diff options
Diffstat (limited to 'third_party/aom/common')
27 files changed, 4151 insertions, 0 deletions
diff --git a/third_party/aom/common/args.c b/third_party/aom/common/args.c new file mode 100644 index 000000000..7131e24de --- /dev/null +++ b/third_party/aom/common/args.c @@ -0,0 +1,297 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "common/args.h" + +#include <stdlib.h> +#include <string.h> +#include <limits.h> + +#include "aom/aom_integer.h" +#include "aom_ports/msvc.h" + +#if defined(__GNUC__) && __GNUC__ +extern void die(const char *fmt, ...) __attribute__((noreturn)); +#else +extern void die(const char *fmt, ...); +#endif + +struct arg arg_init(char **argv) { + struct arg a; + + a.argv = argv; + a.argv_step = 1; + a.name = NULL; + a.val = NULL; + a.def = NULL; + return a; +} + +char *ignore_front_spaces(const char *str) { + while (str[0] == ' ' || str[0] == '\t') ++str; + return (char *)str; +} + +void ignore_end_spaces(char *str) { + char *end = str + strlen(str); + while (end > str && (end[0] == ' ' || end[0] == '\t' || end[0] == '\n' || + end[0] == '\r' || end[0] == '\0')) + --end; + if (end >= str) end[1] = '\0'; +} + +int arg_cfg(int *argc, char ***argv, const char *file) { + char **argv_local = (char **)*argv; + char **argv_org = (char **)*argv; + char line[1024 * 10]; + FILE *f = fopen(file, "r"); + if (!f) return 1; + + while (fgets(line, sizeof(line) - 1, f)) { + char *actual_line = ignore_front_spaces(line); + char *left, *right, *comment; + size_t length = strlen(actual_line); + + if (length == 0 || actual_line[0] == '#') continue; + right = strchr(actual_line, ':'); + if (right == NULL) continue; + right[0] = '\0'; + + left = ignore_front_spaces(actual_line); + right = ignore_front_spaces(right + 1); + + comment = strchr(right, '#'); + if (comment != NULL) comment[0] = '\0'; + + ignore_end_spaces(left); + ignore_end_spaces(right); + + char **new_args = argv_dup(*argc, (const char **)argv_local); + char *new_line = (char *)malloc(sizeof(*new_line) * 128); + + if (argv_local != argv_org) free(argv_local); + + if (!strcmp(right, "ON")) + snprintf(new_line, sizeof(*new_line) * 128, "--%s", left); + else + snprintf(new_line, sizeof(*new_line) * 128, "--%s=%s", left, right); + + new_args[(*argc) - 1] = new_args[(*argc) - 2]; + new_args[(*argc) - 2] = new_line; + argv_local = new_args; + *argv = new_args; + (*argc)++; + } + fclose(f); + return 0; +} + +int arg_match(struct arg *arg_, const struct arg_def *def, char **argv) { + struct arg arg; + + if (!argv[0] || argv[0][0] != '-') return 0; + + arg = arg_init(argv); + + if (def->short_name && strlen(arg.argv[0]) == strlen(def->short_name) + 1 && + !strcmp(arg.argv[0] + 1, def->short_name)) { + arg.name = arg.argv[0] + 1; + arg.val = def->has_val ? arg.argv[1] : NULL; + arg.argv_step = def->has_val ? 2 : 1; + } else if (def->long_name) { + const size_t name_len = strlen(def->long_name); + + if (strlen(arg.argv[0]) >= name_len + 2 && arg.argv[0][1] == '-' && + !strncmp(arg.argv[0] + 2, def->long_name, name_len) && + (arg.argv[0][name_len + 2] == '=' || + arg.argv[0][name_len + 2] == '\0')) { + arg.name = arg.argv[0] + 2; + arg.val = arg.name[name_len] == '=' ? arg.name + name_len + 1 : NULL; + arg.argv_step = 1; + } + } + + if (arg.name && !arg.val && def->has_val) + die("Error: option %s requires argument.\n", arg.name); + + if (arg.name && arg.val && !def->has_val) + die("Error: option %s requires no argument.\n", arg.name); + + if (arg.name && (arg.val || !def->has_val)) { + arg.def = def; + *arg_ = arg; + return 1; + } + + return 0; +} + +const char *arg_next(struct arg *arg) { + if (arg->argv[0]) arg->argv += arg->argv_step; + + return *arg->argv; +} + +char **argv_dup(int argc, const char **argv) { + char **new_argv = malloc((argc + 1) * sizeof(*argv)); + + memcpy(new_argv, argv, argc * sizeof(*argv)); + new_argv[argc] = NULL; + return new_argv; +} + +void arg_show_usage(FILE *fp, const struct arg_def *const *defs) { + char option_text[40] = { 0 }; + + for (; *defs; defs++) { + const struct arg_def *def = *defs; + char *short_val = def->has_val ? " <arg>" : ""; + char *long_val = def->has_val ? "=<arg>" : ""; + + if (def->short_name && def->long_name) { + char *comma = def->has_val ? "," : ", "; + + snprintf(option_text, 37, "-%s%s%s --%s%6s", def->short_name, short_val, + comma, def->long_name, long_val); + } else if (def->short_name) + snprintf(option_text, 37, "-%s%s", def->short_name, short_val); + else if (def->long_name) + snprintf(option_text, 37, " --%s%s", def->long_name, long_val); + + fprintf(fp, " %-37s\t%s\n", option_text, def->desc); + + if (def->enums) { + const struct arg_enum_list *listptr; + + fprintf(fp, " %-37s\t ", ""); + + for (listptr = def->enums; listptr->name; listptr++) + fprintf(fp, "%s%s", listptr->name, listptr[1].name ? ", " : "\n"); + } + } +} + +unsigned int arg_parse_uint(const struct arg *arg) { + char *endptr; + const unsigned long rawval = strtoul(arg->val, &endptr, 10); // NOLINT + + if (arg->val[0] != '\0' && endptr[0] == '\0') { + if (rawval <= UINT_MAX) return (unsigned int)rawval; + + die("Option %s: Value %lu out of range for unsigned int\n", arg->name, + rawval); + } + + die("Option %s: Invalid character '%c'\n", arg->name, *endptr); + return 0; +} + +int arg_parse_int(const struct arg *arg) { + char *endptr; + const long rawval = strtol(arg->val, &endptr, 10); // NOLINT + + if (arg->val[0] != '\0' && endptr[0] == '\0') { + if (rawval >= INT_MIN && rawval <= INT_MAX) return (int)rawval; + + die("Option %s: Value %ld out of range for signed int\n", arg->name, + rawval); + } + + die("Option %s: Invalid character '%c'\n", arg->name, *endptr); + return 0; +} + +struct aom_rational { + int num; /**< fraction numerator */ + int den; /**< fraction denominator */ +}; +struct aom_rational arg_parse_rational(const struct arg *arg) { + long int rawval; + char *endptr; + struct aom_rational rat; + + /* parse numerator */ + rawval = strtol(arg->val, &endptr, 10); + + if (arg->val[0] != '\0' && endptr[0] == '/') { + if (rawval >= INT_MIN && rawval <= INT_MAX) + rat.num = (int)rawval; + else + die("Option %s: Value %ld out of range for signed int\n", arg->name, + rawval); + } else + die("Option %s: Expected / at '%c'\n", arg->name, *endptr); + + /* parse denominator */ + rawval = strtol(endptr + 1, &endptr, 10); + + if (arg->val[0] != '\0' && endptr[0] == '\0') { + if (rawval >= INT_MIN && rawval <= INT_MAX) + rat.den = (int)rawval; + else + die("Option %s: Value %ld out of range for signed int\n", arg->name, + rawval); + } else + die("Option %s: Invalid character '%c'\n", arg->name, *endptr); + + return rat; +} + +int arg_parse_enum(const struct arg *arg) { + const struct arg_enum_list *listptr; + long int rawval; + char *endptr; + + /* First see if the value can be parsed as a raw value */ + rawval = strtol(arg->val, &endptr, 10); + if (arg->val[0] != '\0' && endptr[0] == '\0') { + /* Got a raw value, make sure it's valid */ + for (listptr = arg->def->enums; listptr->name; listptr++) + if (listptr->val == rawval) return (int)rawval; + } + + /* Next see if it can be parsed as a string */ + for (listptr = arg->def->enums; listptr->name; listptr++) + if (!strcmp(arg->val, listptr->name)) return listptr->val; + + die("Option %s: Invalid value '%s'\n", arg->name, arg->val); + return 0; +} + +int arg_parse_enum_or_int(const struct arg *arg) { + if (arg->def->enums) return arg_parse_enum(arg); + return arg_parse_int(arg); +} + +// parse a comma separated list of at most n integers +// return the number of elements in the list +int arg_parse_list(const struct arg *arg, int *list, int n) { + const char *ptr = arg->val; + char *endptr; + int i = 0; + + while (ptr[0] != '\0') { + int32_t rawval = (int32_t)strtol(ptr, &endptr, 10); + if (rawval < INT_MIN || rawval > INT_MAX) { + die("Option %s: Value %ld out of range for signed int\n", arg->name, + rawval); + } else if (i >= n) { + die("Option %s: List has more than %d entries\n", arg->name, n); + } else if (*endptr == ',') { + endptr++; + } else if (*endptr != '\0') { + die("Option %s: Bad list separator '%c'\n", arg->name, *endptr); + } + list[i++] = (int)rawval; + ptr = endptr; + } + return i; +} diff --git a/third_party/aom/common/args.h b/third_party/aom/common/args.h new file mode 100644 index 000000000..d12973666 --- /dev/null +++ b/third_party/aom/common/args.h @@ -0,0 +1,68 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#ifndef ARGS_H_ +#define ARGS_H_ +#include <stdio.h> + +#ifdef __cplusplus +extern "C" { +#endif + +struct arg { + char **argv; + const char *name; + const char *val; + unsigned int argv_step; + const struct arg_def *def; +}; + +struct arg_enum_list { + const char *name; + int val; +}; +#define ARG_ENUM_LIST_END \ + { 0 } + +typedef struct arg_def { + const char *short_name; + const char *long_name; + int has_val; + const char *desc; + const struct arg_enum_list *enums; +} arg_def_t; +#define ARG_DEF(s, l, v, d) \ + { s, l, v, d, NULL } +#define ARG_DEF_ENUM(s, l, v, d, e) \ + { s, l, v, d, e } +#define ARG_DEF_LIST_END \ + { 0 } + +struct arg arg_init(char **argv); +int arg_match(struct arg *arg_, const struct arg_def *def, char **argv); +char *ignore_front_spaces(const char *str); +void ignore_end_spaces(char *str); +int arg_cfg(int *argc, char ***argv, const char *file); +const char *arg_next(struct arg *arg); +void arg_show_usage(FILE *fp, const struct arg_def *const *defs); +char **argv_dup(int argc, const char **argv); + +unsigned int arg_parse_uint(const struct arg *arg); +int arg_parse_int(const struct arg *arg); +struct aom_rational arg_parse_rational(const struct arg *arg); +int arg_parse_enum(const struct arg *arg); +int arg_parse_enum_or_int(const struct arg *arg); +int arg_parse_list(const struct arg *arg, int *list, int n); +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // ARGS_H_ diff --git a/third_party/aom/common/ivfdec.c b/third_party/aom/common/ivfdec.c new file mode 100644 index 000000000..80d73b04c --- /dev/null +++ b/third_party/aom/common/ivfdec.c @@ -0,0 +1,110 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "common/ivfdec.h" + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include "aom_ports/mem_ops.h" +#include "aom_ports/sanitizer.h" + +static const char *IVF_SIGNATURE = "DKIF"; + +static void fix_framerate(int *num, int *den) { + if (*den <= 0 || *den >= 1000000000 || *num <= 0 || *num >= 1000) { + // framerate seems to be invalid, just default to 30fps. + *num = 30; + *den = 1; + } +} + +int file_is_ivf(struct AvxInputContext *input_ctx) { + char raw_hdr[32]; + int is_ivf = 0; + + if (fread(raw_hdr, 1, 32, input_ctx->file) == 32) { + if (memcmp(IVF_SIGNATURE, raw_hdr, 4) == 0) { + is_ivf = 1; + + if (mem_get_le16(raw_hdr + 4) != 0) { + fprintf(stderr, + "Error: Unrecognized IVF version! This file may not" + " decode properly."); + } + + input_ctx->fourcc = mem_get_le32(raw_hdr + 8); + input_ctx->width = mem_get_le16(raw_hdr + 12); + input_ctx->height = mem_get_le16(raw_hdr + 14); + input_ctx->framerate.numerator = mem_get_le32(raw_hdr + 16); + input_ctx->framerate.denominator = mem_get_le32(raw_hdr + 20); + fix_framerate(&input_ctx->framerate.numerator, + &input_ctx->framerate.denominator); + } + } + + if (!is_ivf) { + rewind(input_ctx->file); + input_ctx->detect.buf_read = 0; + } else { + input_ctx->detect.position = 4; + } + return is_ivf; +} + +int ivf_read_frame(FILE *infile, uint8_t **buffer, size_t *bytes_read, + size_t *buffer_size, aom_codec_pts_t *pts) { + char raw_header[IVF_FRAME_HDR_SZ] = { 0 }; + size_t frame_size = 0; + + if (fread(raw_header, IVF_FRAME_HDR_SZ, 1, infile) != 1) { + if (!feof(infile)) warn("Failed to read frame size"); + } else { + frame_size = mem_get_le32(raw_header); + + if (frame_size > 256 * 1024 * 1024) { + warn("Read invalid frame size (%u)", (unsigned int)frame_size); + frame_size = 0; + } + + if (frame_size > *buffer_size) { + uint8_t *new_buffer = (uint8_t *)realloc(*buffer, 2 * frame_size); + + if (new_buffer) { + *buffer = new_buffer; + *buffer_size = 2 * frame_size; + } else { + warn("Failed to allocate compressed data buffer"); + frame_size = 0; + } + } + + if (pts) { + *pts = mem_get_le32(&raw_header[4]); + *pts += ((aom_codec_pts_t)mem_get_le32(&raw_header[8]) << 32); + } + } + + if (!feof(infile)) { + ASAN_UNPOISON_MEMORY_REGION(*buffer, *buffer_size); + if (fread(*buffer, 1, frame_size, infile) != frame_size) { + warn("Failed to read full frame"); + return 1; + } + + ASAN_POISON_MEMORY_REGION(*buffer + frame_size, *buffer_size - frame_size); + *bytes_read = frame_size; + return 0; + } + + return 1; +} diff --git a/third_party/aom/common/ivfdec.h b/third_party/aom/common/ivfdec.h new file mode 100644 index 000000000..9013dea64 --- /dev/null +++ b/third_party/aom/common/ivfdec.h @@ -0,0 +1,30 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef IVFDEC_H_ +#define IVFDEC_H_ + +#include "common/tools_common.h" + +#ifdef __cplusplus +extern "C" { +#endif + +int file_is_ivf(struct AvxInputContext *input); + +typedef int64_t aom_codec_pts_t; +int ivf_read_frame(FILE *infile, uint8_t **buffer, size_t *bytes_read, + size_t *buffer_size, aom_codec_pts_t *pts); + +#ifdef __cplusplus +} /* extern "C" */ +#endif + +#endif // IVFDEC_H_ diff --git a/third_party/aom/common/ivfenc.c b/third_party/aom/common/ivfenc.c new file mode 100644 index 000000000..64715f4d7 --- /dev/null +++ b/third_party/aom/common/ivfenc.c @@ -0,0 +1,52 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "common/ivfenc.h" + +#include "aom/aom_encoder.h" +#include "aom_ports/mem_ops.h" + +void ivf_write_file_header(FILE *outfile, const struct aom_codec_enc_cfg *cfg, + unsigned int fourcc, int frame_cnt) { + char header[32]; + + header[0] = 'D'; + header[1] = 'K'; + header[2] = 'I'; + header[3] = 'F'; + mem_put_le16(header + 4, 0); // version + mem_put_le16(header + 6, 32); // header size + mem_put_le32(header + 8, fourcc); // fourcc + mem_put_le16(header + 12, cfg->g_w); // width + mem_put_le16(header + 14, cfg->g_h); // height + mem_put_le32(header + 16, cfg->g_timebase.den); // rate + mem_put_le32(header + 20, cfg->g_timebase.num); // scale + mem_put_le32(header + 24, frame_cnt); // length + mem_put_le32(header + 28, 0); // unused + + fwrite(header, 1, 32, outfile); +} + +void ivf_write_frame_header(FILE *outfile, int64_t pts, size_t frame_size) { + char header[12]; + + mem_put_le32(header, (int)frame_size); + mem_put_le32(header + 4, (int)(pts & 0xFFFFFFFF)); + mem_put_le32(header + 8, (int)(pts >> 32)); + fwrite(header, 1, 12, outfile); +} + +void