count == 2) {
                     // Mix all channels into 2 channels
-                    if (sink_context.HasDownMixingCoefficients()) {
                     const auto [left, right] = Mix6To2WithCoefficients(
                         fl_sample, fr_sample, fc_sample, lf_sample, bl_sample, br_sample,
                         sink_context.GetDownmixCoefficients());
                     buffer[i * stream_channel_count + 0] = left;
                     buffer[i * stream_channel_count + 1] = right;
-                    } else {
-                        const auto [left, right] = Mix6To2(fl_sample, fr_sample, fc_sample,
-                                                           lf_sample, bl_sample, br_sample);
-                        buffer[i * stream_channel_count + 0] = left;
-                        buffer[i * stream_channel_count + 1] = right;
-                    }
                 } else if (stream_channel_count == 6) {
                     // Pass through
                     buffer[i * stream_channel_count + 0] = fl_sample;

src/audio_core/command_generator.cpp

@@ -31,7 +31,7 @@ constexpr std::array<f32, AudioCommon::I3DL2REVERB_TAPS> EARLY_GAIN{
     0.72867f, 0.69794f, 0.5464f,  0.24563f, 0.45214f, 0.44042f};
 
 template <std::size_t N>
-void ApplyMix(s32* output, const s32* input, s32 gain, s32 sample_count) {
+void ApplyMix(std::span<s32> output, std::span<const s32> input, s32 gain, s32 sample_count) {
     for (std::size_t i = 0; i < static_cast<std::size_t>(sample_count); i += N) {
         for (std::size_t j = 0; j < N; j++) {
             output[i + j] +=
@@ -40,7 +40,17 @@ void ApplyMix(s32* output, const s32* input, s32 gain, s32 sample_count) {
     }
 }
 
-s32 ApplyMixRamp(s32* output, const s32* input, float gain, float delta, s32 sample_count) {
+s32 ApplyMixRamp(std::span<s32> output, std::span<const s32> input, float gain, float delta,
+                 s32 sample_count) {
+    // XC2 passes in NaN mix volumes, causing further issues as we handle everything as s32 rather
+    // than float, so the NaN propogation is lost. As the samples get further modified for
+    // volume etc, they can get out of NaN range, so a later heuristic for catching this is
+    // more difficult. Handle that here by setting these samples to silence.
+    if (std::isnan(gain)) {
+        gain = 0.0f;
+        delta = 0.0f;
+    }
+
     s32 x = 0;
     for (s32 i = 0; i < sample_count; i++) {
         x = static_cast<s32>(static_cast<float>(input[i]) * gain);
@@ -50,20 +60,22 @@ s32 ApplyMixRamp(s32* output, const s32* input, float gain, float delta, s32 sam
     return x;
 }
 
-void ApplyGain(s32* output, const s32* input, s32 gain, s32 delta, s32 sample_count) {
+void ApplyGain(std::span<s32> output, std::span<const s32> input, s32 gain, s32 delta,
+               s32 sample_count) {
     for (s32 i = 0; i < sample_count; i++) {
         output[i] = static_cast<s32>((static_cast<s64>(input[i]) * gain + 0x4000) >> 15);
         gain += delta;
     }
 }
 
-void ApplyGainWithoutDelta(s32* output, const s32* input, s32 gain, s32 sample_count) {
+void ApplyGainWithoutDelta(std::span<s32> output, std::span<const s32> input, s32 gain,
+                           s32 sample_count) {
     for (s32 i = 0; i < sample_count; i++) {
         output[i] = static_cast<s32>((static_cast<s64>(input[i]) * gain + 0x4000) >> 15);
     }
 }
 
-s32 ApplyMixDepop(s32* output, s32 first_sample, s32 delta, s32 sample_count) {
+s32 ApplyMixDepop(std::span<s32> output, s32 first_sample, s32 delta, s32 sample_count) {
     const bool positive = first_sample > 0;
     auto final_sample = std::abs(first_sample);
     for (s32 i = 0; i < sample_count; i++) {
@@ -128,10 +140,10 @@ constexpr std::array<std::size_t, 20> REVERB_TAP_INDEX_6CH{4, 0, 0, 1, 1, 1, 1,
                                                            1, 1, 1, 0, 0, 0, 0, 3, 3, 3};
 
 template <std::size_t CHANNEL_COUNT>
-void ApplyReverbGeneric(I3dl2ReverbState& state,
+void ApplyReverbGeneric(
-                        const std::array<const s32*, AudioCommon::MAX_CHANNEL_COUNT>& input,
+    I3dl2ReverbState& state,
-                        const std::array<s32*, AudioCommon::MAX_CHANNEL_COUNT>& output,
+    const std::array<std::span<const s32>, AudioCommon::MAX_CHANNEL_COUNT>& input,
-                        s32 sample_count) {
+    const std::array<std::span<s32>, AudioCommon::MAX_CHANNEL_COUNT>& output, s32 sample_count) {
 
     auto GetTapLookup = []() {
         if constexpr (CHANNEL_COUNT == 1) {
@@ -457,8 +469,8 @@ void CommandGenerator::GenerateBiquadFilterCommand([[maybe_unused]] s32 mix_buff
                   "input_mix_buffer={}, output_mix_buffer={}",
                   node_id, input_offset, output_offset);
     }
-    const auto* input = GetMixBuffer(input_offset);
+    std::span<const s32> input = GetMixBuffer(input_offset);
-    auto* output = GetMixBuffer(output_offset);
+    std::span<s32> output = GetMixBuffer(output_offset);
 
     // Biquad filter parameters
     const auto [n0, n1, n2] = params.numerator;
@@ -551,8 +563,8 @@ void CommandGenerator::GenerateI3dl2ReverbEffectCommand(s32 mix_buffer_offset, E
         return;
     }
 
-    std::array<const s32*, AudioCommon::MAX_CHANNEL_COUNT> input{};
+    std::array<std::span<const s32>, AudioCommon::MAX_CHANNEL_COUNT> input{};
-    std::array<s32*, AudioCommon::MAX_CHANNEL_COUNT> output{};
+    std::array<std::span<s32>, AudioCommon::MAX_CHANNEL_COUNT> output{};
 
     const auto status = params.status;
     for (s32 i = 0; i < channel_count; i++) {
@@ -587,7 +599,8 @@ void CommandGenerator::GenerateI3dl2ReverbEffectCommand(s32 mix_buffer_offset, E
         for (s32 i = 0; i < channel_count; i++) {
             // Only copy if the buffer input and output do not match!
             if ((mix_buffer_offset + params.input[i]) != (mix_buffer_offset + params.output[i])) {
-                std::memcpy(output[i], input[i], worker_params.sample_count * sizeof(s32));
+                std::memcpy(output[i].data(), input[i].data(),
+                            worker_params.sample_count * sizeof(s32));
             }
         }
     }
@@ -603,8 +616,8 @@ void CommandGenerator::GenerateBiquadFilterEffectCommand(s32 mix_buffer_offset,
     for (s32 i = 0; i < channel_count; i++) {
         // TODO(ogniK): Actually implement biquad filter
         if (params.input[i] != params.output[i]) {
-            const auto* input = GetMixBuffer(mix_buffer_offset + params.input[i]);
+            std::span<const s32> input = GetMixBuffer(mix_buffer_offset + params.input[i]);
-            auto* output = GetMixBuffer(mix_buffer_offset + params.output[i]);
+            std::span<s32> output = GetMixBuffer(mix_buffer_offset + params.output[i]);
             ApplyMix<1>(output, input, 32768, worker_params.sample_count);
         }
     }
@@ -643,14 +656,15 @@ void CommandGenerator::GenerateAuxCommand(s32 mix_buffer_offset, EffectBase* inf
 
                 if (samples_read != static_cast<int>(worker_params.sample_count) &&
                     samples_read <= params.sample_count) {
-                    std::memset(GetMixBuffer(output_index), 0, params.sample_count - samples_read);
+                    std::memset(GetMixBuffer(output_index).data(), 0,
+                                params.sample_count - samples_read);
                 }
             } else {
                 AuxInfoDSP empty{};
                 memory.WriteBlock(aux->GetSendInfo(), &empty, sizeof(AuxInfoDSP));
                 memory.WriteBlock(aux->GetRecvInfo(), &empty, sizeof(AuxInfoDSP));
                 if (output_index != input_index) {
-                    std::memcpy(GetMixBuffer(output_index), GetMixBuffer(input_index),
+                    std::memcpy(GetMixBuffer(output_index).data(), GetMixBuffer(input_index).data(),
                                 worker_params.sample_count * sizeof(s32));
                 }
             }
@@ -668,7 +682,7 @@ ServerSplitterDestinationData* CommandGenerator::GetDestinationData(s32 splitter
 }
 
 s32 CommandGenerator::WriteAuxBuffer(AuxInfoDSP& dsp_info, VAddr send_buffer, u32 max_samples,
-                                     const s32* data, u32 sample_count, u32 write_offset,
+                                     std::span<const s32> data, u32 sample_count, u32 write_offset,
                                      u32 write_count) {
     if (max_samples == 0) {
         return 0;
@@ -678,14 +692,14 @@ s32 CommandGenerator::WriteAuxBuffer(AuxInfoDSP& dsp_info, VAddr send_buffer, u3
         return 0;
     }
 
