github-code/lvgl
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

reduce shadow memory usage by ~30%

Browse Source
This commit is contained in:
Gabor Kiss-Vamosi
2020-01-21 23:59:54 +01:00
parent c8027aa70f
commit 18378e0a08
1 changed files with 18 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
src/lv_draw/lv_draw_rect.c
View File

@@ -890,6 +890,11 @@ static void shadow_draw_corner_buf(const lv_area_t * coords, lv_opa_t * sh_buf,
else sw = sw_ori >> 1; else sw = sw_ori >> 1;
#endif #endif
/* The basic idea of shadow drawing is
*
*/
lv_draw_mask_res_t mask_res; lv_draw_mask_res_t mask_res;
lv_coord_t y; lv_coord_t y;
lv_opa_t * mask_line = lv_mem_buf_get(size); lv_opa_t * mask_line = lv_mem_buf_get(size);
@@ -958,19 +963,17 @@ static void shadow_blur_corner(lv_coord_t size, lv_coord_t sw, lv_opa_t * res_bu
if((sw & 1) == 0) s_left--; if((sw & 1) == 0) s_left--;
/*Horizontal blur*/ /*Horizontal blur*/
uint16_t * sh_ups_hor_buf = lv_mem_buf_get(size * size * sizeof(uint16_t)); uint16_t * sh_ups_hor_buf = lv_mem_buf_get(size * sizeof(uint16_t));
uint16_t * sh_ups_hor_buf_tmp;
lv_coord_t x; lv_coord_t x;
lv_coord_t y; lv_coord_t y;
uint16_t * sh_ups_tmp_buf = sh_ups_buf; uint16_t * sh_ups_tmp_buf = sh_ups_buf;
sh_ups_hor_buf_tmp = sh_ups_hor_buf;
for(y = 0; y < size; y++) { for(y = 0; y < size; y++) {
int32_t v = sh_ups_tmp_buf[size-1] * sw; int32_t v = sh_ups_tmp_buf[size-1] * sw;
for(x = size - 1; x >=0; x--) { for(x = size - 1; x >=0; x--) {
sh_ups_hor_buf_tmp[x] = v; sh_ups_hor_buf[x] = v;
/*Forget the right pixel*/ /*Forget the right pixel*/
uint32_t right_val = 0; uint32_t right_val = 0;
@@ -984,37 +987,35 @@ static void shadow_blur_corner(lv_coord_t size, lv_coord_t sw, lv_opa_t * res_bu
v += left_val; v += left_val;
} }
memcpy(sh_ups_tmp_buf, sh_ups_hor_buf, size * sizeof(uint16_t));
sh_ups_tmp_buf += size; sh_ups_tmp_buf += size;
sh_ups_hor_buf_tmp += size;
} }
/*Vertical blur*/ /*Vertical blur*/
uint32_t i; uint32_t i;
sh_ups_hor_buf[0] = sh_ups_hor_buf[0] / sw; sh_ups_buf[0] = sh_ups_buf[0] / sw;
for(i = 1; i < (uint32_t)size * size; i++) { for(i = 1; i < (uint32_t)size * size; i++) {
if(sh_ups_hor_buf[i] == sh_ups_hor_buf[i-1]) sh_ups_hor_buf[i] = sh_ups_hor_buf[i-1]; if(sh_ups_buf[i] == sh_ups_buf[i-1]) sh_ups_buf[i] = sh_ups_buf[i-1];
else sh_ups_hor_buf[i] = sh_ups_hor_buf[i] / sw; else sh_ups_buf[i] = sh_ups_buf[i] / sw;
} }
for(x = 0; x < size; x++) { for(x = 0; x < size; x++) {
sh_ups_hor_buf_tmp = &sh_ups_hor_buf[x]; sh_ups_tmp_buf = &sh_ups_buf[x];
lv_opa_t * sh_buf_tmp = &res_buf[x]; lv_opa_t * sh_buf_tmp = &res_buf[x];
int32_t v = sh_ups_hor_buf_tmp[0] * sw; int32_t v = sh_ups_tmp_buf[0] * sw;
for(y = 0; y < size ; y++, sh_ups_hor_buf_tmp += size, sh_buf_tmp += size) { for(y = 0; y < size ; y++, sh_ups_tmp_buf += size, sh_buf_tmp += size) {
sh_buf_tmp[0] = v < 0 ? 0 : (v >> SHADOW_UPSACALE_SHIFT); sh_buf_tmp[0] = v < 0 ? 0 : (v >> SHADOW_UPSACALE_SHIFT);
/*Forget the top pixel*/ /*Forget the top pixel*/
uint32_t top_val; uint32_t top_val;
if(y - s_right <= 0) top_val = sh_ups_hor_buf_tmp[0]; if(y - s_right <= 0) top_val = sh_ups_tmp_buf[0];
else top_val = sh_ups_hor_buf[(y - s_right) * size + x]; else top_val = sh_ups_buf[(y - s_right) * size + x];
v -= top_val; v -= top_val;
/*Add the bottom pixel*/ /*Add the bottom pixel*/
uint32_t bottom_val; uint32_t bottom_val;
if(y + s_left + 1 < size) bottom_val = sh_ups_hor_buf[(y + s_left + 1) * size + x]; if(y + s_left + 1 < size) bottom_val = sh_ups_buf[(y + s_left + 1) * size + x];
else bottom_val = sh_ups_hor_buf[(size - 1) * size + x]; else bottom_val = sh_ups_buf[(size - 1) * size + x];
v += bottom_val; v += bottom_val;
} }
} }