2 files changed, 75 insertions, 21 deletions

diff --git a/src/video_core/debug_utils/debug_utils.cpp b/src/video_core/debug_utils/debug_utils.cpp
index c3a9c9598..2f645b441 100644
--- a/src/video_core/debug_utils/debug_utils.cpp
+++ b/src/video_core/debug_utils/debug_utils.cpp
@@ -4,35 +4,41 @@
 
 #include <algorithm>
 #include <condition_variable>
+#include <cstdint>
 #include <cstring>
 #include <fstream>
-#include <list>
 #include <map>
 #include <mutex>
+#include <stdexcept>
 #include <string>
 
 #ifdef HAVE_PNG
 #include <png.h>
+#include <setjmp.h>
 #endif
 
+#include <nihstro/bit_field.h>
 #include <nihstro/float24.h>
 #include <nihstro/shader_binary.h>
 
 #include "common/assert.h"
+#include "common/bit_field.h"
 #include "common/color.h"
 #include "common/common_types.h"
 #include "common/file_util.h"
+#include "common/logging/log.h"
 #include "common/math_util.h"
 #include "common/vector_math.h"
 
-#include "core/settings.h"
-
+#include "video_core/debug_utils/debug_utils.h"
 #include "video_core/pica.h"
 #include "video_core/pica_state.h"
+#include "video_core/pica_types.h"
+#include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
+#include "video_core/shader/shader.h"
 #include "video_core/utils.h"
 #include "video_core/video_core.h"
-#include "video_core/debug_utils/debug_utils.h"
 
 using nihstro::DVLBHeader;
 using nihstro::DVLEHeader;
@@ -40,15 +46,12 @@ using nihstro::DVLPHeader;
 
 namespace Pica {
 
-void DebugContext::OnEvent(Event event, void* data) {
-    if (!breakpoints[event].enabled)
-        return;
-
+void DebugContext::DoOnEvent(Event event, void* data) {
     {
         std::unique_lock<std::mutex> lock(breakpoint_mutex);
 
-        // Commit the hardware renderer's framebuffer so it will show on debug widgets
-        VideoCore::g_renderer->Rasterizer()->FlushFramebuffer();
+        // Commit the rasterizer's caches so framebuffers, render targets, etc. will show on debug widgets
+        VideoCore::g_renderer->Rasterizer()->FlushAll();
 
         // TODO: Should stop the CPU thread here once we multithread emulation.
 
@@ -205,11 +208,12 @@ void DumpShader(const std::string& filename, const Regs::ShaderConfig& config, c
 
     // TODO: Reduce the amount of binary code written to relevant portions
     dvlp.binary_offset = write_offset - dvlp_offset;
-    dvlp.binary_size_words = setup.program_code.size();
-    QueueForWriting(reinterpret_cast<const u8*>(setup.program_code.data()), setup.program_code.size() * sizeof(u32));
+    dvlp.binary_size_words = static_cast<uint32_t>(setup.program_code.size());
+    QueueForWriting(reinterpret_cast<const u8*>(setup.program_code.data()),
+                    static_cast<u32>(setup.program_code.size()) * sizeof(u32));
 
     dvlp.swizzle_info_offset = write_offset - dvlp_offset;
-    dvlp.swizzle_info_num_entries = setup.swizzle_data.size();
+    dvlp.swizzle_info_num_entries = static_cast<uint32_t>(setup.swizzle_data.size());
     u32 dummy = 0;
     for (unsigned int i = 0; i < setup.swizzle_data.size(); ++i) {
         QueueForWriting(reinterpret_cast<const u8*>(&setup.swizzle_data[i]), sizeof(setup.swizzle_data[i]));
@@ -261,7 +265,7 @@ void DumpShader(const std::string& filename, const Regs::ShaderConfig& config, c
             constant_table.emplace_back(constant);
     }
     dvle.constant_table_offset = write_offset - dvlb.dvle_offset;
-    dvle.constant_table_size = constant_table.size();
+    dvle.constant_table_size = static_cast<uint32_t>(constant_table.size());
     for (const auto& constant : constant_table) {
         QueueForWriting(reinterpret_cast<const u8*>(&constant), sizeof(constant));
     }
diff --git a/src/video_core/debug_utils/debug_utils.h b/src/video_core/debug_utils/debug_utils.h
index 7df941619..f628292a4 100644
--- a/src/video_core/debug_utils/debug_utils.h
+++ b/src/video_core/debug_utils/debug_utils.h
@@ -4,23 +4,33 @@
 
