Hi Geert,
I have a small patch to scale-region.c that I would to have your opinion
on. I noticed that the current code sometimes does an unneeded copy
operation. This happens when the scale factor is 2^n. For example when
an image of 800x600 pixels is scaled to 400x300. The function
determine_scale() then decides that a first scale step to 400x300 needs
to be made and the result is then scaled again with a scale factor of
1.0. There's even an optimization n the scale() function for this
special case.
Attached is a patch that changes determine_scale() to avoid this extra
step. Is there anything that I am missing or do you agree that it should
be safe to apply this?
Sven
Index: app/paint-funcs/scale-region.c
===================================================================
--- app/paint-funcs/scale-region.c (revision 26730)
+++ app/paint-funcs/scale-region.c (working copy)
@@ -170,7 +170,7 @@ determine_scale (PixelRegion *srcPR,
*max_progress = ((height % TILE_HEIGHT) + 1) * ((width % TILE_WIDTH) + 1);
/* determine scaling levels */
- while (scalex >= 2)
+ while (scalex > 2)
{
scalex /= 2;
width *=2;
@@ -179,7 +179,7 @@ determine_scale (PixelRegion *srcPR,
((width % TILE_WIDTH) + 1));
}
- while (scaley >= 2)
+ while (scaley > 2)
{
scaley /= 2;
height *= 2;
@@ -188,7 +188,7 @@ determine_scale (PixelRegion *srcPR,
((width % TILE_WIDTH) + 1));
}
- while (scalex <= 0.5)
+ while (scalex < 0.5)
{
scalex *= 2;
width /= 2;
@@ -197,7 +197,7 @@ determine_scale (PixelRegion *srcPR,
((width % TILE_WIDTH) + 1));
}
- while (scaley <= 0.5)
+ while (scaley < 0.5)
{
scaley *= 2;
height *= 2;
@@ -370,6 +370,9 @@ scale_region_tile (PixelRegion
/* determine scaling levels */
determine_scale (srcPR, dstPR, &levelx, &levely, &max_progress);
+ g_printerr ("%d x %d -> %d x %d: %d, %d\n",
+ width, height, dstPR->w, dstPR->h, levelx, levely);
+
if (levelx == 0 && levely == 0)
{
scale (srcTM, dstTM, interpolation,
_______________________________________________
Gimp-developer mailing list
Gimp-developer@xxxxxxxxxxxxxxxxxxxxxx
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
