When playing a video, 55% of CPU used in strlen, callgraph flip_page: ta_talloc_strdup_append , ta_talloc_asprintf_append and ta_talloc_strndup_append . Before the fix: 2 fps when playing 1080p 24 fps video (--vo-kitty-use-shm=no), 24 fps after the fix. video/out/kitty: move bstr cmd to priv
Unified · Split
| @@ -48,7 +48,7 @@ · 47 unmodified lines | ||
| 48 | 48 | #define DEFAULT_WIDTH 80 |
| 49 | 49 | #define DEFAULT_HEIGHT 25 |
| 50 | 50 | |
| 51 | −static inline void write_str(const char *s) | |
| 51 | +static inline void write_bstr(bstr bs) | |
| 52 | 52 | { |
| 53 | 53 | // On POSIX platforms, write() is the fastest method. It also is the only |
| 54 | 54 | // one that allows atomic writes so mpv’s output will not be interrupted |
| @@ -57,25 +57,31 @@ static inline void write_str(const char *s) · 2 unmodified lines | ||
| 57 | 57 | // exceeding PIPE_BUF, but at least Linux does seem to implement it that |
| 58 | 58 | // way. |
| 59 | 59 | #if HAVE_POSIX |
| 60 | − int remain = strlen(s); | |
| 60 | + size_t remain = bs.len; | |
| 61 | + unsigned char *pos = bs.start; | |
| 61 | 62 | while (remain > 0) { |
| 62 | − ssize_t written = write(STDOUT_FILENO, s, remain); | |
| 63 | + ssize_t written = write(STDOUT_FILENO, pos, remain); | |
| 63 | 64 | if (written < 0) |
| 64 | 65 | return; |
| 65 | 66 | remain -= written; |
| 66 | − s += written; | |
| 67 | + pos += written; | |
| 67 | 68 | } |
| 68 | 69 | #else |
| 69 | − printf("%s", s); | |
| 70 | + fwrite(bs.start, 1, bs.len, stdout); | |
| 70 | 71 | fflush(stdout); |
| 71 | 72 | #endif |
| 72 | 73 | } |
| 73 | 74 | |
| 75 | +static inline void write_str(unsigned char* s) | |
| 76 | +{ | |
| 77 | + write_bstr(bstr0(s)); | |
| 78 | +} | |
| 79 | + | |
| 74 | 80 | #define KITTY_ESC_IMG "\033_Ga=T,f=24,s=%d,v=%d,C=1,q=2,m=1;" |
| 75 | 81 | #define KITTY_ESC_IMG_SHM "\033_Ga=T,t=s,f=24,s=%d,v=%d,C=1,q=2,m=1;%s\033\\" |
| 76 | 82 | #define KITTY_ESC_CONTINUE "\033_Gm=%d;" |
| 77 | −#define KITTY_ESC_END "\033\\" | |
| 78 | −#define KITTY_ESC_DELETE_ALL "\033_Ga=d;\033\\" | |
| 83 | +static const bstr KITTY_ESC_END = bstr0_lit("\033\\"); | |
| 84 | +static const bstr KITTY_ESC_DELETE_ALL = bstr0_lit("\033_Ga=d;\033\\"); | |
| 79 | 85 | |
| 80 | 86 | struct vo_kitty_opts { |
| 81 | 87 | int width, height, top, left, rows, cols; |
| @@ -91,6 +97,7 @@ struct priv { · 9 unmodified lines | ||
| 91 | 97 | char *shm_path, *shm_path_b64; |
| 92 | 98 | int buffer_size, output_size; |
| 93 | 99 | int shm_fd; |
| 100 | + bstr cmd; | |
| 94 | 101 | |
| 95 | 102 | int left, top, width, height, cols, rows; |
| 96 | 103 | |
| @@ -126,6 +133,7 @@ static void free_bufs(struct vo* vo) · 29 unmodified lines | ||
| 126 | 133 | |
| 127 | 134 | talloc_free(p->frame); |
| 128 | 135 | talloc_free(p->output); |
| 136 | + talloc_free(p->cmd.start); | |
| 129 | 137 | |
| 130 | 138 | if (p->opts.use_shm) { |
| 131 | 139 | close_shm(p); |