ivf_write_frame_size(FILE *outfile, size_t frame_size) { + char header[4]; + + mem_put_le32(header, (int)frame_size); + fwrite(header, 1, 4, outfile); +} diff --git a/third_party/aom/common/ivfenc.h b/third_party/aom/common/ivfenc.h new file mode 100644 index 000000000..f0cab8178 --- /dev/null +++ b/third_party/aom/common/ivfenc.h @@ -0,0 +1,34 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef IVFENC_H_ +#define IVFENC_H_ + +#include "common/tools_common.h" + +struct aom_codec_enc_cfg; +struct aom_codec_cx_pkt; + +#ifdef __cplusplus +extern "C" { +#endif + +void ivf_write_file_header(FILE *outfile, const struct aom_codec_enc_cfg *cfg, + uint32_t fourcc, int frame_cnt); + +void ivf_write_frame_header(FILE *outfile, int64_t pts, size_t frame_size); + +void ivf_write_frame_size(FILE *outfile, size_t frame_size); + +#ifdef __cplusplus +} /* extern "C" */ +#endif + +#endif // IVFENC_H_ diff --git a/third_party/aom/common/md5_utils.c b/third_party/aom/common/md5_utils.c new file mode 100644 index 000000000..b69e1cc72 --- /dev/null +++ b/third_party/aom/common/md5_utils.c @@ -0,0 +1,249 @@ +/* + * This code implements the MD5 message-digest algorithm. + * The algorithm is due to Ron Rivest. This code was + * written by Colin Plumb in 1993, no copyright is claimed. + * This code is in the public domain; do with it what you wish. + * + * Equivalent code is available from RSA Data Security, Inc. + * This code has been tested against that, and is equivalent, + * except that you don't need to include two pages of legalese + * with every copy. + * + * To compute the message digest of a chunk of bytes, declare an + * MD5Context structure, pass it to MD5Init, call MD5Update as + * needed on buffers full of bytes, and then call MD5Final, which + * will fill a supplied 16-byte array with the digest. + * + * Changed so as no longer to depend on Colin Plumb's `usual.h' header + * definitions + * - Ian Jackson <ian@chiark.greenend.org.uk>. + * Still in the public domain. + */ + +#include <string.h> /* for memcpy() */ + +#include "common/md5_utils.h" + +static void byteSwap(UWORD32 *buf, unsigned words) { + md5byte *p; + + /* Only swap bytes for big endian machines */ + int i = 1; + + if (*(char *)&i == 1) return; + + p = (md5byte *)buf; + + do { + *buf++ = (UWORD32)((unsigned)p[3] << 8 | p[2]) << 16 | + ((unsigned)p[1] << 8 | p[0]); + p += 4; + } while (--words); +} + +/* + * Start MD5 accumulation. Set bit count to 0 and buffer to mysterious + * initialization constants. + */ +void MD5Init(struct MD5Context *ctx) { + ctx->buf[0] = 0x67452301; + ctx->buf[1] = 0xefcdab89; + ctx->buf[2] = 0x98badcfe; + ctx->buf[3] = 0x10325476; + + ctx->bytes[0] = 0; + ctx->bytes[1] = 0; +} + +/* + * Update context to reflect the concatenation of another buffer full + * of bytes. + */ +void MD5Update(struct MD5Context *ctx, md5byte const *buf, unsigned len) { + UWORD32 t; + + /* Update byte count */ + + t = ctx->bytes[0]; + + if ((ctx->bytes[0] = t + len) < t) + ctx->bytes[1]++; /* Carry from low to high */ + + t = 64 - (t & 0x3f); /* Space available in ctx->in (at least 1) */ + + if (t > len) { + memcpy((md5byte *)ctx->in + 64 - t, buf, len); + return; + } + + /* First chunk is an odd size */ + memcpy((md5byte *)ctx->in + 64 - t, buf, t); + byteSwap(ctx->in, 16); + MD5Transform(ctx->buf, ctx->in); + buf += t; + len -= t; + + /* Process data in 64-byte chunks */ + while (len >= 64) { + memcpy(ctx->in, buf, 64); + byteSwap(ctx->in, 16); + MD5Transform(ctx->buf, ctx->in); + buf += 64; + len -= 64; + } + + /* Handle any remaining bytes of data. */ + memcpy(ctx->in, buf, len); +} + +/* + * Final wrapup - pad to 64-byte boundary with the bit pattern + * 1 0* (64-bit count of bits processed, MSB-first) + */ +void MD5Final(md5byte digest[16], struct MD5Context *ctx) { + int count = ctx->bytes[0] & 0x3f; /* Number of bytes in ctx->in */ + md5byte *p = (md5byte *)ctx->in + count; + + /* Set the first char of padding to 0x80. There is always room. */ + *p++ = 0x80; + + /* Bytes of padding needed to make 56 bytes (-8..55) */ + count = 56 - 1 - count; + + if (count < 0) { /* Padding forces an extra block */ + memset(p, 0, count + 8); + byteSwap(ctx->in, 16); + MD5Transform(ctx->buf, ctx->in); + p = (md5byte *)ctx->in; + count = 56; + } + + memset(p, 0, count); + byteSwap(ctx->in, 14); + + /* Append length in bits and transform */ + ctx->in[14] = ctx->bytes[0] << 3; + ctx->in[15] = ctx->bytes[1] << 3 | ctx->bytes[0] >> 29; + MD5Transform(ctx->buf, ctx->in); + + byteSwap(ctx->buf, 4); + memcpy(digest, ctx->buf, 16); + memset(ctx, 0, sizeof(*ctx)); /* In case it's sensitive */ +} + +#ifndef ASM_MD5 + +/* The four core functions - F1 is optimized somewhat */ + +/* #define F1(x, y, z) (x & y | ~x & z) */ +#define F1(x, y, z) (z ^ (x & (y ^ z))) +#define F2(x, y, z) F1(z, x, y) +#define F3(x, y, z) (x ^ y ^ z) +#define F4(x, y, z) (y ^ (x | ~z)) + +/* This is the central step in the MD5 algorithm. */ +#define MD5STEP(f, w, x, y, z, in, s) \ + (w += f(x, y, z) + in, w = (w << s | w >> (32 - s)) + x) + +#if defined(__clang__) && defined(__has_attribute) +#if __has_attribute(no_sanitize) +#define AOM_NO_UNSIGNED_OVERFLOW_CHECK \ + __attribute__((no_sanitize("unsigned-integer-overflow"))) +#endif +#endif + +#ifndef AOM_NO_UNSIGNED_OVERFLOW_CHECK +#define AOM_NO_UNSIGNED_OVERFLOW_CHECK +#endif + +/* + * The core of the MD5 algorithm, this alters an existing MD5 hash to + * reflect the addition of 16 longwords of new data. MD5Update blocks + * the data and converts bytes into longwords for this routine. + */ +AOM_NO_UNSIGNED_OVERFLOW_CHECK void MD5Transform(UWORD32 buf[4], + UWORD32 const in[16]) { + register UWORD32 a, b, c, d; + + a = buf[0]; + b = buf[1]; + c = buf[2]; + d = buf[3]; + + MD5STEP(F1, a, b, c, d, in[0] + 0xd76aa478, 7); + MD5STEP(F1, d, a, b, c, in[1] + 0xe8c7b756, 12); + MD5STEP(F1, c, d, a, b, in[2] + 0x242070db, 17); + MD5STEP(F1, b, c, d, a, in[3] + 0xc1bdceee, 22); + MD5STEP(F1, a, b, c, d, in[4] + 0xf57c0faf, 7); + MD5STEP(F1, d, a, b, c, in[5] + 0x4787c62a, 12); + MD5STEP(F1, c, d, a, b, in[6] + 0xa8304613, 17); + MD5STEP(F1, b, c, d, a, in[7] + 0xfd469501, 22); + MD5STEP(F1, a, b, c, d, in[8] + 0x698098d8, 7); + MD5STEP(F1, d, a, b, c, in[9] + 0x8b44f7af, 12); + MD5STEP(F1, c, d, a, b, in[10] + 0xffff5bb1, 17); + MD5STEP(F1, b, c, d, a, in[11] + 0x895cd7be, 22); + MD5STEP(F1, a, b, c, d, in[12] + 0x6b901122, 7); + MD5STEP(F1, d, a, b, c, in[13] + 0xfd987193, 12); + MD5STEP(F1, c, d, a, b, in[14] + 0xa679438e, 17); + MD5STEP(F1, b, c, d, a, in[15] + 0x49b40821, 22); + + MD5STEP(F2, a, b, c, d, in[1] + 0xf61e2562, 5); + MD5STEP(F2, d, a, b, c, in[6] + 0xc040b340, 9); + MD5STEP(F2, c, d, a, b, in[11] + 0x265e5a51, 14); + MD5STEP(F2, b, c, d, a, in[0] + 0xe9b6c7aa, 20); + MD5STEP(F2, a, b, c, d, in[5] + 0xd62f105d, 5); + MD5STEP(F2, d, a, b, c, in[10] + 0x02441453, 9); + MD5STEP(F2, c, d, a, b, in[15] + 0xd8a1e681, 14); + MD5STEP(F2, b, c, d, a, in[4] + 0xe7d3fbc8, 20); + MD5STEP(F2, a, b, c, d, in[9] + 0x21e1cde6, 5); + MD5STEP(F2, d, a, b, c, in[14] + 0xc33707d6, 9); + MD5STEP(F2, c, d, a, b, in[3] + 0xf4d50d87, 14); + MD5STEP(F2, b, c, d, a, in[8] + 0x455a14ed, 20); + MD5STEP(F2, a, b, c, d, in[13] + 0xa9e3e905, 5); + MD5STEP(F2, d, a, b, c, in[2] + 0xfcefa3f8, 9); + MD5STEP(F2, c, d, a, b, in[7] + 0x676f02d9, 14); + MD5STEP(F2, b, c, d, a, in[12] + 0x8d2a4c8a, 20); + + MD5STEP(F3, a, b, c, d, in[5] + 0xfffa3942, 4); + MD5STEP(F3, d, a, b, c, in[8] + 0x8771f681, 11); + MD5STEP(F3, c, d, a, b, in[11] + 0x6d9d6122, 16); + MD5STEP(F3, b, c, d, a, in[14] + 0xfde5380c, 23); + MD5STEP(F3, a, b, c, d, in[1] + 0xa4beea44, 4); + MD5STEP(F3, d, a, b, c, in[4] + 0x4bdecfa9, 11); + MD5STEP(F3, c, d, a, b, in[7] + 0xf6bb4b60, 16); + MD5STEP(F3, b, c, d, a, in[10] + 0xbebfbc70, 23); + MD5STEP(F3, a, b, c, d, in[13] + 0x289b7ec6, 4); + MD5STEP(F3, d, a, b, c, in[0] + 0xeaa127fa, 11); + MD5STEP(F3, c, d, a, b, in[3] + 0xd4ef3085, 16); + MD5STEP(F3, b, c, d, a, in[6] + 0x04881d05, 23); + MD5STEP(F3, a, b, c, d, in[9] + 0xd9d4d039, 4); + MD5STEP(F3, d, a, b, c, in[12] + 0xe6db99e5, 11); + MD5STEP(F3, c, d, a, b, in[15] + 0x1fa27cf8, 16); + MD5STEP(F3, b, c, d, a, in[2] + 0xc4ac5665, 23); + + MD5STEP(F4, a, b, c, d, in[0] + 0xf4292244, 6); + MD5STEP(F4, d, a, b, c, in[7] + 0x432aff97, 10); + MD5STEP(F4, c, d, a, b, in[14] + 0xab9423a7, 15); + MD5STEP(F4, b, c, d, a, in[5] + 0xfc93a039, 21); + MD5STEP(F4, a, b, c, d, in[12] + 0x655b59c3, 6); + MD5STEP(F4, d, a, b, c, in[3] + 0x8f0ccc92, 10); + MD5STEP(F4, c, d, a, b, in[10] + 0xffeff47d, 15); + MD5STEP(F4, b, c, d, a, in[1] + 0x85845dd1, 21); + MD5STEP(F4, a, b, c, d, in[8] + 0x6fa87e4f, 6); + MD5STEP(F4, d, a, b, c, in[15] + 0xfe2ce6e0, 10); + MD5STEP(F4, c, d, a, b, in[6] + 0xa3014314, 15); + MD5STEP(F4, b, c, d, a, in[13] + 0x4e0811a1, 21); + MD5STEP(F4, a, b, c, d, in[4] + 0xf7537e82, 6); + MD5STEP(F4, d, a, b, c, in[11] + 0xbd3af235, 10); + MD5STEP(F4, c, d, a, b, in[2] + 0x2ad7d2bb, 15); + MD5STEP(F4, b, c, d, a, in[9] + 0xeb86d391, 21); + + buf[0] += a; + buf[1] += b; + buf[2] += c; + buf[3] += d; +} + +#undef AOM_NO_UNSIGNED_OVERFLOW_CHECK + +#endif diff --git a/third_party/aom/common/md5_utils.h b/third_party/aom/common/md5_utils.h new file mode 100644 index 000000000..bd4991b3a --- /dev/null +++ b/third_party/aom/common/md5_utils.h @@ -0,0 +1,49 @@ +/* + * This is the header file for the MD5 message-digest algorithm. + * The algorithm is due to Ron Rivest. This code was + * written by Colin Plumb in 1993, no copyright is claimed. + * This code is in the public domain; do with it what you wish. + * + * Equivalent code is available from RSA Data Security, Inc. + * This code has been tested against that, and is equivalent, + * except that you don't need to include two pages of legalese + * with every copy. + * + * To compute the message digest of a chunk of bytes, declare an + * MD5Context structure, pass it to MD5Init, call MD5Update as + * needed on buffers full of bytes, and then call MD5Final, which + * will fill a supplied 16-byte array with the digest. + * + * Changed so as no longer to depend on Colin Plumb's `usual.h' + * header definitions + * - Ian Jackson <ian@chiark.greenend.org.uk>. + * Still in the public domain. + */ + +#ifndef MD5_UTILS_H_ +#define MD5_UTILS_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +#define md5byte unsigned char +#define UWORD32 unsigned int + +typedef struct MD5Context MD5Context; +struct MD5Context { + UWORD32 buf[4]; + UWORD32 bytes[2]; + UWORD32 in[16]; +}; + +void MD5Init(struct MD5Context *context); +void MD5Update(struct MD5Context *context, md5byte const *buf, unsigned len); +void MD5Final(unsigned char digest[16], struct MD5Context *context); +void MD5Transform(UWORD32 buf[4], UWORD32 const in[16]); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // MD5_UTILS_H_ diff --git a/third_party/aom/common/obudec.c b/third_party/aom/common/obudec.c new file mode 100644 index 000000000..cd88f1648 --- /dev/null +++ b/third_party/aom/common/obudec.c @@ -0,0 +1,448 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include "common/obudec.h" + +#include "aom_ports/mem_ops.h" +#include "av1/common/common.h" +#include "av1/decoder/obu.h" + +#define OBU_BUFFER_SIZE (500 * 1024) + +#define OBU_HEADER_SIZE 1 +#define OBU_EXTENSION_SIZE 1 +#define OBU_MAX_LENGTH_FIELD_SIZE 8 +#define OBU_DETECTION_SIZE \ + (OBU_HEADER_SIZE + OBU_EXTENSION_SIZE + 3 * OBU_MAX_LENGTH_FIELD_SIZE) + +// Reads unsigned LEB128 integer and returns 0 upon successful read and decode. +// Stores raw bytes in 'value_buffer', length of the number in 'value_length', +// and decoded value in 'value'. +static int obudec_read_leb128(FILE *f, uint8_t *value_buffer, + size_t *value_length, uint64_t *value) { + if (!f || !value_buffer || !value_length || !value) return -1; + size_t len; + for (len = 0; len < OBU_MAX_LENGTH_FIELD_SIZE; ++len) { + const size_t num_read = fread(&value_buffer[len], 1, 1, f); + if (num_read == 0) { + if (len == 0 && feof(f)) { + *value_length = 0; + return 0; + } + // Ran out of data before completing read of value. + return -1; + } + if ((value_buffer[len] >> 7) == 0) { + ++len; + *value_length = len; + break; + } + } + + return aom_uleb_decode(value_buffer, len, value, NULL); +} + +// Reads OBU header from 'f'. The 'buffer_capacity' passed in must be large +// enough to store an OBU header with extension (2 bytes). Raw OBU data is +// written to 'obu_data', parsed OBU header values are written to 'obu_header', +// and total bytes read from file are written to 'bytes_read'. Returns 0 for +// success, and non-zero on failure. When end of file is reached, the return +// value is 0 and the 'bytes_read' value is set to 0. +static int obudec_read_obu_header(FILE *f, size_t buffer_capacity, + int is_annexb, uint8_t *obu_data, + ObuHeader *obu_header, size_t *bytes_read) { + if (!f || buffer_capacity < (OBU_HEADER_SIZE + OBU_EXTENSION_SIZE) || + !obu_data || !obu_header || !bytes_read) { + return -1; + } + *bytes_read = fread(obu_data, 1, 1, f); + + if (feof(f) && *bytes_read == 0) { + return 0; + } else if (*bytes_read != 1) { + fprintf(stderr, "obudec: Failure reading OBU header.\n"); + return -1; + } + + const int has_extension = (obu_data[0] >> 2) & 0x1; + if (has_extension) { + if (fread(&obu_data[1], 1, 1, f) != 1) { + fprintf(stderr, "obudec: Failure reading OBU extension."); + return -1; + } + ++*bytes_read; + } + + size_t obu_bytes_parsed = 0; + const aom_codec_err_t parse_result = aom_read_obu_header( + obu_data, *bytes_read, &obu_bytes_parsed, obu_header, is_annexb); + if (parse_result != AOM_CODEC_OK || *bytes_read != obu_bytes_parsed) { + fprintf(stderr, "obudec: Error parsing OBU header.\n"); + return -1; + } + + return 0; +} + +// Reads OBU payload from 'f' and returns 0 for success when all payload bytes +// are read from the file. Payload data is written to 'obu_data', and actual +// bytes read added to 'bytes_read'. +static int obudec_read_obu_payload(FILE *f, size_t payload_length, + uint8_t *obu_data, size_t *bytes_read) { + if (!f || payload_length == 0 || !obu_data || !bytes_read) return -1; + + if (fread(obu_data, 1, payload_length, f) != payload_length) { + fprintf(stderr, "obudec: Failure reading OBU payload.