-    std::size_t data_offset{};
+    s32 data_offset{};
     u32 remaining = sample_count;
     while (remaining > 0) {
         // Get position in buffer
         const auto base = send_buffer + (offset * sizeof(u32));
         const auto samples_to_grab = std::min(max_samples - offset, remaining);
         // Write to output
-        memory.WriteBlock(base, (data + data_offset), samples_to_grab * sizeof(u32));
+        memory.WriteBlock(base, (data.data() + data_offset), samples_to_grab * sizeof(u32));
         offset = (offset + samples_to_grab) % max_samples;
         remaining -= samples_to_grab;
         data_offset += samples_to_grab;
@@ -698,7 +712,7 @@ s32 CommandGenerator::WriteAuxBuffer(AuxInfoDSP& dsp_info, VAddr send_buffer, u3
 }
 
 s32 CommandGenerator::ReadAuxBuffer(AuxInfoDSP& recv_info, VAddr recv_buffer, u32 max_samples,
-                                    s32* out_data, u32 sample_count, u32 read_offset,
+                                    std::span<s32> out_data, u32 sample_count, u32 read_offset,
                                     u32 read_count) {
     if (max_samples == 0) {
         return 0;
@@ -710,15 +724,16 @@ s32 CommandGenerator::ReadAuxBuffer(AuxInfoDSP& recv_info, VAddr recv_buffer, u3
     }
 
     u32 remaining = sample_count;
+    s32 data_offset{};
     while (remaining > 0) {
         const auto base = recv_buffer + (offset * sizeof(u32));
         const auto samples_to_grab = std::min(max_samples - offset, remaining);
         std::vector<s32> buffer(samples_to_grab);
         memory.ReadBlock(base, buffer.data(), buffer.size() * sizeof(u32));
-        std::memcpy(out_data, buffer.data(), buffer.size() * sizeof(u32));
+        std::memcpy(out_data.data() + data_offset, buffer.data(), buffer.size() * sizeof(u32));
-        out_data += samples_to_grab;
         offset = (offset + samples_to_grab) % max_samples;
         remaining -= samples_to_grab;
+        data_offset += samples_to_grab;
     }
 
     if (read_count != 0) {
@@ -965,8 +980,8 @@ void CommandGenerator::GenerateMixCommand(std::size_t output_offset, std::size_t
                   node_id, input_offset, output_offset, volume);
     }
 
-    auto* output = GetMixBuffer(output_offset);
+    std::span<s32> output = GetMixBuffer(output_offset);
-    const auto* input = GetMixBuffer(input_offset);
+    std::span<const s32> input = GetMixBuffer(input_offset);
 
     const s32 gain = static_cast<s32>(volume * 32768.0f);
     // Mix with loop unrolling
@@ -1172,12 +1187,14 @@ s32 CommandGenerator::DecodeAdpcm(ServerVoiceInfo& voice_info, VoiceState& dsp_s
     return samples_processed;
 }
 
-s32* CommandGenerator::GetMixBuffer(std::size_t index) {
+std::span<s32> CommandGenerator::GetMixBuffer(std::size_t index) {
-    return mix_buffer.data() + (index * worker_params.sample_count);
+    return std::span<s32>(mix_buffer.data() + (index * worker_params.sample_count),
+                          worker_params.sample_count);
 }
 
-const s32* CommandGenerator::GetMixBuffer(std::size_t index) const {
+std::span<const s32> CommandGenerator::GetMixBuffer(std::size_t index) const {
-    return mix_buffer.data() + (index * worker_params.sample_count);
+    return std::span<const s32>(mix_buffer.data() + (index * worker_params.sample_count),
+                                worker_params.sample_count);
 }
 
 std::size_t CommandGenerator::GetMixChannelBufferOffset(s32 channel) const {
@@ -1188,15 +1205,15 @@ std::size_t CommandGenerator::GetTotalMixBufferCount() const {
     return worker_params.mix_buffer_count + AudioCommon::MAX_CHANNEL_COUNT;
 }
 
-s32* CommandGenerator::GetChannelMixBuffer(s32 channel) {
+std::span<s32> CommandGenerator::GetChannelMixBuffer(s32 channel) {
     return GetMixBuffer(worker_params.mix_buffer_count + channel);
 }
 
-const s32* CommandGenerator::GetChannelMixBuffer(s32 channel) const {
+std::span<const s32> CommandGenerator::GetChannelMixBuffer(s32 channel) const {
     return GetMixBuffer(worker_params.mix_buffer_count + channel);
 }
 
-void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* output,
+void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, std::span<s32> output,
                                              VoiceState& dsp_state, s32 channel,
                                              s32 target_sample_rate, s32 sample_count,
                                              s32 node_id) {
@@ -1208,7 +1225,7 @@ void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* o
                   node_id, channel, in_params.sample_format, sample_count, in_params.sample_rate,
                   in_params.mix_id, in_params.splitter_info_id);
     }
-    ASSERT_OR_EXECUTE(output != nullptr, { return; });
+    ASSERT_OR_EXECUTE(output.data() != nullptr, { return; });
 
     const auto resample_rate = static_cast<s32>(
         static_cast<float>(in_params.sample_rate) / static_cast<float>(target_sample_rate) *
@@ -1225,6 +1242,7 @@ void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* o
     }
 
     std::size_t temp_mix_offset{};
+    s32 samples_output{};
     auto samples_remaining = sample_count;
     while (samples_remaining > 0) {
         const auto samples_to_output = std::min(samples_remaining, min_required_samples);
@@ -1328,20 +1346,21 @@ void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* o
 
         if (in_params.behavior_flags.is_pitch_and_src_skipped.Value()) {
             // No need to resample
-            std::memcpy(output, sample_buffer.data(), samples_read * sizeof(s32));
+            std::memcpy(output.data() + samples_output, sample_buffer.data(),
+                        samples_read * sizeof(s32));
         } else {
             std::fill(sample_buffer.begin() + temp_mix_offset,
                       sample_buffer.begin() + temp_mix_offset + (samples_to_read - samples_read),
                       0);
-            AudioCore::Resample(output, sample_buffer.data(), resample_rate, dsp_state.fraction,
+            AudioCore::Resample(output.data() + samples_output, sample_buffer.data(), resample_rate,
-                                samples_to_output);
+                                dsp_state.fraction, samples_to_output);
             // Resample
             for (std::size_t i = 0; i < AudioCommon::MAX_SAMPLE_HISTORY; i++) {
                 dsp_state.sample_history[i] = sample_buffer[samples_to_read + i];
             }
         }
-        output += samples_to_output;
         samples_remaining -= samples_to_output;
+        samples_output += samples_to_output;
     }
 }
 

src/audio_core/command_generator.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <span>
 #include "audio_core/common.h"
 #include "audio_core/voice_context.h"
 #include "common/common_types.h"
@@ -41,10 +42,10 @@ public:
     void PreCommand();
     void PostCommand();
 
-    [[nodiscard]] s32* GetChannelMixBuffer(s32 channel);
+    [[nodiscard]] std::span<s32> GetChannelMixBuffer(s32 channel);
-    [[nodiscard]] const s32* GetChannelMixBuffer(s32 channel) const;
+    [[nodiscard]] std::span<const s32> GetChannelMixBuffer(s32 channel) const;
-    [[nodiscard]] s32* GetMixBuffer(std::size_t index);
+    [[nodiscard]] std::span<s32> GetMixBuffer(std::size_t index);
-    [[nodiscard]] const s32* GetMixBuffer(std::size_t index) const;
+    [[nodiscard]] std::span<const s32> GetMixBuffer(std::size_t index) const;
     [[nodiscard]] std::size_t GetMixChannelBufferOffset(s32 channel) const;
 
     [[nodiscard]] std::size_t GetTotalMixBufferCount() const;
@@ -77,10 +78,11 @@ private:
     void GenerateAuxCommand(s32 mix_buffer_offset, EffectBase* info, bool enabled);
     [[nodiscard]] ServerSplitterDestinationData* GetDestinationData(s32 splitter_id, s32 index);
 
-    s32 WriteAuxBuffer(AuxInfoDSP& dsp_info, VAddr send_buffer, u32 max_samples, const s32* data,
+    s32 WriteAuxBuffer(AuxInfoDSP& dsp_info, VAddr send_buffer, u32 max_samples,
-                       u32 sample_count, u32 write_offset, u32 write_count);
+                       std::span<const s32> data, u32 sample_count, u32 write_offset,
-    s32 ReadAuxBuffer(AuxInfoDSP& recv_info, VAddr recv_buffer, u32 max_samples, s32* out_data,
+                       u32 write_count);
-                      u32 sample_count, u32 read_offset, u32 read_count);
+    s32 ReadAuxBuffer(AuxInfoDSP& recv_info, VAddr recv_buffer, u32 max_samples,
+                      std::span<s32> out_data, u32 sample_count, u32 read_offset, u32 read_count);
 
     void InitializeI3dl2Reverb(I3dl2ReverbParams& info, I3dl2ReverbState& state,
                                std::vector<u8>& work_buffer);
@@ -91,8 +93,9 @@ private:
                   s32 sample_end_offset, s32 sample_count, s32 channel, std::size_t mix_offset);
     s32 DecodeAdpcm(ServerVoiceInfo& voice_info, VoiceState& dsp_state, s32 sample_start_offset,
                     s32 sample_end_offset, s32 sample_count, s32 channel, std::size_t mix_offset);
-    void DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, s32* output, VoiceState& dsp_state,
+    void DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, std::span<s32> output,
-                               s32 channel, s32 target_sample_rate, s32 sample_count, s32 node_id);
+                               VoiceState& dsp_state, s32 channel, s32 target_sample_rate,
+                               s32 sample_count, s32 node_id);
 
     AudioCommon::AudioRendererParameter& worker_params;
     VoiceContext& voice_context;

src/audio_core/sink_context.cpp

@@ -15,10 +15,17 @@ std::size_t SinkContext::GetCount() const {
 void SinkContext::UpdateMainSink(const SinkInfo::InParams& in) {
     ASSERT(in.type == SinkTypes::Device);
 
-    has_downmix_coefs = in.device.down_matrix_enabled;
+    if (in.device.down_matrix_enabled) {
-    if (has_downmix_coefs) {
         downmix_coefficients = in.device.down_matrix_coef;
+    } else {
+        downmix_coefficients = {
+            1.0f,   // front
+            0.707f, // center
+            0.0f,   // lfe
+            0.707f, // back
+        };
     }
+
     in_use = in.in_use;
     use_count = in.device.input_count;
     buffers = in.device.input;
@@ -34,10 +41,6 @@ std::vector<u8> SinkContext::OutputBuffers() const {
     return buffer_ret;
 }
 
