Evas GL common: Improve quality of miniature image

Sample in the middle of the "macro pixels" and fool around with the
borders (usually used to limit linear sampling artifacts) to improve
image quality on the edges.

Those miniatures are still 16x16 but MAAAYYYYYBE they will look a bit
less awful.

NOTE: The first row still doesn't scale properly (interpolates with
      garbage above y=0).

diff --git a/src/modules/evas/engines/gl_common/evas_gl_texture.c b/src/modules/evas/engines/gl_common/evas_gl_texture.c
index 2f3852f..6c3b034 100644 (file)
--- a/src/modules/evas/engines/gl_common/evas_gl_texture.c
+++ b/src/modules/evas/engines/gl_common/evas_gl_texture.c
@@ -1283,26 +1283,61 @@ evas_gl_common_texture_update(Evas_GL_Texture *tex, RGBA_Image *im)
 
         out = alloca(bytes_count *  EVAS_GL_TILE_SIZE * EVAS_GL_TILE_SIZE);
         xstep = (float)tex->w / (EVAS_GL_TILE_SIZE - 2);
-        ystep = (float)tex->h / (EVAS_GL_TILE_SIZE - 1);
+        ystep = (float)tex->h / (EVAS_GL_TILE_SIZE - 2);
         in = im->image.data8;
 
-        for (y = 0, j = 0; j < EVAS_GL_TILE_SIZE - 1; y += ystep, j++)
+        // top-left
+        memcpy(&out[0],
+               &in[0],
+               bytes_count);
+
+        // top
+        for (x = xstep * 0.5f, i = 1; i < EVAS_GL_TILE_SIZE - 1; x += xstep, i++)
+          {
+             memcpy(&out[i * bytes_count],
+                    &in[(int)x * bytes_count],
+                    bytes_count);
+          }
+
+        // top-right
+        memcpy(&out[((EVAS_GL_TILE_SIZE - 1) * bytes_count)],
+               &in[(im->cache_entry.w  - 1) * bytes_count],
+               bytes_count);
+
+        for (y = ystep * 0.5f, j = 1; j < EVAS_GL_TILE_SIZE - 1; y += ystep, j++)
           {
+             // left
              memcpy(&out[j * EVAS_GL_TILE_SIZE * bytes_count],
-                    &in[(int)y * im->cache_entry.w * bytes_count],
+                    &in[((int)y * im->cache_entry.w) * bytes_count],
                     bytes_count);
-             for (x = 0, i = 1; i < EVAS_GL_TILE_SIZE - 1; x += xstep, i++)
+             // middle
+             for (x = xstep * 0.5f, i = 1; i < EVAS_GL_TILE_SIZE - 1; x += xstep, i++)
                memcpy(&out[(j * EVAS_GL_TILE_SIZE + i) * bytes_count],
-                      &in[((int)y * im->cache_entry.w + (int)x) * bytes_count],
+                      &in[(((int)y * im->cache_entry.w) + (int)x) * bytes_count],
                       bytes_count);
+             // right
              memcpy(&out[(j * EVAS_GL_TILE_SIZE + i) * bytes_count],
-                    &in[((int)y * im->cache_entry.w + (int)(x - xstep)) * bytes_count],
+                    &in[(((int)y * im->cache_entry.w) + (im->cache_entry.w - 1)) * bytes_count],
                     bytes_count);
           }
 
+        // bottom-left
         memcpy(&out[(j * EVAS_GL_TILE_SIZE) * bytes_count],
-               &out[((j - 1) * EVAS_GL_TILE_SIZE) * bytes_count],
-               EVAS_GL_TILE_SIZE * bytes_count);
+               &in[((im->cache_entry.w * (im->cache_entry.h - 1)) + 1) * bytes_count],
+               bytes_count);
+
+        // bottom
+        for (x = xstep * 0.5f, i = 1; i < EVAS_GL_TILE_SIZE - 1; x += xstep, i++)
+          {
+             memcpy(&out[((EVAS_GL_TILE_SIZE * j) + i) * bytes_count],
+                    &in[((int)x + im->cache_entry.w * (im->cache_entry.h - 1)) * bytes_count],
+                    bytes_count);
+          }
+
+        // bottom-right
+        memcpy(&out[((EVAS_GL_TILE_SIZE * EVAS_GL_TILE_SIZE) - 1) * bytes_count],
+               &in[((im->cache_entry.w * im->cache_entry.h) - 1) * bytes_count],
+               bytes_count);
 
         // out is a miniature of the texture, upload that now and schedule the data for later.