mirror of https://github.com/FFmpeg/FFmpeg.git
Tag:
Branch:
Tree:
a63da4511d
master
oldabi
release/0.10
release/0.11
release/0.5
release/0.6
release/0.7
release/0.8
release/0.9
release/1.0
release/1.1
release/1.2
release/2.0
release/2.1
release/2.2
release/2.3
release/2.4
release/2.5
release/2.6
release/2.7
release/2.8
release/3.0
release/3.1
release/3.2
release/3.3
release/3.4
release/4.0
release/4.1
release/4.2
release/4.3
release/4.4
release/5.0
release/5.1
release/6.0
release/6.1
release/7.0
release/7.1
N
ffmpeg-0.6.3
n0.10
n0.10.1
n0.10.10
n0.10.11
n0.10.12
n0.10.13
n0.10.14
n0.10.15
n0.10.16
n0.10.2
n0.10.3
n0.10.4
n0.10.5
n0.10.6
n0.10.7
n0.10.8
n0.10.9
n0.11
n0.11-dev
n0.11.1
n0.11.2
n0.11.3
n0.11.4
n0.11.5
n0.12-dev
n0.5.10
n0.5.11
n0.5.12
n0.5.13
n0.5.14
n0.5.15
n0.5.5
n0.5.6
n0.5.7
n0.5.8
n0.5.9
n0.6.4
n0.6.5
n0.6.6
n0.6.7
n0.7.1
n0.7.10
n0.7.11
n0.7.12
n0.7.13
n0.7.14
n0.7.15
n0.7.16
n0.7.17
n0.7.2
n0.7.3
n0.7.4
n0.7.5
n0.7.6
n0.7.7
n0.7.8
n0.7.9
n0.8
n0.8.1
n0.8.10
n0.8.11
n0.8.12
n0.8.13
n0.8.14
n0.8.15
n0.8.2
n0.8.3
n0.8.4
n0.8.5
n0.8.6
n0.8.7
n0.8.8
n0.8.9
n0.9
n0.9.1
n0.9.2
n0.9.3
n0.9.4
n1.0
n1.0.1
n1.0.10
n1.0.2
n1.0.3
n1.0.4
n1.0.5
n1.0.6
n1.0.7
n1.0.8
n1.0.9
n1.1
n1.1-dev
n1.1.1
n1.1.10
n1.1.11
n1.1.12
n1.1.13
n1.1.14
n1.1.15
n1.1.16
n1.1.2
n1.1.3
n1.1.4
n1.1.5
n1.1.6
n1.1.7
n1.1.8
n1.1.9
n1.2
n1.2-dev
n1.2.1
n1.2.10
n1.2.11
n1.2.12
n1.2.2
n1.2.3
n1.2.4
n1.2.5
n1.2.6
n1.2.7
n1.2.8
n1.2.9
n1.3-dev
n2.0
n2.0.1
n2.0.2
n2.0.3
n2.0.4
n2.0.5
n2.0.6
n2.0.7
n2.1
n2.1-dev
n2.1.1
n2.1.2
n2.1.3
n2.1.4
n2.1.5
n2.1.6
n2.1.7
n2.1.8
n2.2
n2.2-dev
n2.2-rc1
n2.2-rc2
n2.2.1
n2.2.10
n2.2.11
n2.2.12
n2.2.13
n2.2.14
n2.2.15
n2.2.16
n2.2.2
n2.2.3
n2.2.4
n2.2.5
n2.2.6
n2.2.7
n2.2.8
n2.2.9
n2.3
n2.3-dev
n2.3.1
n2.3.2
n2.3.3
n2.3.4
n2.3.5
n2.3.6
n2.4
n2.4-dev
n2.4.1
n2.4.10
n2.4.11
n2.4.12
n2.4.13
n2.4.14
n2.4.2
n2.4.3
n2.4.4
n2.4.5
n2.4.6
n2.4.7
n2.4.8
n2.4.9
n2.5
n2.5-dev
n2.5.1
n2.5.10
n2.5.11
n2.5.2
n2.5.3
n2.5.4
n2.5.5
n2.5.6
n2.5.7
n2.5.8
n2.5.9
n2.6
n2.6-dev
n2.6.1
n2.6.2
n2.6.3
n2.6.4
n2.6.5
n2.6.6
n2.6.7
n2.6.8
n2.6.9
n2.7
n2.7-dev
n2.7.1
n2.7.2
n2.7.3
n2.7.4
n2.7.5
n2.7.6
n2.7.7
n2.8
n2.8-dev
n2.8.1
n2.8.10
n2.8.11
n2.8.12
n2.8.13
n2.8.14
n2.8.15
n2.8.16
n2.8.17
n2.8.18
n2.8.19
n2.8.2
n2.8.20
n2.8.21
n2.8.22
n2.8.3
n2.8.4
n2.8.5
n2.8.6
n2.8.7
n2.8.8
n2.8.9
n2.9-dev
n3.0
n3.0.1
n3.0.10
n3.0.11
n3.0.12
n3.0.2
n3.0.3
n3.0.4
n3.0.5
n3.0.6
n3.0.7
n3.0.8
n3.0.9
n3.1
n3.1-dev
n3.1.1
n3.1.10
n3.1.11
n3.1.2
n3.1.3
n3.1.4
n3.1.5
n3.1.6
n3.1.7
n3.1.8
n3.1.9
n3.2
n3.2-dev
n3.2.1
n3.2.10
n3.2.11
n3.2.12
