diff --git a/include/PICA/shader.hpp b/include/PICA/shader.hpp
index cf43044a..847391f8 100644
--- a/include/PICA/shader.hpp
+++ b/include/PICA/shader.hpp
@@ -148,8 +148,8 @@ class PICAShader {
 	bool isCondTrue(u32 instruction);
 
 public:
-	std::array<u32, 512> loadedShader; // Currently loaded & active shader
-	std::array<u32, 512> bufferedShader; // Shader to be transferred when the SH_CODETRANSFER_END reg gets written to
+	std::array<u32, 4096> loadedShader; // Currently loaded & active shader
+	std::array<u32, 4096> bufferedShader; // Shader to be transferred when the SH_CODETRANSFER_END reg gets written to
 
 	u32 entrypoint = 0; // Initial shader PC
 	u32 boolUniform;
@@ -164,12 +164,12 @@ public:
 
 	// Theese functions are in the header to be inlined more easily, though with LTO I hope I'll be able to move them
 	void finalize() {
-		std::memcpy(&loadedShader[0], &bufferedShader[0], 512 * sizeof(u32));
+		std::memcpy(&loadedShader[0], &bufferedShader[0], 4096 * sizeof(u32));
 	}
 
 	void setBufferIndex(u32 index) {
-		if (index != 0) Helpers::panic("Is this register 9 or 11 bit?");
-		bufferIndex = (index >> 2) & 0x1ff;
+		if (index != 0) Helpers::panic("How many bits is the shader buffer index reg meant to be?");
+		bufferIndex = (index >> 2) & 0xfff;
 	}
 
 	void setOpDescriptorIndex(u32 index) {
@@ -177,9 +177,9 @@ public:
 	}
 
 	void uploadWord(u32 word) {
-		if (bufferIndex >= 511) Helpers::panic("o no");
+		if (bufferIndex >= 4095) Helpers::panic("o no, shader upload overflew");
 		bufferedShader[bufferIndex++] = word;
-		bufferIndex &= 0x1ff;
+		bufferIndex &= 0xfff;
 	}
 
 	void uploadDescriptor(u32 word) {
diff --git a/include/services/gsp_gpu.hpp b/include/services/gsp_gpu.hpp
index 647a235d..bc8ed287 100644
--- a/include/services/gsp_gpu.hpp
+++ b/include/services/gsp_gpu.hpp
@@ -43,6 +43,7 @@ class GPUService {
 	void processCommandList(u32* cmd);
 	void memoryFill(u32* cmd);
 	void triggerDisplayTransfer(u32* cmd);
+	void flushCacheRegions(u32* cmd);
 
 public:
 	GPUService(Memory& mem, GPU& gpu, u32& currentPID) : mem(mem), gpu(gpu), currentPID(currentPID) {}
diff --git a/src/core/services/gsp_gpu.cpp b/src/core/services/gsp_gpu.cpp
index c3b52a87..fed35b3b 100644
--- a/src/core/services/gsp_gpu.cpp
+++ b/src/core/services/gsp_gpu.cpp
@@ -18,7 +18,8 @@ namespace GXCommands {
 	enum : u32 {
 		ProcessCommandList = 1,
 		MemoryFill = 2,
-		TriggerDisplayTransfer = 3
+		TriggerDisplayTransfer = 3,
+		FlushCacheRegions = 5
 	};
 }
 
@@ -214,6 +215,7 @@ void GPUService::processCommandBuffer() {
 				case GXCommands::ProcessCommandList: processCommandList(cmd); break;
 				case GXCommands::MemoryFill: memoryFill(cmd); break;
 				case GXCommands::TriggerDisplayTransfer: triggerDisplayTransfer(cmd); break;
+				case GXCommands::FlushCacheRegions: flushCacheRegions(cmd); break;
 				default: Helpers::panic("GSP::GPU::ProcessCommands: Unknown cmd ID %d", cmdID);
 			}
 
@@ -252,6 +254,10 @@ void GPUService::triggerDisplayTransfer(u32* cmd) {
 	requestInterrupt(GPUInterrupt::PPF); // Send "Display transfer finished" interrupt
 }
 
+void GPUService::flushCacheRegions(u32* cmd) {
+	log("GSP::GPU::FlushCacheRegions (Stubbed)\n");
+}
+
 // Actually send command list (aka display list) to GPU
 void GPUService::processCommandList(u32* cmd) {
 	u32 address = cmd[1] & ~7; // Buffer address