-bool SinkContext::HasDownMixingCoefficients() const {
-    return has_downmix_coefs;
-}
-
 const DownmixCoefficients& SinkContext::GetDownmixCoefficients() const {
     return downmix_coefficients;
 }

src/audio_core/sink_context.h

@@ -84,7 +84,6 @@ public:
     [[nodiscard]] bool InUse() const;
     [[nodiscard]] std::vector<u8> OutputBuffers() const;
 
-    [[nodiscard]] bool HasDownMixingCoefficients() const;
     [[nodiscard]] const DownmixCoefficients& GetDownmixCoefficients() const;
 
 private:
@@ -92,7 +91,6 @@ private:
     s32 use_count{};
     std::array<u8, AudioCommon::MAX_CHANNEL_COUNT> buffers{};
     std::size_t sink_count{};
-    bool has_downmix_coefs{false};
     DownmixCoefficients downmix_coefficients{};
 };
 } // namespace AudioCore

src/video_core/buffer_cache/buffer_cache.h

@@ -591,7 +591,6 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
         for (auto& interval : intervals) {
             const std::size_t size = interval.upper() - interval.lower();
             const VAddr cpu_addr = interval.lower();
-            const VAddr cpu_addr_end = interval.upper();
             ForEachBufferInRange(cpu_addr, size, [&](BufferId buffer_id, Buffer& buffer) {
                 boost::container::small_vector<BufferCopy, 1> copies;
                 buffer.ForEachDownloadRange(

src/video_core/engines/shader_bytecode.h

@@ -1785,8 +1785,6 @@ public:
         SSY,
         SYNC,
         BRK,
-        CAL,
-        RET,
         DEPBAR,
         VOTE,
         VOTE_VTG,
@@ -2110,8 +2108,6 @@ private:
             INST("1111000011111---", Id::SYNC, Type::Flow, "SYNC"),
             INST("111000110100----", Id::BRK, Type::Flow, "BRK"),
             INST("111000110000----", Id::EXIT, Type::Flow, "EXIT"),
-            INST("111000100110----", Id::CAL, Type::Flow, "CAL"),
-            INST("111000110010----", Id::RET, Type::Flow, "RET"),
             INST("1111000011110---", Id::DEPBAR, Type::Synch, "DEPBAR"),
             INST("0101000011011---", Id::VOTE, Type::Warp, "VOTE"),
             INST("0101000011100---", Id::VOTE_VTG, Type::Warp, "VOTE_VTG"),

src/video_core/renderer_opengl/gl_arb_decompiler.cpp

@@ -491,9 +491,6 @@ private:
     const Registry& registry;
     const ShaderType stage;
 
-    std::shared_ptr<ShaderFunctionIR> context_func;
-    u32 ast_var_base{};
-
     std::size_t num_temporaries = 0;
     std::size_t max_temporaries = 0;
 
@@ -810,33 +807,13 @@ ARBDecompiler::ARBDecompiler(const Device& device_, const ShaderIR& ir_, const R
     : device{device_}, ir{ir_}, registry{registry_}, stage{stage_} {
     DefineGlobalMemory();
 
-    context_func = ir.GetMainFunction();
-    ast_var_base = 0;
-
     AddLine("TEMP RC;");
     AddLine("TEMP FSWZA[4];");
     AddLine("TEMP FSWZB[4];");
-    InitializeVariables();
+    if (ir.IsDecompiled()) {
-    AddLine("main:");
-    if (context_func->IsDecompiled()) {
         DecompileAST();
     } else {
         DecompileBranchMode();
-        AddLine("RET;");
-    }
-
-    const auto& subfunctions = ir.GetSubFunctions();
-    auto it = subfunctions.begin();
-    while (it != subfunctions.end()) {
-        context_func = *it;
-        AddLine("func_{}:", context_func->GetId());
-        if (context_func->IsDecompiled()) {
-            DecompileAST();
-        } else {
-            DecompileBranchMode();
-            AddLine("RET;");
-        }
-        it++;
     }
     AddLine("END");
 
@@ -1083,38 +1060,41 @@ void ARBDecompiler::InitializeVariables() {
 }
 
 void ARBDecompiler::DecompileAST() {
-    const u32 num_flow_variables = context_func->GetASTNumVariables();
+    const u32 num_flow_variables = ir.GetASTNumVariables();
     for (u32 i = 0; i < num_flow_variables; ++i) {
-        AddLine("TEMP F{};", i + ast_var_base);
+        AddLine("TEMP F{};", i);
     }
     for (u32 i = 0; i < num_flow_variables; ++i) {
-        AddLine("MOV.U F{}, {{0, 0, 0, 0}};", i + ast_var_base);
+        AddLine("MOV.U F{}, {{0, 0, 0, 0}};", i);
     }
 
-    VisitAST(context_func->GetASTProgram());
+    InitializeVariables();
-    ast_var_base += num_flow_variables;
+
+    VisitAST(ir.GetASTProgram());
 }
 
 void ARBDecompiler::DecompileBranchMode() {
     static constexpr u32 FLOW_STACK_SIZE = 20;
-    if (!context_func->IsFlowStackDisabled()) {
+    if (!ir.IsFlowStackDisabled()) {
         AddLine("TEMP SSY[{}];", FLOW_STACK_SIZE);
         AddLine("TEMP PBK[{}];", FLOW_STACK_SIZE);
         AddLine("TEMP SSY_TOP;");
         AddLine("TEMP PBK_TOP;");
     }
 
-    AddLine("TEMP PC{};", context_func->GetId());
+    AddLine("TEMP PC;");
 
-    if (!context_func->IsFlowStackDisabled()) {
+    if (!ir.IsFlowStackDisabled()) {
         AddLine("MOV.U SSY_TOP.x, 0;");
         AddLine("MOV.U PBK_TOP.x, 0;");
     }
 
-    const auto basic_block_end = context_func->GetBasicBlocks().end();
+    InitializeVariables();
-    auto basic_block_it = context_func->GetBasicBlocks().begin();
+
+    const auto basic_block_end = ir.GetBasicBlocks().end();
+    auto basic_block_it = ir.GetBasicBlocks().begin();
     const u32 first_address = basic_block_it->first;
-    AddLine("MOV.U PC{}.x, {};", context_func->GetId(), first_address);
+    AddLine("MOV.U PC.x, {};", first_address);
 
     AddLine("REP;");
 
@@ -1123,7 +1103,7 @@ void ARBDecompiler::DecompileBranchMode() {
         const auto& [address, bb] = *basic_block_it;
         ++num_blocks;
 
-        AddLine("SEQ.S.CC RC.x, PC{}.x, {};", context_func->GetId(), address);
+        AddLine("SEQ.S.CC RC.x, PC.x, {};", address);
         AddLine("IF NE.x;");
 
         VisitBlock(bb);
@@ -1134,7 +1114,7 @@ void ARBDecompiler::DecompileBranchMode() {
             const auto op = std::get_if<OperationNode>(&*bb[bb.size() - 1]);
             if (!op || op->GetCode() != OperationCode::Branch) {
                 const u32 next_address = basic_block_it->first;
-                AddLine("MOV.U PC{}.x, {};", context_func->GetId(), next_address);
+                AddLine("MOV.U PC.x, {};", next_address);
                 AddLine("CONT;");
             }
         }
@@ -1172,8 +1152,7 @@ void ARBDecompiler::VisitAST(const ASTNode& node) {
     } else if (const auto decoded = std::get_if<ASTBlockDecoded>(&*node->GetInnerData())) {
         VisitBlock(decoded->nodes);
     } else if (const auto var_set = std::get_if<ASTVarSet>(&*node->GetInnerData())) {
-        AddLine("MOV.U F{}, {};", var_set->index + ast_var_base,
+        AddLine("MOV.U F{}, {};", var_set->index, VisitExpression(var_set->condition));
-                VisitExpression(var_set->condition));
         ResetTemporaries();
     } else if (const auto do_while = std::get_if<ASTDoWhile>(&*node->GetInnerData())) {
         const std::string condition = VisitExpression(do_while->condition);
@@ -1193,11 +1172,7 @@ void ARBDecompiler::VisitAST(const ASTNode& node) {
             ResetTemporaries();
         }
         if (ast_return->kills) {
-            if (stage == ShaderType::Fragment) {
             AddLine("KIL TR;");
-            } else {
-                AddLine("RET;");
-            }
         } else {
             Exit();
         }
@@ -1244,7 +1219,7 @@ std::string ARBDecompiler::VisitExpression(const Expr& node) {
         return Visit(ir.GetConditionCode(expr->cc));
     }
     if (const auto expr = std::get_if<ExprVar>(&*node)) {
-        return fmt::format("F{}.x", expr->var_index + ast_var_base);
+        return fmt::format("F{}.x", expr->var_index);
     }
     if (const auto expr = std::get_if<ExprBoolean>(&*node)) {
         return expr->value ? "0xffffffff" : "0";
@@ -1431,11 +1406,6 @@ std::string ARBDecompiler::Visit(const Node& node) {
         return {};
     }
 