\n"); + return -1; + } + + *bytes_read += payload_length; + return 0; +} + +static int obudec_read_obu_header_and_size(FILE *f, size_t buffer_capacity, + int is_annexb, uint8_t *buffer, + size_t *bytes_read, + size_t *payload_length, + ObuHeader *obu_header) { + const size_t kMinimumBufferSize = + (OBU_HEADER_SIZE + OBU_EXTENSION_SIZE + OBU_MAX_LENGTH_FIELD_SIZE); + if (!f || !buffer || !bytes_read || !payload_length || !obu_header || + buffer_capacity < kMinimumBufferSize) { + return -1; + } + + size_t leb128_length = 0; + uint64_t obu_size = 0; + if (is_annexb) { + if (obudec_read_leb128(f, &buffer[0], &leb128_length, &obu_size) != 0) { + fprintf(stderr, "obudec: Failure reading OBU size length.\n"); + return -1; + } else if (leb128_length == 0) { + *payload_length = 0; + return 0; + } + if (obu_size > UINT32_MAX) { + fprintf(stderr, "obudec: OBU payload length too large.\n"); + return -1; + } + } + + size_t header_size = 0; + if (obudec_read_obu_header(f, buffer_capacity - leb128_length, is_annexb, + buffer + leb128_length, obu_header, + &header_size) != 0) { + return -1; + } else if (header_size == 0) { + *payload_length = 0; + return 0; + } + + if (is_annexb) { + if (obu_size < header_size) { + fprintf(stderr, "obudec: OBU size is too small.\n"); + return -1; + } + *payload_length = (size_t)obu_size - header_size; + } else { + uint64_t u64_payload_length = 0; + if (obudec_read_leb128(f, &buffer[header_size], &leb128_length, + &u64_payload_length) != 0) { + fprintf(stderr, "obudec: Failure reading OBU payload length.\n"); + return -1; + } + if (u64_payload_length > UINT32_MAX) { + fprintf(stderr, "obudec: OBU payload length too large.\n"); + return -1; + } + + *payload_length = (size_t)u64_payload_length; + } + + *bytes_read = leb128_length + header_size; + return 0; +} + +static int obudec_read_one_obu(FILE *f, uint8_t **obu_buffer, + size_t obu_bytes_buffered, + size_t *obu_buffer_capacity, size_t *obu_length, + ObuHeader *obu_header, int is_annexb) { + size_t available_buffer_capacity = *obu_buffer_capacity - obu_bytes_buffered; + + if (!(*obu_buffer)) return -1; + + size_t bytes_read = 0; + size_t obu_payload_length = 0; + const int status = obudec_read_obu_header_and_size( + f, available_buffer_capacity, is_annexb, *obu_buffer + obu_bytes_buffered, + &bytes_read, &obu_payload_length, obu_header); + if (status < 0) return status; + + if (obu_payload_length > SIZE_MAX - bytes_read) return -1; + + if (obu_payload_length > 256 * 1024 * 1024) { + fprintf(stderr, "obudec: Read invalid OBU size (%u)\n", + (unsigned int)obu_payload_length); + *obu_length = bytes_read + obu_payload_length; + return -1; + } + + if (bytes_read + obu_payload_length > available_buffer_capacity) { + // TODO(tomfinegan): Add overflow check. + const size_t new_capacity = + obu_bytes_buffered + bytes_read + 2 * obu_payload_length; + +#if defined AOM_MAX_ALLOCABLE_MEMORY + if (new_capacity > AOM_MAX_ALLOCABLE_MEMORY) { + fprintf(stderr, "obudec: OBU size exceeds max alloc size.\n"); + return -1; + } +#endif + + uint8_t *new_buffer = (uint8_t *)realloc(*obu_buffer, new_capacity); + + if (new_buffer) { + *obu_buffer = new_buffer; + *obu_buffer_capacity = new_capacity; + } else { + fprintf(stderr, "obudec: Failed to allocate compressed data buffer\n"); + *obu_length = bytes_read + obu_payload_length; + return -1; + } + } + + if (obu_payload_length > 0 && + obudec_read_obu_payload(f, obu_payload_length, + *obu_buffer + obu_bytes_buffered + bytes_read, + &bytes_read) != 0) { + return -1; + } + + *obu_length = bytes_read; + return 0; +} + +int file_is_obu(struct ObuDecInputContext *obu_ctx) { + if (!obu_ctx || !obu_ctx->avx_ctx) return 0; + + struct AvxInputContext *avx_ctx = obu_ctx->avx_ctx; + uint8_t detect_buf[OBU_DETECTION_SIZE] = { 0 }; + const int is_annexb = obu_ctx->is_annexb; + FILE *f = avx_ctx->file; + size_t payload_length = 0; + ObuHeader obu_header; + memset(&obu_header, 0, sizeof(obu_header)); + size_t length_of_unit_size = 0; + size_t annexb_header_length = 0; + uint64_t unit_size = 0; + + if (is_annexb) { + // read the size of first temporal unit + if (obudec_read_leb128(f, &detect_buf[0], &length_of_unit_size, + &unit_size) != 0) { + fprintf(stderr, "obudec: Failure reading temporal unit header\n"); + return 0; + } + + // read the size of first frame unit + if (obudec_read_leb128(f, &detect_buf[length_of_unit_size], + &annexb_header_length, &unit_size) != 0) { + fprintf(stderr, "obudec: Failure reading frame unit header\n"); + return 0; + } + annexb_header_length += length_of_unit_size; + } + + size_t bytes_read = 0; + if (obudec_read_obu_header_and_size( + f, OBU_DETECTION_SIZE - annexb_header_length, is_annexb, + &detect_buf[annexb_header_length], &bytes_read, &payload_length, + &obu_header) != 0) { + fprintf(stderr, "obudec: Failure reading first OBU.\n"); + rewind(f); + return 0; + } + + if (is_annexb) { + bytes_read += annexb_header_length; + } + + if (obu_header.type != OBU_TEMPORAL_DELIMITER && + obu_header.type != OBU_SEQUENCE_HEADER) { + return 0; + } + + if (obu_header.has_size_field) { + if (obu_header.type == OBU_TEMPORAL_DELIMITER && payload_length != 0) { + fprintf( + stderr, + "obudec: Invalid OBU_TEMPORAL_DELIMITER payload length (non-zero)."); + rewind(f); + return 0; + } + } else if (!is_annexb) { + fprintf(stderr, "obudec: OBU size fields required, cannot decode input.\n"); + rewind(f); + return 0; + } + + // Appears that input is valid Section 5 AV1 stream. + obu_ctx->buffer = (uint8_t *)malloc(OBU_BUFFER_SIZE); + if (!obu_ctx->buffer) { + fprintf(stderr, "Out of memory.\n"); + rewind(f); + return 0; + } + obu_ctx->buffer_capacity = OBU_BUFFER_SIZE; + + memcpy(obu_ctx->buffer, &detect_buf[0], bytes_read); + obu_ctx->bytes_buffered = bytes_read; + // If the first OBU is a SEQUENCE_HEADER, then it will have a payload. + // We need to read this in so that our buffer only contains complete OBUs. + if (payload_length > 0) { + if (payload_length > (obu_ctx->buffer_capacity - bytes_read)) { + fprintf(stderr, "obudec: First OBU's payload is too large\n"); + rewind(f); + return 0; + } + + size_t payload_bytes = 0; + const int status = obudec_read_obu_payload( + f, payload_length, &obu_ctx->buffer[bytes_read], &payload_bytes); + if (status < 0) { + rewind(f); + return 0; + } + obu_ctx->bytes_buffered += payload_bytes; + } + return 1; +} + +int obudec_read_temporal_unit(struct ObuDecInputContext *obu_ctx, + uint8_t **buffer, size_t *bytes_read, + size_t *buffer_size) { + FILE *f = obu_ctx->avx_ctx->file; + if (!f) return -1; + + *buffer_size = 0; + *bytes_read = 0; + + if (feof(f)) { + return 1; + } + + size_t tu_size; + size_t obu_size = 0; + size_t length_of_temporal_unit_size = 0; + uint8_t tuheader[OBU_MAX_LENGTH_FIELD_SIZE] = { 0 }; + + if (obu_ctx->is_annexb) { + uint64_t size = 0; + + if (obu_ctx->bytes_buffered == 0) { + if (obudec_read_leb128(f, &tuheader[0], &length_of_temporal_unit_size, + &size) != 0) { + fprintf(stderr, "obudec: Failure reading temporal unit header\n"); + return -1; + } + if (size == 0 && feof(f)) { + return 1; + } + } else { + // temporal unit size was already stored in buffer + if (aom_uleb_decode(obu_ctx->buffer, obu_ctx->bytes_buffered, &size, + &length_of_temporal_unit_size) != 0) { + fprintf(stderr, "obudec: Failure reading temporal unit header\n"); + return -1; + } + } + + if (size > UINT32_MAX || size + length_of_temporal_unit_size > UINT32_MAX) { + fprintf(stderr, "obudec: TU too large.\n"); + return -1; + } + + size += length_of_temporal_unit_size; + tu_size = (size_t)size; + } else { + while (1) { + ObuHeader obu_header; + memset(&obu_header, 0, sizeof(obu_header)); + + if (obudec_read_one_obu(f, &obu_ctx->buffer, obu_ctx->bytes_buffered, + &obu_ctx->buffer_capacity, &obu_size, &obu_header, + 0) != 0) { + fprintf(stderr, "obudec: read_one_obu failed in TU loop\n"); + return -1; + } + + if (obu_header.type == OBU_TEMPORAL_DELIMITER || obu_size == 0) { + tu_size = obu_ctx->bytes_buffered; + break; + } else { + obu_ctx->bytes_buffered += obu_size; + } + } + } + +#if defined AOM_MAX_ALLOCABLE_MEMORY + if (tu_size > AOM_MAX_ALLOCABLE_MEMORY) { + fprintf(stderr, "obudec: Temporal Unit size exceeds max alloc size.\n"); + return -1; + } +#endif + uint8_t *new_buffer = (uint8_t *)realloc(*buffer, tu_size); + if (!new_buffer) { + free(*buffer); + fprintf(stderr, "obudec: Out of memory.\n"); + return -1; + } + *buffer = new_buffer; + *bytes_read = tu_size; + *buffer_size = tu_size; + + if (!obu_ctx->is_annexb) { + memcpy(*buffer, obu_ctx->buffer, tu_size); + + // At this point, (obu_ctx->buffer + obu_ctx->bytes_buffered + obu_size) + // points to the end of the buffer. + memmove(obu_ctx->buffer, obu_ctx->buffer + obu_ctx->bytes_buffered, + obu_size); + obu_ctx->bytes_buffered = obu_size; + } else { + if (!feof(f)) { + size_t data_size; + size_t offset; + if (!obu_ctx->bytes_buffered) { + data_size = tu_size - length_of_temporal_unit_size; + memcpy(*buffer, &tuheader[0], length_of_temporal_unit_size); + offset = length_of_temporal_unit_size; + } else { + memcpy(*buffer, obu_ctx->buffer, obu_ctx->bytes_buffered); + offset = obu_ctx->bytes_buffered; + data_size = tu_size - obu_ctx->bytes_buffered; + obu_ctx->bytes_buffered = 0; + } + + if (fread(*buffer + offset, 1, data_size, f) != data_size) { + fprintf(stderr, "obudec: Failed to read full temporal unit\n"); + return -1; + } + } + } + return 0; +} + +void obudec_free(struct ObuDecInputContext *obu_ctx) { free(obu_ctx->buffer); } diff --git a/third_party/aom/common/obudec.h b/third_party/aom/common/obudec.h new file mode 100644 index 000000000..c52a94e9d --- /dev/null +++ b/third_party/aom/common/obudec.h @@ -0,0 +1,48 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef OBUDEC_H_ +#define OBUDEC_H_ + +#include "common/tools_common.h" + +#ifdef __cplusplus +extern "C" { +#endif + +struct ObuDecInputContext { + struct AvxInputContext *avx_ctx; + uint8_t *buffer; + size_t buffer_capacity; + size_t bytes_buffered; + int is_annexb; +}; + +// Returns 1 when file data starts (if Annex B stream, after reading the +// size of the OBU) with what appears to be a Temporal Delimiter +// OBU as defined by Section 5 of the AV1 bitstream specification. +int file_is_obu(struct ObuDecInputContext *obu_ctx); + +// Reads one Temporal Unit from the input file. Returns 0 when a TU is +// successfully read, 1 when end of file is reached, and less than 0 when an +// error occurs. Stores TU data in 'buffer'. Reallocs buffer to match TU size, +// returns buffer capacity via 'buffer_size', and returns size of buffered data +// via 'bytes_read'. +int obudec_read_temporal_unit(struct ObuDecInputContext *obu_ctx, + uint8_t **buffer, size_t *bytes_read, + size_t *buffer_size); + +void obudec_free(struct ObuDecInputContext *obu_ctx); + +#ifdef __cplusplus +} /* extern "C" */ +#endif + +#endif // OBUDEC_H_ diff --git a/third_party/aom/common/tools_common.c b/third_party/aom/common/tools_common.c new file mode 100644 index 000000000..359ec7341 --- /dev/null +++ b/third_party/aom/common/tools_common.c @@ -0,0 +1,423 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "common/tools_common.h" + +#include <math.h> +#include <stdarg.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#if CONFIG_AV1_ENCODER +#include "aom/aomcx.h" +#endif + +#if CONFIG_AV1_DECODER +#include "aom/aomdx.h" +#endif + +#if defined(_WIN32) || defined(__OS2__) +#include <io.h> +#include <fcntl.h> + +#ifdef __OS2__ +#define _setmode setmode +#define _fileno fileno +#define _O_BINARY O_BINARY +#endif +#endif + +#define LOG_ERROR(label) \ + do { \ + const char *l = label; \ + va_list ap; \ + va_start(ap, fmt); \ + if (l) fprintf(stderr, "%s: ", l); \ + vfprintf(stderr, fmt, ap); \ + fprintf(stderr, "\n"); \ + va_end(ap); \ + } while (0) + +FILE *set_binary_mode(FILE *stream) { + (void)stream; +#if defined(_WIN32) || defined(__OS2__) + _setmode(_fileno(stream), _O_BINARY); +#endif + return stream; +} + +void die(const char *fmt, ...) { + LOG_ERROR(NULL); + usage_exit(); +} + +void fatal(const char *fmt, ...) { + LOG_ERROR("Fatal"); + exit(EXIT_FAILURE); +} + +void warn(const char *fmt, ...) { LOG_ERROR("Warning"); } + +void die_codec(aom_codec_ctx_t *ctx, const char *s) { + const char *detail = aom_codec_error_detail(ctx); + + printf("%s: %s\n", s, aom_codec_error(ctx)); + if (detail) printf(" %s\n", detail); + exit(EXIT_FAILURE); +} + +int read_yuv_frame(struct AvxInputContext *input_ctx, aom_image_t *yuv_frame) { + FILE *f = input_ctx->file; + struct FileTypeDetectionBuffer *detect = &input_ctx->detect; + int plane = 0; + int shortread = 0; + const int bytespp = (yuv_frame->fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 2 : 1; + + for (plane = 0; plane < 3; ++plane) { + uint8_t *ptr; + const int w = aom_img_plane_width(yuv_frame, plane); + const int h = aom_img_plane_height(yuv_frame, plane); + int r; + + /* Determine the correct plane based on the image format. The for-loop + * always counts in Y,U,V order, but this may not match the order of + * the data on disk. + */ + switch (plane) { + case 1: + ptr = + yuv_frame->planes[yuv_frame->fmt == AOM_IMG_FMT_YV12 ? AOM_PLANE_V + : AOM_PLANE_U]; + break; + case 2: + ptr = + yuv_frame->planes[yuv_frame->fmt == AOM_IMG_FMT_YV12 ? AOM_PLANE_U + : AOM_PLANE_V]; + break; + default: ptr = yuv_frame->planes[plane]; + } + + for (r = 0; r < h; ++r) { + size_t needed = w * bytespp; + size_t buf_position = 0; + const size_t left = detect->buf_read - detect->position; + if (left > 0) { + const size_t more = (left < needed) ? left : needed; + memcpy(ptr, detect->buf + detect->position, more); + buf_position = more; + needed -= more; + detect->position += more; + } + if (needed > 0) { + shortread |= (fread(ptr + buf_position, 1, needed, f) < needed); + } + + ptr += yuv_frame->stride[plane]; + } + } + + return shortread; +} + +#if CONFIG_AV1_ENCODER +static const AvxInterface aom_encoders[] = { + { "av1", AV1_FOURCC, &aom_codec_av1_cx }, +}; + +int get_aom_encoder_count(void) { + return sizeof(aom_encoders) / sizeof(aom_encoders[0]); +} + +const AvxInterface *get_aom_encoder_by_index(int i) { return &aom_encoders[i]; } + +const AvxInterface *get_aom_encoder_by_name(const char *name) { + int i; + + for (i = 0; i < get_aom_encoder_count(); ++i) { + const AvxInterface *encoder = get_aom_encoder_by_index(i); + if (strcmp(encoder->name, name) == 0) return encoder; + } + + return NULL; +} +#endif // CONFIG_AV1_ENCODER + +#if CONFIG_AV1_DECODER +static const AvxInterface aom_decoders[] = { + { "av1", AV1_FOURCC, &aom_codec_av1_dx }, +}; + +int get_aom_decoder_count(void) { + return sizeof(aom_decoders) / sizeof(aom_decoders[0]); +} + +const AvxInterface *get_aom_decoder_by_index(int i) { return &aom_decoders[i]; } + +const AvxInterface *get_aom_decoder_by_name(const char *name) { + int i; + + for (i = 0; i < get_aom_decoder_count(); ++i) { + const