[FFmpeg-devel] [PATCH] G.729A (floating-point) decoder and ACT demuxer
Michael Niedermayer
michaelni
Sun Feb 24 15:06:05 CET 2008
On Sat, Feb 23, 2008 at 11:08:25PM +0600, Vladimir Voroshilov wrote:
> Hi, All
>
> I've made first draft patches which adds G.729A (floating-point) codec
> and ACT file format support.
>
> ACT format is trivial so patch is short and should not have any issues.
>
> My implementation of G729A does not provide binary identical (with
> reference decoder) files on
> ITU's test vectors but i can't hear any differences between my
> and reference floating point decoder.
>
> I want to collect comments and fix overlooked issues/errors.
>
> Perhaps, code contains bugs/stupid things which i cant see, but
> there should not be any major issues.
First review below, this will need quite a few more patch resend-review passes
though.
>
> All sample ACT files from MPlayer collection works fine on my PC.
>
> P.S. g729a patch contains debugging
> code (under G729A_NATIVE #ifdef which should be removed in final version)
>
> P.P.S. act patch depends on g729a.
[...]
> +/*
> +
> +Validation results on ITU test vectors for Fixed-point G729.A:
> +
> + All vectors are not bit-exactly equal (but reference C code for
> + floating point implementation fails to provide bit-exact equal files
> + while fixed-point code passes all tests). Thus test was based
> + on presence of hearable artefacts/differences
PSNR / standard deviation would be nice instead of just
"hearable artefacts/differences"
[...]
> +#define MA_NP 4
what is MA_NP ?
> +
> +/*
> +-------------------------------------------------------------------------------
> + Formats description
> +-------------------------------------------------------------------------------
> +*/
> +/// Number of pulses in fixed-codebook vector
> +#define FC_PULSE_COUNT 4
> +/**
> + * GA codebook (3.9.2)
> + */
> +#define GA_BITS 3
> +/*
> + * GB codebook (3.9.2)
> + */
> +#define GB_BITS 4
!doxy
[...]
> + int bad_pitch; ///< parity check failed
> + float* exc_base; ///< past excitation signal buffer
vertical align
[...]
> +/**
> + * L1 codebook (10-dimensional, with 128 entries (3.2.4)
> + */
> +static const float cb_L1[128][10] = {
> + {0.1814, 0.2647, 0.4580, 1.1077, 1.4813, 1.7022, 2.1953, 2.3405, 2.5867, 2.6636},
> + {0.2113, 0.3223, 0.4212, 0.5946, 0.7479, 0.9615, 1.9097, 2.1750, 2.4773, 2.6737},
> + {0.1915, 0.2755, 0.3770, 0.5950, 1.3505, 1.6349, 2.2348, 2.3552, 2.5768, 2.6540},
> + {0.2116, 0.3067, 0.4099, 0.5748, 0.8518, 1.2569, 2.0782, 2.1920, 2.3371, 2.4842},
> + {0.2129, 0.2974, 0.4039, 1.0659, 1.2735, 1.4658, 1.9061, 2.0312, 2.6074, 2.6750},
> + {0.2181, 0.2893, 0.4117, 0.5519, 0.8295, 1.5825, 2.1575, 2.3179, 2.5458, 2.6417},
> + {0.1991, 0.2971, 0.4104, 0.7725, 1.3073, 1.4665, 1.6208, 1.6973, 2.3732, 2.5743},
> + {0.1818, 0.2886, 0.4018, 0.7630, 1.1264, 1.2699, 1.6899, 1.8650, 2.1633, 2.6186},
> + {0.2282, 0.3093, 0.4243, 0.5329, 1.1173, 1.7717, 1.9420, 2.0780, 2.5160, 2.6137},
> + {0.2528, 0.3693, 0.5290, 0.7146, 0.9528, 1.1269, 1.2936, 1.9589, 2.4548, 2.6653},
> + {0.2332, 0.3263, 0.4174, 0.5202, 1.3633, 1.8447, 2.0236, 2.1474, 2.3572, 2.4738},
> + {0.1393, 0.2216, 0.3204, 0.5644, 0.7929, 1.1705, 1.7051, 2.0054, 2.3623, 2.5985},
> + {0.2677, 0.3871, 0.5746, 0.7091, 1.3311, 1.5260, 1.7288, 1.9122, 2.5787, 2.6598},
> + {0.1570, 0.2328, 0.3111, 0.4216, 1.1688, 1.4605, 1.9505, 2.1173, 2.4038, 2.7460},
> + {0.2346, 0.3321, 0.5621, 0.8160, 1.4042, 1.5860, 1.7518, 1.8631, 2.0749, 2.5380},
> + {0.2505, 0.3368, 0.4758, 0.6405, 0.8104, 1.2533, 1.9329, 2.0526, 2.2155, 2.6459},
> + {0.2196, 0.3049, 0.6857, 1.3976, 1.6100, 1.7958, 2.0813, 2.2211, 2.4789, 2.5857},