-    if (const auto func_call = std::get_if<FunctionCallNode>(&*node)) {
-        AddLine("CAL func_{};", func_call->GetFuncId());
-        return {};
-    }
-
     if ([[maybe_unused]] const auto cmt = std::get_if<CommentNode>(&*node)) {
         // Uncommenting this will generate invalid code. GLASM lacks comments.
         // AddLine("// {}", cmt->GetText());
@@ -1509,7 +1479,7 @@ std::string ARBDecompiler::GlobalMemoryPointer(const GmemNode& gmem) {
 }
 
 void ARBDecompiler::Exit() {
-    if (!context_func->IsMain() || stage != ShaderType::Fragment) {
+    if (stage != ShaderType::Fragment) {
         AddLine("RET;");
         return;
     }
@@ -2051,13 +2021,13 @@ std::string ARBDecompiler::ImageStore(Operation operation) {
 
 std::string ARBDecompiler::Branch(Operation operation) {
     const auto target = std::get<ImmediateNode>(*operation[0]);
-    AddLine("MOV.U PC{}.x, {};", context_func->GetId(), target.GetValue());
+    AddLine("MOV.U PC.x, {};", target.GetValue());
     AddLine("CONT;");
     return {};
 }
 
 std::string ARBDecompiler::BranchIndirect(Operation operation) {
-    AddLine("MOV.U PC{}.x, {};", context_func->GetId(), Visit(operation[0]));
+    AddLine("MOV.U PC.x, {};", Visit(operation[0]));
     AddLine("CONT;");
     return {};
 }
@@ -2075,7 +2045,7 @@ std::string ARBDecompiler::PopFlowStack(Operation operation) {
     const auto stack = std::get<MetaStackClass>(operation.GetMeta());
     const std::string_view stack_name = StackName(stack);
     AddLine("SUB.S {}_TOP.x, {}_TOP.x, 1;", stack_name, stack_name);
-    AddLine("MOV.U PC{}.x, {}[{}_TOP.x].x;", context_func->GetId(), stack_name, stack_name);
+    AddLine("MOV.U PC.x, {}[{}_TOP.x].x;", stack_name, stack_name);
     AddLine("CONT;");
     return {};
 }
@@ -2086,10 +2056,6 @@ std::string ARBDecompiler::Exit(Operation) {
 }
 
 std::string ARBDecompiler::Discard(Operation) {
-    if (stage != ShaderType::Fragment) {
-        AddLine("RET;");
-        return {};
-    }
     AddLine("KIL TR;");
     return {};
 }

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -79,11 +79,6 @@ const float fswzadd_modifiers_a[] = float[4](-1.0f,  1.0f, -1.0f,  0.0f );
 const float fswzadd_modifiers_b[] = float[4](-1.0f, -1.0f,  1.0f, -1.0f );
 )";
 
-enum class HelperFunction {
-    SignedAtomic = 0,
-    Total,
-};
-
 class ShaderWriter final {
 public:
     void AddExpression(std::string_view text) {
@@ -439,28 +434,6 @@ public:
         DeclareInternalFlags();
         DeclareCustomVariables();
         DeclarePhysicalAttributeReader();
-        DeclareHelpersForward();
-
-        const auto& subfunctions = ir.GetSubFunctions();
-        auto it = subfunctions.rbegin();
-        while (it != subfunctions.rend()) {
-            context_func = *it;
-            code.AddLine("void func_{}() {{", context_func->GetId());
-            ++code.scope;
-
-            if (context_func->IsDecompiled()) {
-                DecompileAST();
-            } else {
-                DecompileBranchMode();
-            }
-
-            --code.scope;
-            code.AddLine("}}");
-
-            it++;
-        }
-
-        context_func = ir.GetMainFunction();
 
         code.AddLine("void main() {{");
         ++code.scope;
@@ -469,7 +442,7 @@ public:
             code.AddLine("gl_Position = vec4(0.0f, 0.0f, 0.0f, 1.0f);");
         }
 
-        if (context_func->IsDecompiled()) {
+        if (ir.IsDecompiled()) {
             DecompileAST();
         } else {
             DecompileBranchMode();
@@ -477,9 +450,6 @@ public:
 
         --code.scope;
         code.AddLine("}}");
-
-        code.AddNewLine();
-        DeclareHelpers();
     }
 
     std::string GetResult() {
@@ -492,13 +462,13 @@ private:
 
     void DecompileBranchMode() {
         // VM's program counter
-        const auto first_address = context_func->GetBasicBlocks().begin()->first;
+        const auto first_address = ir.GetBasicBlocks().begin()->first;
         code.AddLine("uint jmp_to = {}U;", first_address);
 
         // TODO(Subv): Figure out the actual depth of the flow stack, for now it seems
         // unlikely that shaders will use 20 nested SSYs and PBKs.
         constexpr u32 FLOW_STACK_SIZE = 20;
-        if (!context_func->IsFlowStackDisabled()) {
+        if (!ir.IsFlowStackDisabled()) {
             for (const auto stack : std::array{MetaStackClass::Ssy, MetaStackClass::Pbk}) {
                 code.AddLine("uint {}[{}];", FlowStackName(stack), FLOW_STACK_SIZE);
                 code.AddLine("uint {} = 0U;", FlowStackTopName(stack));
@@ -510,7 +480,7 @@ private:
 
         code.AddLine("switch (jmp_to) {{");
 
-        for (const auto& pair : context_func->GetBasicBlocks()) {
+        for (const auto& pair : ir.GetBasicBlocks()) {
             const auto& [address, bb] = pair;
             code.AddLine("case 0x{:X}U: {{", address);
             ++code.scope;
@@ -629,7 +599,7 @@ private:
                 size = limit;
             }
 
-            code.AddLine("shared uint {}[{}];", GetSharedMemory(), size / 4);
+            code.AddLine("shared uint smem[{}];", size / 4);
             code.AddNewLine();
         }
         code.AddLine("layout (local_size_x = {}, local_size_y = {}, local_size_z = {}) in;",
@@ -1013,27 +983,6 @@ private:
         }
     }
 
-    void DeclareHelpersForward() {
-        code.AddLine("int Helpers_AtomicShared(uint offset, int value, bool is_min);");
-        code.AddNewLine();
-    }
-
-    void DeclareHelpers() {
-        if (IsHelperEnabled(HelperFunction::SignedAtomic)) {
-            code.AddLine(
-                R"(int Helpers_AtomicShared(uint offset, int value, bool is_min) {{
-    uint oldValue, newValue;
-    do {{
-        oldValue = {}[offset];
-        newValue = is_min ? uint(min(int(oldValue), value)) : uint(max(int(oldValue), value));
-    }} while (atomicCompSwap({}[offset], newValue, oldValue) != oldValue);
-    return int(oldValue);
-}})",
-                GetSharedMemory(), GetSharedMemory());
-            code.AddNewLine();
-        }
-    }
-
     void VisitBlock(const NodeBlock& bb) {
         for (const auto& node : bb) {
             Visit(node).CheckVoid();
@@ -1160,9 +1109,7 @@ private:
         }
 
         if (const auto smem = std::get_if<SmemNode>(&*node)) {
-            return {
+            return {fmt::format("smem[{} >> 2]", Visit(smem->GetAddress()).AsUint()), Type::Uint};
-                fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
-                Type::Uint};
         }
 
         if (const auto internal_flag = std::get_if<InternalFlagNode>(&*node)) {
@@ -1184,11 +1131,6 @@ private:
             return {};
         }
 
-        if (const auto func_call = std::get_if<FunctionCallNode>(&*node)) {
-            code.AddLine("func_{}();", func_call->GetFuncId());
-            return {};
-        }
-
         if (const auto comment = std::get_if<CommentNode>(&*node)) {
             code.AddLine("// " + comment->GetText());
             return {};
@@ -1656,9 +1598,7 @@ private:
                 Type::Uint};
         } else if (const auto smem = std::get_if<SmemNode>(&*dest)) {
             ASSERT(stage == ShaderType::Compute);
-            target = {
+            target = {fmt::format("smem[{} >> 2]", Visit(smem->GetAddress()).AsUint()), Type::Uint};
-                fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
-                Type::Uint};
         } else if (const auto gmem = std::get_if<GmemNode>(&*dest)) {
             const std::string real = Visit(gmem->GetRealAddress()).AsUint();
             const std::string base = Visit(gmem->GetBaseAddress()).AsUint();
@@ -2175,14 +2115,7 @@ private:
         UNIMPLEMENTED_IF(meta->sampler.is_array);
         const std::size_t count = operation.GetOperandsCount();
 
-        std::string expr = "texelFetch";
+        std::string expr = "texelFetch(";
-
-        if (!meta->aoffi.empty()) {
-            expr += "Offset";
-        }
-
-        expr += '(';
-
         expr += GetSampler(meta->sampler);
         expr += ", ";
 
@@ -2204,20 +2137,6 @@ private:
             expr += ", ";
             expr += Visit(meta->lod).AsInt();
         }
-
-        if (!meta->aoffi.empty()) {
-            expr += ", ";
-            expr += constructors.at(meta->aoffi.size() - 1);
-            expr += '(';
-            for (size_t i = 0; i < meta->aoffi.size(); ++i) {
-                if (i > 0) {
-                    expr += ", ";
-                }
-                expr += Visit(meta->aoffi[i]).AsInt();
-            }
-            expr += ')';
-        }
-
         expr += ')';
         expr += GetSwizzle(meta->element);
 
@@ -2264,11 +2183,8 @@ private:
     template <const std::string_view& opname, Type type>
     Expression Atomic(Operation operation) {
         if ((opname == Func::Min || opname == Func::Max) && type == Type::Int) {
-            // Use a helper as a workaround due to memory being uint
+            UNIMPLEMENTED_MSG("Unimplemented Min & Max for atomic operations");
-            SetHelperEnabled(HelperFunction::SignedAtomic, true);
+            return {};
-            return {fmt::format("Helpers_AtomicShared({}, {}, {})", Visit(operation[0]).AsInt(),
-                                Visit(operation[1]).AsInt(), opname == Func::Min),
-                    Type::Int};
         }
         return {fmt::format("atomic{}({}, {})", opname, Visit(operation[0]).GetCode(),
                             Visit(operation[1]).AsUint()),
@@ -2351,9 +2267,7 @@ private:
     }
 
     Expression Exit(Operation operation) {
-        if (context_func->IsMain()) {
         PreExit();
-        }
         code.AddLine("return;");
         return {};
     }
@@ -2363,11 +2277,7 @@ private:
         // about unexecuted instructions that may follow this.
         code.AddLine("if (true) {{");
         ++code.scope;
-        if (stage != ShaderType::Fragment) {
-            code.AddLine("return;");
-        } else {
         code.AddLine("discard;");
-        }
         --code.scope;
         code.AddLine("}}");
         return {};
@@ -2478,7 +2388,7 @@ private:
     }
 
