| 2008-02-09 |
Loren Merritt | use fewer registers in apply_welch_window_sse2
|
commit | commitdiff | tree |
| 2008-02-06 |
Loren Merritt | sse2 h264 motion compensation. not new code, just separate...
|
commit | commitdiff | tree |
| 2008-02-06 |
Loren Merritt | put loop counter in a register if possible. makes some...
|
commit | commitdiff | tree |
| 2008-02-06 |
Loren Merritt | fix aliasing warnings. simpler too.
|
commit | commitdiff | tree |
| 2008-02-06 |
Loren Merritt | constant was excessively aligned
|
commit | commitdiff | tree |
| 2008-02-05 |
Loren Merritt | ssse3 h264 motion compensation.
|
commit | commitdiff | tree |
| 2008-02-05 |
Loren Merritt | add qpel rounder once during hv rather than twice during...
|
commit | commitdiff | tree |
| 2008-02-05 |
Loren Merritt | clean up an ugliness introduced in r11826. this syntax...
|
commit | commitdiff | tree |
| 2008-02-04 |
Loren Merritt | reduce code duplication
|
commit | commitdiff | tree |
| 2008-02-03 |
Loren Merritt | avg_pixels4_mmx2
|
commit | commitdiff | tree |
| 2008-02-03 |
Loren Merritt | use mmx2/3dnow avg functions in avg_qpel*_mc00
|
commit | commitdiff | tree |
| 2008-02-03 |
Loren Merritt | ff_h264_idct8_add_sse2.
|
commit | commitdiff | tree |
| 2008-02-03 |
Loren Merritt | remove some movq in ff_h264_idct8_add_mmx. 225->217...
|
commit | commitdiff | tree |
| 2008-02-02 |
Loren Merritt | "swap first two elements of L1 if L0 and L1 are identical...
|
commit | commitdiff | tree |
| 2008-02-02 |
Loren Merritt | fix mbaff diagonal neighbor mv
|
commit | commitdiff | tree |
| 2008-02-02 |
Loren Merritt | mbaff spatial direct
|
commit | commitdiff | tree |
| 2007-10-27 |
Loren Merritt | remove a branch from an inner loop.
|
commit | commitdiff | tree |
| 2007-10-01 |
Loren Merritt | div -> mul
|
commit | commitdiff | tree |
| 2007-09-30 |
Loren Merritt | simplify lpc
|
commit | commitdiff | tree |
| 2007-09-30 |
Loren Merritt | 20% faster lpc, 6% overall flac decoding
|
commit | commitdiff | tree |
| 2007-09-30 |
Loren Merritt | replace FIR with finite differences.
|
commit | commitdiff | tree |
| 2007-09-29 |
Loren Merritt | unroll finite differences to avoid swapping registers.
|
commit | commitdiff | tree |
| 2007-09-29 |
Loren Merritt | encode_residual_fixed(): replace FIR with finite differences.
|
commit | commitdiff | tree |
| 2007-09-29 |
Loren Merritt | sse2 version of compute_autocorr().
|
commit | commitdiff | tree |
| 2007-09-29 |
Loren Merritt | was computing one more autocorrelation coefficient...
|
commit | commitdiff | tree |
| 2007-09-29 |
Loren Merritt | replace brute force find_optimal_param() with a closed...
|
commit | commitdiff | tree |
| 2007-09-29 |
Loren Merritt | gcc isn't smart enough to factor out duplicate stores
|
commit | commitdiff | tree |
| 2007-09-28 |
Loren Merritt | 2.5x faster compute_autocorr()
|
commit | commitdiff | tree |
| 2007-09-27 |
Loren Merritt | oops, revert unrelated change
|
commit | commitdiff | tree |
| 2007-09-27 |
Loren Merritt | unroll encode_residual_lpc(). speedup varies between...
|
commit | commitdiff | tree |
| 2007-09-25 |
Loren Merritt | oops, potential overflow on really large blocks
|
commit | commitdiff | tree |
| 2007-09-25 |
Loren Merritt | optimize decode_subframe_lpc()
|
commit | commitdiff | tree |
| 2007-09-25 |
Loren Merritt | optimize encode_residual_lpc()
|
commit | commitdiff | tree |
| 2007-08-23 |
Loren Merritt | simd gmc wasn't impossible.
|
commit | commitdiff | tree |
| 2007-08-05 |
Loren Merritt | prevent huffyuv from generating codewords of length...
|
commit | commitdiff | tree |
| 2007-06-03 |
Andreas Öman | filter_mb_fast() doesn't handle h->deblocking_filter...
|
commit | commitdiff | tree |
| 2007-05-25 |
Loren Merritt | merge huffman tables for rgb too
|
commit | commitdiff | tree |
| 2007-05-24 |
Loren Merritt | cosmetics (reduce nesting)
|
commit | commitdiff | tree |
| 2007-05-24 |
Loren Merritt | use sparse huffman tables. 1.5% faster huffyuv decoding.
|
commit | commitdiff | tree |
| 2007-05-24 |
Loren Merritt | add init_vlc_sparse(). faster than init_vlc() if there...
|
commit | commitdiff | tree |
| 2007-05-23 |
Loren Merritt | merge huffman tables so that we read 2 symbols at a...
|
commit | commitdiff | tree |
| 2007-05-19 |
Loren Merritt | change brute force search to min-heap. 3.6x faster...
|
commit | commitdiff | tree |
| 2007-05-18 |
Loren Merritt | cosmetics: macrofy duplicate code in huffyuv
|
commit | commitdiff | tree |
| 2007-05-18 |
Loren Merritt | r3938 broke 2pass huffyuv (not that anyone uses it)
|
commit | commitdiff | tree |
| 2007-05-12 |
Loren Merritt | sse2 & ssse3 versions of dct_quantize.
|
commit | commitdiff | tree |
| 2007-05-12 |
Loren Merritt | configure check for whether we can compile ssse3
|
commit | commitdiff | tree |
| 2007-05-12 |
Loren Merritt | factor sum_abs_dctelem out of dct_sad, and simd it.
|
commit | commitdiff | tree |
| 2007-05-12 |
Loren Merritt | sse2 & ssse3 versions of hadamard. unroll and inline...
|
commit | commitdiff | tree |
| 2007-05-12 |
Loren Merritt | fix a crash with svq1 + sse2.
|
commit | commitdiff | tree |
| 2007-05-11 |
Loren Merritt | 10l, r8991 broke mmx1 sad
|
commit | commitdiff | tree |
| 2007-05-11 |
Loren Merritt | sse2 version of fullpel sad.
|
commit | commitdiff | tree |
| 2007-05-11 |
Loren Merritt | tweak mmx2 sad.
|
commit | commitdiff | tree |
| 2007-05-10 |
Loren Merritt | tweak mmx2 sad.
|
commit | commitdiff | tree |
| 2007-05-10 |
Loren Merritt | sse2 version of fdct_col.
|
commit | commitdiff | tree |
| 2007-05-09 |
Loren Merritt | cosmetics: remove code duplication in hadamard8_diff_mmx
|
commit | commitdiff | tree |
| 2007-05-08 |
Loren Merritt | cosmetics: remove duplicate transpose macro
|
commit | commitdiff | tree |
| 2007-05-06 |
Loren Merritt | simplify
|
commit | commitdiff | tree |
| 2007-05-06 |
Loren Merritt | simplify
|
commit | commitdiff | tree |
| 2007-05-06 |
Loren Merritt | faster and simpler vp6 bilinear mc
|
commit | commitdiff | tree |
| 2007-03-31 |
Loren Merritt | doxy comment tweaks
|
commit | commitdiff | tree |
| 2007-03-30 |
Loren Merritt | move some stuff out of an inner loop. 12% faster svq1...
|
commit | commitdiff | tree |
| 2007-03-30 |
Loren Merritt | mmx 16-bit ssd. 2.3x faster svq1 encoding.
|
commit | commitdiff | tree |
| 2007-03-20 |
Loren Merritt | oops, encoder was under ifdef decoders.
|
commit | commitdiff | tree |
| 2007-03-18 |
Loren Merritt | huffyuv rgb encoder
|
commit | commitdiff | tree |
| 2007-02-19 |
Loren Merritt | simplify in fill_caches
|
commit | commitdiff | tree |
| 2006-11-27 |
Loren Merritt | 10l in r7174 caused all options to be treated as filenames.
|
commit | commitdiff | tree |
| 2006-10-11 |
Loren Merritt | more prefetches, 2% faster h264
|
commit | commitdiff | tree |
| 2006-10-04 |
Loren Merritt | don't try to inline cabac functions. gcc ignored the...
