RepoMirrors/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2025-03-11 06:58:18 +00:00
Code Issues Projects Releases Wiki Activity

hevc: Add NEON IDCT DC functions for bitdepth 8

Browse Source 
Signed-off-by: Alexandra Hájková <alexandra@khirnov.net>
Signed-off-by: Martin Storsjö <martin@martin.st>
This commit is contained in:
Seppo Tomperi 2017-04-25 19:41:29 +02:00 committed by Martin Storsjö
parent d0ce0634e0
commit 358adef030
2 changed files with 88 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

78
libavcodec/arm/hevc_idct.S
View File

@ -1,5 +1,7 @@
/*
* ARM NEON optimised IDCT functions for HEVC decoding
*
* Copyright (c) 2014 Seppo Tomperi <seppo.tomperi@vtt.fi>
* Copyright (c) 2017 Alexandra Hájková
*
* This file is part of Libav.
@ -28,6 +30,82 @@ const trans, align=4
.short 57, 43, 25, 9
endconst
function ff_hevc_idct_4x4_dc_8_neon, export=1
ldrsh r1, [r0]
ldr r2, =0x20
add r1, #1
asr r1, #1
add r1, r2
asr r1, #6
vdup.16 q0, r1
vdup.16 q1, r1
vst1.16 {q0, q1}, [r0, :128]
bx lr
endfunc
function ff_hevc_idct_8x8_dc_8_neon, export=1
ldrsh r1, [r0]
ldr r2, =0x20
add r1, #1
asr r1, #1
add r1, r2
asr r1, #6
vdup.16 q8, r1
vdup.16 q9, r1
vmov.16 q10, q8
vmov.16 q11, q8
vmov.16 q12, q8
vmov.16 q13, q8
vmov.16 q14, q8
vmov.16 q15, q8
vstm r0, {q8-q15}
bx lr
endfunc
function ff_hevc_idct_16x16_dc_8_neon, export=1
ldrsh r1, [r0]
ldr r2, =0x20
add r1, #1
asr r1, #1
add r1, r2
asr r1, #6
vdup.16 q8, r1
vdup.16 q9, r1
vmov.16 q10, q8
vmov.16 q11, q8
vmov.16 q12, q8
vmov.16 q13, q8
vmov.16 q14, q8
vmov.16 q15, q8
vstm r0!, {q8-q15}
vstm r0!, {q8-q15}
vstm r0!, {q8-q15}
vstm r0, {q8-q15}
bx lr
endfunc
function ff_hevc_idct_32x32_dc_8_neon, export=1
ldrsh r1, [r0]
ldr r2, =0x20
add r1, #1
asr r1, #1
add r1, r2
asr r1, #6
mov r3, #16
vdup.16 q8, r1
vdup.16 q9, r1
vmov.16 q10, q8
vmov.16 q11, q8
vmov.16 q12, q8
vmov.16 q13, q8
vmov.16 q14, q8
vmov.16 q15, q8
1: subs r3, #1
vstm r0!, {q8-q15}
bne 1b
bx lr
endfunc
.macro sum_sub out, in, c, op
.ifc \op, +
vmlal.s16 \out, \in, \c

10
libavcodec/arm/hevcdsp_init_arm.c
View File

@ -25,6 +25,11 @@
#include "libavcodec/hevcdsp.h"
void ff_hevc_idct_4x4_dc_8_neon(int16_t *coeffs);
void ff_hevc_idct_8x8_dc_8_neon(int16_t *coeffs);
void ff_hevc_idct_16x16_dc_8_neon(int16_t *coeffs);
void ff_hevc_idct_32x32_dc_8_neon(int16_t *coeffs);
void ff_hevc_idct_4x4_8_neon(int16_t *coeffs, int col_limit);
void ff_hevc_idct_8x8_8_neon(int16_t *coeffs, int col_limit);
void ff_hevc_idct_16x16_8_neon(int16_t *coeffs, int col_limit);
@ -38,6 +43,11 @@ av_cold void ff_hevc_dsp_init_arm(HEVCDSPContext *c, int bit_depth)
if (have_neon(cpu_flags)) {
if (bit_depth == 8) {
c->idct_dc[0] = ff_hevc_idct_4x4_dc_8_neon;
c->idct_dc[1] = ff_hevc_idct_8x8_dc_8_neon;
c->idct_dc[2] = ff_hevc_idct_16x16_dc_8_neon;
c->idct_dc[3] = ff_hevc_idct_32x32_dc_8_neon;
c->idct[0] = ff_hevc_idct_4x4_8_neon;
c->idct[1] = ff_hevc_idct_8x8_8_neon;
c->idct[2] = ff_hevc_idct_16x16_8_neon;