rsx: Implement separate viewport raster clipping

- Merge viewport raster window and scissor into one clipping region
- Viewport raster clip is different from viewport geometry clipping in
hardware as the latter is configurable separately

rpcs3/Emu/RSX/D3D12/D3D12GSRender.cpp

@@ -476,7 +476,7 @@ void D3D12GSRender::end()
 	};
 	get_current_resource_storage().command_list->RSSetViewports(1, &viewport);
 
-	get_current_resource_storage().command_list->RSSetScissorRects(1, &get_scissor(rsx::method_registers.scissor_origin_x(), rsx::method_registers.scissor_origin_y(),
+	get_current_resource_storage().command_list->RSSetScissorRects(1, &::get_scissor(rsx::method_registers.scissor_origin_x(), rsx::method_registers.scissor_origin_y(),
 		rsx::method_registers.scissor_width(), rsx::method_registers.scissor_height()));
 
 	get_current_resource_storage().command_list->IASetPrimitiveTopology(get_primitive_topology(rsx::method_registers.current_draw_clause.primitive));

rpcs3/Emu/RSX/D3D12/D3D12RenderTargetSets.cpp

@@ -125,12 +125,12 @@ void D3D12GSRender::clear_surface(u32 arg)
 			u32 clear_depth = rsx::method_registers.z_clear_value(depth_format == rsx::surface_depth_format::z24s8);
 			u32 max_depth_value = get_max_depth_value(depth_format);
 			get_current_resource_storage().command_list->ClearDepthStencilView(m_rtts.current_ds_handle, D3D12_CLEAR_FLAG_DEPTH, clear_depth / (float)max_depth_value, 0,
-				1, &get_scissor(rsx::method_registers.scissor_origin_x(), rsx::method_registers.scissor_origin_y(), rsx::method_registers.scissor_width(), rsx::method_registers.scissor_height()));
+				1, &::get_scissor(rsx::method_registers.scissor_origin_x(), rsx::method_registers.scissor_origin_y(), rsx::method_registers.scissor_width(), rsx::method_registers.scissor_height()));
 		}
 
 		if (arg & 0x2)
 			get_current_resource_storage().command_list->ClearDepthStencilView(m_rtts.current_ds_handle, D3D12_CLEAR_FLAG_STENCIL, 0.f, get_clear_stencil(rsx::method_registers.stencil_clear_value()),
-				1, &get_scissor(rsx::method_registers.scissor_origin_x(), rsx::method_registers.scissor_origin_y(), rsx::method_registers.scissor_width(), rsx::method_registers.scissor_height()));
+				1, &::get_scissor(rsx::method_registers.scissor_origin_x(), rsx::method_registers.scissor_origin_y(), rsx::method_registers.scissor_width(), rsx::method_registers.scissor_height()));
 	}
 
 	if (arg & 0xF0)
@@ -147,7 +147,7 @@ void D3D12GSRender::clear_surface(u32 arg)
 		};
 		for (unsigned i = 0; i < rtt_index; i++)
 			get_current_resource_storage().command_list->ClearRenderTargetView(handle.Offset(i, m_descriptor_stride_rtv), clear_color.data(),
-				1, &get_scissor(rsx::method_registers.scissor_origin_x(), rsx::method_registers.scissor_origin_y(), rsx::method_registers.scissor_width(), rsx::method_registers.scissor_height()));
+				1, &::get_scissor(rsx::method_registers.scissor_origin_x(), rsx::method_registers.scissor_origin_y(), rsx::method_registers.scissor_width(), rsx::method_registers.scissor_height()));
 	}
 
 	std::chrono::time_point<steady_clock> end_duration = steady_clock::now();

rpcs3/Emu/RSX/GL/GLGSRender.cpp

@@ -651,43 +651,14 @@ void GLGSRender::set_viewport()
 