> + {0.1232, 0.2011, 0.3527, 0.6969, 1.1647, 1.5081, 1.8593, 2.2576, 2.5594, 2.6896},
> + {0.3682, 0.4632, 0.6600, 0.9118, 1.5245, 1.7071, 1.8712, 1.9939, 2.4356, 2.5380},
> + {0.2690, 0.3711, 0.4635, 0.6644, 1.4633, 1.6495, 1.8227, 1.9983, 2.1797, 2.2954},
> + {0.3555, 0.5240, 0.9751, 1.1685, 1.4114, 1.6168, 1.7769, 2.0178, 2.4420, 2.5724},
> + {0.3493, 0.4404, 0.7231, 0.8587, 1.1272, 1.4715, 1.6760, 2.2042, 2.4735, 2.5604},
> + {0.3747, 0.5263, 0.7284, 0.8994, 1.4017, 1.5502, 1.7468, 1.9816, 2.2380, 2.3404},
> + {0.2972, 0.4470, 0.5941, 0.7078, 1.2675, 1.4310, 1.5930, 1.9126, 2.3026, 2.4208},
> + {0.2467, 0.3180, 0.4712, 1.1281, 1.6206, 1.7876, 1.9544, 2.0873, 2.3521, 2.4721},
> + {0.2292, 0.3430, 0.4383, 0.5747, 1.3497, 1.5187, 1.9070, 2.0958, 2.2902, 2.4301},
> + {0.2573, 0.3508, 0.4484, 0.7079, 1.6577, 1.7929, 1.9456, 2.0847, 2.3060, 2.4208},
> + {0.1968, 0.2789, 0.3594, 0.4361, 1.0034, 1.7040, 1.9439, 2.1044, 2.2696, 2.4558},
> + {0.2955, 0.3853, 0.7986, 1.2470, 1.4723, 1.6522, 1.8684, 2.0084, 2.2849, 2.4268},
> + {0.2036, 0.3189, 0.4314, 0.6393, 1.2834, 1.4278, 1.5796, 2.0506, 2.2044, 2.3656},
> + {0.2916, 0.3684, 0.5907, 1.1394, 1.3933, 1.5540, 1.8341, 1.9835, 2.1301, 2.2800},
> + {0.2289, 0.3402, 0.5166, 0.7716, 1.0614, 1.2389, 1.4386, 2.0769, 2.2715, 2.4366},
> + {0.0829, 0.1723, 0.5682, 0.9773, 1.3973, 1.6174, 1.9242, 2.2128, 2.4855, 2.6327},
> + {0.2244, 0.3169, 0.4368, 0.5625, 0.6897, 1.3763, 1.7524, 1.9393, 2.5121, 2.6556},
> + {0.1591, 0.2387, 0.2924, 0.4056, 1.4677, 1.6802, 1.9389, 2.2067, 2.4635, 2.5919},
> + {0.1756, 0.2566, 0.3251, 0.4227, 1.0167, 1.2649, 1.6801, 2.1055, 2.4088, 2.7276},
> + {0.1050, 0.2325, 0.7445, 0.9491, 1.1982, 1.4658, 1.8093, 2.0397, 2.4155, 2.5797},
> + {0.2043, 0.3324, 0.4522, 0.7477, 0.9361, 1.1533, 1.6703, 1.7631, 2.5071, 2.6528},
> + {0.1522, 0.2258, 0.3543, 0.5504, 0.8815, 1.5516, 1.8110, 1.9915, 2.3603, 2.7735},
> + {0.1862, 0.2759, 0.4715, 0.6908, 0.8963, 1.4341, 1.6322, 1.7630, 2.2027, 2.6043},
> + {0.1460, 0.2254, 0.3790, 0.8622, 1.3394, 1.5754, 1.8084, 2.0798, 2.4319, 2.7632},
> + {0.2621, 0.3792, 0.5463, 0.7948, 1.0043, 1.1921, 1.3409, 1.4845, 2.3159, 2.6002},
> + {0.1935, 0.2937, 0.3656, 0.4927, 1.4015, 1.6086, 1.7724, 1.8837, 2.4374, 2.5971},
> + {0.2171, 0.3282, 0.4412, 0.5713, 1.1554, 1.3506, 1.5227, 1.9923, 2.4100, 2.5391},
> + {0.2274, 0.3157, 0.4263, 0.8202, 1.4293, 1.5884, 1.7535, 1.9688, 2.3939, 2.4934},
> + {0.1704, 0.2633, 0.3259, 0.4134, 1.2948, 1.4802, 1.6619, 2.0393, 2.3165, 2.6083},
> + {0.1763, 0.2585, 0.4012, 0.7609, 1.1503, 1.5847, 1.8309, 1.9352, 2.0982, 2.6681},
> + {0.2447, 0.3535, 0.4618, 0.5979, 0.7530, 0.8908, 1.5393, 2.0075, 2.3557, 2.6203},
> + {0.1826, 0.3496, 0.7764, 0.9888, 1.3915, 1.7421, 1.9412, 2.1620, 2.4999, 2.6931},
> + {0.3033, 0.3802, 0.6981, 0.8664, 1.0254, 1.5401, 1.7180, 1.8124, 2.5068, 2.6119},
> + {0.2960, 0.4001, 0.6465, 0.7672, 1.3782, 1.5751, 1.9559, 2.1373, 2.3601, 2.4760},
> + {0.3132, 0.4613, 0.6544, 0.8532, 1.0721, 1.2730, 1.7566, 1.9217, 2.1693, 2.6531},
> + {0.3329, 0.4131, 0.8073, 1.1297, 1.2869, 1.4937, 1.7885, 1.9150, 2.4505, 2.5760},
> + {0.2340, 0.3605, 0.7659, 0.9874, 1.1854, 1.3337, 1.5128, 2.0062, 2.4427, 2.5859},
> + {0.4131, 0.5330, 0.6530, 0.9360, 1.3648, 1.5388, 1.6994, 1.8707, 2.4294, 2.5335},
> + {0.3754, 0.5229, 0.7265, 0.9301, 1.1724, 1.3440, 1.5118, 1.7098, 2.5218, 2.6242},
> + {0.2138, 0.2998, 0.6283, 1.2166, 1.4187, 1.6084, 1.7992, 2.0106, 2.5377, 2.6558},
> + {0.1761, 0.2672, 0.4065, 0.8317, 1.0900, 1.4814, 1.7672, 1.8685, 2.3969, 2.5079},
> + {0.2801, 0.3535, 0.4969, 0.9809, 1.4934, 1.6378, 1.8021, 2.1200, 2.3135, 2.4034},
> + {0.2365, 0.3246, 0.5618, 0.8176, 1.1073, 1.5702, 1.7331, 1.8592, 1.9589, 2.3044},