     Expression Barrier(Operation) {
-        if (!context_func->IsDecompiled()) {
+        if (!ir.IsDecompiled()) {
             LOG_ERROR(Render_OpenGL, "barrier() used but shader is not decompiled");
             return {};
         }
@@ -2795,10 +2705,6 @@ private:
         }
     }
 
-    constexpr std::string_view GetSharedMemory() const {
-        return "shared_mem";
-    }
-
     std::string GetInternalFlag(InternalFlag flag) const {
         constexpr std::array InternalFlagNames = {"zero_flag", "sign_flag", "carry_flag",
                                                   "overflow_flag"};
@@ -2840,14 +2746,6 @@ private:
         return std::min<u32>(device.GetMaxVaryings(), Maxwell::NumVaryings);
     }
 
-    void SetHelperEnabled(HelperFunction hf, bool enabled) {
-        helper_functions_enabled[static_cast<size_t>(hf)] = enabled;
-    }
-
-    bool IsHelperEnabled(HelperFunction hf) const {
-        return helper_functions_enabled[static_cast<size_t>(hf)];
-    }
-
     const Device& device;
     const ShaderIR& ir;
     const Registry& registry;
@@ -2857,13 +2755,9 @@ private:
     const Header header;
     std::unordered_map<u8, VaryingTFB> transform_feedback;
 
-    std::shared_ptr<ShaderFunctionIR> context_func;
-
     ShaderWriter code;
 
     std::optional<u32> max_input_vertices;
-
-    std::array<bool, static_cast<size_t>(HelperFunction::Total)> helper_functions_enabled{};
 };
 
 std::string GetFlowVariable(u32 index) {
@@ -3008,15 +2902,9 @@ public:
             decomp.code.scope++;
         }
         if (ast.kills) {
-            if (decomp.stage != ShaderType::Fragment) {
-                decomp.code.AddLine("return;");
-            } else {
             decomp.code.AddLine("discard;");
-            }
         } else {
-            if (decomp.context_func->IsMain()) {
             decomp.PreExit();
-            }
             decomp.code.AddLine("return;");
         }
         if (!is_true) {
@@ -3049,13 +2937,13 @@ private:
 };
 
 void GLSLDecompiler::DecompileAST() {
-    const u32 num_flow_variables = context_func->GetASTNumVariables();
+    const u32 num_flow_variables = ir.GetASTNumVariables();
     for (u32 i = 0; i < num_flow_variables; i++) {
         code.AddLine("bool {} = false;", GetFlowVariable(i));
     }
 
     ASTDecompiler decompiler{*this};
-    decompiler.Visit(context_func->GetASTProgram());
+    decompiler.Visit(ir.GetASTProgram());
 }
 
 } // Anonymous namespace

src/video_core/renderer_vulkan/vk_shader_decompiler.cpp

@@ -406,38 +406,10 @@ private:
         binding = DeclareStorageTexels(binding);
         binding = DeclareImages(binding);
 
-        const auto& subfunctions = ir.GetSubFunctions();
-
-        labels.resize(subfunctions.size() + 1);
-        other_functions.resize(subfunctions.size());
-
-        auto it = subfunctions.rbegin();
-        while (it != subfunctions.rend()) {
-            context_func = *it;
-            other_functions[context_func->GetId() - 1] =
-                OpFunction(t_void, {}, TypeFunction(t_void));
-            AddLabel();
-
-            if (context_func->IsDecompiled()) {
-                DeclareFlowVariables();
-                DecompileAST();
-            } else {
-                AllocateLabels();
-                DecompileBranchMode();
-            }
-
-            OpReturn();
-            OpFunctionEnd();
-
-            it++;
-        }
-
-        context_func = ir.GetMainFunction();
-
         const Id main = OpFunction(t_void, {}, TypeFunction(t_void));
         AddLabel();
 
-        if (context_func->IsDecompiled()) {
+        if (ir.IsDecompiled()) {
             DeclareFlowVariables();
             DecompileAST();
         } else {
@@ -469,18 +441,16 @@ private:
     void DecompileAST();
 
     void DecompileBranchMode() {
-        const u32 first_address = context_func->GetBasicBlocks().begin()->first;
+        const u32 first_address = ir.GetBasicBlocks().begin()->first;
-        const u32 func_id = context_func->GetId();
+        const Id loop_label = OpLabel("loop");
-        const std::string func_id_msg = std::to_string(func_id);
+        const Id merge_label = OpLabel("merge");
-        const Id loop_label = OpLabel("loop_" + func_id_msg);
-        const Id merge_label = OpLabel("merge_" + func_id_msg);
         const Id dummy_label = OpLabel();
         const Id jump_label = OpLabel();
-        continue_label = OpLabel("continue_" + func_id_msg);
+        continue_label = OpLabel("continue");
 
         std::vector<Sirit::Literal> literals;
         std::vector<Id> branch_labels;
-        for (const auto& [literal, label] : labels[func_id]) {
+        for (const auto& [literal, label] : labels) {
             literals.push_back(literal);
             branch_labels.push_back(label);
         }
@@ -492,11 +462,11 @@ private:
         std::tie(ssy_flow_stack, ssy_flow_stack_top) = CreateFlowStack();
         std::tie(pbk_flow_stack, pbk_flow_stack_top) = CreateFlowStack();
 
-        Name(jmp_to, "jmp_to_" + func_id_msg);
+        Name(jmp_to, "jmp_to");
-        Name(ssy_flow_stack, "ssy_flow_stack_" + func_id_msg);
+        Name(ssy_flow_stack, "ssy_flow_stack");
-        Name(ssy_flow_stack_top, "ssy_flow_stack_top_" + func_id_msg);
+        Name(ssy_flow_stack_top, "ssy_flow_stack_top");
-        Name(pbk_flow_stack, "pbk_flow_stack_" + func_id_msg);
+        Name(pbk_flow_stack, "pbk_flow_stack");
-        Name(pbk_flow_stack_top, "pbk_flow_stack_top_" + func_id_msg);
+        Name(pbk_flow_stack_top, "pbk_flow_stack_top");
 
         DefinePrologue();
 
@@ -514,14 +484,13 @@ private:
         AddLabel(default_branch);
         OpReturn();
 
-        for (const auto& [address, bb] : context_func->GetBasicBlocks()) {
+        for (const auto& [address, bb] : ir.GetBasicBlocks()) {
-            AddLabel(labels[func_id].at(address));
+            AddLabel(labels.at(address));
 
             VisitBasicBlock(bb);
 
-            const auto next_it = labels[func_id].lower_bound(address + 1);
+            const auto next_it = labels.lower_bound(address + 1);
-            const Id next_label =
+            const Id next_label = next_it != labels.end() ? next_it->second : default_branch;
-                next_it != labels[func_id].end() ? next_it->second : default_branch;
             OpBranch(next_label);
         }
 
@@ -539,10 +508,9 @@ private:
     static constexpr auto INTERNAL_FLAGS_COUNT = static_cast<std::size_t>(InternalFlag::Amount);
 
     void AllocateLabels() {
-        const u32 func_id = context_func->GetId();
+        for (const auto& pair : ir.GetBasicBlocks()) {
-        for (const auto& pair : context_func->GetBasicBlocks()) {
             const u32 address = pair.first;
-            labels[func_id].emplace(address, OpLabel(fmt::format("label_0x{:x}", address)));
+            labels.emplace(address, OpLabel(fmt::format("label_0x{:x}", address)));
         }
     }
 
@@ -621,14 +589,6 @@ private:
         DeclareOutputVertex();
     }
 
-    void SafeKill() {
-        if (stage != ShaderType::Fragment) {
-            OpReturn();
-            return;
-        }
-        OpKill();
-    }
-
     void DeclareFragment() {
         if (stage != ShaderType::Fragment) {
             return;
@@ -696,7 +656,7 @@ private:
     }
 
     void DeclareFlowVariables() {
-        for (u32 i = 0; i < context_func->GetASTNumVariables(); i++) {
+        for (u32 i = 0; i < ir.GetASTNumVariables(); i++) {
             const Id id = OpVariable(t_prv_bool, spv::StorageClass::Private, v_false);
             Name(id, fmt::format("flow_var_{}", static_cast<u32>(i)));
             flow_variables.emplace(i, AddGlobalVariable(id));
@@ -1373,12 +1333,6 @@ private:
             return {};
         }
 
-        if (const auto func_call = std::get_if<FunctionCallNode>(&*node)) {
-            const u32 func_id = func_call->GetFuncId();
-            OpFunctionCall(t_void, other_functions[func_id - 1]);
-            return {};
-        }
-
         if (const auto comment = std::get_if<CommentNode>(&*node)) {
             if (device.HasDebuggingToolAttached()) {
                 // We should insert comments with OpString instead of using named variables
@@ -2170,7 +2124,7 @@ private:
 
         OpBranchConditional(condition, true_label, discard_label);
         AddLabel(discard_label);
-        SafeKill();
+        OpKill();
         AddLabel(true_label);
     }
 
@@ -2221,9 +2175,7 @@ private:
     }
 
     Expression Exit(Operation operation) {
-        if (context_func->IsMain()) {
         PreExit();
-        }
         inside_branch = true;
         if (conditional_branch_set) {
             OpReturn();
@@ -2240,12 +2192,12 @@ private:
     Expression Discard(Operation operation) {
         inside_branch = true;
         if (conditional_branch_set) {
-            SafeKill();
+            OpKill();
         } else {
             const Id dummy = OpLabel();
             OpBranch(dummy);
             AddLabel(dummy);
-            SafeKill();
+            OpKill();
             AddLabel();
         }
         return {};
@@ -2324,7 +2276,7 @@ private:
     }
 
     Expression Barrier(Operation) {
-        if (!context_func->IsDecompiled()) {
+        if (!ir.IsDecompiled()) {
             LOG_ERROR(Render_Vulkan, "OpBarrier used by shader is not decompiled");
             return {};
         }
@@ -2818,8 +2770,6 @@ private:
     const Specialization& specialization;
     std::unordered_map<u8, VaryingTFB> transform_feedback;
 
