GPU/HW: Fix oversized VRAM copies writing out of bounds

Fixes jittering in Duke Nukem - Land of the Babes water effect.

src/core/gpu_hw.h

@@ -127,6 +127,8 @@ protected:
     u32 u_src_y;
     u32 u_dst_x;
     u32 u_dst_y;
+    u32 u_end_x;
+    u32 u_end_y;
     u32 u_width;
     u32 u_height;
     u32 u_set_mask_bit;

src/core/gpu_hw_d3d11.cpp

@@ -759,6 +759,8 @@ void GPU_HW_D3D11::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 widt
                                       src_y * m_resolution_scale,
                                       dst_x * m_resolution_scale,
                                       dst_y * m_resolution_scale,
+                                      ((dst_x + width) % VRAM_WIDTH) * m_resolution_scale,
+                                      ((dst_y + height) % VRAM_HEIGHT) * m_resolution_scale,
                                       width * m_resolution_scale,
                                       height * m_resolution_scale,
                                       m_GPUSTAT.set_mask_while_drawing ? 1u : 0u,

src/core/gpu_hw_opengl.cpp

@@ -864,6 +864,8 @@ void GPU_HW_OpenGL::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 wid
                                 src_y * m_resolution_scale,
                                 dst_x * m_resolution_scale,
                                 dst_y * m_resolution_scale,
+                                ((dst_x + width) % VRAM_WIDTH) * m_resolution_scale,
+                                ((dst_y + height) % VRAM_HEIGHT) * m_resolution_scale,
                                 width * m_resolution_scale,
                                 height * m_resolution_scale,
                                 m_GPUSTAT.set_mask_while_drawing ? 1u : 0u,

src/core/gpu_hw_shadergen.cpp

@@ -136,6 +136,10 @@ void GPU_HW_ShaderGen::WriteHeader(std::stringstream& ss)
     ss << "#define CONSTANT const\n";
     ss << "#define VECTOR_EQ(a, b) ((a) == (b))\n";
     ss << "#define VECTOR_NEQ(a, b) ((a) != (b))\n";
+    ss << "#define VECTOR_LT(a, b) ((a) < (b))\n";
+    ss << "#define VECTOR_LE(a, b) ((a) <= (b))\n";
+    ss << "#define VECTOR_GT(a, b) ((a) > (b))\n";
+    ss << "#define VECTOR_GE(a, b) ((a) >= (b))\n";
     ss << "#define SAMPLE_TEXTURE(name, coords) texture(name, coords)\n";
     ss << "#define LOAD_TEXTURE(name, coords, mip) texelFetch(name, coords, mip)\n";
     ss << "#define LOAD_TEXTURE_OFFSET(name, coords, mip, offset) texelFetchOffset(name, coords, mip, offset)\n";
@@ -148,6 +152,10 @@ void GPU_HW_ShaderGen::WriteHeader(std::stringstream& ss)
     ss << "#define CONSTANT static const\n";
     ss << "#define VECTOR_EQ(a, b) (all((a) == (b)))\n";
     ss << "#define VECTOR_NEQ(a, b) (any((a) != (b)))\n";
+    ss << "#define VECTOR_LT(a, b) (any((a) < (b)))\n";
+    ss << "#define VECTOR_LE(a, b) (any((a) <= (b)))\n";
+    ss << "#define VECTOR_GT(a, b) (any((a) > (b)))\n";
+    ss << "#define VECTOR_GE(a, b) (any((a) >= (b)))\n";
     ss << "#define SAMPLE_TEXTURE(name, coords) name.Sample(name##_ss, coords)\n";
     ss << "#define LOAD_TEXTURE(name, coords, mip) name.Load(int3(coords, mip))\n";
     ss << "#define LOAD_TEXTURE_OFFSET(name, coords, mip, offset) name.Load(int3(coords, mip), offset)\n";
@@ -1104,8 +1112,8 @@ std::string GPU_HW_ShaderGen::GenerateVRAMCopyFragmentShader()
   std::stringstream ss;
   WriteHeader(ss);
   WriteCommonFunctions(ss);
-  DeclareUniformBuffer(
-    ss, {"uint2 u_src_coords", "uint2 u_dst_coords", "uint2 u_size", "bool u_set_mask_bit", "float u_depth_value"});
+  DeclareUniformBuffer(ss, {"uint2 u_src_coords", "uint2 u_dst_coords", "uint2 u_end_coords", "uint2 u_size",
+                            "bool u_set_mask_bit", "float u_depth_value"});
 
   DeclareTexture(ss, "samp0", 0);
   DeclareFragmentEntryPoint(ss, 0, 1, {}, true, 1, true);
@@ -1113,6 +1121,10 @@ std::string GPU_HW_ShaderGen::GenerateVRAMCopyFragmentShader()
 {
   uint2 dst_coords = uint2(v_pos.xy);
 
+  // make sure it's not oversized and out of range
+  if (VECTOR_LT(dst_coords, u_dst_coords) && VECTOR_GE(dst_coords, u_end_coords))
+    discard;
+
   // find offset from the start of the row/column
   uint2 offset;
   offset.x = (dst_coords.x < u_dst_coords.x) ? ((VRAM_SIZE.x - 1u) - u_dst_coords.x + dst_coords.x) : (dst_coords.x - u_dst_coords.x);