 #pragma once
 
+#include <algorithm>
 #include <array>
 #include <condition_variable>
+#include <iterator>
 #include <list>
 #include <map>
 #include <memory>
 #include <mutex>
+#include <string>
+#include <utility>
 #include <vector>
 
+#include "common/common_types.h"
 #include "common/vector_math.h"
 
-#include "core/tracer/recorder.h"
-
 #include "video_core/pica.h"
-#include "video_core/shader/shader.h"
+
+namespace CiTrace {
+class Recorder;
+}
 
 namespace Pica {
 
+namespace Shader {
+struct ShaderSetup;
+}
+
 class DebugContext {
 public:
     enum class Event {
@@ -30,7 +40,7 @@ public:
         PicaCommandProcessed,
         IncomingPrimitiveBatch,
         FinishedPrimitiveBatch,
-        VertexLoaded,
+        VertexShaderInvocation,
         IncomingDisplayTransfer,
         GSPCommandProcessed,
         BufferSwapped,
@@ -114,7 +124,15 @@ public:
      * @param event Event which has happened
      * @param data Optional data pointer (pass nullptr if unused). Needs to remain valid until Resume() is called.
      */
-    void OnEvent(Event event, void* data);
+    void OnEvent(Event event, void* data) {
+        // This check is left in the header to allow the compiler to inline it.
+        if (!breakpoints[(int)event].enabled)
+            return;
+        // For the rest of event handling, call a separate function.
+        DoOnEvent(event, data);
+    }
+
+    void DoOnEvent(Event event, void *data);
 
     /**
      * Resume from the current breakpoint.
@@ -126,12 +144,14 @@ public:
      * Delete all set breakpoints and resume emulation.
      */
     void ClearBreakpoints() {
-        breakpoints.clear();
+        for (auto &bp : breakpoints) {
+            bp.enabled = false;
+        }
         Resume();
     }
 
     // TODO: Evaluate if access to these members should be hidden behind a public interface.
-    std::map<Event, BreakPoint> breakpoints;
+    std::array<BreakPoint, (int)Event::NumEvents> breakpoints;
     Event active_breakpoint;
     bool at_breakpoint = false;
 
@@ -206,6 +226,36 @@ void DumpTexture(const Pica::Regs::TextureConfig& texture_config, u8* data);
 
 void DumpTevStageConfig(const std::array<Pica::Regs::TevStageConfig,6>& stages);
 
+/**
+ * Used in the vertex loader to merge access records. TODO: Investigate if actually useful.
+ */
+class MemoryAccessTracker {
+    /// Combine overlapping and close ranges
+    void SimplifyRanges() {
+        for (auto it = ranges.begin(); it != ranges.end(); ++it) {
+            // NOTE: We add 32 to the range end address to make sure "close" ranges are combined, too
+            auto it2 = std::next(it);
+            while (it2 != ranges.end() && it->first + it->second + 32 >= it2->first) {
+                it->second = std::max(it->second, it2->first + it2->second - it->first);
+                it2 = ranges.erase(it2);
+            }
+        }
+    }
+
+public:
+    /// Record a particular memory access in the list
+    void AddAccess(u32 paddr, u32 size) {
+        // Create new range or extend existing one
+        ranges[paddr] = std::max(ranges[paddr], size);
+
+        // Simplify ranges...
+        SimplifyRanges();
+    }
+
+    /// Map of accessed ranges (mapping start address to range size)
+    std::map<u32, u32> ranges;
+};
+
 } // namespace
 
 } // namespace