-    std::shared_ptr<ShaderFunctionIR> context_func;
-
     const Id t_void = Name(TypeVoid(), "void");
 
     const Id t_bool = Name(TypeBool(), "bool");
@@ -2946,8 +2896,7 @@ private:
     Id ssy_flow_stack{};
     Id pbk_flow_stack{};
     Id continue_label{};
-    std::vector<std::map<u32, Id>> labels;
+    std::map<u32, Id> labels;
-    std::vector<Id> other_functions;
 
     bool conditional_branch_set{};
     bool inside_branch{};
@@ -3098,11 +3047,9 @@ public:
             decomp.OpBranchConditional(condition, then_label, endif_label);
             decomp.AddLabel(then_label);
             if (ast.kills) {
-                decomp.SafeKill();
+                decomp.OpKill();
             } else {
-                if (decomp.context_func->IsMain()) {
                 decomp.PreExit();
-                }
                 decomp.OpReturn();
             }
             decomp.AddLabel(endif_label);
@@ -3111,11 +3058,9 @@ public:
             decomp.OpBranch(next_block);
             decomp.AddLabel(next_block);
             if (ast.kills) {
-                decomp.SafeKill();
+                decomp.OpKill();
             } else {
-                if (decomp.context_func->IsMain()) {
                 decomp.PreExit();
-                }
                 decomp.OpReturn();
             }
             decomp.AddLabel(decomp.OpLabel());
@@ -3152,7 +3097,7 @@ private:
 };
 
 void SPIRVDecompiler::DecompileAST() {
-    const u32 num_flow_variables = context_func->GetASTNumVariables();
+    const u32 num_flow_variables = ir.GetASTNumVariables();
     for (u32 i = 0; i < num_flow_variables; i++) {
         const Id id = OpVariable(t_prv_bool, spv::StorageClass::Private, v_false);
         Name(id, fmt::format("flow_var_{}", i));
@@ -3161,7 +3106,7 @@ void SPIRVDecompiler::DecompileAST() {
 
     DefinePrologue();
 
-    const ASTNode program = context_func->GetASTProgram();
+    const ASTNode program = ir.GetASTProgram();
     ASTDecompiler decompiler{*this};
     decompiler.Visit(program);
 

src/video_core/shader/control_flow.cpp

@@ -7,7 +7,6 @@
 #include <set>
 #include <stack>
 #include <unordered_map>
-#include <unordered_set>
 #include <vector>
 
 #include "common/assert.h"
@@ -27,29 +26,17 @@ using Tegra::Shader::OpCode;
 
 constexpr s32 unassigned_branch = -2;
 
-enum class JumpLabel : u32 {
-    SSYClass = 0,
-    PBKClass = 1,
-};
-
-struct JumpItem {
-    JumpLabel type;
-    u32 address;
-
-    bool operator==(const JumpItem& other) const {
-        return std::tie(type, address) == std::tie(other.type, other.address);
-    }
-};
-
 struct Query {
     u32 address{};
-    std::stack<JumpItem> stack{};
+    std::stack<u32> ssy_stack{};
+    std::stack<u32> pbk_stack{};
 };
 
 struct BlockStack {
     BlockStack() = default;
-    explicit BlockStack(const Query& q) : stack{q.stack} {}
+    explicit BlockStack(const Query& q) : ssy_stack{q.ssy_stack}, pbk_stack{q.pbk_stack} {}
-    std::stack<JumpItem> stack{};
+    std::stack<u32> ssy_stack{};
+    std::stack<u32> pbk_stack{};
 };
 
 template <typename T, typename... Args>
@@ -78,36 +65,20 @@ struct BlockInfo {
     }
 };
 
-struct ProgramControl {
-    std::unordered_set<u32> found_functions{};
-    std::list<u32> pending_functions{};
-
-    void RegisterFunction(u32 address) {
-        if (found_functions.count(address) != 0) {
-            return;
-        }
-        found_functions.insert(address);
-        pending_functions.emplace_back(address);
-    }
-};
-
 struct CFGRebuildState {
-    explicit CFGRebuildState(ProgramControl& control_, const ProgramCode& program_code_, u32 start_,
+    explicit CFGRebuildState(const ProgramCode& program_code_, u32 start_, Registry& registry_)
-                             u32 base_start_, Registry& registry_)
+        : program_code{program_code_}, registry{registry_}, start{start_} {}
-        : control{control_}, program_code{program_code_}, registry{registry_}, start{start_},
-          base_start{base_start_} {}
 
-    ProgramControl& control;
     const ProgramCode& program_code;
     Registry& registry;
     u32 start{};
-    u32 base_start{};
     std::vector<BlockInfo> block_info;
     std::list<u32> inspect_queries;
     std::list<Query> queries;
     std::unordered_map<u32, u32> registered;
     std::set<u32> labels;
-    std::map<u32, JumpItem> jump_labels;
+    std::map<u32, u32> ssy_labels;
+    std::map<u32, u32> pbk_labels;
     std::unordered_map<u32, BlockStack> stacks;
     ASTManager* manager{};
 };
@@ -182,7 +153,7 @@ template <typename Result, typename TestCallable, typename PackCallable>
 std::optional<Result> TrackInstruction(const CFGRebuildState& state, u32& pos, TestCallable test,
                                        PackCallable pack) {
     for (; pos >= state.start; --pos) {
-        if (IsSchedInstruction(pos, state.base_start)) {
+        if (IsSchedInstruction(pos, state.start)) {
             continue;
         }
         const Instruction instr = state.program_code[pos];
@@ -291,7 +262,7 @@ std::pair<ParseResult, ParseInfo> ParseCode(CFGRebuildState& state, u32 address)
             single_branch.ignore = true;
             break;
         }
-        if (IsSchedInstruction(offset, state.base_start)) {
+        if (IsSchedInstruction(offset, state.start)) {
             offset++;
             continue;
         }
@@ -303,7 +274,6 @@ std::pair<ParseResult, ParseInfo> ParseCode(CFGRebuildState& state, u32 address)
         }
 
         switch (opcode->get().GetId()) {
-        case OpCode::Id::RET:
         case OpCode::Id::EXIT: {
             const auto pred_index = static_cast<u32>(instr.pred.pred_index);
             single_branch.condition.predicate = GetPredicate(pred_index, instr.negate_pred != 0);
@@ -441,20 +411,13 @@ std::pair<ParseResult, ParseInfo> ParseCode(CFGRebuildState& state, u32 address)
         case OpCode::Id::SSY: {
             const u32 target = offset + instr.bra.GetBranchTarget();
             insert_label(state, target);
-            JumpItem it = {JumpLabel::SSYClass, target};
+            state.ssy_labels.emplace(offset, target);
-            state.jump_labels.emplace(offset, it);
             break;
         }
         case OpCode::Id::PBK: {
             const u32 target = offset + instr.bra.GetBranchTarget();
             insert_label(state, target);
-            JumpItem it = {JumpLabel::PBKClass, target};
+            state.pbk_labels.emplace(offset, target);
-            state.jump_labels.emplace(offset, it);
-            break;
-        }
-        case OpCode::Id::CAL: {
-            const u32 target = offset + instr.bra.GetBranchTarget();
-            state.control.RegisterFunction(target);
             break;
         }
         case OpCode::Id::BRX: {
@@ -550,7 +513,7 @@ bool TryInspectAddress(CFGRebuildState& state) {
 }
 
 bool TryQuery(CFGRebuildState& state) {
-    const auto gather_labels = [](std::stack<JumpItem>& cc, std::map<u32, JumpItem>& labels,
+    const auto gather_labels = [](std::stack<u32>& cc, std::map<u32, u32>& labels,
                                   BlockInfo& block) {
         auto gather_start = labels.lower_bound(block.start);
         const auto gather_end = labels.upper_bound(block.end);
@@ -559,19 +522,6 @@ bool TryQuery(CFGRebuildState& state) {
             ++gather_start;
         }
     };
-    const auto pop_labels = [](JumpLabel type, SingleBranch* branch, Query& query) -> bool {
-        while (!query.stack.empty() && query.stack.top().type != type) {
-            query.stack.pop();
-        }
-        if (query.stack.empty()) {
-            return false;
-        }
-        if (branch->address == unassigned_branch) {
-            branch->address = query.stack.top().address;
-        }
-        query.stack.pop();
-        return true;
-    };
     if (state.queries.empty()) {
         return false;
     }
@@ -584,7 +534,8 @@ bool TryQuery(CFGRebuildState& state) {
     // consumes a label. Schedule new queries accordingly
     if (block.visited) {
         BlockStack& stack = state.stacks[q.address];
-        const bool all_okay = (stack.stack.empty() || q.stack == stack.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;
     }
@@ -593,7 +544,8 @@ bool TryQuery(CFGRebuildState& state) {
 
     Query q2(q);
     state.queries.pop_front();
-    gather_labels(q2.stack, state.jump_labels, block);
+    gather_labels(q2.ssy_stack, state.ssy_labels, block);
+    gather_labels(q2.pbk_stack, state.pbk_labels, block);
     if (std::holds_alternative<SingleBranch>(*block.branch)) {
         auto* branch = std::get_if<SingleBranch>(block.branch.get());
         if (!branch->condition.IsUnconditional()) {
@@ -603,10 +555,16 @@ bool TryQuery(CFGRebuildState& state) {
 
         auto& conditional_query = state.queries.emplace_back(q2);
         if (branch->is_sync) {
-            pop_labels(JumpLabel::SSYClass, branch, conditional_query);
+            if (branch->address == unassigned_branch) {
+                branch->address = conditional_query.ssy_stack.top();
+            }
+            conditional_query.ssy_stack.pop();
         }
         if (branch->is_brk) {
-            pop_labels(JumpLabel::PBKClass, branch, conditional_query);
+            if (branch->address == unassigned_branch) {
+                branch->address = conditional_query.pbk_stack.top();
+            }
+            conditional_query.pbk_stack.pop();
         }
         conditional_query.address = branch->address;
         return true;
@@ -688,23 +646,25 @@ void DecompileShader(CFGRebuildState& state) {
     state.manager->Decompile();
 }
 
