vid-scratch: Add consolidated versions of my experiments from 2025-02-XXlukeshu/vid-scratch

1 files changed, 252 insertions, 0 deletions

diff --git a/vid-scratch/nutgen-sidedata.c b/vid-scratch/nutgen-sidedata.c
new file mode 100644
index 0000000..62fc680
--- /dev/null
+++ b/vid-scratch/nutgen-sidedata.c
@@ -0,0 +1,252 @@
+/* nut.c - A simple NUT encoder with the following properties:
+ *
+ *  - Each frame is an uncompressed 8-bit RGB (specifically `(msb)2B
+ *    3G 3R(lsb)`, known as the 4CC "BGR\x08") framebuffer.
+ *
+ *  - Each pixel in the framebuffer is non-square; it is twice as tall
+ *    as it is wide.
+ *
+ *  - The display-resolution of each frame may change between 640x480,
+ *    720x480, and 720x576.  This is the display-resolution; because
+ *    of non-square pixels, the framebuffer for each would be 640x240,
+ *    720x240, and 720x288, respectively.
+ *
+ *  - VFR (Variable Frame Rate) - Each frame has its own timestamp.
+ */
+
+#include "common.h"
+
+/* High-level (specific to the properties we want) ****************************/
+
+enum app_res {
+	APP_RES_640_480 = 0,
+	APP_RES_720_480 = 1,
+	APP_RES_720_576 = 2,
+};
+#define APP_RES_MIN APP_RES_640_480
+#define APP_RES_MAX APP_RES_720_576
+
+size_t app_res_w(enum app_res res) {
+	switch (res) {
+	case APP_RES_640_480: return 640;
+	case APP_RES_720_480: return 720;
+	case APP_RES_720_576: return 720;
+	default: assert_notreached("invalid res");
+	}
+}
+size_t app_res_h(enum app_res res) {
+	switch (res) {
+	case APP_RES_640_480: return 480/2;
+	case APP_RES_720_480: return 480/2;
+	case APP_RES_720_576: return 576/2;
+	default: assert_notreached("invalid res");
+	}
+}
+size_t app_res_fb_size(enum app_res res) {
+	return app_res_w(res)*app_res_h(res);
+}
+size_t app_res_sm_size(enum app_res) {
+	/* See app_write_frame() */
+	return 19;
+}
+
+#define APP_FRAME_SIZE_MSB_MUL 0x2000 /* must be less than 0x4000 */
+#define APP_FRAME_MAX_HEADER   16
+
+struct app_state {
+	enum app_res res;
+};
+
+#define APP_COMMON_FLAGS                                                                             \
+	NUT_FRAMEFLAG_KEY       | /* Because they're full framebuffers, all frames are keyframes. */ \
+	NUT_FRAMEFLAG_SIZE_MSB  | /* 640*480/2 > 16384 (the max val of data_size_lsb). */            \
+	NUT_FRAMEFLAG_CODED_PTS | /* framerate is unknown, each frame must have a timestamp. */      \
+	NUT_FRAMEFLAG_CHECKSUM    /* framerate is unknown, guard against exceeding max_pts_distance. */
+
+
+bool app_write_intro(int fd, struct app_state *state, uint64_t time_ns) {
+	struct buf out = {0};
+	struct buf pkt = {0};
+	unsigned int frame_code = 0;
+#define INC_FRAME_CODE() do { frame_code++; if (frame_code == 'N') frame_code++; } while(0)
+
+	append(&out, nut_file_id_string, sizeof(nut_file_id_string));
+
+	/* Changing resolution mid-stream requires per-fame side-data,
+	 * which requires FLAG_SM_DATA, which requires NUT v4.
+	 *
+	 * Unfortunately, libnut / nututils does not support v4, so
+	 * there goes a good debugging tool.
