diff --git a/include/config.hpp b/include/config.hpp
index 25f352e8..52be1af7 100644
--- a/include/config.hpp
+++ b/include/config.hpp
@@ -14,7 +14,7 @@ struct EmulatorConfig {
 #endif
 
 	// For now, use specialized shaders by default on MacOS as M1 drivers are buggy when using the ubershader, and on Android since mobile GPUs are
-	// horrible On other platforms we default to ubershader + shadergen fallback for lights
+	// horrible. On other platforms we default to ubershader + shadergen fallback for lights
 #if defined(__ANDROID__) || defined(__APPLE__)
 	static constexpr bool ubershaderDefault = false;
 #else
@@ -51,4 +51,4 @@ struct EmulatorConfig {
 	EmulatorConfig(const std::filesystem::path& path);
 	void load();
 	void save();
-};
\ No newline at end of file
+};
diff --git a/include/renderer_gl/renderer_gl.hpp b/include/renderer_gl/renderer_gl.hpp
index bfa9922b..d00445ac 100644
--- a/include/renderer_gl/renderer_gl.hpp
+++ b/include/renderer_gl/renderer_gl.hpp
@@ -117,6 +117,7 @@ class RendererGL final : public Renderer {
 	// Note: The caller is responsible for deleting the currently bound FBO before calling this
 	void setFBO(uint handle) { screenFramebuffer.m_handle = handle; }
 	void resetStateManager() { gl.reset(); }
+	void clearShaderCache();
 	void initUbershader(OpenGL::Program& program);
 
 #ifdef PANDA3DS_FRONTEND_QT
diff --git a/src/core/renderer_gl/renderer_gl.cpp b/src/core/renderer_gl/renderer_gl.cpp
index 22750f7d..36827027 100644
--- a/src/core/renderer_gl/renderer_gl.cpp
+++ b/src/core/renderer_gl/renderer_gl.cpp
@@ -24,10 +24,7 @@ void RendererGL::reset() {
 	colourBufferCache.reset();
 	textureCache.reset();
 
-	for (auto& shader : shaderCache) {
-		shader.second.program.free();
-	}
-	shaderCache.clear();
+	clearShaderCache();
 
 	// Init the colour/depth buffer settings to some random defaults on reset
 	colourBufferLoc = 0;
@@ -808,6 +805,8 @@ OpenGL::Program& RendererGL::getSpecializedShader() {
 		program.create({defaultShadergenVs, fragShader});
 		gl.useProgram(program);
 
+		fragShader.free();
+
 		// Init sampler objects. Texture 0 goes in texture unit 0, texture 1 in TU 1, texture 2 in TU 2, and the light maps go in TU 3
 		glUniform1i(OpenGL::uniformLocation(program, "u_tex0"), 0);
 		glUniform1i(OpenGL::uniformLocation(program, "u_tex1"), 1);
@@ -937,16 +936,22 @@ void RendererGL::screenshot(const std::string& name) {
 	stbi_write_png(name.c_str(), width, height, 4, flippedPixels.data(), 0);
 }
 
+void RendererGL::clearShaderCache() {
+	for (auto& shader : shaderCache) {
+		CachedProgram& cachedProgram = shader.second;
+		cachedProgram.program.free();
+		glDeleteBuffers(1, &cachedProgram.uboBinding);
+	}
+
+	shaderCache.clear();
+}
+
 void RendererGL::deinitGraphicsContext() {
 	// Invalidate all surface caches since they'll no longer be valid
 	textureCache.reset();
 	depthBufferCache.reset();
 	colourBufferCache.reset();
-
-	for (auto& shader : shaderCache) {
-		shader.second.program.free();
-	}
-	shaderCache.clear();
+	clearShaderCache();
 
 	// All other GL objects should be invalidated automatically and be recreated by the next call to initGraphicsContext
 	// TODO: Make it so that depth and colour buffers get written back to 3DS memory
diff --git a/src/host_shaders/opengl_fragment_shader.frag b/src/host_shaders/opengl_fragment_shader.frag
index 9f369e39..48b55a4c 100644
--- a/src/host_shaders/opengl_fragment_shader.frag
+++ b/src/host_shaders/opengl_fragment_shader.frag
@@ -307,8 +307,8 @@ void calcLighting(out vec4 primary_color, out vec4 secondary_color) {
 	primary_color = vec4(vec3(0.0), 1.0);
 	secondary_color = vec4(vec3(0.0), 1.0);
 
-	uint GPUREG_LIGHTING_LUTINPUT_SCALE = readPicaReg(0x01D2u);
 	uint GPUREG_LIGHTING_CONFIG0 = readPicaReg(0x01C3u);
+	GPUREG_LIGHTING_LUTINPUT_SCALE = readPicaReg(0x01D2u);
 	GPUREG_LIGHTING_CONFIG1 = readPicaReg(0x01C4u);
 	GPUREG_LIGHTING_LUTINPUT_ABS = readPicaReg(0x01D0u);
 	GPUREG_LIGHTING_LUTINPUT_SELECT = readPicaReg(0x01D1u);
diff --git a/src/libretro_core.cpp b/src/libretro_core.cpp
index 02bf3cd1..3e0436b8 100644
--- a/src/libretro_core.cpp
+++ b/src/libretro_core.cpp
@@ -15,6 +15,8 @@ static retro_input_state_t inputStateCallback;
 static retro_hw_render_callback hw_render;
 static std::filesystem::path savePath;
 
+static bool screenTouched;
+
 std::unique_ptr<Emulator> emulator;
 RendererGL* renderer;
 
@@ -314,7 +316,8 @@ void retro_run() {
 	hid.setCirclepadX((xLeft / +32767) * 0x9C);
 	hid.setCirclepadY((yLeft / -32767) * 0x9C);
 
-	bool touch = inputStateCallback(0, RETRO_DEVICE_MOUSE, 0, RETRO_DEVICE_ID_MOUSE_LEFT);
+	bool touchScreen = false;
+
 	const int posX = inputStateCallback(0, RETRO_DEVICE_POINTER, 0, RETRO_DEVICE_ID_POINTER_X);
 	const int posY = inputStateCallback(0, RETRO_DEVICE_POINTER, 0, RETRO_DEVICE_ID_POINTER_Y);
 
@@ -324,16 +327,23 @@ void retro_run() {
 	const int offsetX = 40;
 	const int offsetY = emulator->height / 2;
 
-	const bool inScreenX = newX >= offsetX && newX < emulator->width - offsetX;
+	const bool inScreenX = newX >= offsetX && newX <= emulator->width - offsetX;
 	const bool inScreenY = newY >= offsetY && newY <= emulator->height;
 
-	if (touch && inScreenX && inScreenY) {
+	if (inScreenX && inScreenY) {
+		touchScreen |= inputStateCallback(0, RETRO_DEVICE_MOUSE, 0, RETRO_DEVICE_ID_MOUSE_LEFT);
+		touchScreen |= inputStateCallback(0, RETRO_DEVICE_POINTER, 0, RETRO_DEVICE_ID_POINTER_PRESSED);
+	}
+
+	if (touchScreen) {
 		u16 x = static_cast<u16>(newX - offsetX);
 		u16 y = static_cast<u16>(newY - offsetY);
 
 		hid.setTouchScreenPress(x, y);
-	} else {
+		screenTouched = true;
+	} else if (screenTouched) {
 		hid.releaseTouchScreen();
+		screenTouched = false;
 	}
 
 	hid.updateInputs(emulator->getTicks());