AvxInterface *const decoder = get_aom_decoder_by_index(i); + if (strcmp(decoder->name, name) == 0) return decoder; + } + + return NULL; +} + +const AvxInterface *get_aom_decoder_by_fourcc(uint32_t fourcc) { + int i; + + for (i = 0; i < get_aom_decoder_count(); ++i) { + const AvxInterface *const decoder = get_aom_decoder_by_index(i); + if (decoder->fourcc == fourcc) return decoder; + } + + return NULL; +} +#endif // CONFIG_AV1_DECODER + +void aom_img_write(const aom_image_t *img, FILE *file) { + int plane; + + for (plane = 0; plane < 3; ++plane) { + const unsigned char *buf = img->planes[plane]; + const int stride = img->stride[plane]; + const int w = aom_img_plane_width(img, plane) * + ((img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 2 : 1); + const int h = aom_img_plane_height(img, plane); + int y; + + for (y = 0; y < h; ++y) { + fwrite(buf, 1, w, file); + buf += stride; + } + } +} + +int aom_img_read(aom_image_t *img, FILE *file) { + int plane; + + for (plane = 0; plane < 3; ++plane) { + unsigned char *buf = img->planes[plane]; + const int stride = img->stride[plane]; + const int w = aom_img_plane_width(img, plane) * + ((img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 2 : 1); + const int h = aom_img_plane_height(img, plane); + int y; + + for (y = 0; y < h; ++y) { + if (fread(buf, 1, w, file) != (size_t)w) return 0; + buf += stride; + } + } + + return 1; +} + +// TODO(dkovalev) change sse_to_psnr signature: double -> int64_t +double sse_to_psnr(double samples, double peak, double sse) { + static const double kMaxPSNR = 100.0; + + if (sse > 0.0) { + const double psnr = 10.0 * log10(samples * peak * peak / sse); + return psnr > kMaxPSNR ? kMaxPSNR : psnr; + } else { + return kMaxPSNR; + } +} + +// TODO(debargha): Consolidate the functions below into a separate file. +static void highbd_img_upshift(aom_image_t *dst, aom_image_t *src, + int input_shift) { + // Note the offset is 1 less than half. + const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0; + int plane; + if (dst->d_w != src->d_w || dst->d_h != src->d_h || + dst->x_chroma_shift != src->x_chroma_shift || + dst->y_chroma_shift != src->y_chroma_shift || dst->fmt != src->fmt || + input_shift < 0) { + fatal("Unsupported image conversion"); + } + switch (src->fmt) { + case AOM_IMG_FMT_I42016: + case AOM_IMG_FMT_I42216: + case AOM_IMG_FMT_I44416: break; + default: fatal("Unsupported image conversion"); break; + } + for (plane = 0; plane < 3; plane++) { + int w = src->d_w; + int h = src->d_h; + int x, y; + if (plane) { + w = (w + src->x_chroma_shift) >> src->x_chroma_shift; + h = (h + src->y_chroma_shift) >> src->y_chroma_shift; + } + for (y = 0; y < h; y++) { + uint16_t *p_src = + (uint16_t *)(src->planes[plane] + y * src->stride[plane]); + uint16_t *p_dst = + (uint16_t *)(dst->planes[plane] + y * dst->stride[plane]); + for (x = 0; x < w; x++) *p_dst++ = (*p_src++ << input_shift) + offset; + } + } +} + +static void lowbd_img_upshift(aom_image_t *dst, aom_image_t *src, + int input_shift) { + // Note the offset is 1 less than half. + const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0; + int plane; + if (dst->d_w != src->d_w || dst->d_h != src->d_h || + dst->x_chroma_shift != src->x_chroma_shift || + dst->y_chroma_shift != src->y_chroma_shift || + dst->fmt != src->fmt + AOM_IMG_FMT_HIGHBITDEPTH || input_shift < 0) { + fatal("Unsupported image conversion"); + } + switch (src->fmt) { + case AOM_IMG_FMT_I420: + case AOM_IMG_FMT_I422: + case AOM_IMG_FMT_I444: break; + default: fatal("Unsupported image conversion"); break; + } + for (plane = 0; plane < 3; plane++) { + int w = src->d_w; + int h = src->d_h; + int x, y; + if (plane) { + w = (w + src->x_chroma_shift) >> src->x_chroma_shift; + h = (h + src->y_chroma_shift) >> src->y_chroma_shift; + } + for (y = 0; y < h; y++) { + uint8_t *p_src = src->planes[plane] + y * src->stride[plane]; + uint16_t *p_dst = + (uint16_t *)(dst->planes[plane] + y * dst->stride[plane]); + for (x = 0; x < w; x++) { + *p_dst++ = (*p_src++ << input_shift) + offset; + } + } + } +} + +void aom_img_upshift(aom_image_t *dst, aom_image_t *src, int input_shift) { + if (src->fmt & AOM_IMG_FMT_HIGHBITDEPTH) { + highbd_img_upshift(dst, src, input_shift); + } else { + lowbd_img_upshift(dst, src, input_shift); + } +} + +void aom_img_truncate_16_to_8(aom_image_t *dst, aom_image_t *src) { + int plane; + if (dst->fmt + AOM_IMG_FMT_HIGHBITDEPTH != src->fmt || dst->d_w != src->d_w || + dst->d_h != src->d_h || dst->x_chroma_shift != src->x_chroma_shift || + dst->y_chroma_shift != src->y_chroma_shift) { + fatal("Unsupported image conversion"); + } + switch (dst->fmt) { + case AOM_IMG_FMT_I420: + case AOM_IMG_FMT_I422: + case AOM_IMG_FMT_I444: break; + default: fatal("Unsupported image conversion"); break; + } + for (plane = 0; plane < 3; plane++) { + int w = src->d_w; + int h = src->d_h; + int x, y; + if (plane) { + w = (w + src->x_chroma_shift) >> src->x_chroma_shift; + h = (h + src->y_chroma_shift) >> src->y_chroma_shift; + } + for (y = 0; y < h; y++) { + uint16_t *p_src = + (uint16_t *)(src->planes[plane] + y * src->stride[plane]); + uint8_t *p_dst = dst->planes[plane] + y * dst->stride[plane]; + for (x = 0; x < w; x++) { + *p_dst++ = (uint8_t)(*p_src++); + } + } + } +} + +static void highbd_img_downshift(aom_image_t *dst, aom_image_t *src, + int down_shift) { + int plane; + if (dst->d_w != src->d_w || dst->d_h != src->d_h || + dst->x_chroma_shift != src->x_chroma_shift || + dst->y_chroma_shift != src->y_chroma_shift || dst->fmt != src->fmt || + down_shift < 0) { + fatal("Unsupported image conversion"); + } + switch (src->fmt) { + case AOM_IMG_FMT_I42016: + case AOM_IMG_FMT_I42216: + case AOM_IMG_FMT_I44416: break; + default: fatal("Unsupported image conversion"); break; + } + for (plane = 0; plane < 3; plane++) { + int w = src->d_w; + int h = src->d_h; + int x, y; + if (plane) { + w = (w + src->x_chroma_shift) >> src->x_chroma_shift; + h = (h + src->y_chroma_shift) >> src->y_chroma_shift; + } + for (y = 0; y < h; y++) { + uint16_t *p_src = + (uint16_t *)(src->planes[plane] + y * src->stride[plane]); + uint16_t *p_dst = + (uint16_t *)(dst->planes[plane] + y * dst->stride[plane]); + for (x = 0; x < w; x++) *p_dst++ = *p_src++ >> down_shift; + } + } +} + +static void lowbd_img_downshift(aom_image_t *dst, aom_image_t *src, + int down_shift) { + int plane; + if (dst->d_w != src->d_w || dst->d_h != src->d_h || + dst->x_chroma_shift != src->x_chroma_shift || + dst->y_chroma_shift != src->y_chroma_shift || + src->fmt != dst->fmt + AOM_IMG_FMT_HIGHBITDEPTH || down_shift < 0) { + fatal("Unsupported image conversion"); + } + switch (dst->fmt) { + case AOM_IMG_FMT_I420: + case AOM_IMG_FMT_I422: + case AOM_IMG_FMT_I444: break; + default: fatal("Unsupported image conversion"); break; + } + for (plane = 0; plane < 3; plane++) { + int w = src->d_w; + int h = src->d_h; + int x, y; + if (plane) { + w = (w + src->x_chroma_shift) >> src->x_chroma_shift; + h = (h + src->y_chroma_shift) >> src->y_chroma_shift; + } + for (y = 0; y < h; y++) { + uint16_t *p_src = + (uint16_t *)(src->planes[plane] + y * src->stride[plane]); + uint8_t *p_dst = dst->planes[plane] + y * dst->stride[plane]; + for (x = 0; x < w; x++) { + *p_dst++ = *p_src++ >> down_shift; + } + } + } +} + +void aom_img_downshift(aom_image_t *dst, aom_image_t *src, int down_shift) { + if (dst->fmt & AOM_IMG_FMT_HIGHBITDEPTH) { + highbd_img_downshift(dst, src, down_shift); + } else { + lowbd_img_downshift(dst, src, down_shift); + } +} diff --git a/third_party/aom/common/tools_common.h b/third_party/aom/common/tools_common.h new file mode 100644 index 000000000..abee4ea63 --- /dev/null +++ b/third_party/aom/common/tools_common.h @@ -0,0 +1,163 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef TOOLS_COMMON_H_ +#define TOOLS_COMMON_H_ + +#include <stdio.h> + +#include "config/aom_config.h" + +#include "aom/aom_codec.h" +#include "aom/aom_image.h" +#include "aom/aom_integer.h" +#include "aom_ports/msvc.h" + +#if CONFIG_AV1_ENCODER +#include "common/y4minput.h" +#endif + +#if defined(_MSC_VER) +/* MSVS uses _f{seek,tell}i64. */ +#define fseeko _fseeki64 +#define ftello _ftelli64 +typedef int64_t FileOffset; +#elif defined(_WIN32) +#include <sys/types.h> /* NOLINT*/ +/* MinGW uses f{seek,tell}o64 for large files. */ +#define fseeko fseeko64 +#define ftello ftello64 +typedef off64_t FileOffset; +#elif CONFIG_OS_SUPPORT +#include <sys/types.h> /* NOLINT*/ +typedef off_t FileOffset; +/* Use 32-bit file operations in WebM file format when building ARM + * executables (.axf) with RVCT. */ +#else +#define fseeko fseek +#define ftello ftell +typedef long FileOffset; /* NOLINT */ +#endif /* CONFIG_OS_SUPPORT */ + +#if CONFIG_OS_SUPPORT +#if defined(_MSC_VER) +#include <io.h> /* NOLINT */ +#define isatty _isatty +#define fileno _fileno +#else +#include <unistd.h> /* NOLINT */ +#endif /* _MSC_VER */ +#endif /* CONFIG_OS_SUPPORT */ + +#define LITERALU64(hi, lo) ((((uint64_t)hi) << 32) | lo) + +#ifndef PATH_MAX +#define PATH_MAX 512 +#endif + +#define IVF_FRAME_HDR_SZ (4 + 8) /* 4 byte size + 8 byte timestamp */ +#define IVF_FILE_HDR_SZ 32 + +#define RAW_FRAME_HDR_SZ sizeof(uint32_t) + +#define AV1_FOURCC 0x31305641 + +enum VideoFileType { + FILE_TYPE_OBU, + FILE_TYPE_RAW, + FILE_TYPE_IVF, + FILE_TYPE_Y4M, + FILE_TYPE_WEBM +}; + +struct FileTypeDetectionBuffer { + char buf[4]; + size_t buf_read; + size_t position; +}; + +struct AvxRational { + int numerator; + int denominator; +}; + +struct AvxInputContext { + const char *filename; + FILE *file; + int64_t length; + struct FileTypeDetectionBuffer detect; + enum VideoFileType file_type; + uint32_t width; + uint32_t height; + struct AvxRational pixel_aspect_ratio; + aom_img_fmt_t fmt; + aom_bit_depth_t bit_depth; + int only_i420; + uint32_t fourcc; + struct AvxRational framerate; +#if CONFIG_AV1_ENCODER + y4m_input y4m; +#endif +}; + +#ifdef __cplusplus +extern "C" { +#endif + +#if defined(__GNUC__) +#define AOM_NO_RETURN __attribute__((noreturn)) +#else +#define AOM_NO_RETURN +#endif + +/* Sets a stdio stream into binary mode */ +FILE *set_binary_mode(FILE *stream); + +void die(const char *fmt, ...) AOM_NO_RETURN; +void fatal(const char *fmt, ...) AOM_NO_RETURN; +void warn(const char *fmt, ...); + +void die_codec(aom_codec_ctx_t *ctx, const char *s) AOM_NO_RETURN; + +/* The tool including this file must define usage_exit() */ +void usage_exit(void) AOM_NO_RETURN; + +#undef AOM_NO_RETURN + +int read_yuv_frame(struct AvxInputContext *input_ctx, aom_image_t *yuv_frame); + +typedef struct AvxInterface { + const char *const name; + const uint32_t fourcc; + aom_codec_iface_t *(*const codec_interface)(); +} AvxInterface; + +int get_aom_encoder_count(void); +const AvxInterface *get_aom_encoder_by_index(int i); +const AvxInterface *get_aom_encoder_by_name(const char *name); + +int get_aom_decoder_count(void); +const AvxInterface *get_aom_decoder_by_index(int i); +const AvxInterface *get_aom_decoder_by_name(const char *name); +const AvxInterface *get_aom_decoder_by_fourcc(uint32_t fourcc); + +void aom_img_write(const aom_image_t *img, FILE *file); +int aom_img_read(aom_image_t *img, FILE *file); + +double sse_to_psnr(double samples, double peak, double mse); +void aom_img_upshift(aom_image_t *dst, aom_image_t *src, int input_shift); +void aom_img_downshift(aom_image_t *dst, aom_image_t *src, int down_shift); +void aom_img_truncate_16_to_8(aom_image_t *dst, aom_image_t *src); + +#ifdef __cplusplus +} /* extern "C" */ +#endif + +#endif // TOOLS_COMMON_H_ diff --git a/third_party/aom/common/video_common.h b/third_party/aom/common/video_common.h new file mode 100644 index 000000000..f96af4b7e --- /dev/null +++ b/third_party/aom/common/video_common.h @@ -0,0 +1,24 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#ifndef VIDEO_COMMON_H_ +#define VIDEO_COMMON_H_ + +#include "common/tools_common.h" + +typedef struct { + uint32_t codec_fourcc; + int frame_width; + int frame_height; + struct AvxRational time_base; +} AvxVideoInfo; + +#endif // VIDEO_COMMON_H_ diff --git a/third_party/aom/common/video_reader.c b/third_party/aom/common/video_reader.c new file mode 100644 index 000000000..f5327c928 --- /dev/null +++ b/third_party/aom/common/video_reader.c @@ -0,0 +1,84 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#include "common/video_reader.h" + +#include <stdlib.h> +#include <string.h> + +#include "aom_ports/mem_ops.h" +#include "common/ivfdec.h" + +static const char *const kIVFSignature = "DKIF"; + +struct AvxVideoReaderStruct { + AvxVideoInfo info; + FILE *file; + uint8_t *buffer; + size_t buffer_size; + size_t frame_size; + aom_codec_pts_t pts; +}; + +AvxVideoReader *aom_video_reader_open(const char *filename) { + char header[32]; + AvxVideoReader *reader = NULL; + FILE *const file = fopen(filename, "rb"); + if (!file) return NULL; // Can't open file + + if (fread(header, 1, 32, file) != 32) return NULL; // Can't read file header + + if (memcmp(kIVFSignature, header, 4) != 0) + return NULL; // Wrong IVF signature + + if (mem_get_le16(header + 4) != 0) return NULL; // Wrong IVF version + + reader = (AvxVideoReader *)calloc(1, sizeof(*reader)); + if (!reader) return NULL; // Can't allocate AvxVideoReader + + reader->file = file; + reader->info.codec_fourcc = mem_get_le32(header + 8); + reader->info.frame_width = mem_get_le16(header + 12); + reader->info.frame_height = mem_get_le16(header + 14); + reader->info.time_base.numerator = mem_get_le32(header + 16); + reader->info.time_base.denominator = mem_get_le32(header + 20); + + return reader; +} + +void aom_video_reader_close(AvxVideoReader *reader) { + if (reader) { + fclose(reader->file); + free(reader->buffer); + free(reader); + } +} + +int aom_video_reader_read_frame(AvxVideoReader *reader) { + return !ivf_read_frame(reader->file, &reader->buffer, &reader->frame_size, + &reader->buffer_size, &reader->pts); +} + +const uint8_t *aom_video_reader_get_frame(AvxVideoReader *reader, + size_t *size) { + if (size) *size = reader->frame_size; + + return reader->buffer; +} + +int64_t aom_video_reader_get_frame_pts(AvxVideoReader *reader) { + return (int64_t)reader->pts; +} + +FILE *aom_video_reader_get_file(AvxVideoReader *reader) { return reader->file; } + +const AvxVideoInfo *aom_video_reader_get_info(AvxVideoReader *reader) { + return &reader->info; +} diff --git a/third_party/aom/common/video_reader.h b/third_party/aom/common/video_reader.h new file mode 100644 index 000000000..1eb33831a --- /dev/null +++ b/third_party/aom/common/video_reader.h @@ -0,0 +1,57 