> + {0.2529, 0.3251, 0.5147, 1.1530, 1.3291, 1.5005, 1.7028, 1.8200, 2.3482, 2.4831},
> + {0.2125, 0.3041, 0.4259, 0.9935, 1.1788, 1.3615, 1.6121, 1.7930, 2.5509, 2.6742},
> + {0.2685, 0.3518, 0.5707, 1.0410, 1.2270, 1.3927, 1.7622, 1.8876, 2.0985, 2.5144},
> + {0.2373, 0.3648, 0.5099, 0.7373, 0.9129, 1.0421, 1.7312, 1.8984, 2.1512, 2.6342},
> + {0.2229, 0.3876, 0.8621, 1.1986, 1.5655, 1.8861, 2.2376, 2.4239, 2.6648, 2.7359},
> + {0.3009, 0.3719, 0.5887, 0.7297, 0.9395, 1.8797, 2.0423, 2.1541, 2.5132, 2.6026},
> + {0.3114, 0.4142, 0.6476, 0.8448, 1.2495, 1.7192, 2.2148, 2.3432, 2.5246, 2.6046},
> + {0.3666, 0.4638, 0.6496, 0.7858, 0.9667, 1.4213, 1.9300, 2.0564, 2.2119, 2.3170},
> + {0.4218, 0.5075, 0.8348, 1.0009, 1.2057, 1.5032, 1.9416, 2.0540, 2.4352, 2.5504},
> + {0.3726, 0.4602, 0.5971, 0.7093, 0.8517, 1.2361, 1.8052, 1.9520, 2.4137, 2.5518},
> + {0.4482, 0.5318, 0.7114, 0.8542, 1.0328, 1.4751, 1.7278, 1.8237, 2.3496, 2.4931},
> + {0.3316, 0.4498, 0.6404, 0.8162, 1.0332, 1.2209, 1.5130, 1.7250, 1.9715, 2.4141},
> + {0.2375, 0.3221, 0.5042, 0.9760, 1.7503, 1.9014, 2.0822, 2.2225, 2.4689, 2.5632},
> + {0.2813, 0.3575, 0.5032, 0.5889, 0.6885, 1.6040, 1.9318, 2.0677, 2.4546, 2.5701},
> + {0.2198, 0.3072, 0.4090, 0.6371, 1.6365, 1.9468, 2.1507, 2.2633, 2.5063, 2.5943},
> + {0.1754, 0.2716, 0.3361, 0.5550, 1.1789, 1.3728, 1.8527, 1.9919, 2.1349, 2.3359},
> + {0.2832, 0.3540, 0.6080, 0.8467, 1.0259, 1.6467, 1.8987, 1.9875, 2.4744, 2.5527},
> + {0.2670, 0.3564, 0.5628, 0.7172, 0.9021, 1.5328, 1.7131, 2.0501, 2.5633, 2.6574},
> + {0.2729, 0.3569, 0.6252, 0.7641, 0.9887, 1.6589, 1.8726, 1.9947, 2.1884, 2.4609},
> + {0.2155, 0.3221, 0.4580, 0.6995, 0.9623, 1.2339, 1.6642, 1.8823, 2.0518, 2.2674},
> + {0.4224, 0.7009, 1.1714, 1.4334, 1.7595, 1.9629, 2.2185, 2.3304, 2.5446, 2.6369},
> + {0.4560, 0.5403, 0.7568, 0.8989, 1.1292, 1.7687, 1.9575, 2.0784, 2.4260, 2.5484},
> + {0.4299, 0.5833, 0.8408, 1.0596, 1.5524, 1.7484, 1.9471, 2.2034, 2.4617, 2.5812},
> + {0.2614, 0.3624, 0.8381, 0.9829, 1.2220, 1.6064, 1.8083, 1.9362, 2.1397, 2.2773},
> + {0.5064, 0.7481, 1.1021, 1.3271, 1.5486, 1.7096, 1.9503, 2.1006, 2.3911, 2.5141},
> + {0.5375, 0.6552, 0.8099, 1.0219, 1.2407, 1.4160, 1.8266, 1.9936, 2.1951, 2.2911},
> + {0.4994, 0.6575, 0.8365, 1.0706, 1.4116, 1.6224, 1.9200, 2.0667, 2.3262, 2.4539},
> + {0.3353, 0.4426, 0.6469, 0.9161, 1.2528, 1.3956, 1.6080, 1.8909, 2.0600, 2.1380},
> + {0.2745, 0.4341, 1.0424, 1.2928, 1.5461, 1.7940, 2.0161, 2.1758, 2.4742, 2.5937},
> + {0.1562, 0.2393, 0.4786, 0.9513, 1.2395, 1.8010, 2.0320, 2.2143, 2.5243, 2.6204},
> + {0.2979, 0.4242, 0.8224, 1.0564, 1.4881, 1.7808, 2.0898, 2.1882, 2.3328, 2.4389},
> + {0.2294, 0.3070, 0.5490, 0.9244, 1.2229, 1.8248, 1.9704, 2.0627, 2.2458, 2.3653},
> + {0.3423, 0.4502, 0.9144, 1.2313, 1.3694, 1.5517, 1.9907, 2.1326, 2.4509, 2.5789},
> + {0.2470, 0.3275, 0.4729, 1.0093, 1.2519, 1.4216, 1.8540, 2.0877, 2.3151, 2.4156},
> + {0.3447, 0.4401, 0.7099, 1.0493, 1.2312, 1.4001, 2.0225, 2.1317, 2.2894, 2.4263},
> + {0.3481, 0.4494, 0.6446, 0.9336, 1.1198, 1.2620, 1.8264, 1.9712, 2.1435, 2.2552},
> + {0.1646, 0.3229, 0.7112, 1.0725, 1.2964, 1.5663, 1.9843, 2.2363, 2.5798, 2.7572},
> + {0.2614, 0.3707, 0.5241, 0.7425, 0.9269, 1.2976, 2.0945, 2.2014, 2.6204, 2.6959},
> + {0.1963, 0.2900, 0.4131, 0.8397, 1.2171, 1.3705, 2.0665, 2.1546, 2.4640, 2.5782},
> + {0.3387, 0.4415, 0.6121, 0.8005, 0.9507, 1.0937, 2.0836, 2.2342, 2.3849, 2.5076},
> + {0.2362, 0.5876, 0.7574, 0.8804, 1.0961, 1.4240, 1.9519, 2.1742, 2.4935, 2.6493},
> + {0.2793, 0.4282, 0.6149, 0.8352, 1.0106, 1.1766, 1.8392, 2.0119, 2.6433, 2.7117},
> + {0.3603, 0.4604, 0.5955, 0.9251, 1.1006, 1.2572, 1.7688, 1.8607, 2.4687, 2.5623},
