Lines Matching full:speed
202 // similar, but applies much more aggressive pruning to get better speed-up
268 // Prune less likely chosen transforms for each intra mode. The speed
269 // feature ranges from 0 to 2, for different speed / compression trade offs.
405 * \brief Sequence/frame level speed vs quality features
488 * Speed features for the first pass.
675 // single_motion_search (assuming no other speed features). Otherwise, reduce
719 // For allintra encode, this speed feature reduces instruction count by 6.4%
720 // for speed=6 with coding performance change less than 0.24%. For AVIF image
721 // encode, this speed feature reduces encode time by 8.14% for speed 6 on a
723 // speed feature is not applicable to speed >= 7.
739 // For allintra encode, this speed feature reduces instruction count by 5.1%
740 // for speed=6 with coding performance change less than 0.22%. For AVIF image
741 // encode, this speed feature reduces encode time by 4.44% for speed 6 on a
743 // For speed >= 7, variance-based logic is used to determine the partition
744 // structure instead of recursive partition search. Therefore, this speed
795 // With this speed feature, the top two sub blocks can directly use rdcost
809 // By default, this feature is turned on to speed up the encoder partition
811 // If disabling it, at speed 0, 30 frames, we could get
875 // Use the rd cost around the best FULLPEL_MV to speed up subpel search
985 // This speed feature equaling 0 means no skipping.
986 // If the speed feature equals 1 or 2, skip the current ref_mv in NEW_MV mode
992 // The speed feature equaling 1 means using subpel mv in the comparison.
993 // The speed feature equaling 2 means using fullpel mv in the comparison.
994 // If the speed feature >= 3, skip the current ref_mv in NEW_MV mode based on
998 // This speed feature checks duplicate ref MVs among NEARESTMV, NEARMV,
1167 // Speed/quality impact:
1168 // Speed 1: 12% faster, 0.1% psnr loss.
1169 // Speed 2: 2% faster, 0.05% psnr loss.
1170 // No change for speed 3 and up, because |disable_onesided_comp| is true.
1230 // For allintra encode, this speed feature reduces instruction count
1231 // by 1.90%, 2.21% and 1.97% for speed 6, 7 and 8 with coding performance
1232 // change less than 0.04%. For AVIF image encode, this speed feature reduces
1233 // encode time by 1.56%, 2.14% and 0.90% for speed 6, 7 and 8 on a typical
1252 // Terminate early in luma palette_size search. Speed feature values indicate
1263 // 2.76%, 2.30%, 1.84%, 2.69%, 2.04%, 2.05% and 1.44% for speed 0, 1, 2, 3, 4,
1265 // than 0.01% for speed <= 2 and less than 0.03% for speed >= 3. For AVIF
1267 // 0.93%, 0.89%, 1.03%, 1.07%, 1.20% and 0.18% for speed 0, 1, 2, 3, 4, 5, 6,
1289 // performance at speed 0.
1295 // intra mode decision. Here, add a speed feature to reduce this number for
1305 // For allintra encode, this speed feature reduces instruction count
1306 // by 4.461%, 3.699% and 3.536% for speed 6, 7 and 8 on a typical video
1308 // encode, this speed feature reduces encode time by 2.849%, 2.471%,
1309 // and 2.051% for speed 6, 7 and 8 on a typical image dataset with coding
1318 // 0.62%, 1.73%, 2.50%, 2.89%, 3.09% and 3.86% for speed 0 to 6 on screen
1321 // 0.85%, 1.05%, 1.45%, 1.66% and 1.95% for speed 0 to 6 on a typical image
1372 // For allintra encode, this speed feature reduces instruction count
1373 // by 4.76%, 8.92% and 11.28% for speed 6, 7 and 8 with coding performance
1374 // change less than 0.32%. For AVIF image encode, this speed feature reduces
1375 // encode time by 4.65%, 9.16% and 10.45% for speed 6, 7 and 8 on a typical
1386 // For allintra encode, this speed feature reduces instruction count
1387 // by 1.11%, 1.08%, 1.02% and 0.93% for speed 3, 6, 7 and 8 with coding
1388 // performance change less than 0.02%. For AVIF image encode, this speed
1390 // for speed 3, 4, 5, 6, 7 and 8 on a typical image dataset with coding
1411 // speed feature tx_domain_dist_thres_level.
1413 // The speed feature tx_domain_dist_level decides which of the above methods
1462 // 0: speed feature OFF
1468 // Different speed feature values (0 to 3) decide the aggressiveness of
1478 // Controls the disabling of winner mode processing. Speed feature levels
1504 // Decoder side speed feature to add penalty for use of dual-sgr filters.
1816 // For allintra encode, this speed feature reduces instruction count by 5.39%
1817 // for speed 9 on a typical video dataset with coding performance gain
1819 // For AVIF image encode, this speed feature reduces encode time
1820 // by 8.44% for speed 9 on a typical image dataset with coding performance
1834 // This speed feature has a substantial gain on coding metrics, with moderate
1835 // increase encoding time. Select threshold based on speed vs quality
1851 // For allintra encode, this speed feature reduces instruction count by 1.10%
1852 // for speed 9 with coding performance change less than 0.04%.
1853 // For AVIF image encode, this speed feature reduces encode time by 1.03% for
1854 // speed 9 on a typical image dataset with coding performance change less than
1862 // For allintra encode, this speed feature reduces instruction count by 3.96%
1863 // for speed 9 with coding performance change less than 0.38%.
1864 // For AVIF image encode, this speed feature reduces encode time by 3.46% for
1865 // speed 9 on a typical image dataset with coding performance change less than
1871 // For allintra encode, this speed feature reduces instruction count by 3.05%
1872 // for speed 9 with coding performance change less than 0.24%.
1873 // For AVIF image encode, this speed feature reduces encode time by 1.87% for
1874 // speed 9 on a typical image dataset with coding performance change less than
1894 // increases with increase in the level set for speed feature.
1907 // can only be turned on when res < 360p and speed >= 9, in which case only
1916 // This generally leads to better coding efficiency but with some speed loss.
1937 * \brief Top level speed vs quality trade off data struture.
1941 * Sequence/frame level speed features:
1946 * Speed features for the first pass.
1951 * Speed features related to how tpl's searches are done.
1956 * Global motion speed features:
1961 * Partition search speed features:
1966 * Motion search speed features:
1971 * Inter mode search speed features:
1976 * Interpolation filter search speed features:
1981 * Intra mode search speed features:
1986 * Transform size/type search speed features:
1991 * RD calculation speed features:
2001 * In-loop filter speed features:
2006 * Real-time mode speed features:
2015 /*!\brief Frame size independent speed vs quality trade off flags
2020 * \param[in] speed Speed setting passed in from the command line
2022 * \remark No return value but configures the various speed trade off flags
2023 * based on the passed in speed setting. (Higher speed gives lower
2027 int speed);
2029 /*!\brief Frame size dependent speed vs quality trade off flags
2034 * \param[in] speed Speed setting passed in from the command line
2036 * \remark No return value but configures the various speed trade off flags
2037 * based on the passed in speed setting and frame size. (Higher speed
2041 int speed);
2042 /*!\brief Q index dependent speed vs quality trade off flags
2047 * \param[in] speed Speed setting passed in from the command line
2049 * \remark No return value but configures the various speed trade off flags
2050 * based on the passed in speed setting and current frame's Q index.
2051 * (Higher speed corresponds to lower quality)
2053 void av1_set_speed_features_qindex_dependent(struct AV1_COMP *cpi, int speed);