-ShaderFunction ScanFunction(ProgramControl& control, const ProgramCode& program_code,
+} // Anonymous namespace
-                            u32 start_address, u32 base_start, const CompilerSettings& settings,
+
+std::unique_ptr<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u32 start_address,
+                                                const CompilerSettings& settings,
                                                 Registry& registry) {
-    ShaderFunction result_out{};
+    auto result_out = std::make_unique<ShaderCharacteristics>();
     if (settings.depth == CompileDepth::BruteForce) {
-        result_out.settings.depth = CompileDepth::BruteForce;
+        result_out->settings.depth = CompileDepth::BruteForce;
         return result_out;
     }
 
-    CFGRebuildState state{control, program_code, start_address, base_start, registry};
+    CFGRebuildState state{program_code, start_address, registry};
     // Inspect Code and generate blocks
     state.labels.clear();
     state.labels.emplace(start_address);
     state.inspect_queries.push_back(state.start);
     while (!state.inspect_queries.empty()) {
         if (!TryInspectAddress(state)) {
-            result_out.settings.depth = CompileDepth::BruteForce;
+            result_out->settings.depth = CompileDepth::BruteForce;
             return result_out;
         }
     }
@@ -715,7 +675,7 @@ ShaderFunction ScanFunction(ProgramControl& control, const ProgramCode& program_
 
     if (settings.depth != CompileDepth::FlowStack) {
         // Decompile Stacks
-        state.queries.push_back(Query{state.start, {}});
+        state.queries.push_back(Query{state.start, {}, {}});
         decompiled = true;
         while (!state.queries.empty()) {
             if (!TryQuery(state)) {
@@ -745,18 +705,19 @@ ShaderFunction ScanFunction(ProgramControl& control, const ProgramCode& program_
             state.manager->ShowCurrentState("Of Shader");
             state.manager->Clear();
         } else {
-            result_out.start = start_address;
+            auto characteristics = std::make_unique<ShaderCharacteristics>();
-            result_out.settings.depth = settings.depth;
+            characteristics->start = start_address;
-            result_out.manager = std::move(manager);
+            characteristics->settings.depth = settings.depth;
-            result_out.end = state.block_info.back().end + 1;
+            characteristics->manager = std::move(manager);
-            return result_out;
+            characteristics->end = state.block_info.back().end + 1;
+            return characteristics;
         }
     }
 
-    result_out.start = start_address;
+    result_out->start = start_address;
-    result_out.settings.depth =
+    result_out->settings.depth =
         use_flow_stack ? CompileDepth::FlowStack : CompileDepth::NoFlowStack;
-    result_out.blocks.clear();
+    result_out->blocks.clear();
     for (auto& block : state.block_info) {
         ShaderBlock new_block{};
         new_block.start = block.start;
@@ -765,20 +726,20 @@ ShaderFunction ScanFunction(ProgramControl& control, const ProgramCode& program_
         if (!new_block.ignore_branch) {
             new_block.branch = block.branch;
         }
-        result_out.end = std::max(result_out.end, block.end);
+        result_out->end = std::max(result_out->end, block.end);
-        result_out.blocks.push_back(new_block);
+        result_out->blocks.push_back(new_block);
     }
     if (!use_flow_stack) {
-        result_out.labels = std::move(state.labels);
+        result_out->labels = std::move(state.labels);
         return result_out;
     }
 
-    auto back = result_out.blocks.begin();
+    auto back = result_out->blocks.begin();
     auto next = std::next(back);
-    while (next != result_out.blocks.end()) {
+    while (next != result_out->blocks.end()) {
         if (!state.labels.contains(next->start) && next->start == back->end + 1) {
             back->end = next->end;
-            next = result_out.blocks.erase(next);
+            next = result_out->blocks.erase(next);
             continue;
         }
         back = next;
@@ -787,22 +748,4 @@ ShaderFunction ScanFunction(ProgramControl& control, const ProgramCode& program_
 
     return result_out;
 }
-
-} // Anonymous namespace
-
-std::unique_ptr<ShaderProgram> ScanFlow(const ProgramCode& program_code, u32 start_address,
-                                        const CompilerSettings& settings, Registry& registry) {
-    ProgramControl control{};
-    auto result_out = std::make_unique<ShaderProgram>();
-    result_out->main =
-        ScanFunction(control, program_code, start_address, start_address, settings, registry);
-    while (!control.pending_functions.empty()) {
-        u32 address = control.pending_functions.front();
-        auto fun = ScanFunction(control, program_code, address, start_address, settings, registry);
-        result_out->subfunctions.emplace(address, std::move(fun));
-        control.pending_functions.pop_front();
-    }
-    return result_out;
-}
-
 } // namespace VideoCommon::Shader

src/video_core/shader/control_flow.h

@@ -5,7 +5,6 @@
 #pragma once
 
 #include <list>
-#include <map>
 #include <optional>
 #include <set>
 #include <variant>
@@ -102,7 +101,7 @@ struct ShaderBlock {
     }
 };
 
-struct ShaderFunction {
+struct ShaderCharacteristics {
     std::list<ShaderBlock> blocks{};
     std::set<u32> labels{};
     u32 start{};
@@ -111,12 +110,8 @@ struct ShaderFunction {
     CompilerSettings settings{};
 };
 
-struct ShaderProgram {
+std::unique_ptr<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u32 start_address,
-    ShaderFunction main;
+                                                const CompilerSettings& settings,
-    std::map<u32, ShaderFunction> subfunctions;
+                                                Registry& registry);
-};
-
-std::unique_ptr<ShaderProgram> ScanFlow(const ProgramCode& program_code, u32 start_address,
-                                        const CompilerSettings& settings, Registry& registry);
 
 } // namespace VideoCommon::Shader

src/video_core/shader/decode.cpp

@@ -64,52 +64,9 @@ std::optional<u32> TryDeduceSamplerSize(const SamplerEntry& sampler_to_deduce,
 
 } // Anonymous namespace
 
-class ExprDecoder {
-public:
-    explicit ExprDecoder(ShaderIR& ir_) : ir(ir_) {}
-
-    void operator()(const ExprAnd& expr) {
-        Visit(expr.operand1);
-        Visit(expr.operand2);
-    }
-
-    void operator()(const ExprOr& expr) {
-        Visit(expr.operand1);
-        Visit(expr.operand2);
-    }
-
-    void operator()(const ExprNot& expr) {
-        Visit(expr.operand1);
-    }
-
-    void operator()(const ExprPredicate& expr) {
-        const auto pred = static_cast<Tegra::Shader::Pred>(expr.predicate);
-        if (pred != Pred::UnusedIndex && pred != Pred::NeverExecute) {
-            ir.used_predicates.insert(pred);
-        }
-    }
-
-    void operator()(const ExprCondCode& expr) {}
-
-    void operator()(const ExprVar& expr) {}
-
-    void operator()(const ExprBoolean& expr) {}
-
-    void operator()(const ExprGprEqual& expr) {
-        ir.used_registers.insert(expr.gpr);
-    }
-
-    void Visit(const Expr& node) {
-        return std::visit(*this, *node);
-    }
-
-private:
-    ShaderIR& ir;
-};
-
 class ASTDecoder {
 public:
-    explicit ASTDecoder(ShaderIR& ir_) : ir(ir_), decoder(ir_) {}
+    explicit ASTDecoder(ShaderIR& ir_) : ir(ir_) {}
 
     void operator()(ASTProgram& ast) {
         ASTNode current = ast.nodes.GetFirst();
@@ -120,7 +77,6 @@ public:
     }
 
     void operator()(ASTIfThen& ast) {
-        decoder.Visit(ast.condition);
         ASTNode current = ast.nodes.GetFirst();
         while (current) {
             Visit(current);
@@ -140,18 +96,13 @@ public:
 
     void operator()(ASTBlockDecoded& ast) {}
 
-    void operator()(ASTVarSet& ast) {
+    void operator()(ASTVarSet& ast) {}
-        decoder.Visit(ast.condition);
-    }
 
     void operator()(ASTLabel& ast) {}
 
-    void operator()(ASTGoto& ast) {
+    void operator()(ASTGoto& ast) {}
-        decoder.Visit(ast.condition);
-    }
 
     void operator()(ASTDoWhile& ast) {
-        decoder.Visit(ast.condition);
         ASTNode current = ast.nodes.GetFirst();
         while (current) {
             Visit(current);
@@ -159,13 +110,9 @@ public:
         }
     }
 
-    void operator()(ASTReturn& ast) {
+    void operator()(ASTReturn& ast) {}
-        decoder.Visit(ast.condition);
-    }
 