n3.2.13
n3.2.14
n3.2.15
n3.2.16
n3.2.17
n3.2.18
n3.2.19
n3.2.2
n3.2.3
n3.2.4
n3.2.5
n3.2.6
n3.2.7
n3.2.8
n3.2.9
n3.3
n3.3-dev
n3.3.1
n3.3.2
n3.3.3
n3.3.4
n3.3.5
n3.3.6
n3.3.7
n3.3.8
n3.3.9
n3.4
n3.4-dev
n3.4.1
n3.4.10
n3.4.11
n3.4.12
n3.4.13
n3.4.2
n3.4.3
n3.4.4
n3.4.5
n3.4.6
n3.4.7
n3.4.8
n3.4.9
n3.5-dev
n4.0
n4.0.1
n4.0.2
n4.0.3
n4.0.4
n4.0.5
n4.0.6
n4.1
n4.1-dev
n4.1.1
n4.1.10
n4.1.11
n4.1.2
n4.1.3
n4.1.4
n4.1.5
n4.1.6
n4.1.7
n4.1.8
n4.1.9
n4.2
n4.2-dev
n4.2.1
n4.2.10
n4.2.2
n4.2.3
n4.2.4
n4.2.5
n4.2.6
n4.2.7
n4.2.8
n4.2.9
n4.3
n4.3-dev
n4.3.1
n4.3.2
n4.3.3
n4.3.4
n4.3.5
n4.3.6
n4.3.7
n4.3.8
n4.4
n4.4-dev
n4.4.1
n4.4.2
n4.4.3
n4.4.4
n4.4.5
n4.5-dev
n5.0
n5.0.1
n5.0.2
n5.0.3
n5.1
n5.1-dev
n5.1.1
n5.1.2
n5.1.3
n5.1.4
n5.1.5
n5.1.6
n5.2-dev
n6.0
n6.0.1
n6.1
n6.1-dev
n6.1.1
n6.1.2
n6.2-dev
n7.0
n7.0.1
n7.0.2
n7.1
n7.1-dev
n7.2-dev
v0.5
v0.5.1
v0.5.2
v0.5.3
v0.6
v0.6.1
${ noResults }
11 Commits (a63da4511d0fee66695ff4afd264ba1dbf1e812d)
Author | SHA1 | Message | Date |
---|---|---|---|
Martin Storsjö | 5eb5aec475 |
arm: vp9itxfm: Do a simpler half/quarter idct16/idct32 when possible
This work is sponsored by, and copyright, Google. This avoids loading and calculating coefficients that we know will be zero, and avoids filling the temp buffer with zeros in places where we know the second pass won't read. This gives a pretty substantial speedup for the smaller subpartitions. The code size increases from 12388 bytes to 19784 bytes. The idct16/32_end macros are moved above the individual functions; the instructions themselves are unchanged, but since new functions are added at the same place where the code is moved from, the diff looks rather messy. Before: Cortex A7 A8 A9 A53 vp9_inv_dct_dct_16x16_sub1_add_neon: 273.0 189.5 212.0 235.8 vp9_inv_dct_dct_16x16_sub2_add_neon: 2102.1 1521.7 1736.2 1265.8 vp9_inv_dct_dct_16x16_sub4_add_neon: 2104.5 1533.0 1736.6 1265.5 vp9_inv_dct_dct_16x16_sub8_add_neon: 2484.8 1828.7 2014.4 1506.5 vp9_inv_dct_dct_16x16_sub12_add_neon: 2851.2 2117.8 2294.8 1753.2 vp9_inv_dct_dct_16x16_sub16_add_neon: 3239.4 2408.3 2543.5 1994.9 vp9_inv_dct_dct_32x32_sub1_add_neon: 758.3 456.7 864.5 553.9 vp9_inv_dct_dct_32x32_sub2_add_neon: 10776.7 7949.8 8567.7 6819.7 vp9_inv_dct_dct_32x32_sub4_add_neon: 10865.6 8131.5 8589.6 6816.3 vp9_inv_dct_dct_32x32_sub8_add_neon: 12053.9 9271.3 9387.7 7564.0 vp9_inv_dct_dct_32x32_sub12_add_neon: 13328.3 10463.2 10217.0 8321.3 vp9_inv_dct_dct_32x32_sub16_add_neon: 14176.4 11509.5 11018.7 9062.3 vp9_inv_dct_dct_32x32_sub20_add_neon: 15301.5 12999.9 11855.1 