 void GLGSRender::set_scissor()
 {
-	if (m_graphics_state & rsx::pipeline_state::scissor_config_state_dirty)
+	areau scissor;
+	if (get_scissor(scissor))
 	{
-		// Optimistic that the new config will allow us to render
+		// NOTE: window origin does not affect scissor region (probably only affects viewport matrix; already applied)
-		framebuffer_status_valid = true;
+		// See LIMBO [NPUB-30373] which uses shader window origin = top
+		glScissor(scissor.x1, scissor.y1, scissor.width(), scissor.height());
+		gl_state.enable(GL_TRUE, GL_SCISSOR_TEST);
 	}
-	else if (!(m_graphics_state & rsx::pipeline_state::scissor_config_state_dirty))
-	{
-		// Nothing to do
-		return;
-	}
-
-	m_graphics_state &= ~(rsx::pipeline_state::scissor_config_state_dirty | rsx::pipeline_state::scissor_setup_invalid);
-
-	const auto clip_width = rsx::apply_resolution_scale(rsx::method_registers.surface_clip_width(), true);
-	const auto clip_height = rsx::apply_resolution_scale(rsx::method_registers.surface_clip_height(), true);
-
-	u16 scissor_x = rsx::apply_resolution_scale(rsx::method_registers.scissor_origin_x(), false);
-	u16 scissor_w = rsx::apply_resolution_scale(rsx::method_registers.scissor_width(), true);
-	u16 scissor_y = rsx::apply_resolution_scale(rsx::method_registers.scissor_origin_y(), false);
-	u16 scissor_h = rsx::apply_resolution_scale(rsx::method_registers.scissor_height(), true);
-
-	// Do not bother drawing anything if output is zero sized
-	// TODO: Clip scissor region
-	if (scissor_x >= clip_width || scissor_y >= clip_height || scissor_w == 0 || scissor_h == 0)
-	{
-		if (!g_cfg.video.strict_rendering_mode)
-		{
-			m_graphics_state |= rsx::pipeline_state::scissor_setup_invalid;
-			framebuffer_status_valid = false;
-			return;
-		}
-	}
-
-	// NOTE: window origin does not affect scissor region (probably only affects viewport matrix; already applied)
-	// See LIMBO [NPUB-30373] which uses shader window origin = top
-	glScissor(scissor_x, scissor_y, scissor_w, scissor_h);
-	gl_state.enable(GL_TRUE, GL_SCISSOR_TEST);
 }
 
 void GLGSRender::on_init_thread()

rpcs3/Emu/RSX/RSXThread.cpp

@@ -1293,6 +1293,56 @@ namespace rsx
 		return layout;
 	}
 
+	bool thread::get_scissor(areau& region)
+	{
+		if (!(m_graphics_state & rsx::pipeline_state::scissor_config_state_dirty))
+		{
+			// Nothing to do
+			return false;
+		}
+
+		m_graphics_state &= ~rsx::pipeline_state::scissor_config_state_dirty;
+
+		u16 scissor_x = rsx::method_registers.scissor_origin_x();
+		u16 scissor_w = rsx::method_registers.scissor_width();
+		u16 scissor_y = rsx::method_registers.scissor_origin_y();
+		u16 scissor_h = rsx::method_registers.scissor_height();
+
+		u16 raster_x = rsx::method_registers.viewport_origin_x();
+		u16 raster_w = rsx::method_registers.viewport_width();
+		u16 raster_y = rsx::method_registers.viewport_origin_y();
+		u16 raster_h = rsx::method_registers.viewport_height();
+
+		// Get the minimum area between these two
+		u16 x1 = std::max(scissor_x, raster_x);
+		u16 y1 = std::max(scissor_y, raster_y);
+		u16 x2 = std::min(scissor_x + scissor_w, raster_x + raster_w);
+		u16 y2 = std::min(scissor_y + scissor_h, raster_y + raster_h);
+
+		if (x2 <= x1 ||
+			y2 <= y1 ||
+			x1 >= rsx::method_registers.window_clip_horizontal() ||
+			y1 >= rsx::method_registers.window_clip_vertical())
+		{
+			m_graphics_state |= rsx::pipeline_state::scissor_setup_invalid;
+			framebuffer_status_valid = false;
+			return false;
+		}
+
+		if (m_graphics_state & rsx::pipeline_state::scissor_setup_invalid)
+		{
+			m_graphics_state &= ~rsx::pipeline_state::scissor_setup_invalid;
+			framebuffer_status_valid = true;
+		}
+
+		region.x1 = rsx::apply_resolution_scale(x1, false);
+		region.x2 = rsx::apply_resolution_scale(x2, true);
+		region.y1 = rsx::apply_resolution_scale(y1, false);
+		region.y2 = rsx::apply_resolution_scale(y2, true);
+
+		return true;
+	}
+
 	void thread::get_current_vertex_program(const std::array<std::unique_ptr<rsx::sampled_image_descriptor_base>, rsx::limits::vertex_textures_count>& sampler_descriptors, bool skip_textures, bool skip_vertex_inputs)
 	{
 		if (!(m_graphics_state & rsx::pipeline_state::vertex_program_dirty))