+	 */
+
+#define BOGUS(n) n
+	/* main_header ********************************************************/
+	pkt.len = 0;
+
+	/* head *******************************************/
+	nut_append_vu(&pkt, 4);                              /*! version */
+	nut_append_vu(&pkt, 0);                              /*! minor_version */
+	nut_append_vu(&pkt, 1);                              /*! stream_count */
+	nut_append_vu(&pkt,                                  /*! max_distance */
+	              APP_FRAME_MAX_HEADER +
+	              app_res_sm_size(APP_RES_MAX) +
+	              app_res_fb_size(APP_RES_MAX));
+
+	/* time bases *************************************/
+	nut_append_vu(&pkt, 1);                              /*! time_base_count */
+
+	/* time_base[0] = 1ns */
+	nut_append_vu(&pkt, 1);                              /*! numerator */
+	nut_append_vu(&pkt, 1000000000ULL);                  /*! denominator */
+
+	/* frame codes ************************************/
+	/* "A muxer SHOULD mark [frame codes] 0x00 and 0xFF as invalid" */
+
+	/* frame_code=0 (invalid) */
+	nut_append_vu(&pkt, NUT_FRAMEFLAG_INVALID);          /*! flags */
+	nut_append_vu(&pkt, 2);                              /*! field_count */
+	nut_append_vu(&pkt, BOGUS(0));                       /*! 0: fields.pts */
+	nut_append_vu(&pkt, 1);                              /*! 1: fields.size_msb_nul */
+	INC_FRAME_CODE();
+
+	for (enum app_res res = APP_RES_MIN; res <= APP_RES_MAX; res++) {
+		/* frame_code=(res*2)+1 (at `res`) */
+		nut_append_vu(&pkt, APP_COMMON_FLAGS);       /*! flags */
+		nut_append_vu(&pkt, 6);                      /*! field_count */
+		nut_append_vu(&pkt, BOGUS(0));               /*! 0: fields.pts */
+		nut_append_vu(&pkt, APP_FRAME_SIZE_MSB_MUL); /*! 1: fields.size_msb_nul */
+		nut_append_vu(&pkt, 0);                      /*! 2: fields.stream */
+		nut_append_vu(&pkt,                          /*! 3: fields.size_lsb */
+		              app_res_fb_size(res) % APP_FRAME_SIZE_MSB_MUL);
+		nut_append_vu(&pkt, 0);                      /*! 4: fields.reserved */
+		nut_append_vu(&pkt, 1);                      /*! 5: fields.count */
+		INC_FRAME_CODE();
+
+		/* frame_code=(res*2)+2 (change to `res`) */
+		nut_append_vu(&pkt, APP_COMMON_FLAGS|        /*! flags */
+		              NUT_FRAMEFLAG_SM_DATA);
+		nut_append_vu(&pkt, 6);                      /*! field_count */
+		nut_append_vu(&pkt, BOGUS(0));               /*! 0: fields.pts */
+		nut_append_vu(&pkt, APP_FRAME_SIZE_MSB_MUL); /*! 1: fields.size_msb_nul */
+		nut_append_vu(&pkt, 0);                      /*! 2: fields.stream */
+		nut_append_vu(&pkt,                          /*! 3: fields.size_lsb */
+		              (app_res_sm_size(res) + app_res_fb_size(res)) % APP_FRAME_SIZE_MSB_MUL);
+		nut_append_vu(&pkt, 0);                      /*! 4: fields.reserved */
+		nut_append_vu(&pkt, 1);                      /*! 5: fields.count */
+		INC_FRAME_CODE();
+	}
+
+	/* frame_code=N-255 (invalid) */
+	nut_append_vu(&pkt, NUT_FRAMEFLAG_INVALID);          /*! flags */
+	nut_append_vu(&pkt, 2);                              /*! field_count */
+	nut_append_vu(&pkt, BOGUS(0));                       /*! 0: fields.pts */
+	nut_append_vu(&pkt,                                  /*! 1: fields.size_msb_nul */
+	              256-frame_code-(frame_code < 'N' ? 1 : 0));
+
+	/* tail *******************************************/
+	nut_append_vu(&pkt, 0);                              /*! header_count_minus_1 */
+	nut_append_vu(&pkt, NUT_MAINFLAG_PIPE_MODE);         /*! main_flags */
+
+	nut_append_packet(&out, nut_startcode_main, pkt.dat, pkt.len);
+
+	/* stream_header ******************************************************/
+	pkt.len = 0;
+
+	nut_append_vu(&pkt, 0);                              /*! stream_id */
+	nut_append_vu(&pkt, NUT_STREAMCLASS_VIDEO);          /*! stream_class */
+	nut_append_vb(&pkt, "BGR\x08", 4);                   /*! fourcc */