> + {0.3975, 0.5849, 0.8059, 0.9182, 1.0552, 1.1850, 1.6356, 1.9627, 2.3318, 2.4719},
> + {0.2231, 0.3192, 0.4256, 0.7373, 1.4831, 1.6874, 1.9765, 2.1097, 2.6152, 2.6906},
> + {0.1221, 0.2081, 0.3665, 0.7734, 1.0341, 1.2818, 1.8162, 2.0727, 2.4446, 2.7377},
> + {0.2010, 0.2791, 0.3796, 0.8845, 1.4030, 1.5615, 2.0538, 2.1567, 2.3171, 2.4686},
> + {0.2086, 0.3053, 0.4047, 0.8224, 1.0656, 1.2115, 1.9641, 2.0871, 2.2430, 2.4313},
> + {0.3203, 0.4285, 0.5467, 0.6891, 1.2039, 1.3569, 1.8578, 2.2055, 2.3906, 2.4881},
> + {0.3074, 0.4192, 0.5772, 0.7799, 0.9866, 1.1335, 1.6068, 2.2441, 2.4194, 2.5089},
> + {0.2108, 0.2910, 0.4993, 0.7695, 0.9528, 1.5681, 1.7838, 2.1495, 2.3522, 2.4636},
> + {0.3492, 0.4560, 0.5906, 0.7379, 0.8855, 1.0257, 1.7128, 1.9997, 2.2019, 2.3694},
> + {0.5185, 0.7316, 0.9708, 1.1954, 1.5066, 1.7887, 2.1396, 2.2918, 2.5429, 2.6489},
> + {0.4276, 0.4946, 0.6934, 0.8308, 0.9944, 1.4582, 2.0324, 2.1294, 2.4891, 2.6324},
> + {0.3847, 0.5973, 0.7202, 0.8787, 1.3938, 1.5959, 1.8463, 2.1574, 2.5050, 2.6687},
> + {0.4835, 0.5919, 0.7235, 0.8862, 1.0756, 1.2853, 1.9118, 2.0215, 2.2213, 2.4638},
> + {0.5492, 0.8062, 0.9810, 1.1293, 1.3189, 1.5415, 1.9385, 2.1378, 2.4439, 2.5691},
> + {0.5190, 0.6764, 0.8123, 1.0154, 1.2085, 1.4266, 1.8433, 2.0866, 2.5113, 2.6474},
> + {0.4602, 0.6503, 0.9602, 1.1427, 1.3043, 1.4427, 1.6676, 1.8758, 2.2868, 2.4271},
> + {0.3764, 0.4845, 0.7627, 0.9914, 1.1961, 1.3421, 1.5129, 1.6707, 2.1836, 2.3322},
> + {0.3334, 0.5701, 0.8622, 1.1232, 1.3851, 1.6767, 2.0600, 2.2946, 2.5375, 2.7295},
> + {0.1449, 0.2719, 0.5783, 0.8807, 1.1746, 1.5422, 1.8804, 2.1934, 2.4734, 2.8728},
> + {0.2333, 0.3024, 0.4780, 1.2327, 1.4180, 1.5815, 1.9804, 2.0921, 2.3524, 2.5304},
> + {0.2154, 0.3075, 0.4746, 0.8477, 1.1170, 1.5369, 1.9847, 2.0733, 2.1880, 2.2504},
> + {0.1709, 0.4486, 0.8705, 1.0643, 1.3047, 1.5269, 1.9175, 2.1621, 2.4073, 2.5718},
> + {0.2835, 0.3752, 0.5234, 0.9898, 1.1484, 1.2974, 1.9363, 2.0378, 2.4065, 2.6214},
> + {0.3211, 0.4077, 0.5809, 1.0206, 1.2542, 1.3835, 1.5723, 2.1209, 2.3464, 2.4336},
> + {0.2101, 0.3146, 0.6779, 0.8783, 1.0561, 1.3045, 1.8395, 2.0695, 2.2831, 2.4328},
> +};
*10000 and it fits in uint16_t
> +/**
> + * L2 codebook (10-dimensional, with 32 entries (3.2.4)
> + */
> +static const float cb_L2[32][5] = {
Same as above, change it to int16_t please.
[...]
> +/**
> + * MA predictor (3.2.4)
> + */
> +static const float ma_predictor[2][4][10] = {
> + {
> + { 0.2570, 0.2780, 0.2800, 0.2736, 0.2757, 0.2764, 0.2675, 0.2678, 0.2779, 0.2647},
> + { 0.2142, 0.2194, 0.2331, 0.2230, 0.2272, 0.2252, 0.2148, 0.2123, 0.2115, 0.2096},
> + { 0.1670, 0.1523, 0.1567, 0.1580, 0.1601, 0.1569, 0.1589, 0.1555, 0.1474, 0.1571},
> + { 0.1238, 0.0925, 0.0798, 0.0923, 0.0890, 0.0828, 0.1010, 0.0988, 0.0872, 0.1060},
> + },
> + {
> + { 0.2360, 0.2405, 0.2499, 0.2495, 0.2517, 0.2591, 0.2636, 0.2625, 0.2551, 0.2310},
> + { 0.1285, 0.0925, 0.0779, 0.1060, 0.1183, 0.1176, 0.1277, 0.1268, 0.1193, 0.1211},
> + { 0.0981, 0.0589, 0.0401, 0.0654, 0.0761, 0.0728, 0.0841, 0.0826, 0.0776, 0.0891},
> + { 0.0923, 0.0486, 0.0287, 0.0498, 0.0526, 0.0482, 0.0621, 0.0636, 0.0584, 0.0794}
> + }
> +};
this also nicely fits in uint16_t
> +
> +/**
> + * ma_predicot_sum[i] := 1-sum{1}{4}{ma_predictor[k][i]}
> + */
> +static const float ma_predictor_sum[2][10] = {
> + { 0.2380000054836, 0.2578000128269, 0.2504000067711, 0.2531000375748, 0.2480000108480, 0.2587000429630, 0.2577999532223, 0.2656000256538, 0.2760000228882, 0.2625999450684},
> + { 0.4451000094414, 0.5595000386238, 0.6034000515938, 0.5292999744415, 0.5012999176979, 0.5023000240326, 0.4625000357628, 0.4645000100136, 0.4895999729633, 0.4793999791145}
> +};
These values do not match the doxygen comment.
