Commits · e3187444692195957eb66989622c7b1ad8448b06 · Alexander Traud / Opus · GitLab

Snippets Groups Projects

Jan 31, 2014
- Skips comb filter overlap when the parameters didn't change. · e3187444
  Jean-Marc Valin authored 11 years ago
  
  e3187444
Jan 30, 2014
- Minor div optimization -- reducing denominators · ad7b28ff
  Jean-Marc Valin authored 11 years ago
  
  ad7b28ff
- Optimizing divisions with a signed numerator · d9fb8a66
  Jean-Marc Valin authored 11 years ago
  
  d9fb8a66
Jan 29, 2014
- Speeding up ec_tell_frac() · 51231750
  Jean-Marc Valin authored 11 years ago
  
  51231750
Jan 23, 2014
- Speeding up extract_collapse_mask() slightly · c543ec1b
  Jean-Marc Valin authored 11 years ago
  
  c543ec1b
- Using SSAT in SIG2WORD16() on ARMv6 · 2e6876c0
  Jean-Marc Valin authored 11 years ago
  
  2e6876c0
Jan 21, 2014
- Unrolled version of the comb filter for ARM (eliminates MOVs) · 9b3a3ad7
  Jean-Marc Valin authored 11 years ago
  
  9b3a3ad7
- Making decode_pulses() compute the L2-norm on the fly · 6218cba2
  Jean-Marc Valin authored 11 years ago
  
  6218cba2
- Speed up the comb filter on ARM by using MAC16_32_Q16() · 17b19783
  Jean-Marc Valin authored 11 years ago
  
  17b19783
- Save more integer divisions on ARM when we know the operands are positive · 29354ff6
  Jean-Marc Valin authored 11 years ago
  
  29354ff6
Jan 20, 2014
- Fixes use of uninitialized values in dynalloc_analysis() · ce1173c7
  Jean-Marc Valin authored 11 years ago
  
  ce1173c7
- Using a table on ARM for unsigned division by small (<=256) integers. · ec5d01cb
  Jean-Marc Valin authored 11 years ago
  
  Saves 0.6% for 64 kb/s and 1.8% for 128 kb/s when decoding on arm7tdmi.
  ec5d01cb
Jan 19, 2014
- Minor cleanup in dynalloc_analysis() · 379af35f
  Jean-Marc Valin authored 11 years ago
  
  379af35f
- Adds a median filter to make dynalloc_analysis() more conservative · 05548fa1
  Jean-Marc Valin authored 11 years ago
  
  This should prevent extreme dynalloc behaviour in cases where some bands are heavily attenuated.
  05548fa1
Jan 08, 2014
- Fix declaration after statement in fixed point. · a65db56e
  Gregory Maxwell authored 11 years ago
  
  a65db56e
- pseudostack instrumentation (off by default) · e7750904
  Jean-Marc Valin authored 11 years ago
  
  e7750904
Jan 07, 2014
- Fixes SMALL_FOOTPRINT for float · 9134e96c
  Jean-Marc Valin authored 11 years ago
  
  9134e96c
- Don't allocate pulses on the stack when calling the SILK PLC. · e17ca256
  Jean-Marc Valin authored 11 years ago
  
  Also minor C89 fix for the previous commit
  e17ca256
- Moves CELT PLC pitch search to a separate function to reduce peak stack · b63e7110
  Jean-Marc Valin authored 11 years ago
  
  b63e7110
- Adds SMALL_FOOTPRINT hack to the CELT PLC too · 5f807c17
  Jean-Marc Valin authored 11 years ago
  
  5f807c17
- Moves deemphasis() call out of celt_decode_lost() to reduce peak stack · 9d1b6fef
  Jean-Marc Valin authored 11 years ago
  
  9d1b6fef
Jan 06, 2014
- Cleaning up leftovers of "freq" in celt_decode_with_ec() · ad8371d1
  Jean-Marc Valin authored 11 years ago
  
  ad8371d1
- Reduces the decoder stack use by removing the pcm_silk buffer in fixed-point · 4d07b135
  Jean-Marc Valin authored 11 years ago
  
  We only keep when concealing less than 10ms with SILK.
  4d07b135
- Moves the remains of compute_inv_mdcts() to celt_synthesis() · 14ca4ed6
  Jean-Marc Valin authored 11 years ago
  
  14ca4ed6
- Hack that makes the SMALL_FOOTPRINT CELT decoder use only 4.25 kB of stack. · 32454dca
  Jean-Marc Valin authored 11 years ago
  
  32454dca
- Reduces decoder stack usage by only storing one channel of denormalized MDCT · bdc7b933
  Jean-Marc Valin authored 11 years ago
  
  bdc7b933
- Some cleaning up of the synthesis code. · 4a6744a4
  Jean-Marc Valin authored 11 years ago
  
  4a6744a4
Jan 05, 2014
- Making exp_rotation1() use MAC16_16(), which saves a few cycles on ARM · ed01a596
  Jean-Marc Valin authored 11 years ago
  
  ed01a596
Jan 04, 2014
- Silences unused parameter warning · ccec752a
  Jean-Marc Valin authored 11 years ago
  
  ccec752a
- Moving the radix-2 to expose trivial twiddle factors · ef0eac49
  Jean-Marc Valin authored 11 years ago
  
  ef0eac49
Dec 31, 2013
- Improving the accuracy of the fixed-point radix-3 and radix-5 · c8f62a4a
  Jean-Marc Valin authored 11 years ago
  
  c8f62a4a
Dec 29, 2013
- Minor cleanup -- nothing to see here · e1f84620
  Jean-Marc Valin authored 11 years ago
  
  e1f84620
- Fixed-point: slight accuracy improvement in the comb filter · 05291fd6
  Jean-Marc Valin authored 11 years ago
  
  05291fd6
- Remove a SAVE_STACK that was pasted accidentally in the previous commit · 30f52cbe
  Jean-Marc Valin authored 11 years ago
  
  30f52cbe
- Unifying scaling of fixed-point and float FFT · e1dc1e22
  Jean-Marc Valin authored 11 years ago
  
  e1dc1e22
- Fixes C89 issue · dbb96ab5
  Jean-Marc Valin authored 11 years ago
  
  dbb96ab5
- Getting rid of some negations · 4c1a90a8
  Jean-Marc Valin authored 11 years ago
  
  Since we're doing two rotations, we can invert the sign on both. Also adding a few comments for optimizing the FFT.
  4c1a90a8
- Slightly improving the accuracy of the fixed-point MDCT downscale · cc344fb8
  Jean-Marc Valin authored 11 years ago
  
  Also simplifying the code
  cc344fb8
Dec 27, 2013
- Commit 99968abb was causing us to allocate too much stack in the MDCT · e0c00e27
  Jean-Marc Valin authored 11 years ago
  
  e0c00e27
- Removes the separate 1/8N rotation in the (I)MDCT and unmerges the MDCT sizes · e43a0abe
  Jean-Marc Valin authored 11 years ago
  
  Undoes commits f7547a4e and 72513f3c
  e43a0abe