Merge 19e8774ad0 into 7c1d2a64f4

speeds up compilation a tiny bit

oh and NOMINMAX too for good measure while we're at it

src/CMakeLists.txt

@@ -127,6 +127,8 @@ if (ENABLE_JIT)
     endif()
 endif()
 
+target_include_directories(core INTERFACE "${CMAKE_CURRENT_SOURCE_DIR}")
+
 set(MELONDS_VERSION_SUFFIX "$ENV{MELONDS_VERSION_SUFFIX}" CACHE STRING "Suffix to add to displayed melonDS version")
 option(MELONDS_EMBED_BUILD_INFO "Embed detailed build info into the binary" OFF)
 set(MELONDS_GIT_BRANCH "$ENV{MELONDS_GIT_BRANCH}" CACHE STRING "The Git branch used for this build")
@@ -178,6 +180,7 @@ endif()
 
 if (WIN32)
     target_link_libraries(core PRIVATE ole32 comctl32 wsock32 ws2_32)
+    target_compile_definitions(core PUBLIC WIN32_LEAN_AND_MEAN NOMINMAX)
 elseif(NOT APPLE AND NOT HAIKU)
     check_library_exists(rt shm_open "" NEED_LIBRT)
     if (NEED_LIBRT)

src/GPU3D_Soft.cpp

@@ -622,7 +622,7 @@ void SoftRenderer::SetupPolygonLeftEdge(SoftRenderer::RendererPolygon* rp, s32 y
 
     rp->XL = rp->SlopeL.Setup(polygon->Vertices[rp->CurVL]->FinalPosition[0], polygon->Vertices[rp->NextVL]->FinalPosition[0],
                               polygon->Vertices[rp->CurVL]->FinalPosition[1], polygon->Vertices[rp->NextVL]->FinalPosition[1],
-                              polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL], y);
+                              polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL], y, polygon->WBuffer);
 }
 
 void SoftRenderer::SetupPolygonRightEdge(SoftRenderer::RendererPolygon* rp, s32 y) const
@@ -649,7 +649,7 @@ void SoftRenderer::SetupPolygonRightEdge(SoftRenderer::RendererPolygon* rp, s32
 
     rp->XR = rp->SlopeR.Setup(polygon->Vertices[rp->CurVR]->FinalPosition[0], polygon->Vertices[rp->NextVR]->FinalPosition[0],
                               polygon->Vertices[rp->CurVR]->FinalPosition[1], polygon->Vertices[rp->NextVR]->FinalPosition[1],
-                              polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR], y);
+                              polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR], y, polygon->WBuffer);
 }
 
 void SoftRenderer::SetupPolygon(SoftRenderer::RendererPolygon* rp, Polygon* polygon) const
@@ -755,8 +755,8 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
     s32 wl = rp->SlopeL.Interp.Interpolate(polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL]);
     s32 wr = rp->SlopeR.Interp.Interpolate(polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR]);
 
-    s32 zl = rp->SlopeL.Interp.InterpolateZ(polygon->FinalZ[rp->CurVL], polygon->FinalZ[rp->NextVL], polygon->WBuffer);
-    s32 zr = rp->SlopeR.Interp.InterpolateZ(polygon->FinalZ[rp->CurVR], polygon->FinalZ[rp->NextVR], polygon->WBuffer);
+    s32 zl = rp->SlopeL.Interp.InterpolateZ(polygon->FinalZ[rp->CurVL], polygon->FinalZ[rp->NextVL]);
+    s32 zr = rp->SlopeR.Interp.InterpolateZ(polygon->FinalZ[rp->CurVR], polygon->FinalZ[rp->NextVR]);
 
     // right vertical edges are pushed 1px to the left as long as either:
     // the left edge slope is not 0, or the span is not 0 pixels wide, and it is not at the leftmost pixel of the screen
@@ -841,7 +841,7 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
     int edge;
 
     s32 x = xstart;
-    Interpolator<0> interpX(xstart, xend+1, wl, wr);
+    Interpolator<0> interpX(xstart, xend+1, wl, wr, polygon->WBuffer, zl, zr);
 
     if (x < 0) x = 0;
     s32 xlimit;
@@ -863,7 +863,7 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
 
         interpX.SetX(x);
 
-        s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer);
+        s32 z = interpX.InterpolateZ(zl, zr);
         u32 dstattr = AttrBuffer[pixeladdr];
 
         if (!fnDepthTest(DepthBuffer[pixeladdr], z, dstattr))
@@ -889,7 +889,7 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
 
         interpX.SetX(x);
 
-        s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer);
+        s32 z = interpX.InterpolateZ(zl, zr);
         u32 dstattr = AttrBuffer[pixeladdr];
 
         if (!fnDepthTest(DepthBuffer[pixeladdr], z, dstattr))
@@ -915,7 +915,7 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
 
         interpX.SetX(x);
 