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#ifndef VIDEO_READER_H_ +#define VIDEO_READER_H_ + +#include "common/video_common.h" + +// The following code is work in progress. It is going to support transparent +// reading of input files. Right now only IVF format is supported for +// simplicity. The main goal the API is to be simple and easy to use in example +// code and in aomenc/aomdec later. All low-level details like memory +// buffer management are hidden from API users. +struct AvxVideoReaderStruct; +typedef struct AvxVideoReaderStruct AvxVideoReader; + +#ifdef __cplusplus +extern "C" { +#endif + +// Opens the input file for reading and inspects it to determine file type. +// Returns an opaque AvxVideoReader* upon success, or NULL upon failure. +// Right now only IVF format is supported. +AvxVideoReader *aom_video_reader_open(const char *filename); + +// Frees all resources associated with AvxVideoReader* returned from +// aom_video_reader_open() call. +void aom_video_reader_close(AvxVideoReader *reader); + +// Reads frame from the file and stores it in internal buffer. +int aom_video_reader_read_frame(AvxVideoReader *reader); + +// Returns the pointer to memory buffer with frame data read by last call to +// aom_video_reader_read_frame(). +const uint8_t *aom_video_reader_get_frame(AvxVideoReader *reader, size_t *size); + +// Returns the pts of the frame. +int64_t aom_video_reader_get_frame_pts(AvxVideoReader *reader); +// Return the reader file. +FILE *aom_video_reader_get_file(AvxVideoReader *reader); + +// Fills AvxVideoInfo with information from opened video file. +const AvxVideoInfo *aom_video_reader_get_info(AvxVideoReader *reader); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // VIDEO_READER_H_ diff --git a/third_party/aom/common/video_writer.c b/third_party/aom/common/video_writer.c new file mode 100644 index 000000000..a7ec309fc --- /dev/null +++ b/third_party/aom/common/video_writer.c @@ -0,0 +1,77 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#include "common/video_writer.h" + +#include <stdlib.h> + +#include "aom/aom_encoder.h" +#include "common/ivfenc.h" + +struct AvxVideoWriterStruct { + AvxVideoInfo info; + FILE *file; + int frame_count; +}; + +static void write_header(FILE *file, const AvxVideoInfo *info, + int frame_count) { + struct aom_codec_enc_cfg cfg; + cfg.g_w = info->frame_width; + cfg.g_h = info->frame_height; + cfg.g_timebase.num = info->time_base.numerator; + cfg.g_timebase.den = info->time_base.denominator; + + ivf_write_file_header(file, &cfg, info->codec_fourcc, frame_count); +} + +AvxVideoWriter *aom_video_writer_open(const char *filename, + AvxContainer container, + const AvxVideoInfo *info) { + if (container == kContainerIVF) { + AvxVideoWriter *writer = NULL; + FILE *const file = fopen(filename, "wb"); + if (!file) return NULL; + + writer = malloc(sizeof(*writer)); + if (!writer) return NULL; + + writer->frame_count = 0; + writer->info = *info; + writer->file = file; + + write_header(writer->file, info, 0); + + return writer; + } + + return NULL; +} + +void aom_video_writer_close(AvxVideoWriter *writer) { + if (writer) { + // Rewriting frame header with real frame count + rewind(writer->file); + write_header(writer->file, &writer->info, writer->frame_count); + + fclose(writer->file); + free(writer); + } +} + +int aom_video_writer_write_frame(AvxVideoWriter *writer, const uint8_t *buffer, + size_t size, int64_t pts) { + ivf_write_frame_header(writer->file, pts, size); + if (fwrite(buffer, 1, size, writer->file) != size) return 0; + + ++writer->frame_count; + + return 1; +} diff --git a/third_party/aom/common/video_writer.h b/third_party/aom/common/video_writer.h new file mode 100644 index 000000000..16655d3a6 --- /dev/null +++ b/third_party/aom/common/video_writer.h @@ -0,0 +1,45 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#ifndef VIDEO_WRITER_H_ +#define VIDEO_WRITER_H_ + +#include "common/video_common.h" + +typedef enum { kContainerIVF } AvxContainer; + +struct AvxVideoWriterStruct; +typedef struct AvxVideoWriterStruct AvxVideoWriter; + +#ifdef __cplusplus +extern "C" { +#endif + +// Finds and opens writer for specified container format. +// Returns an opaque AvxVideoWriter* upon success, or NULL upon failure. +// Right now only IVF format is supported. +AvxVideoWriter *aom_video_writer_open(const char *filename, + AvxContainer container, + const AvxVideoInfo *info); + +// Frees all resources associated with AvxVideoWriter* returned from +// aom_video_writer_open() call. +void aom_video_writer_close(AvxVideoWriter *writer); + +// Writes frame bytes to the file. +int aom_video_writer_write_frame(AvxVideoWriter *writer, const uint8_t *buffer, + size_t size, int64_t pts); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // VIDEO_WRITER_H_ diff --git a/third_party/aom/common/warnings.c b/third_party/aom/common/warnings.c new file mode 100644 index 000000000..2facee252 --- /dev/null +++ b/third_party/aom/common/warnings.c @@ -0,0 +1,97 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "common/warnings.h" + +#include <assert.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include "aom/aom_encoder.h" +#include "apps/aomenc.h" +#include "common/tools_common.h" + +static const char quantizer_warning_string[] = + "Bad quantizer values. Quantizer values should not be equal, and should " + "differ by at least 8."; + +struct WarningListNode { + const char *warning_string; + struct WarningListNode *next_warning; +}; + +struct WarningList { + struct WarningListNode *warning_node; +}; + +static void add_warning(const char *warning_string, + struct WarningList *warning_list) { + struct WarningListNode **node = &warning_list->warning_node; + + struct WarningListNode *new_node = malloc(sizeof(*new_node)); + if (new_node == NULL) { + fatal("Unable to allocate warning node."); + } + + new_node->warning_string = warning_string; + new_node->next_warning = NULL; + + while (*node != NULL) node = &(*node)->next_warning; + + *node = new_node; +} + +static void free_warning_list(struct WarningList *warning_list) { + while (warning_list->warning_node != NULL) { + struct WarningListNode *const node = warning_list->warning_node; + warning_list->warning_node = node->next_warning; + free(node); + } +} + +static int continue_prompt(int num_warnings) { + int c; + fprintf(stderr, + "%d encoder configuration warning(s). Continue? (y to continue) ", + num_warnings); + c = getchar(); + return c == 'y'; +} + +static void check_quantizer(int min_q, int max_q, + struct WarningList *warning_list) { + const int lossless = min_q == 0 && max_q == 0; + if (!lossless && (min_q == max_q || abs(max_q - min_q) < 8)) + add_warning(quantizer_warning_string, warning_list); +} + +void check_encoder_config(int disable_prompt, + const struct AvxEncoderConfig *global_config, + const struct aom_codec_enc_cfg *stream_config) { + int num_warnings = 0; + struct WarningListNode *warning = NULL; + struct WarningList warning_list = { 0 }; + (void)global_config; + check_quantizer(stream_config->rc_min_quantizer, + stream_config->rc_max_quantizer, &warning_list); + /* Count and print warnings. */ + for (warning = warning_list.warning_node; warning != NULL; + warning = warning->next_warning, ++num_warnings) { + warn(warning->warning_string); + } + + free_warning_list(&warning_list); + + if (num_warnings) { + if (!disable_prompt && !continue_prompt(num_warnings)) exit(EXIT_FAILURE); + } +} diff --git a/third_party/aom/common/warnings.h b/third_party/aom/common/warnings.h new file mode 100644 index 000000000..61db2dcf8 --- /dev/null +++ b/third_party/aom/common/warnings.h @@ -0,0 +1,34 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef WARNINGS_H_ +#define WARNINGS_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +struct aom_codec_enc_cfg; +struct AvxEncoderConfig; + +/* + * Checks config for improperly used settings. Warns user upon encountering + * settings that will lead to poor output quality. Prompts user to continue + * when warnings are issued. + */ +void check_encoder_config(int disable_prompt, + const struct AvxEncoderConfig *global_config, + const struct aom_codec_enc_cfg *stream_config); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // WARNINGS_H_ diff --git a/third_party/aom/common/webmdec.cc b/third_party/aom/common/webmdec.cc new file mode 100644 index 000000000..17ac53c93 --- /dev/null +++ b/third_party/aom/common/webmdec.cc @@ -0,0 +1,229 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "common/webmdec.h" + +#include <cassert> +#include <cstring> +#include <cstdio> + +#include "third_party/libwebm/mkvparser/mkvparser.h" +#include "third_party/libwebm/mkvparser/mkvreader.h" + +namespace { + +void reset(struct WebmInputContext *const webm_ctx) { + if (webm_ctx->reader != NULL) { + mkvparser::MkvReader *const reader = + reinterpret_cast<mkvparser::MkvReader *>(webm_ctx->reader); + delete reader; + } + if (webm_ctx->segment != NULL) { + mkvparser::Segment *const segment = + reinterpret_cast<mkvparser::Segment *>(webm_ctx->segment); + delete segment; + } + if (webm_ctx->buffer != NULL) { + delete[] webm_ctx->buffer; + } + webm_ctx->reader = NULL; + webm_ctx->segment = NULL; + webm_ctx->buffer = NULL; + webm_ctx->cluster = NULL; + webm_ctx->block_entry = NULL; + webm_ctx->block = NULL; + webm_ctx->block_frame_index = 0; + webm_ctx->video_track_index = 0; + webm_ctx->timestamp_ns = 0; + webm_ctx->is_key_frame = false; +} + +void get_first_cluster(struct WebmInputContext *const webm_ctx) { + mkvparser::Segment *const segment = + reinterpret_cast<mkvparser::Segment *>(webm_ctx->segment); + const mkvparser::Cluster *const cluster = segment->GetFirst(); + webm_ctx->cluster = cluster; +} + +void rewind_and_reset(struct WebmInputContext *const webm_ctx, + struct AvxInputContext *const aom_ctx) { + rewind(aom_ctx->file); + reset(webm_ctx); +} + +} // namespace + +int file_is_webm(struct WebmInputContext *webm_ctx, + struct AvxInputContext *aom_ctx) { + mkvparser::MkvReader *const reader = new mkvparser::MkvReader(aom_ctx->file); + webm_ctx->reader = reader; + webm_ctx->reached_eos = 0; + + mkvparser::EBMLHeader header; + long long pos = 0; + if (header.Parse(reader, pos) < 0) { + rewind_and_reset(webm_ctx, aom_ctx); + return 0; + } + + mkvparser::Segment *segment; + if (mkvparser::Segment::CreateInstance(reader, pos, segment)) { + rewind_and_reset(webm_ctx, aom_ctx); + return 0; + } + webm_ctx->segment = segment; + if (segment->Load() < 0) { + rewind_and_reset(webm_ctx, aom_ctx); + return 0; + } + + const mkvparser::Tracks *const tracks = segment->GetTracks(); + const mkvparser::VideoTrack *video_track = NULL; + for (unsigned long i = 0; i < tracks->GetTracksCount(); ++i) { + const mkvparser::Track *const track = tracks->GetTrackByIndex(i); + if (track->GetType() == mkvparser::Track::kVideo) { + video_track = static_cast<const mkvparser::VideoTrack *>(track); + webm_ctx->video_track_index = static_cast<int>(track->GetNumber()); + break; + } + } + + if (video_track == NULL || video_track->GetCodecId() == NULL) { + rewind_and_reset(webm_ctx, aom_ctx); + return 0; + } + + if (!strncmp(video_track->GetCodecId(), "V_AV1", 5)) { + aom_ctx->fourcc = AV1_FOURCC; + } else { + rewind_and_reset(webm_ctx, aom_ctx); + return 0; + } + + aom_ctx->framerate.denominator = 0; + aom_ctx->framerate.numerator = 0; + aom_ctx->width = static_cast<uint32_t>(video_track->GetWidth()); + aom_ctx->height = static_cast<uint32_t>(video_track->GetHeight()); + + get_first_cluster(webm_ctx); + + return 1; +} + +int webm_read_frame(struct WebmInputContext *webm_ctx, uint8_t **buffer, + size_t *bytes_read, size_t *buffer_size) { + assert(webm_ctx->buffer == *buffer); + // This check is needed for frame parallel decoding, in which case this + // function could be called even after it has reached end of input stream. + if (webm_ctx->reached_eos) { + return 1; + } + mkvparser::Segment *const segment = + reinterpret_cast<mkvparser::Segment *>(webm_ctx->segment); + const mkvparser::Cluster *cluster = + reinterpret_cast<const mkvparser::Cluster *>(webm_ctx->cluster); + const mkvparser::Block *block = + reinterpret_cast<const mkvparser::Block *>(webm_ctx->block); + const mkvparser::BlockEntry *block_entry = + reinterpret_cast<const mkvparser::BlockEntry *>(webm_ctx->block_entry); + bool block_entry_eos = false; + do { + long status = 0; + bool get_new_block = false; + if (block_entry == NULL && !block_entry_eos) { + status = cluster->GetFirst(block_entry); + get_new_block = true; + } else if (block_entry_eos || block_entry->EOS()) { + cluster = segment->GetNext(cluster); + if (cluster == NULL || cluster->EOS()) { + *bytes_read = 0; + webm_ctx->reached_eos = 1; + return 1; + } + status = cluster->GetFirst(block_entry); + block_entry_eos = false; + get_new_block = true; + } else if (block == NULL || + webm_ctx->block_frame_index == block->GetFrameCount() || + block->GetTrackNumber() != webm_ctx->video_track_index) { + status = cluster->GetNext(block_entry, block_entry); + if (block_entry == NULL || block_entry->EOS()) { + block_entry_eos = true; + continue; + } + get_new_block = true; + } + if (status || block_entry == NULL) { + return -1; + } + if (get_new_block) { + block = block_entry->GetBlock(); + if (block == NULL) return -1; + webm_ctx->block_frame_index = 0; + } + } while (block_entry_eos || + block->GetTrackNumber() != webm_ctx->video_track_index); + + webm_ctx->cluster = cluster; + webm_ctx->block_entry = block_entry; + webm_ctx->block = block; + + const mkvparser::Block::Frame &frame = + block->GetFrame(webm_ctx->block_frame_index); + ++webm_ctx->block_frame_index; + if (frame.len > static_cast<long>(*buffer_size)) { + delete[] * buffer; + *buffer = new uint8_t[frame.len]; + webm_ctx->buffer = *buffer; + if (*buffer == NULL) { + return -1; + } + *buffer_size = frame.len; + } + *bytes_read = frame.len; + webm_ctx->timestamp_ns = block->GetTime(cluster); + webm_ctx->is_key_frame = block->IsKey(); + + mkvparser::MkvReader *const reader = + reinterpret_cast<mkvparser::MkvReader *>(webm_ctx->reader); + return frame.Read(reader, *buffer) ? -1 : 0; +} + +int webm_guess_framerate(struct WebmInputContext *webm_ctx, + struct AvxInputContext *aom_ctx) { + uint32_t i = 0; + uint8_t *buffer = NULL; + size_t buffer_size = 0; + size_t bytes_read = 0; + assert(webm_ctx->buffer == NULL); + while (webm_ctx->timestamp_ns < 1000000000 && i < 50) { + if (webm_read_frame(webm_ctx, &buffer, &bytes_read, &buffer_size)) { + break; + } + ++i; + } + aom_ctx->framerate.numerator = (i - 1) * 1000000; + aom_ctx->framerate.denominator = + static_cast<int>(webm_ctx->timestamp_ns / 1000); + delete[] buffer; + webm_ctx->buffer = NULL; + + get_first_cluster(webm_ctx); + webm_ctx->block = NULL; + webm_ctx->block_entry = NULL; + webm_ctx->block_frame_index = 0; + webm_ctx->timestamp_ns = 0; + webm_ctx->reached_eos = 0; + + return 0; +} + +void webm_free(struct WebmInputContext *webm_ctx) { reset(webm_ctx); } diff --git a/third_party/aom/common/webmdec.h b/third_party/aom/common/webmdec.h new file mode 100644 index 000000000..d5b472a01 --- /dev/null +++ b/third_party/aom/common/webmdec.h @@ -0,0 +1,71 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef WEBMDEC_H_ +#define WEBMDEC_H_ + +#include "common/tools_common.h" + +#ifdef __cplusplus +extern "C" { +#endif + +struct AvxInputContext; + +struct WebmInputContext { + void *reader; + void *segment; + uint8_t *buffer; + const void *cluster; + const void *block_entry; + const void *block; + int block_frame_index; + int video_track_index; + uint64_t timestamp_ns; + int is_key_frame; + int reached_eos; +}; + +// Checks if the input is a WebM file. If so, initializes WebMInputContext so +// that webm_read_frame can be called to retrieve a video frame. +// Returns 1 on success and 0 on failure or input is not WebM file. +// TODO(vigneshv): Refactor this function into two smaller functions specific +// to their task. +int file_is_webm(struct WebmInputContext *webm_ctx, + struct AvxInputContext *aom_ctx); + +// Reads a WebM Video Frame. Memory for the buffer is created, owned and managed +// by this function. For the first call, |buffer| should be NULL and +// |*buffer_size| should be 0. Once all the frames are read and used, +// webm_free() should be called, otherwise there will be a leak. +// Parameters: +// webm_ctx - WebmInputContext object +// buffer - pointer where the frame data will be filled. +// bytes_read - pointer to bytes read. +// buffer_size - pointer to buffer size. +// Return values: +// 0 - Success +// 1 - End of Stream +// -1 - Error +int webm_read_frame(struct WebmInputContext *webm_ctx, uint8_t **buffer, + size_t *bytes_read, size_t *buffer_size); + +// Guesses the frame rate of the input file based on the container timestamps. +int webm_guess_framerate(struct WebmInputContext *webm_ctx, + struct AvxInputContext *aom_ctx); + +// Resets the WebMInputContext. +void webm_free(struct WebmInputContext *webm_ctx); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // WEBMDEC_H_ diff --git a/third_party/aom/common/webmenc.cc b/third_party/aom/common/webmenc.cc new file mode 100644 index 000000000..58ab33670 --- /dev/null +++ b/third_party/aom/common/webmenc.cc @@ -0,0 +1,96 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include "common/webmenc.h" + +#include <string> + +#include "third_party/libwebm/mkvmuxer/mkvmuxer.h" +#include "third_party/libwebm/mkvmuxer/mkvmuxerutil.h" +#include "third_party/libwebm/mkvmuxer/mkvwriter.h" + +namespace { +const uint64_t kDebugTrackUid = 0xDEADBEEF; +const int kVideoTrackNumber = 1; +} // namespace + +void write_webm_file_header(struct WebmOutputContext *webm_ctx, + const aom_codec_enc_cfg_t *cfg, + stereo_format_t stereo_fmt, unsigned int fourcc, + const struct AvxRational *par) { + mkvmuxer::MkvWriter *const writer = new mkvmuxer::MkvWriter(webm_ctx->stream); + mkvmuxer::Segment *const segment = new mkvmuxer::Segment(); + segment->Init(writer); + segment->set_mode(mkvmuxer::Segment::kFile); + segment->OutputCues(true); + + mkvmuxer::SegmentInfo *const info = segment->GetSegmentInfo(); + const uint64_t kTimecodeScale = 1000000; + info->set_timecode_scale(kTimecodeScale); + std::string version = "aomenc"; + if (!webm_ctx->debug) { + version.append(std::string(" ") + aom_codec_version_str()); + } + info->set_writing_app(version.c_str()); + + const uint64_t video_track_id = + segment->AddVideoTrack(static_cast<int>(cfg->g_w), + static_cast<int>(cfg->g_h), kVideoTrackNumber); + mkvmuxer::VideoTrack *const video_track = static_cast<mkvmuxer::VideoTrack *>( + segment->GetTrackByNumber(video_track_id)); + video_track->SetStereoMode(stereo_fmt); + const char *codec_id; + switch (fourcc) { + case AV1_FOURCC: codec_id = "V_AV1"; break; + default: codec_id = "V_AV1"; break; + } + video_track->set_codec_id(codec_id); + if (par->numerator > 1 || par->denominator > 1) { + // TODO(fgalligan): Add support of DisplayUnit, Display Aspect Ratio type + // to WebM format. + const uint64_t display_width = static_cast<uint64_t>( + ((cfg->g_w * par->numerator * 1.0) / par->denominator) + .5); + video_track->set_display_width(display_width); + video_track->set_display_height(cfg->g_h); + } + if (webm_ctx->debug) { + video_track->set_uid(kDebugTrackUid); + } + webm_ctx->writer = writer; + webm_ctx->segment = segment; +} + +void write_webm_block(struct WebmOutputContext *webm_ctx, + const aom_codec_enc_cfg_t *cfg, + const aom_codec_cx_pkt_t *pkt) { + mkvmuxer::Segment *const segment = + reinterpret_cast<mkvmuxer::Segment *>(webm_ctx->segment); + int64_t pts_ns = pkt->data.frame.pts * 1000000000ll * cfg->g_timebase.num / + cfg->g_timebase.den; + if (pts_ns <= webm_ctx->last_pts_ns) pts_ns = webm_ctx->last_pts_ns + 1000000; + webm_ctx->last_pts_ns = pts_ns; + + segment->AddFrame(static_cast<uint8_t *>(pkt->data.frame.buf), + pkt->data.frame.sz, kVideoTrackNumber, pts_ns, + pkt->data.frame.flags & AOM_FRAME_IS_KEY); +} + +void write_webm_file_footer(struct WebmOutputContext *webm_ctx) { + mkvmuxer::MkvWriter *const writer = + reinterpret_cast<mkvmuxer::MkvWriter *>(webm_ctx->writer); + mkvmuxer::Segment *const segment = + reinterpret_cast<mkvmuxer::Segment *>(webm_ctx->segment); + segment->Finalize(); + delete segment; + delete writer; + webm_ctx->writer = NULL; + webm_ctx->segment = NULL; +} diff --git a/third_party/aom/common/webmenc.h b/third_party/aom/common/webmenc.h new file mode 100644 index 000000000..74387fb8d --- /dev/null +++ b/third_party/aom/common/webmenc.h @@ -0,0 +1,56 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ +#ifndef WEBMENC_H_ +#define WEBMENC_H_ + +#include <stdio.h> +#include <stdlib.h> + +#include "tools_common.h" +#include "aom/aom_encoder.h" + +#ifdef __cplusplus +extern "C" { +#endif + +struct WebmOutputContext { + int debug; + FILE *stream; + int64_t last_pts_ns; + void *writer; + void *segment; +}; + +/* Stereo 3D packed frame format */ +typedef enum stereo_format { + STEREO_FORMAT_MONO = 0, + STEREO_FORMAT_LEFT_RIGHT = 1, + STEREO_FORMAT_BOTTOM_TOP = 2, + STEREO_FORMAT_TOP_BOTTOM = 3, + STEREO_FORMAT_RIGHT_LEFT = 11 +} stereo_format_t; + +void write_webm_file_header(struct WebmOutputContext *webm_ctx, + const aom_codec_enc_cfg_t *cfg, + stereo_format_t stereo_fmt, unsigned int fourcc, + const struct AvxRational *par); + +void write_webm_block(struct WebmOutputContext *webm_ctx, + const aom_codec_enc_cfg_t *cfg, + const aom_codec_cx_pkt_t *pkt); + +void write_webm_file_footer(struct WebmOutputContext *webm_ctx); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // WEBMENC_H_ diff --git a/third_party/aom/common/y4menc.c b/third_party/aom/common/y4menc.c new file mode 100644 index 000000000..d33d0313e --- /dev/null +++ b/third_party/aom/common/y4menc.c @@ -0,0 +1,66 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include <assert.h> + +#include "common/y4menc.h" + +int y4m_write_file_header(char *buf, size_t len, int width, int height, + const struct AvxRational *framerate, + aom_img_fmt_t fmt, unsigned int bit_depth) { + const char *color; + switch (bit_depth) { + case 8: + color = fmt == AOM_IMG_FMT_444A + ? "C444alpha\n" + : fmt == AOM_IMG_FMT_I444 + ? "C444\n" + : fmt == AOM_IMG_FMT_I422 ? "C422\n" : "C420jpeg\n"; + break; + case 9: + color = fmt == AOM_IMG_FMT_I44416 + ? "C444p9 XYSCSS=444P9\n" + : fmt == AOM_IMG_FMT_I42216 ? "C422p9 XYSCSS=422P9\n" + : "C420p9 XYSCSS=420P9\n"; + break; + case 10: + color = fmt == AOM_IMG_FMT_I44416 + ? "C444p10 XYSCSS=444P10\n" + : fmt == AOM_IMG_FMT_I42216 ? "C422p10 XYSCSS=422P10\n" + : "C420p10 XYSCSS=420P10\n"; + break; + case 12: + color = fmt == AOM_IMG_FMT_I44416 + ? "C444p12 XYSCSS=444P12\n" + : fmt == AOM_IMG_FMT_I42216 ? "C422p12 XYSCSS=422P12\n" + : "C420p12 XYSCSS=420P12\n"; + break; + case 14: + color = fmt == AOM_IMG_FMT_I44416 + ? "C444p14 XYSCSS=444P14\n" + : fmt == AOM_IMG_FMT_I42216 ? "C422p14 XYSCSS=422P14\n" + : "C420p14 XYSCSS=420P14\n"; + break; + case 16: + color = fmt == AOM_IMG_FMT_I44416 + ? "C444p16 XYSCSS=444P16\n" + : fmt == AOM_IMG_FMT_I42216 ? "C422p16 XYSCSS=422P16\n" + : "C420p16 XYSCSS=420P16\n"; + break; + default: color = NULL; assert(0); + } + return snprintf(buf, len, "YUV4MPEG2 W%u H%u F%u:%u I%c %s", width, height, + framerate->numerator, framerate->denominator, 'p', color); +} + +int y4m_write_frame_header(char *buf, size_t len) { + return snprintf(buf, len, "FRAME\n"); +} diff --git a/third_party/aom/common/y4menc.h b/third_party/aom/common/y4menc.h new file mode 100644 index 000000000..6344176ba --- /dev/null +++ b/third_party/aom/common/y4menc.h @@ -0,0 +1,33 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#ifndef Y4MENC_H_ +#define Y4MENC_H_ + +#include "aom/aom_decoder.h" +#include "common/tools_common.h" + +#ifdef __cplusplus +extern "C" { +#endif + +#define Y4M_BUFFER_SIZE 128 + +int y4m_write_file_header(char *buf, size_t len, int width, int height, + const struct AvxRational *framerate, + aom_img_fmt_t fmt, unsigned int bit_depth); +int y4m_write_frame_header(char *buf, size_t len); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // Y4MENC_H_ diff --git a/third_party/aom/common/y4minput.c b/third_party/aom/common/y4minput.c new file mode 100644 index 000000000..a1dca10cd --- /dev/null +++ b/third_party/aom/common/y4minput.c @@ -0,0 +1,1142 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + * + * Based on code from the OggTheora software codec source code, + * Copyright (C) 2002-2010 The Xiph.Org Foundation and contributors. + */ +#include <errno.h> +#include <stdlib.h> +#include <string.h> + +#include "aom/aom_integer.h" +#include "y4minput.h" + +// Reads 'size' bytes from 'file' into 'buf' with some fault tolerance. +// Returns true on success. +static int file_read(void *buf, size_t size, FILE *file) { + const int kMaxRetries = 5; + int retry_count = 0; + int file_error; + size_t len = 0; + do { + const size_t n = fread((uint8_t *)buf + len, 1, size - len, file); + len += n; + file_error = ferror(file); + if (file_error) { + if (errno == EINTR || errno == EAGAIN) { + clearerr(file); + continue; + } else { + fprintf(stderr, "Error reading file: %u of %u bytes read, %d: %s\n", + (uint32_t)len, (uint32_t)size, errno, strerror(errno)); + return 0; + } + } + } while (!feof(file) && len < size && ++retry_count < kMaxRetries); + + if (!feof(file) && len != size) { + fprintf(stderr, + "Error reading file: %u of %u bytes read," + " error: %d, retries: %d, %d: %s\n", + (uint32_t)len, (uint32_t)size, file_error, retry_count, errno, + strerror(errno)); + } + return len == size; +} + +static int y4m_parse_tags(y4m_input *_y4m, char *_tags) { + int got_w; + int got_h; + int got_fps; + int got_interlace; + int got_par; + int got_chroma; + char *p; + char *q; + got_w = got_h = got_fps = got_interlace = got_par = got_chroma = 0; + for (p = _tags;; p = q) { + /*Skip any leading spaces.*/ + while (*p == ' ') p++; + /*If that's all we have, stop.*/ + if (p[0] == '\0') break; + /*Find the end of this tag.*/ + for (q = p + 1; *q != '\0' && *q != ' '; q++) { + } + /*Process the tag.*/ + switch (p[0]) { + case 'W': { + if (sscanf(p + 1, "%d", &_y4m->pic_w) != 1) return -1; + got_w = 1; + } break; + case 'H': { + if (sscanf(p + 1, "%d", &_y4m->pic_h) != 1) return -1; + got_h = 1; + } break; + case 'F': { + if (sscanf(p + 1, "%d:%d", &_y4m->fps_n, &_y4m->fps_d) != 2) { + return -1; + } + got_fps = 1; + } break; + case 'I': { + _y4m->interlace = p[1]; + got_interlace = 1; + } break; + case 'A': { + if (sscanf(p + 1, "%d:%d", &_y4m->par_n, &_y4m->par_d) != 2) { + return -1; + } + got_par = 1; + } break; + case 'C': { + if (q - p > 16) return -1; + memcpy(_y4m->chroma_type, p + 1, q - p - 1); + _y4m->chroma_type[q - p - 1] = '\0'; + got_chroma = 1; + } break; + /*Ignore unknown tags.*/ + } + } + if (!got_w || !got_h || !got_fps) return -1; + if (!got_interlace) _y4m->interlace = '?'; + if (!got_par) _y4m->par_n = _y4m->par_d = 0; + /*Chroma-type is not specified in older files, e.g., those generated by + mplayer.*/ + if (!got_chroma) strcpy(_y4m->chroma_type, "420"); + return 0; +} + +/*All anti-aliasing filters in the following conversion functions are based on + one of two window functions: + The 6-tap Lanczos window (for down-sampling and shifts): + sinc(\pi*t)*sinc(\pi*t/3), |t|<3 (sinc(t)==sin(t)/t) + 0, |t|>=3 + The 4-tap Mitchell window (for up-sampling): + 7|t|^3-12|t|^2+16/3, |t|<1 + -(7/3)|x|^3+12|x|^2-20|x|+32/3, |t|<2 + 0, |t|>=2 + The number of taps is intentionally kept small to reduce computational + overhead and limit ringing. + + The taps from these filters are scaled so that their sum is 1, and the + result is scaled by 128 and rounded to integers to create a filter whose + intermediate values fit inside 16 bits. + Coefficients are rounded in such a way as to ensure their sum is still 128, + which is usually equivalent to normal rounding. + + Conversions which require both horizontal and vertical filtering could + have these steps pipelined, for less memory consumption and better cache + performance, but we do them separately for simplicity.*/ +#define OC_MINI(_a, _b) ((_a) > (_b) ? (_b) : (_a)) +#define OC_MAXI(_a, _b) ((_a) < (_b) ? (_b) : (_a)) +#define OC_CLAMPI(_a, _b, _c) (OC_MAXI(_a, OC_MINI(_b, _c))) + +/*420jpeg chroma samples are sited like: + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + + 420mpeg2 chroma samples are sited like: + Y-------Y-------Y-------Y------- + | | | | + BR | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + Y-------Y-------Y-------Y------- + | | | | + BR | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + + We use a resampling filter to shift the site locations one quarter pixel (at + the chroma plane's resolution) to the right. + The 4:2:2 modes look exactly the same, except there are twice as many chroma + lines, and they are vertically co-sited with the luma samples in both the + mpeg2 and jpeg cases (thus requiring no vertical resampling).*/ +static void y4m_42xmpeg2_42xjpeg_helper(unsigned char *_dst, + const unsigned char *_src, int _c_w, + int _c_h) { + int y; + int x; + for (y = 0; y < _c_h; y++) { + /*Filter: [4 -17 114 35 -9 1]/128, derived from a 6-tap Lanczos + window.