[PATCH v4 0/2] MR7428: wined3d: Use temporary buffer when stretching a surface to itself with cpu blit

[Yuri Hérouard (＠yurih)]

Fixes https://bugs.winehq.org/show_bug.cgi?id=57283.

I believe some sort of smart two pass stretching is possible instead, but not worth it.

[dlls/wined3d/surface.c#L1570](https://gitlab.winehq.org/riyu12383/wine/-/blob/f48645249a3402801f1ca942c35e...) not exactly sure if it's related and should be removed then ?

--
  v4: wined3d: Use temporary buffer when stretching a surface to itself with cpu blit.
      ddraw/tests: Add tests for surface blitting to itself with stretching.

https://gitlab.winehq.org/wine/wine/-/merge_requests/7428

[PATCH v4 1/2] ddraw/tests: Add tests for surface blitting to itself with stretching.

[Yuri Hérouard]

From: Yuri Hérouard

---
 dlls/ddraw/tests/ddraw1.c | 148 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 148 insertions(+)

diff --git a/dlls/ddraw/tests/ddraw1.c b/dlls/ddraw/tests/ddraw1.c
index e62d9283aaa..fc546856c65 100644
--- a/dlls/ddraw/tests/ddraw1.c
+++ b/dlls/ddraw/tests/ddraw1.c
@@ -9582,6 +9582,153 @@ done:
     DestroyWindow(window);
 }

