Merge pull request #702 from neobrain/citrace

Add CiTrace recording support.

3 files changed, 91 insertions, 27 deletions

diff --git a/src/core/hw/gpu.cpp b/src/core/hw/gpu.cpp
index a1789f9c7..a3a7d128f 100644
--- a/src/core/hw/gpu.cpp
+++ b/src/core/hw/gpu.cpp
@@ -21,12 +21,17 @@
 #include "core/hw/hw.h"
 #include "core/hw/gpu.h"
 
+#include "core/tracer/recorder.h"
+
 #include "video_core/command_processor.h"
 #include "video_core/hwrasterizer_base.h"
 #include "video_core/renderer_base.h"
 #include "video_core/utils.h"
 #include "video_core/video_core.h"
 
+#include "video_core/debug_utils/debug_utils.h"
+
+
 namespace GPU {
 
 Regs g_regs;
@@ -101,39 +106,43 @@ inline void Write(u32 addr, const T data) {
         const bool is_second_filler = (index != GPU_REG_INDEX(memory_fill_config[0].trigger));
         auto& config = g_regs.memory_fill_config[is_second_filler];
 
-        if (config.address_start && config.trigger) {
-            u8* start = Memory::GetPhysicalPointer(config.GetStartAddress());
-            u8* end = Memory::GetPhysicalPointer(config.GetEndAddress());
-
-            if (config.fill_24bit) {
-                // fill with 24-bit values
-                for (u8* ptr = start; ptr < end; ptr += 3) {
-                    ptr[0] = config.value_24bit_r;
-                    ptr[1] = config.value_24bit_g;
-                    ptr[2] = config.value_24bit_b;
+        if (config.trigger) {
+            if (config.address_start) { // Some games pass invalid values here
+                u8* start = Memory::GetPhysicalPointer(config.GetStartAddress());
+                u8* end = Memory::GetPhysicalPointer(config.GetEndAddress());
+
+                if (config.fill_24bit) {
+                    // fill with 24-bit values
+                    for (u8* ptr = start; ptr < end; ptr += 3) {
+                        ptr[0] = config.value_24bit_r;
+                        ptr[1] = config.value_24bit_g;
+                        ptr[2] = config.value_24bit_b;
+                    }
+                } else if (config.fill_32bit) {
+                    // fill with 32-bit values
+                    for (u32* ptr = (u32*)start; ptr < (u32*)end; ++ptr)
+                        *ptr = config.value_32bit;
+                } else {
+                    // fill with 16-bit values
+                    for (u16* ptr = (u16*)start; ptr < (u16*)end; ++ptr)
+                        *ptr = config.value_16bit;
                 }
-            } else if (config.fill_32bit) {
-                // fill with 32-bit values
-                for (u32* ptr = (u32*)start; ptr < (u32*)end; ++ptr)
-                    *ptr = config.value_32bit;
-            } else {
-                // fill with 16-bit values
-                for (u16* ptr = (u16*)start; ptr < (u16*)end; ++ptr)
-                    *ptr = config.value_16bit;
-            }
 
-            LOG_TRACE(HW_GPU, "MemoryFill from 0x%08x to 0x%08x", config.GetStartAddress(), config.GetEndAddress());
+                LOG_TRACE(HW_GPU, "MemoryFill from 0x%08x to 0x%08x", config.GetStartAddress(), config.GetEndAddress());
 
-            config.trigger = 0;
-            config.finished = 1;
+                if (!is_second_filler) {
+                    GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PSC0);
+                } else {
+                    GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PSC1);
+                }
 
-            if (!is_second_filler) {
-                GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PSC0);
-            } else {
-                GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PSC1);
+                VideoCore::g_renderer->hw_rasterizer->NotifyFlush(config.GetStartAddress(), config.GetEndAddress() - config.GetStartAddress());
             }
 
-            VideoCore::g_renderer->hw_rasterizer->NotifyFlush(config.GetStartAddress(), config.GetEndAddress() - config.GetStartAddress());
+            // Reset "trigger" flag and set the "finish" flag
+            // NOTE: This was confirmed to happen on hardware even if "address_start" is zero.
+            config.trigger = 0;
+            config.finished = 1;
         }
         break;
     }
@@ -270,6 +279,7 @@ inline void Write(u32 addr, const T data) {
                       config.GetPhysicalOutputAddress(), output_width, output_height,
                       config.output_format.Value(), config.flags);
 
+            g_regs.display_transfer_config.trigger = 0;
             GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PPF);
 
