Suggested change to lv_refr_join_area()

Bug (archived)
1

I use lvgl v6.0, but as far as I can tell, this still is usable in v7 too.

Following change is suggested to lv_refr_join_area():

      static void lv_refr_join_area(void)  {
      uint32_t join_from;
      uint32_t join_in;
      uint32_t diff;
      lv_area_t joined_area;

      if (disp_refr->inv_p == 0) return; // sanity

      for(join_in = 0; join_in < disp_refr->inv_p - 1; join_in++) {

          diff = 0;
          /*Check all areas to join them in 'join_in'*/
          for(join_from = join_in + 1; join_from < disp_refr->inv_p; join_from++) {

              // move if there was some joining previously
              if (diff) {
                lv_area_copy(&disp_refr->inv_areas[join_from - diff], &disp_refr->inv_areas[join_from]);
              }

              /*Check if the areas are on each other*/
              if(lv_area_is_on(&disp_refr->inv_areas[join_in], &disp_refr->inv_areas[join_from]) == false) {
                  continue;
              }

              lv_area_join(&joined_area, &disp_refr->inv_areas[join_in], &disp_refr->inv_areas[join_from]);

              /*Join two area only if the joined area size is smaller*/
              if(lv_area_get_size(&joined_area) < (lv_area_get_size(&disp_refr->inv_areas[join_in]) +
                                                   lv_area_get_size(&disp_refr->inv_areas[join_from]))) {
                  lv_area_copy(&disp_refr->inv_areas[join_in], &joined_area);
                  diff++;
              }
          }
          disp_refr->inv_p -= diff;
      }
  }

This removes the need for inv_area_joined array in lv_disp_t, sparing some precious RAM space. All code related to that array may be removed. Whether this reduces execution time may depend on context but I believe it mostly does. In v7 it should also simplify the finding of the “last” area to trivial (my motivation was different - I have a modified refresher, which allows function calls (cooperative multitasking) during refresh, leaving the refresh process during DMAing from the partial buffer; so I needed a way to be able to add new areas to the refresh FIFO while refresh process is running).

JW

2

Hi,

Thanks for the suggestion.

Could you send a pull request for v7 and add some comments about meaning of diff variable?

3

I don’t use git. With a help of a friend, I concocted something resembling a pull request, although I don’t have a working v7 setup so the patch is untested other than compiling (for a 32-bit mcu, obviously failing with warning-turned-error on the implicit arithmetic conversions, on what is probably 64-bit compiler).

diff is a badly named variable keeping track of the number of removed (joined-in) areas within one pass of the outer cycle.

JW