| @@ -177,7 +185,7 @@ static int reconfig(struct vo *vo, struct mp_image_params *params) · 45 unmodified lines | ||
| 177 | 185 | struct priv *p = vo->priv; |
| 178 | 186 | |
| 179 | 187 | vo->want_redraw = true; |
| 180 | − write_str(KITTY_ESC_DELETE_ALL); | |
| 188 | + write_bstr(KITTY_ESC_DELETE_ALL); | |
| 181 | 189 | if (p->opts.config_clear) |
| 182 | 190 | write_str(TERM_ESC_CLEAR_SCREEN); |
| 183 | 191 | |
| @@ -292,36 +300,40 @@ done: · 108 unmodified lines | ||
| 292 | 300 | |
| 293 | 301 | static void flip_page(struct vo *vo) |
| 294 | 302 | { |
| 295 | − struct priv* p = vo->priv; | |
| 296 | − | |
| 297 | − if (p->buffer == NULL) | |
| 303 | + struct priv *p = vo->priv; | |
| 304 | + if (!p->buffer) | |
| 298 | 305 | return; |
| 299 | 306 | |
| 300 | − char *cmd = talloc_asprintf(NULL, TERM_ESC_GOTO_YX, p->top, p->left); | |
| 307 | + p->cmd.len = 0; | |
| 308 | + | |
| 309 | + // Start with ESC to position the cursor | |
| 310 | + bstr_xappend_asprintf(NULL, &p->cmd, TERM_ESC_GOTO_YX, p->top, p->left); | |
| 301 | 311 | |
| 302 | 312 | if (p->opts.use_shm) { |
| 303 | − cmd = talloc_asprintf_append(cmd, KITTY_ESC_IMG_SHM, p->width, p->height, p->shm_path_b64); | |
| 313 | + bstr_xappend_asprintf(NULL, &p->cmd, KITTY_ESC_IMG_SHM, | |
| 314 | + p->width, p->height, p->shm_path_b64); | |
| 304 | 315 | } else { |
| 305 | − if (p->output == NULL) { | |
| 306 | − talloc_free(cmd); | |
| 316 | + if (!p->output) { | |
| 307 | 317 | return; |
| 308 | 318 | } |
| 309 | 319 | |
| 310 | − cmd = talloc_asprintf_append(cmd, KITTY_ESC_IMG, p->width, p->height); | |
| 311 | − for (int offset = 0, noffset;; offset += noffset) { | |
| 312 | − if (offset) | |
| 313 | − cmd = talloc_asprintf_append(cmd, KITTY_ESC_CONTINUE, offset < p->output_size); | |
| 314 | − noffset = MPMIN(4096, p->output_size - offset); | |
| 315 | − cmd = talloc_strndup_append(cmd, p->output + offset, noffset); | |
| 316 | − cmd = talloc_strdup_append(cmd, KITTY_ESC_END); | |
| 320 | + bstr_xappend_asprintf(NULL, &p->cmd, KITTY_ESC_IMG, p->width, p->height); | |
| 321 | + | |
| 322 | + for (int offset = 0; offset < p->output_size; ) { | |
| 323 | + int chunk = MPMIN(4096, p->output_size - offset); | |
| 324 | + | |
| 325 | + if (offset > 0) | |
| 326 | + bstr_xappend_asprintf(NULL, &p->cmd, KITTY_ESC_CONTINUE, | |
| 327 | + offset + chunk < p->output_size); | |
| 317 | 328 | |
| 318 | − if (offset >= p->output_size) | |
| 319 | − break; | |
| 329 | + // Append at max chunk bytes | |
| 330 | + bstr_xappend(NULL, &p->cmd, (bstr){p->output + offset, chunk}); | |
| 331 | + bstr_xappend(NULL, &p->cmd, KITTY_ESC_END); | |
| 332 | + offset += chunk; | |
| 320 | 333 | } |
| 321 | 334 | } |
| 322 | 335 | |
| 323 | − write_str(cmd); | |
| 324 | − talloc_free(cmd); | |
| 336 | + write_bstr(p->cmd); | |
| 325 | 337 | |
| 326 | 338 | #if HAVE_POSIX |
| 327 | 339 | if (p->opts.use_shm) |