             VideoCore::g_renderer->hw_rasterizer->NotifyFlush(config.GetPhysicalOutputAddress(), output_size);
@@ -284,7 +294,14 @@ inline void Write(u32 addr, const T data) {
         if (config.trigger & 1)
         {
             u32* buffer = (u32*)Memory::GetPhysicalPointer(config.GetPhysicalAddress());
+
+            if (Pica::g_debug_context && Pica::g_debug_context->recorder) {
+                Pica::g_debug_context->recorder->MemoryAccessed((u8*)buffer, config.size * sizeof(u32), config.GetPhysicalAddress());
+            }
+
             Pica::CommandProcessor::ProcessCommandList(buffer, config.size);
+
+            g_regs.command_processor_config.trigger = 0;
         }
         break;
     }
@@ -292,6 +309,13 @@ inline void Write(u32 addr, const T data) {
     default:
         break;
     }
+
+    // Notify tracer about the register write
+    // This is happening *after* handling the write to make sure we properly catch all memory reads.
+    if (Pica::g_debug_context && Pica::g_debug_context->recorder) {
+        // addr + GPU VBase - IO VBase + IO PBase
+        Pica::g_debug_context->recorder->RegisterWritten<T>(addr + 0x1EF00000 - 0x1EC00000 + 0x10100000, data);
+    }
 }
 
 // Explicitly instantiate template functions because we aren't defining this in the header:
diff --git a/src/core/hw/hw.cpp b/src/core/hw/hw.cpp
index c7006a498..b5fdbf9c1 100644
--- a/src/core/hw/hw.cpp
+++ b/src/core/hw/hw.cpp
@@ -15,6 +15,21 @@ template <typename T>
 inline void Read(T &var, const u32 addr) {
     switch (addr & 0xFFFFF000) {
     case VADDR_GPU:
+    case VADDR_GPU + 0x1000:
+    case VADDR_GPU + 0x2000:
+    case VADDR_GPU + 0x3000:
+    case VADDR_GPU + 0x4000:
+    case VADDR_GPU + 0x5000:
+    case VADDR_GPU + 0x6000:
+    case VADDR_GPU + 0x7000:
+    case VADDR_GPU + 0x8000:
+    case VADDR_GPU + 0x9000:
+    case VADDR_GPU + 0xA000:
+    case VADDR_GPU + 0xB000:
+    case VADDR_GPU + 0xC000:
+    case VADDR_GPU + 0xD000:
+    case VADDR_GPU + 0xE000:
+    case VADDR_GPU + 0xF000:
         GPU::Read(var, addr);
         break;
     case VADDR_LCD:
@@ -29,6 +44,21 @@ template <typename T>
 inline void Write(u32 addr, const T data) {
     switch (addr & 0xFFFFF000) {
     case VADDR_GPU:
+    case VADDR_GPU + 0x1000:
+    case VADDR_GPU + 0x2000:
+    case VADDR_GPU + 0x3000:
+    case VADDR_GPU + 0x4000:
+    case VADDR_GPU + 0x5000:
+    case VADDR_GPU + 0x6000:
+    case VADDR_GPU + 0x7000:
+    case VADDR_GPU + 0x8000:
+    case VADDR_GPU + 0x9000:
+    case VADDR_GPU + 0xA000:
+    case VADDR_GPU + 0xB000:
+    case VADDR_GPU + 0xC000:
+    case VADDR_GPU + 0xD000:
+    case VADDR_GPU + 0xE000:
+    case VADDR_GPU + 0xF000:
         GPU::Write(addr, data);
         break;
     case VADDR_LCD:
diff --git a/src/core/hw/lcd.cpp b/src/core/hw/lcd.cpp
index cdb757a18..6f93709e3 100644
--- a/src/core/hw/lcd.cpp
+++ b/src/core/hw/lcd.cpp
@@ -10,6 +10,9 @@
 #include "core/hw/hw.h"
 #include "core/hw/lcd.h"
 
+#include "core/tracer/recorder.h"
+#include "video_core/debug_utils/debug_utils.h"
+
 namespace LCD {
 
 Regs g_regs;
@@ -40,6 +43,13 @@ inline void Write(u32 addr, const T data) {
     }
 
     g_regs[index] = static_cast<u32>(data);
+
+    // Notify tracer about the register write
+    // This is happening *after* handling the write to make sure we properly catch all memory reads.
+    if (Pica::g_debug_context && Pica::g_debug_context->recorder) {
+        // addr + GPU VBase - IO VBase + IO PBase
+        Pica::g_debug_context->recorder->RegisterWritten<T>(addr + HW::VADDR_LCD - 0x1EC00000 + 0x10100000, data);
+    }
 }
 
 // Explicitly instantiate template functions because we aren't defining this in the header: