Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Whitespace fixesexp_mips_opt | Jean-Marc Valin | 2014-06-19 | 3 | -3/+1 |
| | |||||
* | MIPS optimizations | Rhishikesh Agashe | 2014-06-19 | 13 | -0/+1040 |
| | | | | Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> | ||||
* | Adds SIG2WORD16() to fixed_debug.h | Jean-Marc Valin | 2014-06-18 | 1 | -0/+10 |
| | |||||
* | Suppress some coverity false positives. | Gregory Maxwell | 2014-04-16 | 2 | -1/+2 |
| | |||||
* | Fix iOS builds with assembly. | Timothy B. Terriberry | 2014-03-26 | 1 | -3/+3 |
| | | | | | | | | | | | | The patch in 76e831d917ff got us most of the way there, but out-of-tree builds required a second Makefile.am rule, which was missing @ARM2GNU_PARAMS@. Also, the arm2gnu.pl was terminating argument processing on any argument beginning with --, rather than an argument that was just -- by itself (as is the normal convention in GNU programs). That meant it never saw the --apple flag even when it was passed. Thanks to Jonathan Lennox for the report and for testing. | ||||
* | Make the arm2gnu.pl converter handle apple specific details | Martin Storsjo | 2014-03-19 | 1 | -6/+17 |
| | | | | | | | | | This allows building the arm assembly for iOS. This checks for the __APPLE__ preprocessor built-in define to determine whether this extra handling should be enabled. Signed-off-by: Timothy B. Terriberry <tterribe@xiph.org> | ||||
* | Add separate labels for the start of public functions | Martin Storsjo | 2014-03-19 | 1 | -2/+4 |
| | | | | | | | This avoids having to use the public symbol name when jumping here, on platforms where the public symbols have an underscore prefix. Signed-off-by: Timothy B. Terriberry <tterribe@xiph.org> | ||||
* | arm: Use the UAL syntax for instructions | Martin Storsjo | 2014-02-24 | 2 | -9/+11 |
| | | | | | | | This is required in order to build using the built-in assembler in clang. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> | ||||
* | Fixes an aliasing bug in the MDCT when the frame size isn't a multiple of 4. | Marcello Caramma (mcaramma) | 2014-02-24 | 1 | -2/+2 |
| | | | | | | | In that case, the yp0 and yp1 ended up aliasing for the last element, despite being marked as restrict. Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca> | ||||
* | Adds missing license for arm2gnu.pl | Jean-Marc Valin | 2014-02-24 | 1 | -0/+24 |
| | |||||
* | Skips comb filter overlap when the parameters didn't change. | Jean-Marc Valin | 2014-01-31 | 1 | -0/+3 |
| | |||||
* | Minor div optimization -- reducing denominators | Jean-Marc Valin | 2014-01-30 | 2 | -2/+2 |
| | |||||
* | Optimizing divisions with a signed numerator | Jean-Marc Valin | 2014-01-30 | 2 | -2/+15 |
| | |||||
* | Speeding up ec_tell_frac() | Jean-Marc Valin | 2014-01-29 | 1 | -0/+22 |
| | |||||
* | Speeding up extract_collapse_mask() slightly | Jean-Marc Valin | 2014-01-23 | 1 | -1/+3 |
| | |||||
* | Using SSAT in SIG2WORD16() on ARMv6 | Jean-Marc Valin | 2014-01-22 | 4 | -11/+29 |
| | |||||
* | Unrolled version of the comb filter for ARM (eliminates MOVs) | Jean-Marc Valin | 2014-01-21 | 1 | -2/+32 |
| | |||||
* | Making decode_pulses() compute the L2-norm on the fly | Jean-Marc Valin | 2014-01-21 | 3 | -19/+31 |
| | |||||
* | Speed up the comb filter on ARM by using MAC16_32_Q16() | Jean-Marc Valin | 2014-01-21 | 6 | -1/+56 |
| | |||||
* | Save more integer divisions on ARM when we know the operands are positive | Jean-Marc Valin | 2014-01-21 | 5 | -13/+18 |
| | |||||
* | Fixes use of uninitialized values in dynalloc_analysis() | Jean-Marc Valin | 2014-01-20 | 1 | -3/+3 |
| | |||||
* | Using a table on ARM for unsigned division by small (<=256) integers. | Jean-Marc Valin | 2014-01-20 | 4 | -2/+62 |
| | | | | Saves 0.6% for 64 kb/s and 1.8% for 128 kb/s when decoding on arm7tdmi. | ||||
* | Minor cleanup in dynalloc_analysis() | Jean-Marc Valin | 2014-01-19 | 1 | -9/+11 |
| | |||||
* | Adds a median filter to make dynalloc_analysis() more conservative | Jean-Marc Valin | 2014-01-19 | 1 | -0/+76 |
| | | | | | This should prevent extreme dynalloc behaviour in cases where some bands are heavily attenuated. | ||||
* | Fix declaration after statement in fixed point. | Gregory Maxwell | 2014-01-08 | 1 | -2/+2 |
| | |||||
* | pseudostack instrumentation (off by default) | Jean-Marc Valin | 2014-01-07 | 1 | -1/+7 |
| | |||||
* | Fixes SMALL_FOOTPRINT for float | Jean-Marc Valin | 2014-01-07 | 1 | -4/+7 |
| | |||||
* | Don't allocate pulses on the stack when calling the SILK PLC. | Jean-Marc Valin | 2014-01-07 | 1 | -1/+1 |
| | | | | Also minor C89 fix for the previous commit | ||||
* | Moves CELT PLC pitch search to a separate function to reduce peak stack | Jean-Marc Valin | 2014-01-07 | 1 | -9/+17 |
| | |||||
* | Adds SMALL_FOOTPRINT hack to the CELT PLC too | Jean-Marc Valin | 2014-01-07 | 1 | -2/+10 |
| | |||||
* | Moves deemphasis() call out of celt_decode_lost() to reduce peak stack | Jean-Marc Valin | 2014-01-07 | 1 | -11/+7 |
| | |||||
* | Cleaning up leftovers of "freq" in celt_decode_with_ec() | Jean-Marc Valin | 2014-01-06 | 1 | -2/+0 |
| | |||||
* | Reduces the decoder stack use by removing the pcm_silk buffer in fixed-point | Jean-Marc Valin | 2014-01-06 | 3 | -19/+50 |
| | | | | We only keep when concealing less than 10ms with SILK. | ||||
* | Moves the remains of compute_inv_mdcts() to celt_synthesis() | Jean-Marc Valin | 2014-01-06 | 1 | -37/+24 |
| | |||||
* | Hack that makes the SMALL_FOOTPRINT CELT decoder use only 4.25 kB of stack. | Jean-Marc Valin | 2014-01-06 | 1 | -8/+16 |
| | |||||
* | Reduces decoder stack usage by only storing one channel of denormalized MDCT | Jean-Marc Valin | 2014-01-06 | 5 | -98/+125 |
| | |||||
* | Some cleaning up of the synthesis code. | Jean-Marc Valin | 2014-01-05 | 4 | -31/+24 |
| | |||||
* | Making exp_rotation1() use MAC16_16(), which saves a few cycles on ARM | Jean-Marc Valin | 2014-01-04 | 1 | -4/+6 |
| | |||||
* | Silences unused parameter warning | Jean-Marc Valin | 2014-01-04 | 1 | -0/+1 |
| | |||||
* | Moving the radix-2 to expose trivial twiddle factors | Jean-Marc Valin | 2014-01-03 | 3 | -46/+73 |
| | |||||
* | Improving the accuracy of the fixed-point radix-3 and radix-5 | Jean-Marc Valin | 2013-12-31 | 1 | -3/+14 |
| | |||||
* | Minor cleanup -- nothing to see here | Jean-Marc Valin | 2013-12-29 | 3 | -10/+2 |
| | |||||
* | Fixed-point: slight accuracy improvement in the comb filter | Jean-Marc Valin | 2013-12-29 | 1 | -6/+6 |
| | |||||
* | Remove a SAVE_STACK that was pasted accidentally in the previous commit | Jean-Marc Valin | 2013-12-29 | 1 | -1/+0 |
| | |||||
* | Unifying scaling of fixed-point and float FFT | Jean-Marc Valin | 2013-12-29 | 5 | -33/+32 |
| | |||||
* | Fixes C89 issue | Jean-Marc Valin | 2013-12-29 | 1 | -2/+4 |
| | |||||
* | Getting rid of some negations | Jean-Marc Valin | 2013-12-28 | 2 | -10/+15 |
| | | | | | Since we're doing two rotations, we can invert the sign on both. Also adding a few comments for optimizing the FFT. | ||||
* | Slightly improving the accuracy of the fixed-point MDCT downscale | Jean-Marc Valin | 2013-12-28 | 1 | -13/+9 |
| | | | | Also simplifying the code | ||||
* | Commit 99968ab was causing us to allocate too much stack in the MDCT | Jean-Marc Valin | 2013-12-27 | 1 | -1/+1 |
| | |||||
* | Removes the separate 1/8N rotation in the (I)MDCT and unmerges the MDCT sizes | Jean-Marc Valin | 2013-12-27 | 4 | -260/+794 |
| | | | | Undoes commits f7547a4e and 72513f3c |