*/ + for (x = 0; x < OC_MINI(_c_w, 2); x++) { + _dst[x] = (unsigned char)OC_CLAMPI( + 0, + (4 * _src[0] - 17 * _src[OC_MAXI(x - 1, 0)] + 114 * _src[x] + + 35 * _src[OC_MINI(x + 1, _c_w - 1)] - + 9 * _src[OC_MINI(x + 2, _c_w - 1)] + _src[OC_MINI(x + 3, _c_w - 1)] + + 64) >> + 7, + 255); + } + for (; x < _c_w - 3; x++) { + _dst[x] = (unsigned char)OC_CLAMPI( + 0, + (4 * _src[x - 2] - 17 * _src[x - 1] + 114 * _src[x] + + 35 * _src[x + 1] - 9 * _src[x + 2] + _src[x + 3] + 64) >> + 7, + 255); + } + for (; x < _c_w; x++) { + _dst[x] = (unsigned char)OC_CLAMPI( + 0, + (4 * _src[x - 2] - 17 * _src[x - 1] + 114 * _src[x] + + 35 * _src[OC_MINI(x + 1, _c_w - 1)] - + 9 * _src[OC_MINI(x + 2, _c_w - 1)] + _src[_c_w - 1] + 64) >> + 7, + 255); + } + _dst += _c_w; + _src += _c_w; + } +} + +/*Handles both 422 and 420mpeg2 to 422jpeg and 420jpeg, respectively.*/ +static void y4m_convert_42xmpeg2_42xjpeg(y4m_input *_y4m, unsigned char *_dst, + unsigned char *_aux) { + int c_w; + int c_h; + int c_sz; + int pli; + /*Skip past the luma data.*/ + _dst += _y4m->pic_w * _y4m->pic_h; + /*Compute the size of each chroma plane.*/ + c_w = (_y4m->pic_w + _y4m->dst_c_dec_h - 1) / _y4m->dst_c_dec_h; + c_h = (_y4m->pic_h + _y4m->dst_c_dec_v - 1) / _y4m->dst_c_dec_v; + c_sz = c_w * c_h; + for (pli = 1; pli < 3; pli++) { + y4m_42xmpeg2_42xjpeg_helper(_dst, _aux, c_w, c_h); + _dst += c_sz; + _aux += c_sz; + } +} + +/*This format is only used for interlaced content, but is included for + completeness. + + 420jpeg chroma samples are sited like: + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + + 420paldv chroma samples are sited like: + YR------Y-------YR------Y------- + | | | | + | | | | + | | | | + YB------Y-------YB------Y------- + | | | | + | | | | + | | | | + YR------Y-------YR------Y------- + | | | | + | | | | + | | | | + YB------Y-------YB------Y------- + | | | | + | | | | + | | | | + + We use a resampling filter to shift the site locations one quarter pixel (at + the chroma plane's resolution) to the right. + Then we use another filter to move the C_r location down one quarter pixel, + and the C_b location up one quarter pixel.*/ +static void y4m_convert_42xpaldv_42xjpeg(y4m_input *_y4m, unsigned char *_dst, + unsigned char *_aux) { + unsigned char *tmp; + int c_w; + int c_h; + int c_sz; + int pli; + int y; + int x; + /*Skip past the luma data.*/ + _dst += _y4m->pic_w * _y4m->pic_h; + /*Compute the size of each chroma plane.*/ + c_w = (_y4m->pic_w + 1) / 2; + c_h = (_y4m->pic_h + _y4m->dst_c_dec_h - 1) / _y4m->dst_c_dec_h; + c_sz = c_w * c_h; + tmp = _aux + 2 * c_sz; + for (pli = 1; pli < 3; pli++) { + /*First do the horizontal re-sampling. + This is the same as the mpeg2 case, except that after the horizontal + case, we need to apply a second vertical filter.*/ + y4m_42xmpeg2_42xjpeg_helper(tmp, _aux, c_w, c_h); + _aux += c_sz; + switch (pli) { + case 1: { + /*Slide C_b up a quarter-pel. + This is the same filter used above, but in the other order.*/ + for (x = 0; x < c_w; x++) { + for (y = 0; y < OC_MINI(c_h, 3); y++) { + _dst[y * c_w] = (unsigned char)OC_CLAMPI( + 0, + (tmp[0] - 9 * tmp[OC_MAXI(y - 2, 0) * c_w] + + 35 * tmp[OC_MAXI(y - 1, 0) * c_w] + 114 * tmp[y * c_w] - + 17 * tmp[OC_MINI(y + 1, c_h - 1) * c_w] + + 4 * tmp[OC_MINI(y + 2, c_h - 1) * c_w] + 64) >> + 7, + 255); + } + for (; y < c_h - 2; y++) { + _dst[y * c_w] = (unsigned char)OC_CLAMPI( + 0, + (tmp[(y - 3) * c_w] - 9 * tmp[(y - 2) * c_w] + + 35 * tmp[(y - 1) * c_w] + 114 * tmp[y * c_w] - + 17 * tmp[(y + 1) * c_w] + 4 * tmp[(y + 2) * c_w] + 64) >> + 7, + 255); + } + for (; y < c_h; y++) { + _dst[y * c_w] = (unsigned char)OC_CLAMPI( + 0, + (tmp[(y - 3) * c_w] - 9 * tmp[(y - 2) * c_w] + + 35 * tmp[(y - 1) * c_w] + 114 * tmp[y * c_w] - + 17 * tmp[OC_MINI(y + 1, c_h - 1) * c_w] + + 4 * tmp[(c_h - 1) * c_w] + 64) >> + 7, + 255); + } + _dst++; + tmp++; + } + _dst += c_sz - c_w; + tmp -= c_w; + } break; + case 2: { + /*Slide C_r down a quarter-pel. + This is the same as the horizontal filter.*/ + for (x = 0; x < c_w; x++) { + for (y = 0; y < OC_MINI(c_h, 2); y++) { + _dst[y * c_w] = (unsigned char)OC_CLAMPI( + 0, + (4 * tmp[0] - 17 * tmp[OC_MAXI(y - 1, 0) * c_w] + + 114 * tmp[y * c_w] + 35 * tmp[OC_MINI(y + 1, c_h - 1) * c_w] - + 9 * tmp[OC_MINI(y + 2, c_h - 1) * c_w] + + tmp[OC_MINI(y + 3, c_h - 1) * c_w] + 64) >> + 7, + 255); + } + for (; y < c_h - 3; y++) { + _dst[y * c_w] = (unsigned char)OC_CLAMPI( + 0, + (4 * tmp[(y - 2) * c_w] - 17 * tmp[(y - 1) * c_w] + + 114 * tmp[y * c_w] + 35 * tmp[(y + 1) * c_w] - + 9 * tmp[(y + 2) * c_w] + tmp[(y + 3) * c_w] + 64) >> + 7, + 255); + } + for (; y < c_h; y++) { + _dst[y * c_w] = (unsigned char)OC_CLAMPI( + 0, + (4 * tmp[(y - 2) * c_w] - 17 * tmp[(y - 1) * c_w] + + 114 * tmp[y * c_w] + 35 * tmp[OC_MINI(y + 1, c_h - 1) * c_w] - + 9 * tmp[OC_MINI(y + 2, c_h - 1) * c_w] + tmp[(c_h - 1) * c_w] + + 64) >> + 7, + 255); + } + _dst++; + tmp++; + } + } break; + } + /*For actual interlaced material, this would have to be done separately on + each field, and the shift amounts would be different. + C_r moves down 1/8, C_b up 3/8 in the top field, and C_r moves down 3/8, + C_b up 1/8 in the bottom field. + The corresponding filters would be: + Down 1/8 (reverse order for up): [3 -11 125 15 -4 0]/128 + Down 3/8 (reverse order for up): [4 -19 98 56 -13 2]/128*/ + } +} + +/*Perform vertical filtering to reduce a single plane from 4:2:2 to 4:2:0. + This is used as a helper by several converation routines.*/ +static void y4m_422jpeg_420jpeg_helper(unsigned char *_dst, + const unsigned char *_src, int _c_w, + int _c_h) { + int y; + int x; + /*Filter: [3 -17 78 78 -17 3]/128, derived from a 6-tap Lanczos window.*/ + for (x = 0; x < _c_w; x++) { + for (y = 0; y < OC_MINI(_c_h, 2); y += 2) { + _dst[(y >> 1) * _c_w] = + OC_CLAMPI(0, + (64 * _src[0] + 78 * _src[OC_MINI(1, _c_h - 1) * _c_w] - + 17 * _src[OC_MINI(2, _c_h - 1) * _c_w] + + 3 * _src[OC_MINI(3, _c_h - 1) * _c_w] + 64) >> + 7, + 255); + } + for (; y < _c_h - 3; y += 2) { + _dst[(y >> 1) * _c_w] = + OC_CLAMPI(0, + (3 * (_src[(y - 2) * _c_w] + _src[(y + 3) * _c_w]) - + 17 * (_src[(y - 1) * _c_w] + _src[(y + 2) * _c_w]) + + 78 * (_src[y * _c_w] + _src[(y + 1) * _c_w]) + 64) >> + 7, + 255); + } + for (; y < _c_h; y += 2) { + _dst[(y >> 1) * _c_w] = OC_CLAMPI( + 0, + (3 * (_src[(y - 2) * _c_w] + _src[(_c_h - 1) * _c_w]) - + 17 * (_src[(y - 1) * _c_w] + _src[OC_MINI(y + 2, _c_h - 1) * _c_w]) + + 78 * (_src[y * _c_w] + _src[OC_MINI(y + 1, _c_h - 1) * _c_w]) + + 64) >> + 7, + 255); + } + _src++; + _dst++; + } +} + +/*420jpeg chroma samples are sited like: + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + + 422jpeg chroma samples are sited like: + Y---BR--Y-------Y---BR--Y------- + | | | | + | | | | + | | | | + Y---BR--Y-------Y---BR--Y------- + | | | | + | | | | + | | | | + Y---BR--Y-------Y---BR--Y------- + | | | | + | | | | + | | | | + Y---BR--Y-------Y---BR--Y------- + | | | | + | | | | + | | | | + + We use a resampling filter to decimate the chroma planes by two in the + vertical direction.*/ +static void y4m_convert_422jpeg_420jpeg(y4m_input *_y4m, unsigned char *_dst, + unsigned char *_aux) { + int c_w; + int c_h; + int c_sz; + int dst_c_w; + int dst_c_h; + int dst_c_sz; + int pli; + /*Skip past the luma data.*/ + _dst += _y4m->pic_w * _y4m->pic_h; + /*Compute the size of each chroma plane.*/ + c_w = (_y4m->pic_w + _y4m->src_c_dec_h - 1) / _y4m->src_c_dec_h; + c_h = _y4m->pic_h; + dst_c_w = (_y4m->pic_w + _y4m->dst_c_dec_h - 1) / _y4m->dst_c_dec_h; + dst_c_h = (_y4m->pic_h + _y4m->dst_c_dec_v - 1) / _y4m->dst_c_dec_v; + c_sz = c_w * c_h; + dst_c_sz = dst_c_w * dst_c_h; + for (pli = 1; pli < 3; pli++) { + y4m_422jpeg_420jpeg_helper(_dst, _aux, c_w, c_h); + _aux += c_sz; + _dst += dst_c_sz; + } +} + +/*420jpeg chroma samples are sited like: + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + + 422 chroma samples are sited like: + YBR-----Y-------YBR-----Y------- + | | | | + | | | | + | | | | + YBR-----Y-------YBR-----Y------- + | | | | + | | | | + | | | | + YBR-----Y-------YBR-----Y------- + | | | | + | | | | + | | | | + YBR-----Y-------YBR-----Y------- + | | | | + | | | | + | | | | + + We use a resampling filter to shift the original site locations one quarter + pixel (at the original chroma resolution) to the right. + Then we use a second resampling filter to decimate the chroma planes by two + in the vertical direction.*/ +static void y4m_convert_422_420jpeg(y4m_input *_y4m, unsigned char *_dst, + unsigned char *_aux) { + unsigned char *tmp; + int c_w; + int c_h; + int c_sz; + int dst_c_h; + int dst_c_sz; + int pli; + /*Skip past the luma data.*/ + _dst += _y4m->pic_w * _y4m->pic_h; + /*Compute the size of each chroma plane.*/ + c_w = (_y4m->pic_w + _y4m->src_c_dec_h - 1) / _y4m->src_c_dec_h; + c_h = _y4m->pic_h; + dst_c_h = (_y4m->pic_h + _y4m->dst_c_dec_v - 1) / _y4m->dst_c_dec_v; + c_sz = c_w * c_h; + dst_c_sz = c_w * dst_c_h; + tmp = _aux + 2 * c_sz; + for (pli = 1; pli < 3; pli++) { + /*In reality, the horizontal and vertical steps could be pipelined, for + less memory consumption and better cache performance, but we do them + separately for simplicity.*/ + /*First do horizontal filtering (convert to 422jpeg)*/ + y4m_42xmpeg2_42xjpeg_helper(tmp, _aux, c_w, c_h); + /*Now do the vertical filtering.*/ + y4m_422jpeg_420jpeg_helper(_dst, tmp, c_w, c_h); + _aux += c_sz; + _dst += dst_c_sz; + } +} + +/*420jpeg chroma samples are sited like: + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | BR | | BR | + | | | | + Y-------Y-------Y-------Y------- + | | | | + | | | | + | | | | + + 411 chroma samples are sited like: + YBR-----Y-------Y-------Y------- + | | | | + | | | | + | | | | + YBR-----Y-------Y-------Y------- + | | | | + | | | | + | | | | + YBR-----Y-------Y-------Y------- + | | | | + | | | | + | | | | + YBR-----Y-------Y-------Y------- + | | | | + | | | | + | | | | + + We use a filter to resample at site locations one eighth pixel (at the source + chroma plane's horizontal resolution) and five eighths of a pixel to the + right. + Then we use another filter to decimate the planes by 2 in the vertical + direction.*/ +static void y4m_convert_411_420jpeg(y4m_input *_y4m, unsigned char *_dst, + unsigned char *_aux) { + unsigned char *tmp; + int c_w; + int c_h; + int c_sz; + int dst_c_w; + int dst_c_h; + int dst_c_sz; + int tmp_sz; + int pli; + int y; + int x; + /*Skip past the luma data.*/ + _dst += _y4m->pic_w * _y4m->pic_h; + /*Compute the size of each chroma plane.*/ + c_w = (_y4m->pic_w + _y4m->src_c_dec_h - 1) / _y4m->src_c_dec_h; + c_h = _y4m->pic_h; + dst_c_w = (_y4m->pic_w + _y4m->dst_c_dec_h - 1) / _y4m->dst_c_dec_h; + dst_c_h = (_y4m->pic_h + _y4m->dst_c_dec_v - 1) / _y4m->dst_c_dec_v; + c_sz = c_w * c_h; + dst_c_sz = dst_c_w * dst_c_h; + tmp_sz = dst_c_w * c_h; + tmp = _aux + 2 * c_sz; + for (pli = 1; pli < 3; pli++) { + /*In reality, the horizontal and vertical steps could be pipelined, for + less memory consumption and better cache performance, but we do them + separately for simplicity.*/ + /*First do horizontal filtering (convert to 422jpeg)*/ + for (y = 0; y < c_h; y++) { + /*Filters: [1 110 18 -1]/128 and [-3 50 86 -5]/128, both derived from a + 4-tap Mitchell window.*/ + for (x = 0; x < OC_MINI(c_w, 1); x++) { + tmp[x << 1] = (unsigned char)OC_CLAMPI( + 0, + (111 * _aux[0] + 18 * _aux[OC_MINI(1, c_w - 1)] - + _aux[OC_MINI(2, c_w - 1)] + 64) >> + 7, + 255); + tmp[x << 1 | 1] = (unsigned char)OC_CLAMPI( + 0, + (47 * _aux[0] + 86 * _aux[OC_MINI(1, c_w - 1)] - + 5 * _aux[OC_MINI(2, c_w - 1)] + 64) >> + 7, + 255); + } + for (; x < c_w - 2; x++) { + tmp[x << 1] = + (unsigned char)OC_CLAMPI(0, + (_aux[x - 1] + 110 * _aux[x] + + 18 * _aux[x + 1] - _aux[x + 2] + 64) >> + 7, + 255); + tmp[x << 1 | 1] = (unsigned char)OC_CLAMPI( + 0, + (-3 * _aux[x - 1] + 50 * _aux[x] + 86 * _aux[x + 1] - + 5 * _aux[x + 2] + 64) >> + 7, + 255); + } + for (; x < c_w; x++) { + tmp[x << 1] = (unsigned char)OC_CLAMPI( + 0, + (_aux[x - 1] + 110 * _aux[x] + 18 * _aux[OC_MINI(x + 1, c_w - 1)] - + _aux[c_w - 1] + 64) >> + 7, + 255); + if ((x << 1 | 1) < dst_c_w) { + tmp[x << 1 | 1] = (unsigned char)OC_CLAMPI( + 0, + (-3 * _aux[x - 1] + 50 * _aux[x] + + 86 * _aux[OC_MINI(x + 1, c_w - 1)] - 5 * _aux[c_w - 1] + 64) >> + 7, + 255); + } + } + tmp += dst_c_w; + _aux += c_w; + } + tmp -= tmp_sz; + /*Now do the vertical filtering.*/ + y4m_422jpeg_420jpeg_helper(_dst, tmp, dst_c_w, c_h); + _dst += dst_c_sz; + } +} + +/*Convert 444 to 420jpeg.*/ +static void y4m_convert_444_420jpeg(y4m_input *_y4m, unsigned char *_dst, + unsigned char *_aux) { + unsigned char *tmp; + int c_w; + int c_h; + int c_sz; + int dst_c_w; + int dst_c_h; + int dst_c_sz; + int tmp_sz; + int pli; + int y; + int x; + /*Skip past the luma data.*/ + _dst += _y4m->pic_w * _y4m->pic_h; + /*Compute the size of each chroma plane.*/ + c_w = (_y4m->pic_w + _y4m->src_c_dec_h - 1) / _y4m->src_c_dec_h; + c_h = _y4m->pic_h; + dst_c_w = (_y4m->pic_w + _y4m->dst_c_dec_h - 1) / _y4m->dst_c_dec_h; + dst_c_h = (_y4m->pic_h + _y4m->dst_c_dec_v - 1) / _y4m->dst_c_dec_v; + c_sz = c_w * c_h; + dst_c_sz = dst_c_w * dst_c_h; + tmp_sz = dst_c_w * c_h; + tmp = _aux + 2 * c_sz; + for (pli = 1; pli < 3; pli++) { + /*Filter: [3 -17 78 78 -17 3]/128, derived from a 6-tap Lanczos window.*/ + for (y = 0; y < c_h; y++) { + for (x = 0; x < OC_MINI(c_w, 2); x += 2) { + tmp[x >> 1] = OC_CLAMPI(0, + (64 * _aux[0] + 78 * _aux[OC_MINI(1, c_w - 1)] - + 17 * _aux[OC_MINI(2, c_w - 1)] + + 3 * _aux[OC_MINI(3, c_w - 1)] + 64) >> + 7, + 255); + } + for (; x < c_w - 3; x += 2) { + tmp[x >> 1] = OC_CLAMPI(0, + (3 * (_aux[x - 2] + _aux[x + 3]) - + 17 * (_aux[x - 1] + _aux[x + 2]) + + 78 * (_aux[x] + _aux[x + 1]) + 64) >> + 7, + 255); + } + for (; x < c_w; x += 2) { + tmp[x >> 1] = + OC_CLAMPI(0, + (3 * (_aux[x - 2] + _aux[c_w - 1]) - + 17 * (_aux[x - 1] + _aux[OC_MINI(x + 2, c_w - 1)]) + + 78 * (_aux[x] + _aux[OC_MINI(x + 1, c_w - 1)]) + 64) >> + 7, + 255); + } + tmp += dst_c_w; + _aux += c_w; + } + tmp -= tmp_sz; + /*Now do the vertical filtering.*/ + y4m_422jpeg_420jpeg_helper(_dst, tmp, dst_c_w, c_h); + _dst += dst_c_sz; + } +} + +/*The image is padded with empty chroma components at 4:2:0.*/ +static void y4m_convert_mono_420jpeg(y4m_input *_y4m, unsigned char *_dst, + unsigned char *_aux) { + int c_sz; + (void)_aux; + _dst += _y4m->pic_w * _y4m->pic_h; + c_sz = ((_y4m->pic_w + _y4m->dst_c_dec_h - 1) / _y4m->dst_c_dec_h) * + ((_y4m->pic_h + _y4m->dst_c_dec_v - 1) / _y4m->dst_c_dec_v); + memset(_dst, 128, c_sz * 2); +} + +/*No conversion function needed.