diff options
| author | Martin Storsjö <martin@martin.st> | 2023-10-17 14:27:17 +0300 |
|---|---|---|
| committer | Martin Storsjö <martin@martin.st> | 2023-10-21 23:25:54 +0300 |
| commit | a76b409dd030a461b0c5dd1ead8b22d1be560afd (patch) | |
| tree | 76d2d19d10c17857f664a70d5c155def96ee440f /libswresample/aarch64 | |
| parent | cada4597ca873b7636e145783dd3ae3725862005 (diff) | |
| download | ffmpeg-a76b409dd030a461b0c5dd1ead8b22d1be560afd.tar.gz | |
aarch64: Reindent all assembly to 8/24 column indentation
libavcodec/aarch64/vc1dsp_neon.S is skipped here, as it intentionally
uses a layered indentation style to visually show how different
unrolled/interleaved phases fit together.
Signed-off-by: Martin Storsjö <martin@martin.st>
Diffstat (limited to 'libswresample/aarch64')
| -rw-r--r-- | libswresample/aarch64/resample.S | 80 |
1 files changed, 40 insertions, 40 deletions
diff --git a/libswresample/aarch64/resample.S b/libswresample/aarch64/resample.S index 114d1216fb..6d9eaaeb23 100644 --- a/libswresample/aarch64/resample.S +++ b/libswresample/aarch64/resample.S @@ -21,57 +21,57 @@ #include "libavutil/aarch64/asm.S" function ff_resample_common_apply_filter_x4_float_neon, export=1 - movi v0.4s, #0 // accumulator -1: ld1 {v1.4s}, [x1], #16 // src[0..3] - ld1 {v2.4s}, [x2], #16 // filter[0..3] - fmla v0.4s, v1.4s, v2.4s // accumulator += src[0..3] * filter[0..3] - subs w3, w3, #4 // filter_length -= 4 - b.gt 1b // loop until filter_length - faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values - faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values - st1 {v0.s}[0], [x0], #4 // write accumulator + movi v0.4s, #0 // accumulator +1: ld1 {v1.4s}, [x1], #16 // src[0..3] + ld1 {v2.4s}, [x2], #16 // filter[0..3] + fmla v0.4s, v1.4s, v2.4s // accumulator += src[0..3] * filter[0..3] + subs w3, w3, #4 // filter_length -= 4 + b.gt 1b // loop until filter_length + faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values + faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values + st1 {v0.s}[0], [x0], #4 // write accumulator ret endfunc function ff_resample_common_apply_filter_x8_float_neon, export=1 - movi v0.4s, #0 // accumulator -1: ld1 {v1.4s}, [x1], #16 // src[0..3] - ld1 {v2.4s}, [x2], #16 // filter[0..3] - ld1 {v3.4s}, [x1], #16 // src[4..7] - ld1 {v4.4s}, [x2], #16 // filter[4..7] - fmla v0.4s, v1.4s, v2.4s // accumulator += src[0..3] * filter[0..3] - fmla v0.4s, v3.4s, v4.4s // accumulator += src[4..7] * filter[4..7] - subs w3, w3, #8 // filter_length -= 8 - b.gt 1b // loop until filter_length - faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values - faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values - st1 {v0.s}[0], [x0], #4 // write accumulator + movi v0.4s, #0 // accumulator +1: ld1 {v1.4s}, [x1], #16 // src[0..3] + ld1 {v2.4s}, [x2], #16 // filter[0..3] + ld1 {v3.4s}, [x1], #16 // src[4..7] + ld1 {v4.4s}, [x2], #16 // filter[4..7] + fmla v0.4s, v1.4s, v2.4s // accumulator += src[0..3] * filter[0..3] + fmla v0.4s, v3.4s, v4.4s // accumulator += src[4..7] * filter[4..7] + subs w3, w3, #8 // filter_length -= 8 + b.gt 1b // loop until filter_length + faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values + faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values + st1 {v0.s}[0], [x0], #4 // write accumulator ret endfunc function ff_resample_common_apply_filter_x4_s16_neon, export=1 - movi v0.4s, #0 // accumulator -1: ld1 {v1.4h}, [x1], #8 // src[0..3] - ld1 {v2.4h}, [x2], #8 // filter[0..3] - smlal v0.4s, v1.4h, v2.4h // accumulator += src[0..3] * filter[0..3] - subs w3, w3, #4 // filter_length -= 4 - b.gt 1b // loop until filter_length - addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values - addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values - st1 {v0.s}[0], [x0], #4 // write accumulator + movi v0.4s, #0 // accumulator +1: ld1 {v1.4h}, [x1], #8 // src[0..3] + ld1 {v2.4h}, [x2], #8 // filter[0..3] + smlal v0.4s, v1.4h, v2.4h // accumulator += src[0..3] * filter[0..3] + subs w3, w3, #4 // filter_length -= 4 + b.gt 1b // loop until filter_length + addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values + addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values + st1 {v0.s}[0], [x0], #4 // write accumulator ret endfunc function ff_resample_common_apply_filter_x8_s16_neon, export=1 - movi v0.4s, #0 // accumulator -1: ld1 {v1.8h}, [x1], #16 // src[0..7] - ld1 {v2.8h}, [x2], #16 // filter[0..7] - smlal v0.4s, v1.4h, v2.4h // accumulator += src[0..3] * filter[0..3] - smlal2 v0.4s, v1.8h, v2.8h // accumulator += src[4..7] * filter[4..7] - subs w3, w3, #8 // filter_length -= 8 - b.gt 1b // loop until filter_length - addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values - addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values - st1 {v0.s}[0], [x0], #4 // write accumulator + movi v0.4s, #0 // accumulator +1: ld1 {v1.8h}, [x1], #16 // src[0..7] + ld1 {v2.8h}, [x2], #16 // filter[0..7] + smlal v0.4s, v1.4h, v2.4h // accumulator += src[0..3] * filter[0..3] + smlal2 v0.4s, v1.8h, v2.8h // accumulator += src[4..7] * filter[4..7] + subs w3, w3, #8 // filter_length -= 8 + b.gt 1b // loop until filter_length + addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values + addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values + st1 {v0.s}[0], [x0], #4 // write accumulator ret endfunc |