-        s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer);
+        s32 z = interpX.InterpolateZ(zl, zr);
         u32 dstattr = AttrBuffer[pixeladdr];
 
         if (!fnDepthTest(DepthBuffer[pixeladdr], z, dstattr))
@@ -980,8 +980,8 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
     s32 wl = rp->SlopeL.Interp.Interpolate(polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL]);
     s32 wr = rp->SlopeR.Interp.Interpolate(polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR]);
 
-    s32 zl = rp->SlopeL.Interp.InterpolateZ(polygon->FinalZ[rp->CurVL], polygon->FinalZ[rp->NextVL], polygon->WBuffer);
-    s32 zr = rp->SlopeR.Interp.InterpolateZ(polygon->FinalZ[rp->CurVR], polygon->FinalZ[rp->NextVR], polygon->WBuffer);
+    s32 zl = rp->SlopeL.Interp.InterpolateZ(polygon->FinalZ[rp->CurVL], polygon->FinalZ[rp->NextVL]);
+    s32 zr = rp->SlopeR.Interp.InterpolateZ(polygon->FinalZ[rp->CurVR], polygon->FinalZ[rp->NextVR]);
 
     // right vertical edges are pushed 1px to the left as long as either:
     // the left edge slope is not 0, or the span is not 0 pixels wide, and it is not at the leftmost pixel of the screen
@@ -1091,7 +1091,7 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
     int edge;
 
     s32 x = xstart;
-    Interpolator<0> interpX(xstart, xend+1, wl, wr);
+    Interpolator<0> interpX(xstart, xend+1, wl, wr, polygon->WBuffer, zl, zr);
 
     if (x < 0) x = 0;
     s32 xlimit;
@@ -1130,7 +1130,7 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
 
         interpX.SetX(x);
 
-        s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer);
+        s32 z = interpX.InterpolateZ(zl, zr);
 
         // if depth test against the topmost pixel fails, test
         // against the pixel underneath
@@ -1226,7 +1226,7 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
 
         interpX.SetX(x);
 
-        s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer);
+        s32 z = interpX.InterpolateZ(zl, zr);
 
         // if depth test against the topmost pixel fails, test
         // against the pixel underneath
@@ -1318,7 +1318,7 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
 
         interpX.SetX(x);
 
-        s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer);
+        s32 z = interpX.InterpolateZ(zl, zr);
 
         // if depth test against the topmost pixel fails, test
         // against the pixel underneath

src/GPU3D_Soft.h

@@ -69,23 +69,24 @@ private:
     {
     public:
         constexpr Interpolator() {}
-        constexpr Interpolator(s32 x0, s32 x1, s32 w0, s32 w1)
+        constexpr Interpolator(s32 x0, s32 x1, s32 w0, s32 w1, bool wbuffer, s32 z0, s32 z1)
         {
-            Setup(x0, x1, w0, w1);
+            Setup(x0, x1, w0, w1, wbuffer, z0, z1);
         }
 
-        constexpr void Setup(s32 x0, s32 x1, s32 w0, s32 w1)
+        constexpr void Setup(s32 x0, s32 x1, s32 w0, s32 w1, bool wbuffer, s32 z0 = 0, s32 z1 = 0)
         {
             this->x0 = x0;
             this->x1 = x1;
             this->xdiff = x1 - x0;
+            this->wbuffer = wbuffer;
 
-            // calculate reciprocal for Z interpolation
-            // TODO eventually: use a faster reciprocal function?
-            if (this->xdiff != 0)
-                this->xrecip_z = (1<<22) / this->xdiff;
-            else
-                this->xrecip_z = 0;
+            // calculate increment and init counter for Z interpolation
+            if (!dir && !wbuffer && xdiff != 0)
+            {
+                this->zincr = ((z1 - z0) >> 1) / xdiff << 1;
+                this->zcounter = z0;
+            }
 
             // linear mode is used if both W values are equal and have
             // low-order bits cleared (0-6 along X, 1-6 along Y)
@@ -144,7 +145,7 @@ private:
 
         constexpr s32 Interpolate(s32 y0, s32 y1) const
         {
-            if (xdiff == 0 || y0 == y1) return y0;
+            if (x == 0 || xdiff == 0 ||  y0 == y1) return y0;
 
             if (!linear)
             {
@@ -164,9 +165,9 @@ private:
             }
         }
 