*/ +static void y4m_convert_null(y4m_input *_y4m, unsigned char *_dst, + unsigned char *_aux) { + (void)_y4m; + (void)_dst; + (void)_aux; +} + +int y4m_input_open(y4m_input *_y4m, FILE *_fin, char *_skip, int _nskip, + int only_420) { + char buffer[80] = { 0 }; + int ret; + int i; + /*Read until newline, or 80 cols, whichever happens first.*/ + for (i = 0; i < 79; i++) { + if (_nskip > 0) { + buffer[i] = *_skip++; + _nskip--; + } else { + if (!file_read(buffer + i, 1, _fin)) return -1; + } + if (buffer[i] == '\n') break; + } + /*We skipped too much header data.*/ + if (_nskip > 0) return -1; + if (i == 79) { + fprintf(stderr, "Error parsing header; not a YUV2MPEG2 file?\n"); + return -1; + } + buffer[i] = '\0'; + if (memcmp(buffer, "YUV4MPEG", 8)) { + fprintf(stderr, "Incomplete magic for YUV4MPEG file.\n"); + return -1; + } + if (buffer[8] != '2') { + fprintf(stderr, "Incorrect YUV input file version; YUV4MPEG2 required.\n"); + } + ret = y4m_parse_tags(_y4m, buffer + 5); + if (ret < 0) { + fprintf(stderr, "Error parsing YUV4MPEG2 header.\n"); + return ret; + } + if (_y4m->interlace == '?') { + fprintf(stderr, + "Warning: Input video interlacing format unknown; " + "assuming progressive scan.\n"); + } else if (_y4m->interlace != 'p') { + fprintf(stderr, + "Input video is interlaced; " + "Only progressive scan handled.\n"); + return -1; + } + _y4m->aom_fmt = AOM_IMG_FMT_I420; + _y4m->bps = 12; + _y4m->bit_depth = 8; + if (strcmp(_y4m->chroma_type, "420") == 0 || + strcmp(_y4m->chroma_type, "420jpeg") == 0) { + _y4m->src_c_dec_h = _y4m->dst_c_dec_h = _y4m->src_c_dec_v = + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = + _y4m->pic_w * _y4m->pic_h + + 2 * ((_y4m->pic_w + 1) / 2) * ((_y4m->pic_h + 1) / 2); + /* Natively supported: no conversion required. */ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + } else if (strcmp(_y4m->chroma_type, "420p10") == 0) { + _y4m->src_c_dec_h = 2; + _y4m->dst_c_dec_h = 2; + _y4m->src_c_dec_v = 2; + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = + 2 * (_y4m->pic_w * _y4m->pic_h + + 2 * ((_y4m->pic_w + 1) / 2) * ((_y4m->pic_h + 1) / 2)); + /* Natively supported: no conversion required. */ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + _y4m->bit_depth = 10; + _y4m->bps = 15; + _y4m->aom_fmt = AOM_IMG_FMT_I42016; + if (only_420) { + fprintf(stderr, "Unsupported conversion from 420p10 to 420jpeg\n"); + return -1; + } + } else if (strcmp(_y4m->chroma_type, "420p12") == 0) { + _y4m->src_c_dec_h = 2; + _y4m->dst_c_dec_h = 2; + _y4m->src_c_dec_v = 2; + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = + 2 * (_y4m->pic_w * _y4m->pic_h + + 2 * ((_y4m->pic_w + 1) / 2) * ((_y4m->pic_h + 1) / 2)); + /* Natively supported: no conversion required. */ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + _y4m->bit_depth = 12; + _y4m->bps = 18; + _y4m->aom_fmt = AOM_IMG_FMT_I42016; + if (only_420) { + fprintf(stderr, "Unsupported conversion from 420p12 to 420jpeg\n"); + return -1; + } + } else if (strcmp(_y4m->chroma_type, "420mpeg2") == 0) { + _y4m->src_c_dec_h = _y4m->dst_c_dec_h = _y4m->src_c_dec_v = + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h; + /*Chroma filter required: read into the aux buf first.*/ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = + 2 * ((_y4m->pic_w + 1) / 2) * ((_y4m->pic_h + 1) / 2); + _y4m->convert = y4m_convert_42xmpeg2_42xjpeg; + } else if (strcmp(_y4m->chroma_type, "420paldv") == 0) { + _y4m->src_c_dec_h = _y4m->dst_c_dec_h = _y4m->src_c_dec_v = + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h; + /*Chroma filter required: read into the aux buf first. + We need to make two filter passes, so we need some extra space in the + aux buffer.*/ + _y4m->aux_buf_sz = 3 * ((_y4m->pic_w + 1) / 2) * ((_y4m->pic_h + 1) / 2); + _y4m->aux_buf_read_sz = + 2 * ((_y4m->pic_w + 1) / 2) * ((_y4m->pic_h + 1) / 2); + _y4m->convert = y4m_convert_42xpaldv_42xjpeg; + } else if (strcmp(_y4m->chroma_type, "422jpeg") == 0) { + _y4m->src_c_dec_h = _y4m->dst_c_dec_h = 2; + _y4m->src_c_dec_v = 1; + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h; + /*Chroma filter required: read into the aux buf first.*/ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = + 2 * ((_y4m->pic_w + 1) / 2) * _y4m->pic_h; + _y4m->convert = y4m_convert_422jpeg_420jpeg; + } else if (strcmp(_y4m->chroma_type, "422") == 0) { + _y4m->src_c_dec_h = 2; + _y4m->src_c_dec_v = 1; + if (only_420) { + _y4m->dst_c_dec_h = 2; + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h; + /*Chroma filter required: read into the aux buf first. + We need to make two filter passes, so we need some extra space in the + aux buffer.*/ + _y4m->aux_buf_read_sz = 2 * ((_y4m->pic_w + 1) / 2) * _y4m->pic_h; + _y4m->aux_buf_sz = + _y4m->aux_buf_read_sz + ((_y4m->pic_w + 1) / 2) * _y4m->pic_h; + _y4m->convert = y4m_convert_422_420jpeg; + } else { + _y4m->aom_fmt = AOM_IMG_FMT_I422; + _y4m->bps = 16; + _y4m->dst_c_dec_h = _y4m->src_c_dec_h; + _y4m->dst_c_dec_v = _y4m->src_c_dec_v; + _y4m->dst_buf_read_sz = + _y4m->pic_w * _y4m->pic_h + 2 * ((_y4m->pic_w + 1) / 2) * _y4m->pic_h; + /*Natively supported: no conversion required.*/ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + } + } else if (strcmp(_y4m->chroma_type, "422p10") == 0) { + _y4m->src_c_dec_h = 2; + _y4m->src_c_dec_v = 1; + _y4m->aom_fmt = AOM_IMG_FMT_I42216; + _y4m->bps = 20; + _y4m->bit_depth = 10; + _y4m->dst_c_dec_h = _y4m->src_c_dec_h; + _y4m->dst_c_dec_v = _y4m->src_c_dec_v; + _y4m->dst_buf_read_sz = 2 * (_y4m->pic_w * _y4m->pic_h + + 2 * ((_y4m->pic_w + 1) / 2) * _y4m->pic_h); + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + if (only_420) { + fprintf(stderr, "Unsupported conversion from 422p10 to 420jpeg\n"); + return -1; + } + } else if (strcmp(_y4m->chroma_type, "422p12") == 0) { + _y4m->src_c_dec_h = 2; + _y4m->src_c_dec_v = 1; + _y4m->aom_fmt = AOM_IMG_FMT_I42216; + _y4m->bps = 24; + _y4m->bit_depth = 12; + _y4m->dst_c_dec_h = _y4m->src_c_dec_h; + _y4m->dst_c_dec_v = _y4m->src_c_dec_v; + _y4m->dst_buf_read_sz = 2 * (_y4m->pic_w * _y4m->pic_h + + 2 * ((_y4m->pic_w + 1) / 2) * _y4m->pic_h); + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + if (only_420) { + fprintf(stderr, "Unsupported conversion from 422p12 to 420jpeg\n"); + return -1; + } + } else if (strcmp(_y4m->chroma_type, "411") == 0) { + _y4m->src_c_dec_h = 4; + _y4m->dst_c_dec_h = 2; + _y4m->src_c_dec_v = 1; + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h; + /*Chroma filter required: read into the aux buf first. + We need to make two filter passes, so we need some extra space in the + aux buffer.*/ + _y4m->aux_buf_read_sz = 2 * ((_y4m->pic_w + 3) / 4) * _y4m->pic_h; + _y4m->aux_buf_sz = + _y4m->aux_buf_read_sz + ((_y4m->pic_w + 1) / 2) * _y4m->pic_h; + _y4m->convert = y4m_convert_411_420jpeg; + } else if (strcmp(_y4m->chroma_type, "444") == 0) { + _y4m->src_c_dec_h = 1; + _y4m->src_c_dec_v = 1; + if (only_420) { + _y4m->dst_c_dec_h = 2; + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h; + /*Chroma filter required: read into the aux buf first. + We need to make two filter passes, so we need some extra space in the + aux buffer.*/ + _y4m->aux_buf_read_sz = 2 * _y4m->pic_w * _y4m->pic_h; + _y4m->aux_buf_sz = + _y4m->aux_buf_read_sz + ((_y4m->pic_w + 1) / 2) * _y4m->pic_h; + _y4m->convert = y4m_convert_444_420jpeg; + } else { + _y4m->aom_fmt = AOM_IMG_FMT_I444; + _y4m->bps = 24; + _y4m->dst_c_dec_h = _y4m->src_c_dec_h; + _y4m->dst_c_dec_v = _y4m->src_c_dec_v; + _y4m->dst_buf_read_sz = 3 * _y4m->pic_w * _y4m->pic_h; + /*Natively supported: no conversion required.*/ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + } + } else if (strcmp(_y4m->chroma_type, "444p10") == 0) { + _y4m->src_c_dec_h = 1; + _y4m->src_c_dec_v = 1; + _y4m->aom_fmt = AOM_IMG_FMT_I44416; + _y4m->bps = 30; + _y4m->bit_depth = 10; + _y4m->dst_c_dec_h = _y4m->src_c_dec_h; + _y4m->dst_c_dec_v = _y4m->src_c_dec_v; + _y4m->dst_buf_read_sz = 2 * 3 * _y4m->pic_w * _y4m->pic_h; + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + if (only_420) { + fprintf(stderr, "Unsupported conversion from 444p10 to 420jpeg\n"); + return -1; + } + } else if (strcmp(_y4m->chroma_type, "444p12") == 0) { + _y4m->src_c_dec_h = 1; + _y4m->src_c_dec_v = 1; + _y4m->aom_fmt = AOM_IMG_FMT_I44416; + _y4m->bps = 36; + _y4m->bit_depth = 12; + _y4m->dst_c_dec_h = _y4m->src_c_dec_h; + _y4m->dst_c_dec_v = _y4m->src_c_dec_v; + _y4m->dst_buf_read_sz = 2 * 3 * _y4m->pic_w * _y4m->pic_h; + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + if (only_420) { + fprintf(stderr, "Unsupported conversion from 444p12 to 420jpeg\n"); + return -1; + } + } else if (strcmp(_y4m->chroma_type, "444alpha") == 0) { + _y4m->src_c_dec_h = 1; + _y4m->src_c_dec_v = 1; + if (only_420) { + _y4m->dst_c_dec_h = 2; + _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h; + /*Chroma filter required: read into the aux buf first. + We need to make two filter passes, so we need some extra space in the + aux buffer. + The extra plane also gets read into the aux buf. + It will be discarded.*/ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 3 * _y4m->pic_w * _y4m->pic_h; + _y4m->convert = y4m_convert_444_420jpeg; + } else { + _y4m->aom_fmt = AOM_IMG_FMT_444A; + _y4m->bps = 32; + _y4m->dst_c_dec_h = _y4m->src_c_dec_h; + _y4m->dst_c_dec_v = _y4m->src_c_dec_v; + _y4m->dst_buf_read_sz = 4 * _y4m->pic_w * _y4m->pic_h; + /*Natively supported: no conversion required.*/ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_null; + } + } else if (strcmp(_y4m->chroma_type, "mono") == 0) { + _y4m->src_c_dec_h = _y4m->src_c_dec_v = 0; + _y4m->dst_c_dec_h = _y4m->dst_c_dec_v = 2; + _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h; + /*No extra space required, but we need to clear the chroma planes.*/ + _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0; + _y4m->convert = y4m_convert_mono_420jpeg; + } else { + fprintf(stderr, "Unknown chroma sampling type: %s\n", _y4m->chroma_type); + return -1; + } + /*The size of the final frame buffers is always computed from the + destination chroma decimation type.*/ + _y4m->dst_buf_sz = + _y4m->pic_w * _y4m->pic_h + + 2 * ((_y4m->pic_w + _y4m->dst_c_dec_h - 1) / _y4m->dst_c_dec_h) * + ((_y4m->pic_h + _y4m->dst_c_dec_v - 1) / _y4m->dst_c_dec_v); + if (_y4m->bit_depth == 8) + _y4m->dst_buf = (unsigned char *)malloc(_y4m->dst_buf_sz); + else + _y4m->dst_buf = (unsigned char *)malloc(2 * _y4m->dst_buf_sz); + + if (_y4m->aux_buf_sz > 0) + _y4m->aux_buf = (unsigned char *)malloc(_y4m->aux_buf_sz); + return 0; +} + +void y4m_input_close(y4m_input *_y4m) { + free(_y4m->dst_buf); + free(_y4m->aux_buf); +} + +int y4m_input_fetch_frame(y4m_input *_y4m, FILE *_fin, aom_image_t *_img) { + char frame[6]; + int pic_sz; + int c_w; + int c_h; + int c_sz; + int bytes_per_sample = _y4m->bit_depth > 8 ? 2 : 1; + /*Read and skip the frame header.*/ + if (!file_read(frame, 6, _fin)) return 0; + if (memcmp(frame, "FRAME", 5)) { + fprintf(stderr, "Loss of framing in Y4M input data\n"); + return -1; + } + if (frame[5] != '\n') { + char c; + int j; + for (j = 0; j < 79 && file_read(&c, 1, _fin) && c != '\n'; j++) { + } + if (j == 79) { + fprintf(stderr, "Error parsing Y4M frame header\n"); + return -1; + } + } + /*Read the frame data that needs no conversion.*/ + if (!file_read(_y4m->dst_buf, _y4m->dst_buf_read_sz, _fin)) { + fprintf(stderr, "Error reading Y4M frame data.\n"); + return -1; + } + /*Read the frame data that does need conversion.*/ + if (!file_read(_y4m->aux_buf, _y4m->aux_buf_read_sz, _fin)) { + fprintf(stderr, "Error reading Y4M frame data.\n"); + return -1; + } + /*Now convert the just read frame.*/ + (*_y4m->convert)(_y4m, _y4m->dst_buf, _y4m->aux_buf); + /*Fill in the frame buffer pointers. + We don't use aom_img_wrap() because it forces padding for odd picture + sizes, which would require a separate fread call for every row.*/ + memset(_img, 0, sizeof(*_img)); + /*Y4M has the planes in Y'CbCr order, which libaom calls Y, U, and V.*/ + _img->fmt = _y4m->aom_fmt; + _img->w = _img->d_w = _y4m->pic_w; + _img->h = _img->d_h = _y4m->pic_h; + _img->x_chroma_shift = _y4m->dst_c_dec_h >> 1; + _img->y_chroma_shift = _y4m->dst_c_dec_v >> 1; + _img->bps = _y4m->bps; + + /*Set up the buffer pointers.*/ + pic_sz = _y4m->pic_w * _y4m->pic_h * bytes_per_sample; + c_w = (_y4m->pic_w + _y4m->dst_c_dec_h - 1) / _y4m->dst_c_dec_h; + c_w *= bytes_per_sample; + c_h = (_y4m->pic_h + _y4m->dst_c_dec_v - 1) / _y4m->dst_c_dec_v; + c_sz = c_w * c_h; + _img->stride[AOM_PLANE_Y] = _img->stride[AOM_PLANE_ALPHA] = + _y4m->pic_w * bytes_per_sample; + _img->stride[AOM_PLANE_U] = _img->stride[AOM_PLANE_V] = c_w; + _img->planes[AOM_PLANE_Y] = _y4m->dst_buf; + _img->planes[AOM_PLANE_U] = _y4m->dst_buf + pic_sz; + _img->planes[AOM_PLANE_V] = _y4m->dst_buf + pic_sz + c_sz; + _img->planes[AOM_PLANE_ALPHA] = _y4m->dst_buf + pic_sz + 2 * c_sz; + return 1; +} diff --git a/third_party/aom/common/y4minput.h b/third_party/aom/common/y4minput.h new file mode 100644 index 000000000..db20190db --- /dev/null +++ b/third_party/aom/common/y4minput.h @@ -0,0 +1,69 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + * + * Based on code from the OggTheora software codec source code, + * Copyright (C) 2002-2010 The Xiph.Org Foundation and contributors. + */ + +#ifndef Y4MINPUT_H_ +#define Y4MINPUT_H_ + +#include <stdio.h> +#include "aom/aom_image.h" + +#ifdef __cplusplus +extern "C" { +#endif + +typedef struct y4m_input y4m_input; + +/*The function used to perform chroma conversion.*/ +typedef void (*y4m_convert_func)(y4m_input *_y4m, unsigned char *_dst, + unsigned char *_src); + +struct y4m_input { + int pic_w; + int pic_h; + int fps_n; + int fps_d; + int par_n; + int par_d; + char interlace; + int src_c_dec_h; + int src_c_dec_v; + int dst_c_dec_h; + int dst_c_dec_v; + char chroma_type[16]; + /*The size of each converted frame buffer.*/ + size_t dst_buf_sz; + /*The amount to read directly into the converted frame buffer.*/ + size_t dst_buf_read_sz; + /*The size of the auxilliary buffer.*/ + size_t aux_buf_sz; + /*The amount to read into the auxilliary buffer.*/ + size_t aux_buf_read_sz; + y4m_convert_func convert; + unsigned char *dst_buf; + unsigned char *aux_buf; + enum aom_img_fmt aom_fmt; + int bps; + unsigned int bit_depth; +}; + +int y4m_input_open(y4m_input *_y4m, FILE *_fin, char *_skip, int _nskip, + int only_420); +void y4m_input_close(y4m_input *_y4m); +int y4m_input_fetch_frame(y4m_input *_y4m, FILE *_fin, aom_image_t *img); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // Y4MINPUT_H_ |