rpcs3/Emu/RSX/RSXThread.h

@@ -520,6 +520,7 @@ namespace rsx
 		u32 get_zeta_surface_address() const;
 
 		framebuffer_layout get_framebuffer_layout(rsx::framebuffer_creation_context context);
+		bool get_scissor(areau& region);
 
 		/**
 		 * Analyze vertex inputs and group all interleaved blocks

rpcs3/Emu/RSX/VK/VKGSRender.cpp

@@ -1796,37 +1796,13 @@ void VKGSRender::set_viewport()
 
 void VKGSRender::set_scissor()
 {
-	if (m_graphics_state & rsx::pipeline_state::scissor_config_state_dirty)
+	areau scissor;
+	if (get_scissor(scissor))
 	{
-		// Optimistic that the new config will allow us to render
+		m_scissor.extent.height = scissor.height();
-		framebuffer_status_valid = true;
+		m_scissor.extent.width = scissor.width();
-	}
+		m_scissor.offset.x = scissor.x1;
-	else if (!(m_graphics_state & rsx::pipeline_state::scissor_config_state_dirty))
+		m_scissor.offset.y = scissor.y1;
-	{
-		// Nothing to do
-		return;
-	}
-
-	m_graphics_state &= ~(rsx::pipeline_state::scissor_config_state_dirty | rsx::pipeline_state::scissor_setup_invalid);
-
-	u16 scissor_x = rsx::apply_resolution_scale(rsx::method_registers.scissor_origin_x(), false);
-	u16 scissor_w = rsx::apply_resolution_scale(rsx::method_registers.scissor_width(), true);
-	u16 scissor_y = rsx::apply_resolution_scale(rsx::method_registers.scissor_origin_y(), false);
-	u16 scissor_h = rsx::apply_resolution_scale(rsx::method_registers.scissor_height(), true);
-
-	m_scissor.extent.height = scissor_h;
-	m_scissor.extent.width = scissor_w;
-	m_scissor.offset.x = scissor_x;
-	m_scissor.offset.y = scissor_y;
-
-	if (scissor_x >= m_viewport.width || scissor_y >= m_viewport.height || scissor_w == 0 || scissor_h == 0)
-	{
-		if (!g_cfg.video.strict_rendering_mode)
-		{
-			m_graphics_state |= rsx::pipeline_state::scissor_setup_invalid;
-			framebuffer_status_valid = false;
-			return;
-		}
 	}
 }
 

rpcs3/Emu/RSX/rsx_methods.cpp

@@ -2790,6 +2790,8 @@ namespace rsx
 		bind<NV4097_SET_FOG_MODE, nv4097::set_ROP_state_dirty_bit>();
 		bind<NV4097_SET_SCISSOR_HORIZONTAL, nv4097::set_scissor_dirty_bit>();
 		bind<NV4097_SET_SCISSOR_VERTICAL, nv4097::set_scissor_dirty_bit>();
+		bind<NV4097_SET_VIEWPORT_HORIZONTAL, nv4097::set_scissor_dirty_bit>();
+		bind<NV4097_SET_VIEWPORT_VERTICAL, nv4097::set_scissor_dirty_bit>();
 		bind_array<NV4097_SET_FOG_PARAMS, 1, 2, nv4097::set_ROP_state_dirty_bit>();
 		bind_range<NV4097_SET_VIEWPORT_SCALE, 1, 3, nv4097::set_viewport_dirty_bit>();
 		bind_range<NV4097_SET_VIEWPORT_OFFSET, 1, 3, nv4097::set_viewport_dirty_bit>();