Janne Grunau authored Oct 29, 2014 and Anton Mitrofanov committed Dec 16, 2014

x264_mbtree_propagate_cost_neon is ~7 times faster.
x264_mbtree_propagate_list_neon is 33% faster.

Janne Grunau authored Oct 21, 2014 and Anton Mitrofanov committed Dec 16, 2014

3.5 times faster.

Janne Grunau authored Oct 20, 2014 and Anton Mitrofanov committed Dec 16, 2014

All functions ~33% faster.

Janne Grunau authored Oct 14, 2014 and Anton Mitrofanov committed Dec 16, 2014

deblock_luma_intra[0]_neon is 2 times fastes,
deblock_luma_intra[1]_neon is ~4 times faster.

Janne Grunau authored Oct 13, 2014 and Anton Mitrofanov committed Dec 16, 2014

deblock_h_chroma_422 2.5 times faster

Janne Grunau authored Oct 13, 2014 and Anton Mitrofanov committed Dec 16, 2014

deblock_chroma_420_mbaff_neon  2 times faster

Janne Grunau authored Oct 10, 2014 and Anton Mitrofanov committed Dec 16, 2014

deblock_h_chroma_420_intra, deblock_h_chroma_422_intra and
x264_deblock_h_chroma_intra_mbaff_neon are ~3 times faster.
deblock_chroma_intra[1] is ~4 times faster than C.

Janne Grunau authored Aug 14, 2014 and Anton Mitrofanov committed Dec 16, 2014

integral_init4h_neon and integral_init8h_neon are 3-4 times faster than
C. integral_init8v_neon is 6 times faster and integral_init4v_neon is 10
times faster.

Janne Grunau authored Aug 13, 2014 and Anton Mitrofanov committed Dec 16, 2014

Between 10% and 40% faster than C.

Janne Grunau authored Aug 12, 2014 and Anton Mitrofanov committed Dec 16, 2014

decimate_score15 and 16 are 60% faster, decimate_score64 is 4 times
faster than C.

Janne Grunau authored Aug 08, 2014 and Anton Mitrofanov committed Dec 16, 2014

4 times faster than C.

Janne Grunau authored Aug 07, 2014 and Anton Mitrofanov committed Dec 16, 2014

7 times faster than C.

Janne Grunau authored Aug 07, 2014 and Anton Mitrofanov committed Dec 16, 2014

pixel_sad_4x16_neon: 33% faster than C
pixel_satd_4x16_neon: 5 times faster
pixel_ssd_4x16_neon:  4 times faster

Janne Grunau authored Jul 30, 2014 and Anton Mitrofanov committed Dec 16, 2014

13 times faster than C.

Janne Grunau authored Jul 29, 2014 and Anton Mitrofanov committed Dec 16, 2014

35 times faster than C.

Janne Grunau authored Jul 29, 2014 and Anton Mitrofanov committed Dec 16, 2014

zigzag_scan_4x4_field_neon, zigzag_sub_4x4_field_neon,
zigzag_sub_4x4ac_field_neon, zigzag_sub_4x4_frame_neon,
igzag_sub_4x4ac_frame_neon more than 2 times faster

zigzag_scan_8x8_frame_neon, zigzag_scan_8x8_field_neon,
zigzag_sub_8x8_field_neon, zigzag_sub_8x8_frame_neon 4-5 times faster

zigzag_interleave_8x8_cavlc_neon 6 times faster

Janne Grunau authored Jul 25, 2014 and Anton Mitrofanov committed Dec 16, 2014

~20% faster than calling pixel_sa8d_16x16 and pixel_satd_16x16
separately.

Janne Grunau authored Aug 14, 2014 and Anton Mitrofanov committed Dec 16, 2014

25% faster than the previous version.

All CPUs with AVX2 supports FMA3 (but not the other way around).

Didn't affect output due to the incorrect values either not being used in the
code path or producing equal results compared to the correct values.

Also deduplicate hpel_ref arrays.

It would previously report FAILED if any of the earlier plane_copy tests failed.

40->27 cycles on Haswell.

Improves the accuracy of benchmarks, especially in short functions.

To quote the Intel 64 and IA-32 Architectures Software Developer's Manual:
"The RDTSC instruction is not a serializing instruction. It does not necessarily
wait until all previous instructions have been executed before reading the counter.
Similarly, subsequent instructions may begin execution before the read operation
is performed. If software requires RDTSC to be executed only after all previous
instructions have completed locally, it can either use RDTSCP (if the processor
supports that instruction) or execute the sequence LFENCE;RDTSC."

RDTSCP would accomplish the same task, but it's only available since Nehalem.

This change makes SSE2 a requirement to run checkasm.

Previously there was a limit of two cpuflags.

Reduce the number of vector registers used from 7 to 5.
Eliminate some moves in the AVX implementation.
Avoid bypass delays for transitioning between int and float domains.