mirrors/kitty
1
0
Fork 0
mirror of https://github.com/kovidgoyal/kitty synced 2026-07-22 08:18:08 +02:00
Code Issues Packages Projects Releases Wiki Activity

Reduce data sent to GPU per draw by 30%

Browse Source 
Split up the Cell structure into a CPUCell and a GPUCell. Only the
GPUCell part needs to be sent to the GPU. Should make kitty use even
less system resources and make a performance difference on systems where
the GPU bandwidth is constrained.

Also allows adding more CPU only data in the future without affecting
GPU bandwidth. For example, hyperlinks or more combining characters.
This commit is contained in:
Kovid Goyal
2018-05-27 21:25:09 +05:30
parent c8756612f7
commit 8dea5b3e3e
13 changed files with 256 additions and 212 deletions
Download Patch File Download Diff File Expand all files Collapse all files

37
kitty/screen.c
View File

@@ -280,9 +280,10 @@ screen_designate_charset(Screen *self, uint32_t which, uint32_t as) {
}
static inline void
move_widened_char(Screen *self, Cell* cell, index_type xpos, index_type ypos) {
move_widened_char(Screen *self, CPUCell* cpu_cell, GPUCell *gpu_cell, index_type xpos, index_type ypos) {
self->cursor->x = xpos; self->cursor->y = ypos;
Cell src = *cell, *dest;
CPUCell src_cpu = *cpu_cell, *dest_cpu;
GPUCell src_gpu = *gpu_cell, *dest_gpu;
line_clear_text(self->linebuf->line, xpos, 1, BLANK_CHAR);
if (self->modes.mDECAWM) { // overflow goes onto next line
@@ -290,14 +291,17 @@ move_widened_char(Screen *self, Cell* cell, index_type xpos, index_type ypos) {
screen_linefeed(self);
self->linebuf->line_attrs[self->cursor->y] |= CONTINUED_MASK;
linebuf_init_line(self->linebuf, self->cursor->y);
dest = self->linebuf->line->cells;
dest_cpu = self->linebuf->line->cpu_cells;
dest_gpu = self->linebuf->line->gpu_cells;
self->cursor->x = MIN(2, self->columns);
linebuf_mark_line_dirty(self->linebuf, self->cursor->y);
} else {
dest = cell - 1;
dest_cpu = cpu_cell - 1;
dest_gpu = gpu_cell - 1;
self->cursor->x = self->columns;
}
*dest = src;
*dest_cpu = src_cpu;
*dest_gpu = src_gpu;
}
static inline void
@@ -320,10 +324,11 @@ draw_combining_char(Screen *self, char_type ch) {
self->is_dirty = true;
linebuf_mark_line_dirty(self->linebuf, ypos);
if (ch == 0xfe0f) { // emoji presentation variation marker makes default text presentation emoji (narrow emoji) into wide emoji
Cell *cell = self->linebuf->line->cells + xpos;
if ((cell->attrs & WIDTH_MASK) != 2 && cell->cc_idx[0] == VS16 && is_emoji_presentation_base(cell->ch)) {
cell->attrs = (cell->attrs & !WIDTH_MASK) | 2;
if (xpos == self->columns - 1) move_widened_char(self, cell, xpos, ypos);
CPUCell *cpu_cell = self->linebuf->line->cpu_cells + xpos;
GPUCell *gpu_cell = self->linebuf->line->gpu_cells + xpos;
if ((gpu_cell->attrs & WIDTH_MASK) != 2 && cpu_cell->cc_idx[0] == VS16 && is_emoji_presentation_base(cpu_cell->ch)) {
gpu_cell->attrs = (gpu_cell->attrs & !WIDTH_MASK) | 2;
if (xpos == self->columns - 1) move_widened_char(self, cpu_cell, gpu_cell, xpos, ypos);
else self->cursor->x++;
}
}
@@ -409,7 +414,7 @@ select_graphic_rendition(Screen *self, unsigned int *params, unsigned int count,
num = MIN(num, self->columns - x);
for (index_type y = region.top; y < MIN(region.bottom + 1, self->lines); y++) {
linebuf_init_line(self->linebuf, y);
apply_sgr_to_cells(self->linebuf->line->cells + x, num, params, count);
apply_sgr_to_cells(self->linebuf->line->gpu_cells + x, num, params, count);
}
} else {
index_type x, num;
@@ -418,7 +423,7 @@ select_graphic_rendition(Screen *self, unsigned int *params, unsigned int count,
else if (y == region.bottom) { x = 0; num = MIN(region.right + 1, self->columns); }
else { x = 0; num = self->columns; }
linebuf_init_line(self->linebuf, y);
apply_sgr_to_cells(self->linebuf->line->cells + x, num, params, count);
apply_sgr_to_cells(self->linebuf->line->gpu_cells + x, num, params, count);
}
}
} else cursor_from_sgr(self->cursor, params, count);
@@ -1280,8 +1285,8 @@ screen_request_capabilities(Screen *self, char c, PyObject *q) {
// Rendering {{{
static inline void
update_line_data(Line *line, unsigned int dest_y, uint8_t *data) {
size_t base = dest_y * line->xnum * sizeof(Cell);
memcpy(data + base, line->cells, line->xnum * sizeof(Cell));
size_t base = dest_y * line->xnum * sizeof(GPUCell);
memcpy(data + base, line->gpu_cells, line->xnum * sizeof(GPUCell));
}
@@ -1718,8 +1723,8 @@ screen_selection_range_for_line(Screen *self, index_type y, index_type *start, i
if (y >= self->lines) { return false; }
Line *line = visual_line_(self, y);
index_type xlimit = line->xnum, xstart = 0;
while (xlimit > 0 && CHAR_IS_BLANK(line->cells[xlimit - 1].ch)) xlimit--;
while (xstart < xlimit && CHAR_IS_BLANK(line->cells[xstart].ch)) xstart++;
while (xlimit > 0 && CHAR_IS_BLANK(line->cpu_cells[xlimit - 1].ch)) xlimit--;
while (xstart < xlimit && CHAR_IS_BLANK(line->cpu_cells[xstart].ch)) xstart++;
*start = xstart; *end = xlimit;
return true;
}
@@ -1738,7 +1743,7 @@ screen_selection_range_for_word(Screen *self, index_type x, index_type *y1, inde
index_type start, end;
Line *line = visual_line_(self, *y1);
*y2 = *y1;
#define is_ok(x) (is_word_char((line->cells[x].ch)) || is_opt_word_char(line->cells[x].ch))
#define is_ok(x) (is_word_char((line->cpu_cells[x].ch)) || is_opt_word_char(line->cpu_cells[x].ch))
if (!is_ok(x)) {
start = x; end = x + 1;
} else {