|
commit | commitdiff | tree |
| 2006-10-04 |
Loren Merritt | simplify write_back_motion for skip mbs
|
commit | commitdiff | tree |
| 2006-10-04 |
Loren Merritt | 10l in r6412
|
commit | commitdiff | tree |
| 2006-10-01 |
Loren Merritt | approximate qpel functions: sacrifice some quality...
|
commit | commitdiff | tree |
| 2006-09-21 |
Loren Merritt | remove a division from STOP_TIMER
|
commit | commitdiff | tree |
| 2006-09-21 |
Loren Merritt | simplify asm mid_pred, as suggested by michaelni.
|
commit | commitdiff | tree |
| 2006-09-21 |
Loren Merritt | cosmetics (indentation)
|
commit | commitdiff | tree |
| 2006-09-21 |
Loren Merritt | tweak ff_imdct_calc_3dn2
|
commit | commitdiff | tree |
| 2006-09-21 |
Loren Merritt | sse implementation of imdct.
|
commit | commitdiff | tree |
| 2006-09-15 |
Loren Merritt | asm implementation of mid_pred.
|
commit | commitdiff | tree |
| 2006-09-01 |
Loren Merritt | snow cosmetics: merge the sliced and non-sliced versions...
|
commit | commitdiff | tree |
| 2006-09-01 |
Loren Merritt | fix snow 1pass rc: the qlog used and the one written...
|
commit | commitdiff | tree |
| 2006-08-29 |
Loren Merritt | slightly faster deblock
|
commit | commitdiff | tree |
| 2006-08-28 |
Loren Merritt | mmx implementation of deblocking strength decision.
|
commit | commitdiff | tree |
| 2006-08-27 |
Loren Merritt | tweak cabac. 0.5% faster h264.
|
commit | commitdiff | tree |
| 2006-08-19 |
Loren Merritt | use ff_fft_calc_3dn2 instead of sse on a K8.
|
commit | commitdiff | tree |
| 2006-08-18 |
Loren Merritt | ff_fft_calc_3dn/3dn2/sse: convert intrinsics to inline...
|
commit | commitdiff | tree |
| 2006-08-11 |
Loren Merritt | vorbis simd tweaks
|
commit | commitdiff | tree |
| 2006-08-10 |
Loren Merritt | vorbis cosmetics: mdct0,mdct1 => mdct[2]
|
commit | commitdiff | tree |
| 2006-08-10 |
Loren Merritt | misc tweaks in vorbis_residue_decode().
|
commit | commitdiff | tree |
| 2006-08-10 |
Loren Merritt | sse and 3dnow implementations of float->int conversion...
|
commit | commitdiff | tree |
| 2006-08-10 |
Loren Merritt | r5954 broke fft on cpus with 3dnow but without mm3dnow.h
|
commit | commitdiff | tree |
| 2006-08-09 |
Loren Merritt | emms -> femms
|
commit | commitdiff | tree |
| 2006-08-09 |
Loren Merritt | gcc 2.95 and 3.4.x on x86 32bit without fomit-frame...
|
commit | commitdiff | tree |
| 2006-08-08 |
Loren Merritt | slightly faster ff_imdct_calc_3dn2() on amd64. (gcc...
|
commit | commitdiff | tree |
| 2006-08-08 |
Loren Merritt | 3dnow2 implementation of imdct.
|
commit | commitdiff | tree |
| 2006-08-03 |
Loren Merritt | 10l, vorbis_inverse_coupling_sse() was really 3dnow
|
commit | commitdiff | tree |
| 2006-08-03 |
Loren Merritt | another div -> fastdiv, another 2% faster vorbis.
|
commit | commitdiff | tree |
| 2006-08-03 |
Loren Merritt | sse & sse2 implementations of vorbis channel coupling.
|
commit | commitdiff | tree |
| 2006-08-03 |
Loren Merritt | int16_t is faster than int_fast16_t for division.
|
commit | commitdiff | tree |
| 2006-07-27 |
Loren Merritt | 10l in r5768 (broke mbaff)
|
commit | commitdiff | tree |
| 2006-07-18 |
Loren Merritt | sync to x264 r537
|
commit | commitdiff | tree |
| 2006-07-16 |
Loren Merritt | slightly faster h264 fill_caches()
|
commit | commitdiff | tree |
| next |
projects / ffmpeg / search