RSS feed
f11d6f49
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 19:13
optimisation: giving more hints to the compiler about the sizes in
3fe0db17
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 17:00
removed unneeded variable shifts from alg_quant()
784959e8
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 15:35
separated the two passes from interp_bits2pulses()
05778451
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 15:34
optimisation: removed the shifts from the vq_index() inner loop
7adca0e6
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 13:54
optimisation: better indexing/looping in vq_index()
f7cec83c
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 13:29
pseudo-stack no longer checks on every function entry whether it has been
ba829524
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 12:58
optimisation: one less conditional branch in pulse2comb()
e6ce0c6a
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 12:46
optimisation: Removed a bunch of conditional branches from comb2pulse()
fa147077
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 06:59
Removed implicit 32=>16 conversion (changed to EXTRACT16)
ed90daef
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/18 06:33
optimisation: Got rid of the 32-bit mul in find_spectral_pitch()
309d6420
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/17 03:41
fixed an issue (lacking parentheses) in the no-op version of BITREV
558c50eb
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/16 10:15
optimisation: Making it clear to the compiler that many of the loops in cwrs
7f28f768
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/16 07:47
Make use of CELT_MEMSET() in find_spectral_pitch()
b2b71a58
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/16 06:45
Fixed incorrect energy calculation in stereo intra prediction
a66c3208
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/16 06:43
optimisation: reworked intra_prediction() so that yy is computed fully only
0a864645
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/16 06:29
optimisation: intra_prediction() uses a 16-bit numerator for the search
96069fd0
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/15 17:14
optimisation: managed to avoid dividing in the "full gain" case of alg_quant()
d5683031
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/15 14:04
optimisation: simplified the "full gain" case of alg_quant() to remove some
ed317c94
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/15 13:31
optimisation: another bunch of simplifications to the "simple case" of the
3d152a5d
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/15 03:46
optimisation: merged the init loop of vq_quant().
b18ec0b0
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/11 00:07
Reworked the static modes. Now, if all static modes have the same frame size,
c1a4c2ec
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/10 14:54
some index work (simplifications for dumb compilers) on IMDCT
6d3289c7
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/10 10:43
Added a missing RESTORE_STACK in intra_prediction()
df9275b3
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/10 10:38
Not all compilers are equal -- making it clearer how the MDCT indexing is done
821945d9
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/10 09:24
Defining IMUL32 for 32x32=>32 int multiplications and using it in the range
381b29c2
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/10 07:00
fixed ordering of the channels in the intra prediction.
05e56c4d
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/10 05:13
Defining DISABLE_STEREO now optimises for the mono case
15588ad6
– Jean-Marc Valin <
[email protected]>
authored
on 2008/04/10 05:00
Fixed a stereo regression introduced in e28f25f0d14959d521fda0cdb8f1220995bc50e8
ae76e553
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/28 13:59
Changed the rules for using the pulse spreading. It should be used less often
45a7ab52
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/27 11:41
Optimisation: got rid of about 10% of the 32-bit divisions by using ec_enc_uint
e28f25f0
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/27 10:18
Removed a few int divisions from the intra prediction code.
0828b650
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/27 06:22
fixed-point: using MULT16_16 instead of * in compute_band_energies()
98c86c78
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/27 04:40
Trying to clean up celt_ilog2() vs. EC_ILOG a bit.
fed97d58
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 17:31
optimisation: changed some for() loops to do-while() to give the compiler
df7ab430
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 14:03
Making it obvious to the compiler how to generate a dual-MAC in
95088d45
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 13:57
mix_pitch_and_residual() no longer computing Ryp twice
189acec5
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 12:42
optimisation: defined a reciprocal square root (celt_rsqrt) for use in
385795ed
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 11:56
Fixed the rcp() testcase for new assumptions (x is positive)
233e3172
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 11:46
optimisation: shaving a few cycles off prev_cwrs* by not computed the values
a56c34f3
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 10:36
optimisations: faster handling of the zero for compute_band_energies() and
89c5fd16
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 08:16
optimisation: intra_prediction() no longer needs to divide inside the search
35a1f88e
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 06:34
optimisation: The "simple" Rxy/sqrt(Ryy) case in alg_quant no longer requires
bd2828f6
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/26 04:10
Decision on whether to use pitch is now taken only based on energy in the
5de868c6
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/25 18:38
A bunch of pointers marked as "restrict" to ease the job of the compiler
1bc8a2e4
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/25 18:09
optimisation: spreading_func now in-place with no branch in the loop and half
ac72c282
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/25 17:31
oops. find_max32() now uses VERY_LARGE32 (instead of VERY_LARGE16)
44c63350
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/25 17:28
optimisations: Another bunch of simplifications to alg_quant(), mainly to
05974935
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/25 14:04
optimisation: Making use of restrict in find_spectral_pitch() to disambiguate
49134381
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/25 12:07
optimisations: caching sign of x in alg_quant(), changed celt_div()/celt_rcp()
208ae6e3
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/25 11:25
Optimised intra prediction a bit -- removed a conditional branch and replaced
b886ddc0
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/25 10:38
Removed support for more than one MDCT blocks per frame. I don't think there's
bd718ba5
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/25 10:15
Removed the "pitch compression" in the residual quantisation. Also, removed
7cf79a7a
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/24 17:39
Unrolled the inner loop in vq_index() so that the codebook unpacking doesn't
0de9d8ab
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/24 13:06
Making bits2pulses() use a fixed number of iterations to allow further
23e82b26
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/24 04:15
replaced divisions by recoprocals in intra prediction and folding
c2decd39
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/22 18:58
Made a second version of ec_{en|de}code optimised for encoding bits (no div
dc767f67
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/22 18:23
No longer trying to save bits when encoding integers near the upper limit
22823834
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/22 17:17
fixed-point: added cheap celt_div() division using a reciprocal
32034747
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/22 07:23
Using restrict to make it clear there's no aliasing issues in the mdct.
a536f772
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/22 05:01
Added a few "restrict" keywords and changed some divisions to shifts
c163fc62
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/21 08:53
local var name maxval was shadowing the TI function used to compute it
70162a4f
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/21 08:09
fixed-point: Wrapper for the 32-bit complex FFT used in the MDCT so we can use
9901cb9e
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/21 07:13
fixed-point: defined celt_maxabs16() as basic operator
948dabc7
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/21 06:53
fixed-point: MULT16_32_Q15 for TI DSP (not entirely happy with it)
83006eec
– Jean-Marc Valin <
[email protected]>
authored
on 2008/03/21 06:40
fixed-point: using TI intrinsic for celt_ilog2() if available.
Next