9828.2 vp9_inv_dct_dct_32x32_sub24_add_neon: 16482.7 14931.5 12650.1 10575.0 vp9_inv_dct_dct_32x32_sub28_add_neon: 17589.5 15811.9 13482.8 11333.4 vp9_inv_dct_dct_32x32_sub32_add_neon: 18696.2 17049.2 14355.6 12089.7 After: vp9_inv_dct_dct_16x16_sub1_add_neon: 273.0 189.5 211.7 235.8 vp9_inv_dct_dct_16x16_sub2_add_neon: 1203.5 998.2 1035.3 763.0 vp9_inv_dct_dct_16x16_sub4_add_neon: 1203.5 998.1 1035.5 760.8 vp9_inv_dct_dct_16x16_sub8_add_neon: 1926.1 1610.6 1722.1 1271.7 vp9_inv_dct_dct_16x16_sub12_add_neon: 2873.2 2129.7 2285.1 1757.3 vp9_inv_dct_dct_16x16_sub16_add_neon: 3221.4 2520.3 2557.6 2002.1 vp9_inv_dct_dct_32x32_sub1_add_neon: 753.0 457.5 866.6 554.6 vp9_inv_dct_dct_32x32_sub2_add_neon: 7554.6 5652.4 6048.4 4920.2 vp9_inv_dct_dct_32x32_sub4_add_neon: 7549.9 5685.0 6046.9 4925.7 vp9_inv_dct_dct_32x32_sub8_add_neon: 8336.9 6704.5 6604.0 5478.0 vp9_inv_dct_dct_32x32_sub12_add_neon: 10914.0 9777.2 9240.4 7416.9 vp9_inv_dct_dct_32x32_sub16_add_neon: 11859.2 11223.3 9966.3 8095.1 vp9_inv_dct_dct_32x32_sub20_add_neon: 15237.1 13029.4 11838.3 9829.4 vp9_inv_dct_dct_32x32_sub24_add_neon: 16293.2 14379.8 12644.9 10572.0 vp9_inv_dct_dct_32x32_sub28_add_neon: 17424.3 15734.7 13473.0 11326.9 vp9_inv_dct_dct_32x32_sub32_add_neon: 18531.3 17457.0 14298.6 12080.0 Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Martin Storsjö | 47b3c2c18d |
arm: vp9itxfm: Move the load_add_store macro out from the itxfm16 pass2 function
This allows reusing the macro for a separate implementation of the pass2 function. Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Martin Storsjö | 0331c3f5e8 |
arm: vp9itxfm: Make the larger core transforms standalone functions
This work is sponsored by, and copyright, Google. This reduces the code size of libavcodec/arm/vp9itxfm_neon.o from 15324 to 12388 bytes. This gives a small slowdown of a couple tens of cycles, up to around 150 cycles for the full case of the largest transform, but makes it more feasible to add more optimized versions of these transforms. Before: Cortex A7 A8 A9 A53 vp9_inv_dct_dct_16x16_sub4_add_neon: 2063.4 1516.0 1719.5 1245.1 vp9_inv_dct_dct_16x16_sub16_add_neon: 3279.3 2454.5 2525.2 1982.3 vp9_inv_dct_dct_32x32_sub4_add_neon: 10750.0 7955.4 8525.6 6754.2 vp9_inv_dct_dct_32x32_sub32_add_neon: 18574.0 17108.4 14216.7 12010.2 After: vp9_inv_dct_dct_16x16_sub4_add_neon: 2060.8 1608.5 1735.7 1262.0 vp9_inv_dct_dct_16x16_sub16_add_neon: 3211.2 2443.5 2546.1 1999.5 vp9_inv_dct_dct_32x32_sub4_add_neon: 10682.0 8043.8 8581.3 6810.1 vp9_inv_dct_dct_32x32_sub32_add_neon: 18522.4 17277.4 14286.7 12087.9 Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Martin Storsjö | 3bc5b28d5a |