> +
> +/**
> + * MA prediction coefficients (3.9.1, near Equation 69)
> + */
> +static const float ma_prediction_coeff[4] =
> +{
> + 0.68, 0.58, 0.34, 0.19
> +};
these even fit in uint8_t
[...]
> +/**
> + * \brief rouding function from reference code
> + *
> + * FIXME: found system replacement for it
> + */
> +static inline int g729_round(float f)
> +{
> + int i=ceil(f*2);
> + return i>>1;
> +}
remove this function
> +
> +/**
> + * \brief pseudo random number generator
> + */
> +static inline uint16_t g729_random(G729A_Context* ctx)
> +{
> + return ctx->rand_seed = (uint16_t)(31821 * (uint32_t)ctx->rand_seed + 13849 + ctx->rand_seed);
> +}
loose the casts
rand_seed is a bad name its a state not seed
and its wrong,its
state= state*31821 + 13849;
> +
> +
> +/**
> + * \brief Check parity bit (3.7.2)
> + * \param P1 Pitch delay first subframe
> + * \param P0 Parity bit for Pitch delay
> + *
> + * \return 1 if parity check is ok, 0 - otherwise
> + */
> +int g729_parity_check(int P1, int P0)
> +{
> + int P=P1>>2;
> + int S=P0&1;
> + int i;
> +
> + for(i=0; i<6; i++)
> + {
> + S ^= P&1;
> + P >>= 1;
> + }
> + S ^= 1;
> + return (!S);
> +}
ehh?
you have 6 bits, to find the parity of 6bit you do:
P1>>=2;
return ((0xMAGIC ULL>>P1)^P0)&1;
> +
> +/**
> + * \brief Decoding of the adaptive-codebook vector delay for first subframe (4.1.3)
> + * \param ctx private data structure
> + * \param P1 Pitch delay first subframe
> + * \param intT [out] integer part of delay
> + * \param frac [out] fractional part of delay [-1, 0, 1]
> + */
> +static void g729_decode_ac_delay_subframe1(G729A_Context* ctx, int P1, int* intT, int* frac)
> +{
> + /* if no parity error */
> + if(!ctx->bad_pitch)
> + {
> + if(P1<197)
> + {
> + *intT=1.0*(P1+2)/3+19;
> + *frac=P1-3*(*intT)+58;
remove the float (1.0) use proper / and %
> + }
> + else
> + {
> + *intT=P1-112;
> + *frac=0;
> + }
> + }
> + else{
> + *intT=ctx->intT2_prev;
> + *frac=0;
> + }
> + ctx->intT1=*intT;
> +}
your {} placement is inconsistant
Use the return value of your functions, avoid returning values through
pointers. Its harder to read
Also i must say that the reference implemetation is cleaner than
your code
if intT and frac are integer and frac of something they should have
matching names, the ref impl uses T0 T0_frac, now we shouldnt copy
that but still.
P1 which you document as "Pitch delay first subframe"
is called index and documented as "input : received pitch index"
in the ref impl.
> +
> +/**
> + * \brief Decoding of the adaptive-codebook vector delay for second subframe (4.1.3)
> + * \param ctx private data structure
> + * \param P1 Pitch delay second subframe
> + * \param T1 first subframe's vector delay integer part
> + * \param intT [out] integer part of delay
> + * \param frac [out] fractional part of delay [-1, 0, 1]
> + */
> +static void g729_decode_ac_delay_subframe2(G729A_Context* ctx, int P2, int intT1, int* intT, int* frac)
> +{
> +
> + int tmin=FFMIN(FFMAX(intT1-5, PITCH_MIN)+9, PITCH_MAX)-9;
int tmin=FFMIN(FFMAX(intT1-5, PITCH_MIN), PITCH_MAX-9);
> +
> + if(ctx->data_error)
> + {
> + *intT=intT1;
> + *frac=0;
> + ctx->intT2_prev=FFMIN(intT1+1, PITCH_MAX);
> + return;
> + }
> +
> + *intT=(P2+2)/3-1;
> + *frac=P2-2-3*(*intT);
*T_int = (P2+2)/3 - 1;
*T_frac= (P2+2)%3 - 1;
> +
> + *intT+=tmin;
> +
> + ctx->intT2_prev=*intT;
> +}
as already said use the return value!
> +
> +/**
> + * \brief Decoding of the adaptive-codebook vector (4.1.3)
> + * \param ctx private data structure
> + * \param k pitch delay, integer part
> + * \param t pitch delay, fraction part [-1, 0, 1]
Do NOT use meaningless single letter variable names, if you mean integer
and fraction choose names somehow related to the 2 words.