+	nut_append_vu(&pkt, 0);                              /*! time_base_id */
+	nut_append_vu(&pkt, BOGUS(0));                       /*! msb_pts_shift (only relevant if FRAMEFLAG_CODED_PTS) */
+	nut_append_vu(&pkt, BOGUS(0));                       /*! max_pts_distance (all frames have a checksum)  */
+	nut_append_vu(&pkt, 0);                              /*! decode_delay */
+	nut_append_vu(&pkt, 0);                              /*! stream_flags */
+	nut_append_vb(&pkt, NULL, 0);                        /*! codec_specific_data */
+	state->res = APP_RES_640_480;
+	nut_append_vu(&pkt, app_res_w(state->res));          /*! width */
+	nut_append_vu(&pkt, app_res_h(state->res));          /*! height */
+	nut_append_vu(&pkt, 1);                              /*! sample_width */
+	nut_append_vu(&pkt, 2);                              /*! sample_height */
+	nut_append_vu(&pkt, NUT_COLORSPACE_UNKNOWN);         /*! colorspace_type */
+
+	nut_append_packet(&out, nut_startcode_stream, pkt.dat, pkt.len);
+
+	/* syncpoint **********************************************************/
+	pkt.len = 0;
+
+	nut_append_vu(&pkt, time_ns);                        /*! global_key_pts */
+	nut_append_vu(&pkt, 0);                              /*! back_ptr_div16 */
+
+	nut_append_packet(&out, nut_startcode_syncpoint, pkt.dat, pkt.len);
+
+	/* flush **************************************************************/
+#undef BOGUS
+
+	bool ret = xwrite(fd, out.dat, out.len);
+	free(out.dat);
+	free(pkt.dat);
+	return ret;
+}
+
+bool app_write_frame(int fd, struct app_state *state, uint64_t time_ns, enum app_res res, void *framebuffer) {
+	assert(framebuffer);
+
+	struct buf out = {0};
+	struct buf pkt = {0};
+
+	/* frame header ( 1+10+1+4 = 16 bytes) */
+	uint64_t frame_beg = out.len;
+	nut_append_u8(&out, 1+(2*res)+(res != state->res ? 1 : 0));                /*! frame_code    (1 byte) */
+	nut_append_vu(&out, time_ns);                                              /*! coded_pts     (<=10 bytes) */
+	nut_append_vu(&out,                                                        /*! data_size_msb (1 byte) */
+	              (app_res_fb_size(res) + (res != state->res ? app_res_sm_size(res) : 0)) / APP_FRAME_SIZE_MSB_MUL);
+	nut_append_u32(&out, nut_crc32(&out.dat[frame_beg], out.len - frame_beg)); /*! checksum      (4 bytes) */
+	assert(out.len <= APP_FRAME_MAX_HEADER);
+
+	/* side/meta data ( 1+(6+2)+(7+2)+1 = 19 bytes) */
+	if (res != state->res) {
+		uint64_t data_beg = out.len;
+		nut_append_vu(&out, 2);              /*! side_data_count    (1 byte) */
+		nut_append_vb_str(&out, "Width");    /*! side_data[0].name  (1+5 bytes) */
+		nut_append_vu(&out, app_res_w(res)); /*! side_data[0].value (2 bytes) */
+		nut_append_vb_str(&out, "Height");   /*! side_data[1].name  (1+6 bytes) */
+		nut_append_vu(&out, app_res_h(res)); /*! side_data[1].value (2 bytes) */
+		nut_append_vu(&out, 0);              /*! meta_data_count    (1 byte) */
+		assert(out.len - data_beg == app_res_sm_size(res));
+	}
+
+	/* flush */
+	bool err = xwrite(fd, out.dat, out.len);
+	free(out.dat);
+	free(pkt.dat);
+	if (err)
+		return true;
+
+	/* frame data */
+	if (xwrite(fd, framebuffer, app_res_fb_size(res)))
+		return true;
+
+	/* return */
+	state->res = res;
+	return false;
+}
+
+/* Demo application ***********************************************************/
+
+int main() {
+	struct app_state state;
+	if (app_write_intro(1, &state, 0))
+		return 1;
+
+	uint8_t framebuffer[app_res_fb_size(APP_RES_MAX)];
+
+#define SCALE 10
+	for (int i = 0; i < 10; i++) {
+		memset(framebuffer, 0, sizeof(framebuffer));
+		for (int y = 0; y < 4*SCALE; y++)
+			for (int x = 0; x < 8*SCALE; x++)
+				framebuffer[(y*app_res_w(i%3))+x] = font[i][((y/SCALE)*8)+(x/SCALE)] == ' ' ? 0b11000000 : 0b00000011;
+		if (app_write_frame(1, &state, ((uint64_t)i)*1000000000ULL, i%3, framebuffer))
+			return 1;
+	}
+	return 0;
+}