GPU: Fix mismatched scanout between hardware/software renderers

src/core/gpu_hw_d3d11.cpp

@@ -598,7 +598,7 @@ void GPU_HW_D3D11::UpdateDisplay()
     }
     else if (!m_GPUSTAT.display_area_color_depth_24 && !interlaced &&
              (scaled_vram_offset_x + scaled_display_width) <= m_vram_texture.GetWidth() &&
-             (scaled_vram_offset_y + scaled_vram_offset_y <= m_vram_texture.GetHeight()))
+             (scaled_vram_offset_y + scaled_display_height) <= m_vram_texture.GetHeight())
     {
       m_host_display->SetDisplayTexture(m_vram_texture.GetD3DSRV(), m_vram_texture.GetWidth(),
                                         m_vram_texture.GetHeight(), scaled_vram_offset_x, scaled_vram_offset_y,
@@ -612,13 +612,13 @@ void GPU_HW_D3D11::UpdateDisplay()
 
       const u32 reinterpret_field_offset = GetInterlacedField();
       const u32 reinterpret_start_x = m_crtc_state.regs.X * m_resolution_scale;
-      const u32 reinterpret_width =
+      const u32 reinterpret_crop_left = (m_crtc_state.display_vram_left - m_crtc_state.regs.X) * m_resolution_scale;
-        scaled_display_width + ((m_crtc_state.display_vram_left - m_crtc_state.regs.X) * m_resolution_scale);
+      const u32 uniforms[4] = {reinterpret_start_x, scaled_vram_offset_y, reinterpret_crop_left,
-      const u32 uniforms[4] = {reinterpret_start_x, scaled_vram_offset_y, reinterpret_field_offset};
+                               reinterpret_field_offset};
       ID3D11PixelShader* display_pixel_shader =
         m_display_pixel_shaders[BoolToUInt8(m_GPUSTAT.display_area_color_depth_24)][BoolToUInt8(interlaced)].Get();
 
-      SetViewportAndScissor(0, reinterpret_field_offset, reinterpret_width, scaled_display_height);
+      SetViewportAndScissor(0, reinterpret_field_offset, scaled_display_width, scaled_display_height);
       DrawUtilityShader(display_pixel_shader, uniforms, sizeof(uniforms));
 
       m_host_display->SetDisplayTexture(m_display_texture.GetD3DSRV(), m_display_texture.GetWidth(),

src/core/gpu_hw_opengl.cpp

@@ -599,7 +599,7 @@ void GPU_HW_OpenGL::UpdateDisplay()
     }
     else if (!m_GPUSTAT.display_area_color_depth_24 && !interlaced &&
              (scaled_vram_offset_x + scaled_display_width) <= m_vram_texture.GetWidth() &&
-             (scaled_vram_offset_y + scaled_vram_offset_y <= m_vram_texture.GetHeight()))
+             (scaled_vram_offset_y + scaled_display_height) <= m_vram_texture.GetHeight())
     {
       m_host_display->SetDisplayTexture(reinterpret_cast<void*>(static_cast<uintptr_t>(m_vram_texture.GetGLId())),
                                         m_vram_texture.GetWidth(), m_vram_texture.GetHeight(), scaled_vram_offset_x,
@@ -621,13 +621,13 @@ void GPU_HW_OpenGL::UpdateDisplay()
         m_vram_texture.GetHeight() - scaled_vram_offset_y - scaled_display_height;
       const u32 reinterpret_field_offset = GetInterlacedField();
       const u32 reinterpret_start_x = m_crtc_state.regs.X * m_resolution_scale;
-      const u32 reinterpret_width =
+      const u32 reinterpret_crop_left = (m_crtc_state.display_vram_left - m_crtc_state.regs.X) * m_resolution_scale;
-        scaled_display_width + ((m_crtc_state.display_vram_left - m_crtc_state.regs.X) * m_resolution_scale);
+      const u32 uniforms[4] = {reinterpret_start_x, scaled_flipped_vram_offset_y, reinterpret_crop_left,
-      const u32 uniforms[4] = {reinterpret_start_x, scaled_flipped_vram_offset_y, reinterpret_field_offset};
+                               reinterpret_field_offset};
       UploadUniformBuffer(uniforms, sizeof(uniforms));
       m_batch_ubo_dirty = true;
 
-      glViewport(0, reinterpret_field_offset, reinterpret_width, scaled_display_height);
+      glViewport(0, reinterpret_field_offset, scaled_display_width, scaled_display_height);
       glBindVertexArray(m_attributeless_vao_id);
       glDrawArrays(GL_TRIANGLES, 0, 3);
 

src/core/gpu_hw_shadergen.cpp

@@ -995,13 +995,13 @@ std::string GPU_HW_ShaderGen::GenerateDisplayFragmentShader(bool depth_24bit, bo
   DefineMacro(ss, "INTERLACED", interlaced);
 
   WriteCommonFunctions(ss);
-  DeclareUniformBuffer(ss, {"uint2 u_vram_offset", "uint u_field_offset"});
+  DeclareUniformBuffer(ss, {"uint2 u_vram_offset", "uint u_crop_left", "uint u_field_offset"});
   DeclareTexture(ss, "samp0", 0);
 