> + * \param ac_v buffer to store decoded vector into
> + */
> +static void g729_decode_ac_vector(G729A_Context* ctx, int k, int t, float* ac_v)
> +{
> + int n, i;
> + float v;
> +
> + //Make sure that t will be always positive
> + t=-t;
> + if(t<0)
> + {
> + t+=3;
> + k++;
> + }
> +
> + //t [0, 1, 2]
> + //k [PITCH_MIN-1; PITCH_MAX]
> + for(n=0; n<ctx->subframe_size; n++)
> + {
> + /* 3.7.1, Equation 40 */
> + v=0;
> + for(i=0; i<10; i++)
> + {
> + /* R(x):=ac_v[-k+x] */
> + v+=ac_v[n-k-i]*b30[t+3*i]; //R(n-i)*b30(t+3i)
> + v+=ac_v[n-k+i+1]*b30[3-t+3*i]; //R(n+i+1)*b30(3-t+3i)
vertical align please, and rename b30 to something remotely sane.
hammsinc would be a random idea, interpol_filter would be another
> + }
> + ac_v[n]=v;
> + }
> +}
> +
> +/**
> + * \brief Decoding fo the fixed-codebook vector (3.8)
> + * \param ctx private data structure
> + * \param C Fixed codebook
> + * \param S Signs of fixed-codebook pulses (0 bit value means negative sign)
use meaningfull variable names and upper case should only be used for #defines
> + * \param fc_v [out] decoded fixed codebook vector
> + *
> + * bit allocations:
> + * 8k mode: 3+3+3+1+3
> + * 4.4k mode: 4+4+4+1+4 (non-standard)
> + *
> + * FIXME: error handling required
> + */
> +static void g729_decode_fc_vector(G729A_Context* ctx, int C, int S, float* fc_v)
> +{
> + int accC=C;
> + int accS=S;
> + int i;
> + int index;
> + int bits=formats[ctx->format].fc_index_bits;
all values which are the same in all formats should be replaced by named
constants (#define FC_INDEX_BITS or a better name) theres no sense in
having them in these structs
[...]
> +/**
> + * \brief fixed codebook vector modification if delay is less than 40 (4.1.4 and 3.8)
> + * \param T pitch delay to check
> + * \param fc_v [in/out] fixed codebook vector to change
> + *
> + * \remark if T>=subframe_size no changes to vector are made
> + */
> +static void g729_fix_fc_vector(G729A_Context *ctx, int T, float* fc_v)
> +{
> + int i;
> +
> + if(T>=ctx->subframe_size)
> + return;
> +
> + for(i=T; i<ctx->subframe_size;i++)
> + fc_v[i]+=fc_v[i-T]*ctx->gain_pitch;
> +}
please use more spaces in your expressions
fc_v[i] += fc_v[i-T] * ctx->gain_pitch;
is more readable IMHO
[...]
> +/**
> + * \brief Memory update (3.10)
> + * \param ctx private data structure
> + * \param fc_v fixed-codebook vector
> + * \param gp quantized fixed-codebook gain (gain pitch)
> + * \param gc quantized adaptive-codebook gain (gain code)
> + * \param exc last excitation signal buffer for current subframe
> + */
> +static void g729_mem_update(G729A_Context *ctx, float *fc_v, float gp, float gc, float* exc)
> +{
> + int i;
> +
> + for(i=0; i<ctx->subframe_size; i++)
> + exc[i]=exc[i]*gp+fc_v[i]*gc;
spaces, please!
exc[i] = exc[i] * gp + fc_v[i] * gc;
or
exc[i]= exc[i]*gp + fc_v[i]*gc;
Just examples, pick what you like but please dont pick random spaces
or no spaces ;)
> +}
> +
> +/**
> + * \brief LP synthesis filter
> + * \param ctx private data structure
> + * \param lp filter coefficients
> + * \param in input signal
> + * \param out output (filtered) signal
> + * \param filter_data filter data array (previous synthesis data)
> + *
> + * Routine applies 1/A(z) filter to given speech data
> + *
> + */
> +static void g729_lp_synthesis_filter(G729A_Context *ctx, float* lp, float *in, float *out, float *filter_data)
> +{
> + float* tmp_buf=av_mallocz((10+ctx->subframe_size)*sizeof(float));
> + float* tmp=tmp_buf+10;
> + int i,n;
> +
> + memcpy(tmp_buf, filter_data, 10*sizeof(float));
> +
> + for(n=0; n<ctx->subframe_size; n++)
> + {
> + tmp[n]=in[n];
> + for(i=0; i<10; i++)
> + tmp[n]-= lp[i]*tmp[n-i-1];
> + }
> + memcpy(filter_data, tmp+ctx->subframe_size-10, 10*sizeof(float));
> + memcpy(out, tmp, ctx->subframe_size*sizeof(float));
> +}
did you plan to free the allocated memory?
float tmp_buf[0+ctx->subframe_size];
solves this nicely
> +
> +/**
> + * \brief Calculates gain value of speech signal
> + * \param ctx private data structure
> + * \param speech signal buffer
> + *
> + * \return squared gain value
> + */
> +static float g729_get_signal_gain(G729A_Context *ctx, float *speech)
> +{
> + int n;
> + float gain;
> +
> + gain=0;
can be merged
[...]
> +/**
> + * \brief Adaptive gain control (4.2.4)
> + * \param ctx private data structure
> + * \param gain_before gain of speech before applying postfilters
> + * \param gain_after gain of speech after applying postfilters
> + * \param speech signal buffer
> + */
> +static void g729a_adaptive_gain_control(G729A_Context *ctx, float gain_before, float gain_after, float *speech)
> +{
> + float gain;
> + int n;
> +
> + if(!gain_after)
> + return;
testing floats for equallity is risky ...
[...]
> +/**
> + * \brief long-term postfilter (4.2.1)
> + * \param ctx private data structure
> + * \param residual_filt speech signal with applied A(z/GAMMA_N) filter
> + */
> +static void g729a_long_term_filter(G729A_Context *ctx, float *residual_filt)
> +{
> + int k, n, intT0;
> + float gl; ///< gain coefficient for long-term postfilter
> + float corr_t0; ///< corellation of residual signal with delay intT0
> + float corr_0; ///< corellation of residual signal with delay 0
> + float corellation, corr_max;
> + float inv_glgp;///< 1.0/(1+gl*GAMMA_P)
> + float glgp_inv_glgp; ///< gl*GAMMA_P/(1+gl*GAMMA_P);
> +
i dont think doxy works with local vars
[...]
