RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
ffmpeg/libswscale/loongarch/input.S

781 lines
24 KiB
ArmAsm
Raw Normal View History

swscale/la: Optimize the functions of the swscale series with lsx. ./configure --disable-lasx ffmpeg -i ~/media/1_h264_1080p_30fps_3Mbps.mp4 -f rawvideo -s 640x480 -pix_fmt bgra -y /dev/null -an before: 91fps after: 160fps Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2023-05-25 07:24:30 +00:00
/*
* Loongson LSX optimized swscale
*
* Copyright (c) 2023 Loongson Technology Corporation Limited
* Contributed by Lu Wang <wanglu@loongson.cn>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavcodec/loongarch/loongson_asm.S"
/* void planar_rgb_to_y_lsx(uint8_t *_dst, const uint8_t *src[4],
* int width, int32_t *rgb2yuv)
*/
function planar_rgb_to_y_lsx
ld.d a5, a1, 0
ld.d a6, a1, 8
ld.d a7, a1, 16
ld.w t1, a3, 0 // ry
ld.w t2, a3, 4 // gy
ld.w t3, a3, 8 // by
li.w t4, 9
li.w t5, 524544
li.w t7, 4
li.w t8, 8
vldi vr7, 0
vreplgr2vr.w vr1, t1
vreplgr2vr.w vr2, t2
vreplgr2vr.w vr3, t3
vreplgr2vr.w vr4, t4
vreplgr2vr.w vr5, t5
bge a2, t8, .WIDTH8
bge a2, t7, .WIDTH4
blt zero, a2, .WIDTH
b .END
.WIDTH8:
vld vr8, a5, 0
vld vr9, a6, 0
vld vr10, a7, 0
vilvl.b vr11, vr7, vr8
vilvl.b vr12, vr7, vr9
vilvl.b vr13, vr7, vr10
vilvl.h vr14, vr7, vr11
vilvl.h vr15, vr7, vr12
vilvl.h vr16, vr7, vr13
vilvh.h vr17, vr7, vr11
vilvh.h vr18, vr7, vr12
vilvh.h vr19, vr7, vr13
vmul.w vr20, vr1, vr16
vmul.w vr21, vr1, vr19
vmadd.w vr20, vr2, vr14
vmadd.w vr20, vr3, vr15
vmadd.w vr21, vr2, vr17
vmadd.w vr21, vr3, vr18
vadd.w vr20, vr20, vr5
vadd.w vr21, vr21, vr5
vsra.w vr20, vr20, vr4
vsra.w vr21, vr21, vr4
vpickev.h vr20, vr21, vr20
vst vr20, a0, 0
addi.d a2, a2, -8
addi.d a5, a5, 8
addi.d a6, a6, 8
addi.d a7, a7, 8
addi.d a0, a0, 16
bge a2, t8, .WIDTH8
bge a2, t7, .WIDTH4
blt zero, a2, .WIDTH
b .END
.WIDTH4:
vld vr8, a5, 0
vld vr9, a6, 0
vld vr10, a7, 0
vilvl.b vr11, vr7, vr8
vilvl.b vr12, vr7, vr9
vilvl.b vr13, vr7, vr10
vilvl.h vr14, vr7, vr11
vilvl.h vr15, vr7, vr12
vilvl.h vr16, vr7, vr13
vmul.w vr17, vr1, vr16
vmadd.w vr17, vr2, vr14
vmadd.w vr17, vr3, vr15
vadd.w vr17, vr17, vr5
vsra.w vr17, vr17, vr4
vpickev.h vr17, vr17, vr17
vstelm.d vr17, a0, 0, 0
addi.d a2, a2, -4
addi.d a5, a5, 4
addi.d a6, a6, 4
addi.d a7, a7, 4
addi.d a0, a0, 8
bge a2, t7, .WIDTH4
blt zero, a2, .WIDTH
b .END
.WIDTH:
ld.bu t0, a5, 0
ld.bu t4, a6, 0
ld.bu t6, a7, 0
mul.w t8, t6, t1
mul.w t7, t0, t2
add.w t8, t8, t7
mul.w t7, t4, t3
add.w t8, t8, t7
add.w t8, t8, t5
srai.w t8, t8, 9
st.h t8, a0, 0
addi.d a2, a2, -1
addi.d a5, a5, 1
addi.d a6, a6, 1
addi.d a7, a7, 1
addi.d a0, a0, 2
blt zero, a2, .WIDTH
.END:
endfunc
/* void planar_rgb_to_uv_lsx(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *src[4],
* int width, int32_t *rgb2yuv)
*/
function planar_rgb_to_uv_lsx
addi.d sp, sp, -24
st.d s1, sp, 0
st.d s2, sp, 8
st.d s3, sp, 16
ld.d a5, a2, 0
ld.d a6, a2, 8
ld.d a7, a2, 16
ld.w t1, a4, 12 // ru
ld.w t2, a4, 16 // gu
ld.w t3, a4, 20 // bu
ld.w s1, a4, 24 // rv
ld.w s2, a4, 28 // gv
ld.w s3, a4, 32 // bv
li.w t4, 9
li.w t5, 4194560
li.w t7, 4
li.w t8, 8
vldi vr0, 0
vreplgr2vr.w vr1, t1
vreplgr2vr.w vr2, t2
vreplgr2vr.w vr3, t3
vreplgr2vr.w vr4, s1
vreplgr2vr.w vr5, s2
vreplgr2vr.w vr6, s3
vreplgr2vr.w vr7, t4
vreplgr2vr.w vr8, t5
bge a2, t8, .LOOP_WIDTH8
bge a2, t7, .LOOP_WIDTH4
blt zero, a2, .LOOP_WIDTH
b .LOOP_END
.LOOP_WIDTH8:
vld vr9, a5, 0
vld vr10, a6, 0
vld vr11, a7, 0
vilvl.b vr9, vr0, vr9
vilvl.b vr10, vr0, vr10
vilvl.b vr11, vr0, vr11
vilvl.h vr12, vr0, vr9
vilvl.h vr13, vr0, vr10
vilvl.h vr14, vr0, vr11
vilvh.h vr15, vr0, vr9
vilvh.h vr16, vr0, vr10
vilvh.h vr17, vr0, vr11
vmul.w vr18, vr1, vr14
vmul.w vr19, vr1, vr17
vmul.w vr20, vr4, vr14
vmul.w vr21, vr4, vr17
vmadd.w vr18, vr2, vr12
vmadd.w vr18, vr3, vr13
vmadd.w vr19, vr2, vr15
vmadd.w vr19, vr3, vr16
vmadd.w vr20, vr5, vr12
vmadd.w vr20, vr6, vr13
vmadd.w vr21, vr5, vr15
vmadd.w vr21, vr6, vr16
vadd.w vr18, vr18, vr8
vadd.w vr19, vr19, vr8
vadd.w vr20, vr20, vr8
vadd.w vr21, vr21, vr8
vsra.w vr18, vr18, vr7
vsra.w vr19, vr19, vr7
vsra.w vr20, vr20, vr7
vsra.w vr21, vr21, vr7
vpickev.h vr18, vr19, vr18
vpickev.h vr20, vr21, vr20
vst vr18, a0, 0
vst vr20, a1, 0
addi.d a3, a3, -8
addi.d a5, a5, 8
addi.d a6, a6, 8
addi.d a7, a7, 8
addi.d a0, a0, 16
addi.d a1, a1, 16
bge a3, t8, .LOOP_WIDTH8
bge a3, t7, .LOOP_WIDTH4
blt zero, a3, .LOOP_WIDTH
b .LOOP_END
.LOOP_WIDTH4:
vld vr9, a5, 0
vld vr10, a6, 0
vld vr11, a7, 0
vilvl.b vr9, vr0, vr9
vilvl.b vr10, vr0, vr10
vilvl.b vr11, vr0, vr11
vilvl.h vr12, vr0, vr9
vilvl.h vr13, vr0, vr10
vilvl.h vr14, vr0, vr11
vmul.w vr18, vr1, vr14
vmul.w vr19, vr4, vr14
vmadd.w vr18, vr2, vr12
vmadd.w vr18, vr3, vr13
vmadd.w vr19, vr5, vr12
vmadd.w vr19, vr6, vr13
vadd.w vr18, vr18, vr8
vadd.w vr19, vr19, vr8
vsra.w vr18, vr18, vr7
vsra.w vr19, vr19, vr7
vpickev.h vr18, vr18, vr18
vpickev.h vr19, vr19, vr19
vstelm.d vr18, a0, 0, 0
vstelm.d vr19, a1, 0, 0
addi.d a3, a3, -4
addi.d a5, a5, 4
addi.d a6, a6, 4
addi.d a7, a7, 4
addi.d a0, a0, 8
addi.d a1, a1, 8
bge a3, t7, .LOOP_WIDTH4
blt zero, a3, .LOOP_WIDTH
b .LOOP_END
.LOOP_WIDTH:
ld.bu t0, a5, 0
ld.bu t4, a6, 0
ld.bu t6, a7, 0
mul.w t8, t6, t1
mul.w t7, t0, t2
add.w t8, t8, t7
mul.w t7, t4, t3
add.w t8, t8, t7
add.w t8, t8, t5
srai.w t8, t8, 9
st.h t8, a0, 0
mul.w t8, t6, s1
mul.w t7, t0, s2
add.w t8, t8, t7
mul.w t7, t4, s3
add.w t8, t8, t7
add.w t8, t8, t5
srai.w t8, t8, 9
st.h t8, a1, 0
addi.d a3, a3, -1
addi.d a5, a5, 1
addi.d a6, a6, 1
addi.d a7, a7, 1
addi.d a0, a0, 2
addi.d a1, a1, 2
blt zero, a3, .LOOP_WIDTH
.LOOP_END:
ld.d s1, sp, 0
ld.d s2, sp, 8
ld.d s3, sp, 16
addi.d sp, sp, 24
endfunc
swscale: [LA] Optimize swscale funcs in input.c Optimized 7 funcs with LSX and LASX: 1. yuy2ToUV_c 2. yvy2ToUV_c 3. uyvyToUV_c 4. nv12ToUV_c 5. nv21ToUV_c 6. abgrToA_c 7. rgbaToA_c Reviewed-by: colleague of Shiyou Yin Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2024-03-16 03:03:33 +00:00
/*
* void yuy2ToUV_lsx(uint8_t *dstU, uint8_t *dstV, const uint8_t *unused0, const uint8_t *src1,
* const uint8_t *src2, int width, uint32_t *unused, void *opq)
*/
function yuy2ToUV_lsx
andi t0, a5, 7
srli.d a5, a5, 3
beqz a5, 2f
1:
vld vr0, a3, 1
vld vr1, a3, 17
addi.d a5, a5, -1
addi.d a3, a3, 32
vpickev.b vr2, vr1, vr0
vpickev.b vr0, vr2, vr2
vpickod.b vr1, vr2, vr2
fst.d f0, a0, 0
fst.d f1, a1, 0
addi.d a0, a0, 8
addi.d a1, a1, 8
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 1
ld.b t2, a3, 3
addi.d a3, a3, 4
addi.d t0, t0, -1
st.b t1, a0, 0
st.b t2, a1, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
function yuy2ToUV_lasx
andi t0, a5, 15
srli.d a5, a5, 4
beqz a5, 2f
1:
xvld xr0, a3, 1
xvld xr1, a3, 33
addi.d a5, a5, -1
addi.d a3, a3, 64
xvpickev.b xr2, xr1, xr0
xvpermi.d xr2, xr2, 0xd8
xvpickev.b xr0, xr2, xr2
xvpermi.d xr0, xr0, 0xd8
xvpickod.b xr1, xr2, xr2
xvpermi.d xr1, xr1, 0xd8
vst vr0, a0, 0
vst vr1, a1, 0
addi.d a0, a0, 16
addi.d a1, a1, 16
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 1
ld.b t2, a3, 3
addi.d a3, a3, 4
addi.d t0, t0, -1
st.b t1, a0, 0
st.b t2, a1, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
/*
* void yvy2ToUV_lsx(uint8_t *dstU, uint8_t *dstV, const uint8_t *unused0, const uint8_t *src1,
* const uint8_t *src2, int width, uint32_t *unused, void *opq)
*/
function yvy2ToUV_lsx
andi t0, a5, 7
srli.d a5, a5, 3
beqz a5, 2f
1:
vld vr0, a3, 1
vld vr1, a3, 17
addi.d a5, a5, -1
addi.d a3, a3, 32
vpickev.b vr2, vr1, vr0
vpickev.b vr0, vr2, vr2
vpickod.b vr1, vr2, vr2
fst.d f0, a1, 0
fst.d f1, a0, 0
addi.d a0, a0, 8
addi.d a1, a1, 8
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 1
ld.b t2, a3, 3
addi.d a3, a3, 4
addi.d t0, t0, -1
st.b t1, a1, 0
st.b t2, a0, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
function yvy2ToUV_lasx
andi t0, a5, 15
srli.d a5, a5, 4
beqz a5, 2f
1:
xvld xr0, a3, 1
xvld xr1, a3, 33
addi.d a5, a5, -1
addi.d a3, a3, 64
xvpickev.b xr2, xr1, xr0
xvpermi.d xr2, xr2, 0xd8
xvpickev.b xr0, xr2, xr2
xvpermi.d xr0, xr0, 0xd8
xvpickod.b xr1, xr2, xr2
xvpermi.d xr1, xr1, 0xd8
vst vr0, a1, 0
vst vr1, a0, 0
addi.d a0, a0, 16
addi.d a1, a1, 16
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 1
ld.b t2, a3, 3
addi.d a3, a3, 4
addi.d t0, t0, -1
st.b t1, a1, 0
st.b t2, a0, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
/*
* void uyvyToUV_lsx(uint8_t *dstU, uint8_t *dstV, const uint8_t *unused0, const uint8_t *src1,
* const uint8_t *src2, int width, uint32_t *unused, void *opq)
*/
function uyvyToUV_lsx
andi t0, a5, 7
srli.d a5, a5, 3
beqz a5, 2f
1:
vld vr0, a3, 0
vld vr1, a3, 16
addi.d a5, a5, -1
addi.d a3, a3, 32
vpickev.b vr2, vr1, vr0
vpickev.b vr0, vr2, vr2
vpickod.b vr1, vr2, vr2
fst.d f0, a0, 0
fst.d f1, a1, 0
addi.d a0, a0, 8
addi.d a1, a1, 8
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 1
ld.b t2, a3, 3
addi.d a3, a3, 4
addi.d t0, t0, -1
st.b t1, a0, 0
st.b t2, a1, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
function uyvyToUV_lasx
andi t0, a5, 15
srli.d a5, a5, 4
beqz a5, 2f
1:
xvld xr0, a3, 0
xvld xr1, a3, 32
addi.d a5, a5, -1
addi.d a3, a3, 64
xvpickev.b xr2, xr1, xr0
xvpermi.d xr2, xr2, 0xd8
xvpickev.b xr0, xr2, xr2
xvpermi.d xr0, xr0, 0xd8
xvpickod.b xr1, xr2, xr2
xvpermi.d xr1, xr1, 0xd8
vst vr0, a0, 0
vst vr1, a1, 0
addi.d a0, a0, 16
addi.d a1, a1, 16
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 1
ld.b t2, a3, 3
addi.d a3, a3, 4
addi.d t0, t0, -1
st.b t1, a0, 0
st.b t2, a1, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
/*
* void nv12ToUV_lsx(uint8_t *dstU, uint8_t *dstV, const uint8_t *unused0, const uint8_t *src1,
* const uint8_t *src2, int width, uint32_t *unused, void *opq)
*/
function nv12ToUV_lsx
andi t0, a5, 15
srli.d a5, a5, 4
beqz a5, 2f
1:
vld vr0, a3, 0
vld vr1, a3, 16
addi.d a5, a5, -1
addi.d a3, a3, 32
vpickev.b vr2, vr1, vr0
vpickod.b vr3, vr1, vr0
vst vr2, a0, 0
vst vr3, a1, 0
addi.d a0, a0, 16
addi.d a1, a1, 16
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 0
ld.b t2, a3, 1
addi.d a3, a3, 2
addi.d t0, t0, -1
st.b t1, a0, 0
st.b t2, a1, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
function nv12ToUV_lasx
andi t0, a5, 31
srli.d a5, a5, 5
beqz a5, 2f
1:
xvld xr0, a3, 0
xvld xr1, a3, 32
addi.d a5, a5, -1
addi.d a3, a3, 64
xvpickev.b xr2, xr1, xr0
xvpickod.b xr3, xr1, xr0
xvpermi.d xr2, xr2, 0xd8
xvpermi.d xr3, xr3, 0xd8
xvst xr2, a0, 0
xvst xr3, a1, 0
addi.d a0, a0, 32
addi.d a1, a1, 32
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 0
ld.b t2, a3, 1
addi.d a3, a3, 2
addi.d t0, t0, -1
st.b t1, a0, 0
st.b t2, a1, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
/*
* void nv21ToUV_lsx(uint8_t *dstU, uint8_t *dstV, const uint8_t *unused0, const uint8_t *src1,
* const uint8_t *src2, int width, uint32_t *unused, void *opq)
*/
function nv21ToUV_lsx
andi t0, a5, 15
srli.d a5, a5, 4
beqz a5, 2f
1:
vld vr0, a3, 0
vld vr1, a3, 16
addi.d a5, a5, -1
addi.d a3, a3, 32
vpickev.b vr2, vr1, vr0
vpickod.b vr3, vr1, vr0
vst vr2, a1, 0
vst vr3, a0, 0
addi.d a0, a0, 16
addi.d a1, a1, 16
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 0
ld.b t2, a3, 1
addi.d a3, a3, 2
addi.d t0, t0, -1
st.b t1, a1, 0
st.b t2, a0, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
function nv21ToUV_lasx
andi t0, a5, 31
srli.d a5, a5, 5
beqz a5, 2f
1:
xvld xr0, a3, 0
xvld xr1, a3, 32
addi.d a5, a5, -1
addi.d a3, a3, 64
xvpickev.b xr2, xr1, xr0
xvpickod.b xr3, xr1, xr0
xvpermi.d xr2, xr2, 0xd8
xvpermi.d xr3, xr3, 0xd8
xvst xr2, a1, 0
xvst xr3, a0, 0
addi.d a0, a0, 32
addi.d a1, a1, 32
bnez a5, 1b
2:
beqz t0, 4f
3:
ld.b t1, a3, 0
ld.b t2, a3, 1
addi.d a3, a3, 2
addi.d t0, t0, -1
st.b t1, a1, 0
st.b t2, a0, 0
addi.d a0, a0, 1
addi.d a1, a1, 1
bnez t0, 3b
4:
endfunc
/*