   DeclareFragmentEntryPoint(ss, 0, 1, {}, true, 1);
   ss << R"(
 {
-  uint2 icoords = uint2(v_pos.xy) + u_vram_offset;
+  uint2 icoords = uint2(v_pos.xy);
 
   #if INTERLACED
     if (((icoords.y / RESOLUTION_SCALE) & 1u) != u_field_offset)
@@ -1010,12 +1010,12 @@ std::string GPU_HW_ShaderGen::GenerateDisplayFragmentShader(bool depth_24bit, bo
 
   #if DEPTH_24BIT
     // relative to start of scanout
-    uint relative_x = (icoords.x - u_vram_offset.x) / RESOLUTION_SCALE;
+    uint relative_x = (icoords.x + u_crop_left) / RESOLUTION_SCALE;
-    icoords.x = u_vram_offset.x + ((relative_x * 3u) / 2u) * RESOLUTION_SCALE;
+    uint2 vram_coords = u_vram_offset + uint2(((relative_x * 3u) / 2u) * RESOLUTION_SCALE, icoords.y);
 
     // load adjacent 16-bit texels
-    uint s0 = RGBA8ToRGBA5551(LOAD_TEXTURE(samp0, int2(icoords % VRAM_SIZE), 0));
+    uint s0 = RGBA8ToRGBA5551(LOAD_TEXTURE(samp0, int2(vram_coords % VRAM_SIZE), 0));
-    uint s1 = RGBA8ToRGBA5551(LOAD_TEXTURE(samp0, int2((icoords + uint2(RESOLUTION_SCALE, 0)) % VRAM_SIZE), 0));
+    uint s1 = RGBA8ToRGBA5551(LOAD_TEXTURE(samp0, int2((vram_coords + uint2(RESOLUTION_SCALE, 0)) % VRAM_SIZE), 0));
     
     // select which part of the combined 16-bit texels we are currently shading
     uint s1s0 = ((s1 << 16) | s0) >> ((relative_x & 1u) * 8u);
@@ -1025,7 +1025,8 @@ std::string GPU_HW_ShaderGen::GenerateDisplayFragmentShader(bool depth_24bit, bo
                     float((s1s0 >> 16u) & 0xFFu) / 255.0, 1.0);
   #else
     // load and return
-    o_col0 = LOAD_TEXTURE(samp0, int2(icoords % VRAM_SIZE), 0);
+    uint2 vram_coords = u_vram_offset + uint2(icoords.x + u_crop_left, icoords.y);
+    o_col0 = LOAD_TEXTURE(samp0, int2(vram_coords % VRAM_SIZE), 0);
   #endif
 }
 )";

src/core/gpu_sw.cpp

@@ -154,36 +154,36 @@ void GPU_SW::UpdateDisplay()
     const u32 vram_offset_y = m_crtc_state.display_vram_top;
     const u32 display_width = m_crtc_state.display_vram_width;
     const u32 display_height = m_crtc_state.display_vram_height;
+    const u32 texture_offset_x = m_crtc_state.display_vram_left - m_crtc_state.regs.X;
     if (IsInterlacedDisplayEnabled())
     {
       const u32 field = GetInterlacedField();
       if (m_GPUSTAT.display_area_color_depth_24)
       {
         CopyOut24Bit(m_crtc_state.regs.X, vram_offset_y + field, m_display_texture_buffer.data() + field * VRAM_WIDTH,
-                     VRAM_WIDTH, display_width, display_height, true);
+                     VRAM_WIDTH, display_width + texture_offset_x, display_height, true);
       }
       else
       {
         CopyOut15Bit(m_crtc_state.regs.X, vram_offset_y + field, m_display_texture_buffer.data() + field * VRAM_WIDTH,
-                     VRAM_WIDTH, display_width, display_height, true);
+                     VRAM_WIDTH, display_width + texture_offset_x, display_height, true);
       }
     }
     else
     {
       if (m_GPUSTAT.display_area_color_depth_24)
       {
-        CopyOut24Bit(m_crtc_state.regs.X, vram_offset_y, m_display_texture_buffer.data(), VRAM_WIDTH, display_width,
+        CopyOut24Bit(m_crtc_state.regs.X, vram_offset_y, m_display_texture_buffer.data(), VRAM_WIDTH,
-                     display_height, false);
+                     display_width + texture_offset_x, display_height, false);
       }
       else
       {
-        CopyOut15Bit(m_crtc_state.regs.X, vram_offset_y, m_display_texture_buffer.data(), VRAM_WIDTH, display_width,
+        CopyOut15Bit(m_crtc_state.regs.X, vram_offset_y, m_display_texture_buffer.data(), VRAM_WIDTH,
-                     display_height, false);
+                     display_width + texture_offset_x, display_height, false);
       }
     }
 
-    const u32 texture_offset_x = m_crtc_state.display_vram_left - m_crtc_state.regs.X;
+    m_host_display->UpdateTexture(m_display_texture.get(), 0, 0, display_width, display_height,
-    m_host_display->UpdateTexture(m_display_texture.get(), texture_offset_x, 0, display_width, display_height,
                                   m_display_texture_buffer.data(), VRAM_WIDTH * sizeof(u32));
     m_host_display->SetDisplayTexture(m_display_texture->GetHandle(), VRAM_WIDTH, VRAM_HEIGHT, texture_offset_x, 0,
                                       display_width, display_height);