> + /*
> + First pass: searching the best T0 (pitch delay)
> + Second pass is not used in G.729A: fractional part is always zero
> + */
> + k=minT0;
> + corellation=0;
> + /* 4.2.1, Equation 80 */
> + for(n=0; n<ctx->subframe_size; n++)
> + corellation+=ctx->residual[n+PITCH_MAX]*ctx->residual[n+PITCH_MAX-k];
> +
> + corr_max=corellation;
> + intT0=k;
> +
> + for(; k<=maxT0; k++)
> + {
> + corellation=0;
> + /* 4.2.1, Equation 80 */
> + for(n=0; n<ctx->subframe_size; n++)
> + corellation+=ctx->residual[n+PITCH_MAX]*ctx->residual[n+PITCH_MAX-k];
> + if(corellation>corr_max)
> + {
> + corr_max=corellation;
> + intT0=k;
> + }
> + }
the first loop does nothing, remove it
[...]
> + inv_glgp=1.0/(1+gl*GAMMA_P);
> + glgp_inv_glgp=gl*GAMMA_P/(1+gl*GAMMA_P);
glgp_inv_glgp= gl * GAMMA_P * inv_glgp;
[...]
> + for(i=11; i<22;i++)
> + hf[i]=0;
> +
> + /* Applying 1/A(z/GAMMA_D) to hf */
> + for(i=0; i<10; i++)
> + tmp_buf[i]=hf[i+11];
for(i=0; i<11; i++)
tmp_buf[i]= hf[i+11]= 0;
[...]
> +/**
> + * \brief high-pass filtering and upscaling (4.2.5)
> + * \param ctx private data structure
> + * \param speech reconstructed speech signal for applying filter to
> + *
> + * Filter has cut-off frequency 100Hz
> + */
> +static void g729_high_pass_filter(G729A_Context* ctx, float* speech)
> +{
> + const float az[3] = {0.93980581, -1.8795834, 0.93980581};
> + const float af[3] = {1.00000000, 1.9330735, -0.93589199};
> + float z_2=0;
> +
> + float f_0=0;
> + int i;
> +
> + for(i=0; i<ctx->subframe_size; i++)
> + {
> + z_2=ctx->hpf_z1;
> + ctx->hpf_z1=ctx->hpf_z0;
> + ctx->hpf_z0=speech[i];
> +
> + f_0 = ctx->hpf_f1*af[1]+ctx->hpf_f2*af[2] + ctx->hpf_z0*az[0]+ctx->hpf_z1*az[1]+z_2*az[2];
the az/af arrays are unneeded you can just write the constants here
[...]
> + for(i=0; i<ctx->subframe_size; i++)
> + {
> + tmp_speech[i] = FFMIN(tmp_speech[i],32767.0);
> + tmp_speech[i] = FFMAX(tmp_speech[i],-32768.0);
> + speech[i]=g729_round(tmp_speech[i]);
lrintf() or float_to_int16()
[...]
> +/**
> + * \brief Decode LP coefficients from L0-L3 (3.2.4)
> + * \param ctx private data structure
> + * \param L0 Switched MA predictor of LSP quantizer
> + * \param L1 First stage vector of quantizer
> + * \param L2 Second stage lower vector of LSP quantizer
> + * \param L3 Second stage higher vector of LSP quantizer
> + * \param lsfq Decoded LSP coefficients
> + */
> +static void g729_lsp_decode(G729A_Context* ctx, int16_t L0, int16_t L1, int16_t L2, int16_t L3, float* lsfq)
> +{
> + int i,j,k;
> + float J[2]={0.0012, 0.0006};
> + float lq[10];
> + float diff;
> + float* tmp;
> +
> + /* 3.2.4 Equation 19 */
> + for(i=0;i<5; i++)
> + {
> + lq[i] =cb_L1[L1][i ] + cb_L2[L2][i];
> + lq[i+5]=cb_L1[L1][i+5] + cb_L3[L3][i];
> + }
> +
> + /* 3.2.4 rearrangement routine */
> + for(j=0; j<2; j++)
> + {
> + for(i=1; i<10; i++)
> + {
> + diff=(lq[i-1]-lq[i]+J[j])/2;
> + if(diff>0)
> + {
> + lq[i-1]-= diff;
> + lq[i] += diff;
> + }
> + }
> + }
the above can be done exactly in integers
[...]
> + /* sorting lsfq in ascending order. float bubble agorithm*/
> + for(i=0; i<5; i++)
> + {
> + float min=lsfq[i];
> + int mini=i;
> + float max=lsfq[i];
> + int maxi=i;
> + for(j=i; j< 10-i; j++)
> + {
> + if(lsfq[j]>max)
> + {
> + maxi=j;
> + max=lsfq[j];
> + }
> + if(lsfq[j]<min)
> + {
> + mini=j;
> + min=lsfq[j];
> + }
> + }
> + FFSWAP(float, lsfq[i], lsfq[mini]);
> + FFSWAP(float, lsfq[10-i-1], lsfq[maxi]);
> + }
for(j=9; j>0; j--)
for(i=0; i<j; i++)
if(lsfq[i] > lsfq[i+1])
FSWAP(float, lsfq[i], lsfq[i+1]);
[...]
> + /* 3.2.6, Equation 25*/
> + for(i=0;i<5;i++)
> + {
> + ff1[i]=f1[i+1]+f1[i];
> + ff2[i]=f2[i+1]-f2[i];
> + }
> +
> + /* 3.2.6, Equation 26*/
> + for(i=0;i<5;i++)
> + {
> + a[i]=(ff1[i]+ff2[i])/2;
> + a[i+5]=(ff1[4-i]-ff2[4-i])/2;
> + }
/* 3.2.6, Equation 25 and 26 */
for(i=0;i<5;i++){
a[ i]=
a[9-i]= (f1[i+1] + f1[i] - f2[i+1] + f2[i])*0.5;
}
[...]