*void abgrToA_lsx(uint8_t *_dst, const uint8_t *src, const uint8_t *unused1,
* const uint8_t *unused2, int width, uint32_t *unused, void *opq)
*/
function abgrToA_lsx
andi t0, a4, 7
srli.d a4, a4, 3
vxor.v vr0, vr0, vr0
beqz a4, 2f
1:
vld vr1, a1, 0
vld vr2, a1, 16
addi.d a4, a4, -1
addi.d a1, a1, 32
vpickev.b vr3, vr2, vr1
vpackev.b vr3, vr0, vr3
vslli.h vr1, vr3, 6
vsrli.h vr2, vr3, 2
vor.v vr3, vr2, vr1
vst vr3, a0, 0
addi.d a0, a0, 16
bnez a4, 1b
2:
beqz t0, 4f
3:
ld.b t1, a1, 3
addi.d t0, t0, -1
addi.d a1, a1, 4
andi t1, t1, 0xff
slli.w t2, t1, 6
srli.w t3, t1, 2
or t1, t2, t3
st.h t1, a0, 0
addi.d a0, a0, 2
bnez t0, 3b
4:
endfunc
function abgrToA_lasx
andi t0, a4, 15
srli.d a4, a4, 4
xvxor.v xr0, xr0, xr0
beqz a4, 2f
1:
xvld xr1, a1, 0
xvld xr2, a1, 32
addi.d a4, a4, -1
addi.d a1, a1, 64
xvpickev.b xr3, xr2, xr1
xvpermi.d xr3, xr3, 0xd8
xvpackev.b xr3, xr0, xr3
xvslli.h xr1, xr3, 6
xvsrli.h xr2, xr3, 2
xvor.v xr3, xr2, xr1
xvst xr3, a0, 0
addi.d a0, a0, 32
bnez a4, 1b
2:
beqz t0, 4f
3:
ld.b t1, a1, 3
addi.d t0, t0, -1
addi.d a1, a1, 4
andi t1, t1, 0xff
slli.w t2, t1, 6
srli.w t3, t1, 2
or t1, t2, t3
st.h t1, a0, 0
addi.d a0, a0, 2
bnez t0, 3b
4:
endfunc
/*
*void rgbaToA_lsx(uint8_t *_dst, const uint8_t *src, const uint8_t *unused1,
* const uint8_t *unused2, int width, uint32_t *unused, void *opq)
*/
function rgbaToA_lsx
andi t0, a4, 7
srli.d a4, a4, 3
vxor.v vr0, vr0, vr0
beqz a4, 2f
1:
vld vr1, a1, 3
vld vr2, a1, 19
addi.d a4, a4, -1
addi.d a1, a1, 32
vpickev.b vr3, vr2, vr1
vpackev.b vr3, vr0, vr3
vslli.h vr1, vr3, 6
vsrli.h vr2, vr3, 2
vor.v vr3, vr2, vr1
vst vr3, a0, 0
addi.d a0, a0, 16
bnez a4, 1b
2:
beqz t0, 4f
3:
ld.b t1, a1, 3
addi.d t0, t0, -1
addi.d a1, a1, 4
andi t1, t1, 0xff
slli.w t2, t1, 6
srli.w t3, t1, 2
or t1, t2, t3
st.h t1, a0, 0
addi.d a0, a0, 2
bnez t0, 3b
4:
endfunc
function rgbaToA_lasx
andi t0, a4, 15
srli.d a4, a4, 4
xvxor.v xr0, xr0, xr0
beqz a4, 2f
1:
xvld xr1, a1, 3
xvld xr2, a1, 35
addi.d a4, a4, -1
addi.d a1, a1, 64
xvpickev.b xr3, xr2, xr1
xvpermi.d xr3, xr3, 0xd8
xvpackev.b xr3, xr0, xr3
xvslli.h xr1, xr3, 6
xvsrli.h xr2, xr3, 2
xvor.v xr3, xr2, xr1
xvst xr3, a0, 0
addi.d a0, a0, 32
bnez a4, 1b
2:
beqz t0, 4f
3:
ld.b t1, a1, 3
addi.d t0, t0, -1
addi.d a1, a1, 4
andi t1, t1, 0xff
slli.w t2, t1, 6
srli.w t3, t1, 2
or t1, t2, t3
st.h t1, a0, 0
addi.d a0, a0, 2
bnez t0, 3b
4:
endfunc