-        constexpr s32 InterpolateZ(s32 z0, s32 z1, bool wbuffer) const
+        constexpr s32 InterpolateZ(s32 z0, s32 z1)
         {
-            if (xdiff == 0 || z0 == z1) return z0;
+            if (x == 0 || xdiff == 0 || z0 == z1) return z0;
 
             if (wbuffer)
             {
@@ -179,37 +180,22 @@ private:
             else
             {
                 // Z-buffering: linear interpolation
-                // still doesn't quite match hardware...
-                s32 base = 0, disp = 0, factor = 0;
-
-                if (z0 < z1)
-                {
-                    base = z0;
-                    disp = z1 - z0;
-                    factor = x;
-                }
-                else
-                {
-                    base = z1;
-                    disp = z0 - z1,
-                    factor = xdiff - x;
-                }
+                // not perfect, but close
 
                 if (dir)
                 {
-                    int shift = 0;
-                    while (disp > 0x3FF)
-                    {
-                        disp >>= 1;
-                        shift++;
-                    }
-
-                    return base + ((((s64)disp * factor * xrecip_z) >> 22) << shift);
+                    // interpolating along y uses a different algorithm than x
+                    // this algo probably isn't quite right though...
+                    if (z0 < z1)
+                        return z0 + (s64)(z1-z0) * x / xdiff;
+                    else
+                        return z1 + (s64)(z0-z1) * (xdiff-x) / xdiff;
                 }
                 else
                 {
-                    disp >>= 9;
-                    return base + (((s64)disp * factor * xrecip_z) >> 13);
+                    // unoptimized algorithm is: z0 + ((z1-z0 >> 1) / xdiff * x << 1);
+                    // or alternatively there's: z0 + (z1-z0) / (xdiff<<1) * (x<<1);
+                    return zcounter += zincr;
                 }
             }
         }
@@ -219,8 +205,10 @@ private:
 
         int shift;
         bool linear;
+        bool wbuffer;
 
-        s32 xrecip_z;
+        s32 zincr;
+        s32 zcounter;
         s32 w0n, w0d, w1d;
 
         u32 yfactor;
@@ -244,7 +232,7 @@ private:
             Increment = 0;
             XMajor = false;
 
-            Interp.Setup(0, 0, 0, 0);
+            Interp.Setup(0, 0, 0, 0, false);
             Interp.SetX(0);
 
             xcov_incr = 0;
@@ -252,7 +240,7 @@ private:
             return x0;
         }
 
-        constexpr s32 Setup(s32 x0, s32 x1, s32 y0, s32 y1, s32 w0, s32 w1, s32 y)
+        constexpr s32 Setup(s32 x0, s32 x1, s32 y0, s32 y1, s32 w0, s32 w1, s32 y, bool wbuffer)
         {
             this->x0 = x0;
             this->y = y;
@@ -318,7 +306,7 @@ private:
             s32 x = XVal();
 
             int interpoffset = (Increment >= 0x40000) && (side ^ Negative);
-            Interp.Setup(y0-interpoffset, y1-interpoffset, w0, w1);
+            Interp.Setup(y0-interpoffset, y1-interpoffset, w0, w1, wbuffer);
             Interp.SetX(y);
 
             // used for calculating AA coverage

src/frontend/qt_sdl/CMakeLists.txt

@@ -91,8 +91,7 @@ add_compile_definitions(ARCHIVE_SUPPORT_ENABLED)
 add_executable(melonDS ${SOURCES_QT_SDL})
 
 add_subdirectory("../../net"
-        "${CMAKE_BINARY_DIR}/net"
-)
+    ${CMAKE_BINARY_DIR}/net)
 
 target_link_libraries(melonDS PRIVATE net-utils)
 
@@ -171,10 +170,10 @@ if (BUILD_STATIC)
     endif()
 endif()
 
-target_include_directories(melonDS PUBLIC "${CMAKE_CURRENT_SOURCE_DIR}")
-target_include_directories(melonDS PUBLIC "${CMAKE_CURRENT_SOURCE_DIR}/..")
-target_include_directories(melonDS PUBLIC "${CMAKE_CURRENT_SOURCE_DIR}/../..")
-target_include_directories(melonDS PUBLIC "${CMAKE_CURRENT_SOURCE_DIR}/../../net")
+target_include_directories(melonDS PUBLIC
+    "${CMAKE_CURRENT_SOURCE_DIR}"
+    "${CMAKE_CURRENT_SOURCE_DIR}/..")
+
 if (USE_QT6)
     target_include_directories(melonDS PUBLIC ${Qt6Gui_PRIVATE_INCLUDE_DIRS})
 else()

src/net/CMakeLists.txt

@@ -11,9 +11,9 @@ add_library(net-utils STATIC
     MPInterface.cpp
 )
 
-target_include_directories(net-utils PUBLIC "${CMAKE_CURRENT_SOURCE_DIR}")
-target_include_directories(net-utils PUBLIC "${CMAKE_CURRENT_SOURCE_DIR}/..")
-
+target_include_directories(net-utils PUBLIC
+    "${CMAKE_CURRENT_SOURCE_DIR}"
+    "${CMAKE_CURRENT_SOURCE_DIR}/..")
 
 option(USE_SYSTEM_LIBSLIRP "Use system libslirp instead of the bundled version" OFF)
 if (USE_SYSTEM_LIBSLIRP)