+static void test_stretched_blit(void)
+{
+    IDirectDrawSurface *surface1, *surface2, *surface3;
+    DDSURFACEDESC surface_desc;
+    IDirectDraw *ddraw;
+    HRESULT hr;
+    IDirect3DDevice *device;
+    ULONG color1, color2, color, refcount;
+    HWND window;
+    RECT rect;
+    BYTE* pSurface, *buffer;
+    color1 = 0xffffffff;
+    color2 = 0x00000000;
+    window = create_window();
+    ddraw = create_ddraw();
+    ok(!!ddraw, "Failed to create a ddraw object.\n");
+    if (!(device = create_device(ddraw, window, DDSCL_EXCLUSIVE | DDSCL_FULLSCREEN)))
+    {
+        skip("Failed to create a 3D device.\n");
+        IDirectDraw_Release(ddraw);
+        DestroyWindow(window);
+        return;
+    }
+    hr = IDirect3DDevice_QueryInterface(device, &IID_IDirectDrawSurface, (void **)&surface1);
+    ok(SUCCEEDED(hr), "Failed to get render target, hr %#lx.\n", hr);
+    memset(&surface_desc, 0, sizeof(surface_desc));
+    surface_desc.dwSize = sizeof(surface_desc);
+    surface_desc.dwFlags = DDSD_WIDTH | DDSD_HEIGHT | DDSD_CAPS;
+    surface_desc.dwWidth = 640;
+    surface_desc.dwHeight = 480;
+    surface_desc.ddsCaps.dwCaps = DDSCAPS_OFFSCREENPLAIN | DDSCAPS_SYSTEMMEMORY;
+    hr = IDirectDraw_CreateSurface(ddraw, &surface_desc, &surface1, NULL);
+    ok(SUCCEEDED(hr), "Failed to create surface, hr %#lx.\n", hr);
+    hr = IDirectDraw_CreateSurface(ddraw, &surface_desc, &surface2, NULL);
+    ok(SUCCEEDED(hr), "Failed to create surface, hr %#lx.\n", hr);
+    hr = IDirectDraw_CreateSurface(ddraw, &surface_desc, &surface3, NULL);
+    ok(SUCCEEDED(hr), "Failed to create surface, hr %#lx.\n", hr);
+
+    //Setting 1px checkerboard texture
+    hr = IDirectDrawSurface_Lock(surface1, NULL, &surface_desc, DDLOCK_WAIT, NULL);
+    ok(SUCCEEDED(hr), "Failed to lock surface, hr %#lx.\n", hr);
+    pSurface = (BYTE*)surface_desc.lpSurface;
+    for (int row = 0; row < 480; row++)
+    {
+        buffer = pSurface + row*surface_desc.lPitch;
+        for (int col = 0; col < 640; col++)
+        {
+            ((DWORD *)buffer)[col] = (row % 2 == 0) == (col % 2 == 0) ? color1 : color2;
+        }
+    }
+    hr = IDirectDrawSurface_Unlock(surface1, 0);
+    ok(SUCCEEDED(hr), "Failed to unlock surface, hr %#lx.\n", hr);
+
+#define CHECK_BLITTED_TEXTURE(surface_ptr, expected_color) \
+do { \
+BOOL color_ok = TRUE; \
+hr = IDirectDrawSurface_Lock(surface_ptr, NULL, &surface_desc, DDLOCK_WAIT, NULL); \
+ok(SUCCEEDED(hr), "Failed to lock surface, hr %#lx.\n", hr); \
+pSurface = (BYTE*)surface_desc.lpSurface; \
+for (int row = 0; row < surface_desc.dwHeight; row++) \
+{ \
+    for (int col = 0; col < surface_desc.dwWidth; col++) \
+    { \
+        buffer = pSurface + row*surface_desc.lPitch; \
+        color = ((DWORD *)buffer)[col]; \
+        if (color != expected_color) \
+        { \
+            color_ok = FALSE; \
+            break; \
+        } \
+        if(!color_ok) break; \
+    } \
+} \
+ok(color_ok, "Blitted texture color is incorrect.\n"); \
+hr = IDirectDrawSurface_Unlock(surface_ptr, 0); \
+ok(SUCCEEDED(hr), "Failed to unlock surface, hr %#lx.\n", hr); \
+} while(0)
+
+    rect.bottom = 240;
+    rect.right = 640;
+    rect.left = 0;
+    rect.top = 0;
+    hr = IDirectDrawSurface_Blt(surface2, NULL, surface1, &rect, DDBLT_WAIT, NULL);
+    ok(SUCCEEDED(hr), "Failed to blit, hr %#lx.\n", hr);
+
+    //Checking that dst is a 2px vertical checkerboard
+    CHECK_BLITTED_TEXTURE(surface2, ((row % 4 < 2) == (col % 2 == 0) ? color1 : color2));
+
+    rect.bottom = 480;
+    rect.right = 320;
+    rect.left = 0;
+    rect.top = 0;
+    hr = IDirectDrawSurface_Blt(surface2, NULL, surface1, &rect, DDBLT_WAIT, NULL);
+    ok(SUCCEEDED(hr), "Failed to blit, hr %#lx.\n", hr);
+
+    //Checking that dst is a 2px horizontal checkerboard
+    CHECK_BLITTED_TEXTURE(surface2, ((row % 2 == 0) == (col % 4 < 2) ? color1 : color2));
+
+    rect.bottom = 240;
+    rect.right = 320;
+    rect.left = 0;
+    rect.top = 0;
+    hr = IDirectDrawSurface_Blt(surface3, NULL, surface1, &rect, DDBLT_WAIT, NULL);
+    ok(SUCCEEDED(hr), "Failed to blit, hr %#lx.\n", hr);
+
+    //Checking that src is a 4px checkerboard
+    CHECK_BLITTED_TEXTURE(surface3, ((row % 4 < 2) == (col % 4 < 2) ? color1 : color2));
+
+    rect.bottom = 240;
+    rect.right = 320;
+    rect.left = 0;
+    rect.top = 0;
+    //Blitting surface to itself ! Both axis stretched
+    hr = IDirectDrawSurface_Blt(surface1, NULL, surface1, &rect, DDBLT_WAIT, NULL);
+    ok(SUCCEEDED(hr), "Failed to blit, hr %#lx.\n", hr);
+
+    //Checking that src is a 4px checkerboard
+    CHECK_BLITTED_TEXTURE(surface1, ((row % 4 < 2) == (col % 4 < 2) ? color1 : color2));
+
+    rect.bottom = 240;
+    rect.right = 640;
+    rect.left = 0;
+    rect.top = 0;
+    //Blitting surface to itself ! Vertical stretch
+    hr = IDirectDrawSurface_Blt(surface2, NULL, surface2, &rect, DDBLT_WAIT, NULL);
+    ok(SUCCEEDED(hr), "Failed to blit, hr %#lx.\n", hr);
+
+    //Checking that src is a 4px checkerboard
+    CHECK_BLITTED_TEXTURE(surface2, ((row % 4 < 2) == (col % 4 < 2) ? color1 : color2));
+
+    rect.bottom = 480;
+    rect.right = 320;
+    rect.left = 0;
+    rect.top = 0;
+    //Blitting surface to itself ! Horizontal stretch
+    hr = IDirectDrawSurface_Blt(surface3, NULL, surface3, &rect, DDBLT_WAIT, NULL);
+    ok(SUCCEEDED(hr), "Failed to blit, hr %#lx.\n", hr);
+
+    //Checking that src is a 2x4 px checkerboard
+    CHECK_BLITTED_TEXTURE(surface3, ((row % 4 < 2) == (col % 8 < 4) ? color1 : color2));
+
+    refcount = IDirectDraw_Release(ddraw);
+    ok(!refcount, "Unexpected refcount %lu.\n", refcount);
+    if (window) DestroyWindow(window);
+#undef CHECK_BLITTED_TEXTURE
+}
+
 static void test_colorkey_precision(void)
 {
     static D3DTLVERTEX quad[] =
@@ -16242,4 +16389,5 @@ START_TEST(ddraw1)
     test_multiple_devices();
     test_sysmem_x_channel();
     test_yuv_blit();
+    test_stretched_blit();
 }