-    void operator()(ASTBreak& ast) {
+    void operator()(ASTBreak& ast) {}
-        decoder.Visit(ast.condition);
-    }
 
     void Visit(ASTNode& node) {
         std::visit(*this, *node->GetInnerData());
@@ -178,12 +125,16 @@ public:
 
 private:
     ShaderIR& ir;
-    ExprDecoder decoder;
 };
 
 void ShaderIR::Decode() {
-    const auto decode_function = ([this](ShaderFunction& shader_info) {
+    std::memcpy(&header, program_code.data(), sizeof(Tegra::Shader::Header));
-        coverage_end = std::max<u32>(0, shader_info.end);
+
+    decompiled = false;
+    auto info = ScanFlow(program_code, main_offset, settings, registry);
+    auto& shader_info = *info;
+    coverage_begin = shader_info.start;
+    coverage_end = shader_info.end;
     switch (shader_info.settings.depth) {
     case CompileDepth::FlowStack: {
         for (const auto& block : shader_info.blocks) {
@@ -224,7 +175,7 @@ void ShaderIR::Decode() {
         disable_flow_stack = true;
         decompiled = true;
         ASTDecoder decoder{*this};
-            ASTNode program = program_manager.GetProgram();
+        ASTNode program = GetASTProgram();
         decoder.Visit(program);
         break;
     }
@@ -243,48 +194,8 @@ void ShaderIR::Decode() {
     }
     if (settings.depth != shader_info.settings.depth) {
         LOG_WARNING(
-                HW_GPU,
+            HW_GPU, "Decompiling to this setting \"{}\" failed, downgrading to this setting \"{}\"",
-                "Decompiling to this setting \"{}\" failed, downgrading to this setting \"{}\"",
+            CompileDepthAsString(settings.depth), CompileDepthAsString(shader_info.settings.depth));
-                CompileDepthAsString(settings.depth),
-                CompileDepthAsString(shader_info.settings.depth));
-        }
-    });
-    const auto gen_function =
-        ([this](ShaderFunction& shader_info, u32 id) -> std::shared_ptr<ShaderFunctionIR> {
-            std::shared_ptr<ShaderFunctionIR> result;
-            if (decompiled) {
-                result = std::make_shared<ShaderFunctionIR>(std::move(program_manager), id,
-                                                            shader_info.start, shader_info.end);
-            } else {
-                result =
-                    std::make_shared<ShaderFunctionIR>(std::move(basic_blocks), disable_flow_stack,
-                                                       id, shader_info.start, shader_info.end);
-            }
-            decompiled = false;
-            disable_flow_stack = false;
-            basic_blocks.clear();
-            program_manager.Clear();
-            return result;
-        });
-    std::memcpy(&header, program_code.data(), sizeof(Tegra::Shader::Header));
-
-    decompiled = false;
-    auto info = ScanFlow(program_code, main_offset, settings, registry);
-    u32 id_start = 1;
-    for (auto& pair : info->subfunctions) {
-        func_map.emplace(pair.first, id_start);
-        id_start++;
-    }
-    coverage_begin = info->main.start;
-    coverage_end = 0;
-    decode_function(info->main);
-    main_function = gen_function(info->main, 0);
-    subfunctions.resize(info->subfunctions.size());
-    for (auto& pair : info->subfunctions) {
-        auto& func_info = pair.second;
-        decode_function(func_info);
-        u32 id = func_map[pair.first];
-        subfunctions[id - 1] = gen_function(func_info, id);
     }
 }
 

src/video_core/shader/decode/other.cpp

@@ -33,7 +33,6 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
         // With the previous preconditions, this instruction is a no-operation.
         break;
     }
-    case OpCode::Id::RET:
     case OpCode::Id::EXIT: {
         const ConditionCode cc = instr.flow_condition_code;
         UNIMPLEMENTED_IF_MSG(cc != ConditionCode::T, "EXIT condition code used: {}", cc);
@@ -313,16 +312,6 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
         LOG_DEBUG(HW_GPU, "DEPBAR instruction is stubbed");
         break;
     }
-    case OpCode::Id::CAL: {
-        const u32 target = pc + instr.bra.GetBranchTarget();
-        const auto it = func_map.find(target);
-        if (it == func_map.end()) {
-            UNREACHABLE();
-            break;
-        }
-        bb.push_back(FunctionCall(it->second));
-        break;
-    }
     default:
         UNIMPLEMENTED_MSG("Unhandled instruction: {}", opcode->get().GetName());
     }

src/video_core/shader/decode/texture.cpp

@@ -339,6 +339,8 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         const 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");
 
         const Node4 components = GetTldsCode(instr, texture_type, is_array);
@@ -820,7 +822,7 @@ Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is
     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 && !aoffi_enabled ? last_coord_register : (coord_register + i)));
+            GetRegister(last && !aoffi_enabled ? last_coord_register : coord_register + i));
     }
 
     const Node array = is_array ? GetRegister(array_register) : nullptr;

src/video_core/shader/node.h

@@ -267,11 +267,10 @@ class PatchNode;
 class SmemNode;
 class GmemNode;
 class CommentNode;
-class FunctionCallNode;
 
 using NodeData = std::variant<OperationNode, ConditionalNode, GprNode, CustomVarNode, ImmediateNode,
                               InternalFlagNode, PredicateNode, AbufNode, PatchNode, CbufNode,
-                              LmemNode, SmemNode, GmemNode, FunctionCallNode, CommentNode>;
+                              LmemNode, SmemNode, GmemNode, CommentNode>;
 using Node = std::shared_ptr<NodeData>;
 using Node4 = std::array<Node, 4>;
 using NodeBlock = std::vector<Node>;
@@ -495,18 +494,6 @@ private:
     std::vector<Node> code; ///< Code to execute
 };
 
-class FunctionCallNode final : public AmendNode {
-public:
-    explicit FunctionCallNode(u32 func_id_) : func_id{func_id_} {}
-
-    [[nodiscard]] u32 GetFuncId() const {
-        return func_id;
-    }
-
-private:
-    u32 func_id; ///< Id of the function to call
-};
-
 /// A general purpose register
 class GprNode final {
 public:

src/video_core/shader/node_helper.cpp

@@ -19,11 +19,6 @@ Node Comment(std::string text) {
     return MakeNode<CommentNode>(std::move(text));
 }
 
-/// Creates a function call
-Node FunctionCall(u32 func_id) {
-    return MakeNode<FunctionCallNode>(func_id);
-}
-
 Node Immediate(u32 value) {
     return MakeNode<ImmediateNode>(value);
 }

src/video_core/shader/node_helper.h

@@ -27,9 +27,6 @@ Node Conditional(Node condition, std::vector<Node> code);
 /// Creates a commentary node
 Node Comment(std::string text);
 
-/// Creates a function call
-Node FunctionCall(u32 func_id);
-
 /// Creates an u32 immediate
 Node Immediate(u32 value);
 

src/video_core/shader/shader_ir.h

@@ -26,7 +26,7 @@ namespace VideoCommon::Shader {
 
 struct ShaderBlock;
 
-constexpr u32 MAX_PROGRAM_LENGTH = 0x2000;
+constexpr u32 MAX_PROGRAM_LENGTH = 0x1000;
 
 struct ConstBuffer {
     constexpr explicit ConstBuffer(u32 max_offset_, bool is_indirect_)
@@ -64,68 +64,16 @@ struct GlobalMemoryUsage {
     bool is_written{};
 };
 
-class ShaderFunctionIR final {
-public:
-    explicit ShaderFunctionIR(std::map<u32, NodeBlock>&& basic_blocks_, bool disable_flow_stack_,
-                              u32 id_, u32 coverage_begin_, u32 coverage_end_)
-        : basic_blocks{std::move(basic_blocks_)}, decompiled{false},
-          disable_flow_stack{disable_flow_stack_}, id{id_}, coverage_begin{coverage_begin_},
-          coverage_end{coverage_end_} {}
-    explicit ShaderFunctionIR(ASTManager&& program_manager_, u32 id_, u32 coverage_begin_,
-                              u32 coverage_end_)
-        : program_manager{std::move(program_manager_)}, decompiled{true}, disable_flow_stack{true},
-          id{id_}, coverage_begin{coverage_begin_}, coverage_end{coverage_end_} {}
-
-    const std::map<u32, NodeBlock>& GetBasicBlocks() const {
-        return basic_blocks;
-    }
-
-    [[nodiscard]] bool IsFlowStackDisabled() const {
-        return disable_flow_stack;
-    }
-
-    [[nodiscard]] bool IsDecompiled() const {
-        return decompiled;
-    }
-
-    const ASTManager& GetASTManager() const {
-        return program_manager;
-    }
-
-    [[nodiscard]] ASTNode GetASTProgram() const {
-        return program_manager.GetProgram();
-    }
-
-    [[nodiscard]] u32 GetASTNumVariables() const {
-        return program_manager.GetVariables();
-    }
-
-    [[nodiscard]] bool IsMain() const {
-        return id == 0;
-    }
-
-    [[nodiscard]] u32 GetId() const {
-        return id;
-    }
-
-private:
-    std::map<u32, NodeBlock> basic_blocks;
-    ASTManager program_manager{true, true};
-
-    bool decompiled{};
-    bool disable_flow_stack{};
-    u32 id{};
-
-    u32 coverage_begin{};
-    u32 coverage_end{};
-};
-
 class ShaderIR final {
 public:
     explicit ShaderIR(const ProgramCode& program_code_, u32 main_offset_,
                       CompilerSettings settings_, Registry& registry_);
     ~ShaderIR();
 
+    const std::map<u32, NodeBlock>& GetBasicBlocks() const {
+        return basic_blocks;
+    }
+
     const std::set<u32>& GetRegisters() const {
         return used_registers;
     }
@@ -207,6 +155,26 @@ public:
         return header;
     }
 
+    bool IsFlowStackDisabled() const {
+        return disable_flow_stack;
+    }
+
+    bool IsDecompiled() const {
+        return decompiled;
+    }
+
+    const ASTManager& GetASTManager() const {
+        return program_manager;
+    }
+
+    ASTNode GetASTProgram() const {
+        return program_manager.GetProgram();
+    }
+
+    u32 GetASTNumVariables() const {
+        return program_manager.GetVariables();
+    }
+
     u32 ConvertAddressToNvidiaSpace(u32 address) const {
         return (address - main_offset) * static_cast<u32>(sizeof(Tegra::Shader::Instruction));
     }
@@ -222,16 +190,7 @@ public:
         return num_custom_variables;
     }
 
-    std::shared_ptr<ShaderFunctionIR> GetMainFunction() const {
-        return main_function;
-    }
-
-    const std::vector<std::shared_ptr<ShaderFunctionIR>>& GetSubFunctions() const {
-        return subfunctions;
-    }
-
 private:
-    friend class ExprDecoder;
     friend class ASTDecoder;
 
     struct SamplerInfo {
@@ -494,10 +453,6 @@ private:
     std::vector<Node> amend_code;
     u32 num_custom_variables{};
 
-    std::shared_ptr<ShaderFunctionIR> main_function;
-    std::vector<std::shared_ptr<ShaderFunctionIR>> subfunctions;
-    std::unordered_map<u32, u32> func_map;
-
     std::set<u32> used_registers;
     std::set<Tegra::Shader::Pred> used_predicates;
     std::set<Tegra::Shader::Attribute::Index> used_input_attributes;