arm: vp9itxfm: Avoid .irp when it doesn't save any lines
This makes it more readable. Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Martin Storsjö | 9c8bc74c2b |
arm: vp9itxfm: Skip empty slices in the first pass of idct_idct 16x16 and 32x32
This work is sponsored by, and copyright, Google. Previously all subpartitions except the eob=1 (DC) case ran with the same runtime: Cortex A7 A8 A9 A53 vp9_inv_dct_dct_16x16_sub16_add_neon: 3188.1 2435.4 2499.0 1969.0 vp9_inv_dct_dct_32x32_sub32_add_neon: 18531.7 16582.3 14207.6 12000.3 By skipping individual 4x16 or 4x32 pixel slices in the first pass, we reduce the runtime of these functions like this: vp9_inv_dct_dct_16x16_sub1_add_neon: 274.6 189.5 211.7 235.8 vp9_inv_dct_dct_16x16_sub2_add_neon: 2064.0 1534.8 1719.4 1248.7 vp9_inv_dct_dct_16x16_sub4_add_neon: 2135.0 1477.2 1736.3 1249.5 vp9_inv_dct_dct_16x16_sub8_add_neon: 2446.7 1828.7 1993.6 1494.7 vp9_inv_dct_dct_16x16_sub12_add_neon: 2832.4 2118.3 2266.5 1735.1 vp9_inv_dct_dct_16x16_sub16_add_neon: 3211.7 2475.3 2523.5 1983.1 vp9_inv_dct_dct_32x32_sub1_add_neon: 756.2 456.7 862.0 553.9 vp9_inv_dct_dct_32x32_sub2_add_neon: 10682.2 8190.4 8539.2 6762.5 vp9_inv_dct_dct_32x32_sub4_add_neon: 10813.5 8014.9 8518.3 6762.8 vp9_inv_dct_dct_32x32_sub8_add_neon: 11859.6 9313.0 9347.4 7514.5 vp9_inv_dct_dct_32x32_sub12_add_neon: 12946.6 10752.4 10192.2 8280.2 vp9_inv_dct_dct_32x32_sub16_add_neon: 14074.6 11946.5 11001.4 9008.6 vp9_inv_dct_dct_32x32_sub20_add_neon: 15269.9 13662.7 11816.1 9762.6 vp9_inv_dct_dct_32x32_sub24_add_neon: 16327.9 14940.1 12626.7 10516.0 vp9_inv_dct_dct_32x32_sub28_add_neon: 17462.7 15776.1 13446.2 11264.7 vp9_inv_dct_dct_32x32_sub32_add_neon: 18575.5 17157.0 14249.3 12015.1 I.e. in general a very minor overhead for the full subpartition case due to the additional loads and cmps, but a significant speedup for the cases when we only need to process a small part of the actual input data. In common VP9 content in a few inspected clips, 70-90% of the non-dc-only 16x16 and 32x32 IDCTs only have nonzero coefficients in the upper left 8x8 or 16x16 subpartitions respectively. Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Martin Storsjö | 3c87039a40 |
arm: vp9itxfm: Only reload the idct coeffs for the iadst_idct combination
This avoids reloading them if they haven't been clobbered, if the first pass also was idct. This is similar to what was done in the aarch64 version. Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Martin Storsjö | 79566ec8c7 |