[PATCH v4 2/2] wined3d: Use temporary buffer when stretching a surface to itself with cpu blit.

[Yuri Hérouard]

From: Yuri Hérouard

---
 dlls/wined3d/surface.c | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)

diff --git a/dlls/wined3d/surface.c b/dlls/wined3d/surface.c
index 14cfebd7074..a054caba15a 100644
--- a/dlls/wined3d/surface.c
+++ b/dlls/wined3d/surface.c
@@ -616,6 +616,7 @@ static HRESULT surface_cpu_blt(struct wined3d_texture *dst_texture, unsigned int
     const BYTE *sbase;
     const BYTE *sbuf;
     BYTE *dbuf;
+    BYTE *tmp_buffer = NULL;

TRACE("dst_texture %p, dst_sub_resource_idx %u, dst_box %s, src_texture %p, "
             "src_sub_resource_idx %u, src_box %s, flags %#x, fx %p, filter %s.\n",
@@ -808,18 +809,45 @@ static HRESULT surface_cpu_blt(struct wined3d_texture *dst_texture, unsigned int
             else
             {
                 /* Stretching in y direction only. */
+
+                if (same_sub_resource)
+                {
+                    /* Use a temporary buffer if blitting surface to itself */
+                    tmp_buffer = malloc(src_height * src_map.row_pitch);
+                    if (tmp_buffer != NULL)
+                    {
+                        memcpy(tmp_buffer, sbase, src_height * src_map.row_pitch);
+                        sbase = tmp_buffer;
+                    }
+                }
+
                 for (y = sy = 0; y < dst_height; ++y, sy += yinc)
                 {
                     sbuf = sbase + (sy >> 16) * src_map.row_pitch;
                     memcpy(dbuf, sbuf, row_byte_count);
                     dbuf += dst_map.row_pitch;
                 }
+
+                if(same_sub_resource && tmp_buffer != NULL)
+                    free(tmp_buffer);
             }
         }
         else
         {
             /* Stretching in X direction. */
             unsigned int last_sy = ~0u;
+
+            if (same_sub_resource)
+            {
+                /* Use a temporary buffer if blitting surface to itself */
+                tmp_buffer = malloc(src_height * src_map.row_pitch);
+                if (tmp_buffer != NULL)
+                {
+                    memcpy(tmp_buffer, sbase, src_height * src_map.row_pitch);
+                    sbase = tmp_buffer;
+                }
+            }
+
             for (y = sy = 0; y < dst_height; ++y, sy += yinc)
             {
                 sbuf = sbase + (sy >> 16) * src_map.row_pitch;
@@ -878,6 +906,9 @@ do { \
                 dbuf += dst_map.row_pitch;
                 last_sy = sy;
             }
+
+            if(same_sub_resource && tmp_buffer != NULL)
+                free(tmp_buffer);
         }
     }
     else