> + for(ctx->format=0; formats[ctx->format].name; ctx->format++)
> + if(formats[ctx->format].sample_rate==avctx->sample_rate)
> + break;
> + if(!formats[ctx->format].name){
> + av_log(avctx, AV_LOG_ERROR, "Sample rate %d is not supported\n", avctx->sample_rate);
> + return -1;
> + }
if(avctx->sample_rate == 4400) ctx->format=1;
else ctx->format=0;
and remove formats[].name its not used
[...]
> + /* LSP coefficients */
> + for(k=0; k<MA_NP; k++)
> + ctx->lq_prev[k]=av_malloc(sizeof(float)*frame_size);
IIRC frame_size is either 10 or 11 no need for a malloc()
[...]
> + // Two subframes + PITCH_MAX inetries for last excitation signal data + ???
> + ctx->exc_base=av_mallocz((frame_size*8+PITCH_MAX+INTERPOL_LEN)*sizeof(float));
> + if(!ctx->exc_base)
> + return AVERROR(ENOMEM);
> +
> + ctx->exc=ctx->exc_base+PITCH_MAX+INTERPOL_LEN;
> +
> + ctx->residual=av_mallocz((PITCH_MAX+ctx->subframe_size)*sizeof(float));
Also same here, whatever the size, put it in the context.
[...]
> + avctx->frame_size=frame_size;
this doesnt look correct, this doesnt match the number of samples
returned from "decode"
> + return 0;
> +}
> +
> +/**
> + * G.729A decoder uninitialization
> + * \param ctx private data structure
> + */
> +static int ff_g729a_decoder_close(AVCodecContext *avctx)
> +{
> + G729A_Context *ctx=avctx->priv_data;
> + int k;
> +
> + av_free(ctx->residual);
> + ctx->residual=NULL;
av_freep
[...]
> +/**
> + * \brief decode one G.729 frame into PCM samples
> + * \param serial array if bits (0x81 - 1, 0x7F -0)
> + * \param serial_size number of items in array
> + * \param out_frame array for output PCM samples
> + * \param out_frame_size maximum number of elements in output array
> + */
> +static int g729a_decode_frame_internal(void* context, short* out_frame, int out_frame_size, int *parm)
> +{
there is no need for g729a_ prefixes
[...]
> + /* Return reconstructed speech to caller */
> + memcpy(out_frame, speech_buf, 2*ctx->subframe_size*sizeof(short));
this cpy must be removed, directly decode into the output buffer
[...]
> +/**
> + * \brief decodes one G.729 frame (10 bytes long) into parameters vector
> + * \param ctx private data structure
> + * \param buf 10 bytes of decoder parameters
> + * \param buf_size size of input buffer
> + * \param int parm output vector of decoded parameters
> + *
> + * \return 0 if success, nonzero - otherwise
> + */
> +static int g729_bytes2parm(G729A_Context *ctx, const uint8_t *buf, int buf_size, int *parm)
> +{
> + GetBitContext gb;
> + int i;
> + int l_frame=formats[ctx->format].frame_size>>3;
> +
> + if(buf_size<l_frame)
> + return AVERROR(EIO);
> +
> + init_get_bits(&gb, buf, buf_size);
> +
> + for(i=0; i<VECTOR_SIZE; i++)
> + {
> + if(formats[ctx->format].vector_bits[i]>16)
> + return AVERROR(EIO);
how can this happen? the table is const
> + parm[i]=get_bits(&gb, formats[ctx->format].vector_bits[i]);
> + }
> + return 0;
> +}
> +
> +
> +static int ff_g729a_decode_frame(AVCodecContext *avctx,
> + void *data, int *data_size,
> + const uint8_t *buf, int buf_size)
> +{
> + int parm[VECTOR_SIZE];
> + G729A_Context *ctx=avctx->priv_data;
> + int l_frame=formats[ctx->format].frame_size;
> +
> + g729_bytes2parm(ctx, buf, buf_size, parm);
> +
> + *data_size=0;
you must check if there is enough space available, simply ignoring the buffer
size is not ok
> + g729a_decode_frame_internal(ctx,(short*)data, l_frame*sizeof(short), parm);
> + *data_size+=l_frame*sizeof(short);
its int16_t not short
[...]
> +static int act_probe(AVProbeData *p)
> +{
> + ACTHeader* hdr=(ACTHeader*)&p->buf[256];
> +
> + if ((AV_RL32(&p->buf[0]) != RIFF_TAG) ||
> + (AV_RL32(&p->buf[8]) != WAVE_TAG) ||
> + (AV_RL32(&p->buf[16]) != 16))
> + return 0;
> +
> + if(hdr->tag!=0x84)
> + return 0;
> +
> + return AVPROBE_SCORE_MAX;
> +}
if its riff wave use wav.c
[...]
> +static int act_read_packet(AVFormatContext *s,
> + AVPacket *pkt)
> +{
> + ACTContext* ctx = s->priv_data;
> + ByteIOContext *pb = s->pb;
> + uint8_t bFrame[22];
> + uint8_t *pkt_buf;
> + int i, bytes_read;
> + int frame_size=s->streams[0]->codec->frame_size;
> +
> + bytes_read = get_buffer(pb, bFrame, frame_size);
> +
> + if(bytes_read != frame_size || av_new_packet(pkt, frame_size))
> + return AVERROR(EIO);
> +
> + pkt_buf=(uint16_t*)pkt->data;
> +
> + pkt_buf[1]=bFrame[0];
> + pkt_buf[3]=bFrame[1];
> + pkt_buf[5]=bFrame[2];
> + pkt_buf[7]=bFrame[3];
> + pkt_buf[9]=bFrame[4];
> + pkt_buf[0]=bFrame[5];
> + pkt_buf[2]=bFrame[6];
> + pkt_buf[4]=bFrame[7];
> + pkt_buf[6]=bFrame[8];
> + pkt_buf[8]=bFrame[9];
reimar already commented this IIRC ...
[...]
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
Asymptotically faster algorithms should always be preferred if you have
asymptotical amounts of data
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20080224/8b6753ed/attachment.pgp>
More information about the ffmpeg-devel
mailing list