arm: vp9itxfm: Rename a macro parameter to fit better
Since the same parameter is used for both input and output, the name inout is more fitting. This matches the naming used below in the dmbutterfly macro. Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Martin Storsjö | 721bc37522 |
arm/aarch64: vp9itxfm: Fix indentation of macro arguments
Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Janne Grunau | e5b0fc170f |
arm: vp9itxfm: Simplify the stack alignment code
This is one instruction less for thumb, and only have got 1/2 arm/thumb specific instructions. Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Martin Storsjö | 52d196fb30 |
arm: vp9itxfm: Simplify txfm string comparisons
Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |
Martin Storsjö | a67ae67083 |
arm: vp9: Add NEON itxfm routines
This work is sponsored by, and copyright, Google. For the transforms up to 8x8, we can fit all the data (including temporaries) in registers and just do a straightforward transform of all the data. For 16x16, we do a transform of 4x16 pixels in 4 slices, using a temporary buffer. For 32x32, we transform 4x32 pixels at a time, in two steps of 4x16 pixels each. Examples of relative speedup compared to the C version, from checkasm: Cortex A7 A8 A9 A53 vp9_inv_adst_adst_4x4_add_neon: 3.39 5.83 4.17 4.01 vp9_inv_adst_adst_8x8_add_neon: 3.79 4.86 4.23 3.98 vp9_inv_adst_adst_16x16_add_neon: 3.33 4.36 4.11 4.16 vp9_inv_dct_dct_4x4_add_neon: 4.06 6.16 4.59 4.46 vp9_inv_dct_dct_8x8_add_neon: 4.61 6.01 4.98 4.86 vp9_inv_dct_dct_16x16_add_neon: 3.35 3.44 3.36 3.79 vp9_inv_dct_dct_32x32_add_neon: 3.89 3.50 3.79 4.42 vp9_inv_wht_wht_4x4_add_neon: 3.22 5.13 3.53 3.77 Thus, the speedup vs C code is around 3-6x. This is mostly marginally faster than the corresponding routines in libvpx on most cores, tested with their 32x32 idct (compared to vpx_idct32x32_1024_add_neon). These numbers are slightly in libvpx's favour since their version doesn't clear the input buffer like ours do (although the effect of that on the total runtime probably is negligible.) Cortex A7 A8 A9 A53 vp9_inv_dct_dct_32x32_add_neon: 18436.8 16874.1 14235.1 11988.9 libvpx vpx_idct32x32_1024_add_neon 20789.0 13344.3 15049.9 13030.5 Only on the Cortex A8, the libvpx function is faster. On the other cores, ours is slightly faster even though ours has got source block clearing integrated. Signed-off-by: Martin Storsjö <martin@martin.st> |
8 years ago |