mirror of https://git.ffmpeg.org/ffmpeg.git
128 lines
3.6 KiB
ArmAsm
128 lines
3.6 KiB
ArmAsm
/*
|
|
* Copyright © 2024 Rémi Denis-Courmont.
|
|
*
|
|
* 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 "libavutil/riscv/asm.S"
|
|
|
|
.equ BASIS_SHIFT, 16
|
|
.equ RECON_SHIFT, 6
|
|
|
|
func ff_try_8x8basis_rvv, zve32x, b
|
|
li t1, 64
|
|
csrwi vxrm, 0
|
|
vsetvli t0, t1, e32, m8, ta, ma
|
|
vmv.v.x v24, zero
|
|
vmv.s.x v1, zero
|
|
1:
|
|
vsetvli zero, zero, e16, m4, ta, ma
|
|
vle16.v v4, (a2)
|
|
sub t1, t1, t0
|
|
vwmul.vx v16, v4, a3
|
|
sh1add a2, t0, a2
|
|
vle16.v v8, (a0)
|
|
sh1add a0, t0, a0
|
|
vnclip.wi v4, v16, BASIS_SHIFT - RECON_SHIFT
|
|
vle16.v v12, (a1)
|
|
sh1add a1, t0, a1
|
|
vadd.vv v4, v8, v4
|
|
vsra.vi v4, v4, RECON_SHIFT
|
|
vwmul.vv v16, v12, v4
|
|
vsetvli zero, zero, e32, m8, ta, ma
|
|
vmul.vv v16, v16, v16
|
|
vsra.vi v16, v16, 4
|
|
vadd.vv v24, v24, v16
|
|
bnez t1, 1b
|
|
|
|
vredsum.vs v1, v24, v1
|
|
vmv.x.s a0, v1
|
|
srai a0, a0, 2
|
|
ret
|
|
endfunc
|
|
|
|
func ff_add_8x8basis_rvv, zve32x, b
|
|
li t1, 64
|
|
csrwi vxrm, 0
|
|
1:
|
|
vsetvli t0, t1, e16, m4, ta, ma
|
|
vle16.v v4, (a1)
|
|
sub t1, t1, t0
|
|
vwmul.vx v16, v4, a2
|
|
sh1add a1, t0, a1
|
|
vle16.v v8, (a0)
|
|
vnclip.wi v4, v16, BASIS_SHIFT - RECON_SHIFT
|
|
vadd.vv v4, v8, v4
|
|
vse16.v v4, (a0)
|
|
sh1add a0, t0, a0
|
|
bnez t1, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_pix_sum_rvv, zve64x, b
|
|
lpad 0
|
|
vsetivli t0, 16, e16, m1, ta, ma
|
|
slli t1, t0, 3
|
|
vlsseg2e64.v v8, (a0), a1 # poor man's vlse128.v
|
|
li t2, 16
|
|
vmv.s.x v1, zero
|
|
vsetvli zero, t1, e8, m4, ta, ma
|
|
vwaddu.vv v16, v8, v12
|
|
beq t0, t2, 2f # VLEN >= 256
|
|
1:
|
|
sh3add a0, a1, a0
|
|
vsetivli zero, 8, e16, m1, ta, ma
|
|
vlsseg2e64.v v8, (a0), a1
|
|
vsetvli zero, t1, e8, m4, ta, ma
|
|
vwaddu.wv v16, v16, v8
|
|
vwaddu.wv v16, v16, v12
|
|
2:
|
|
vsetvli zero, zero, e16, m8, ta, ma
|
|
vredsum.vs v1, v16, v1
|
|
vmv.x.s a0, v1
|
|
zext.h a0, a0
|
|
ret
|
|
endfunc
|
|
|
|
func ff_pix_norm1_rvv, zve64x
|
|
lpad 0
|
|
li t1, 16
|
|
vsetvli t0, zero, e32, m8, ta, ma
|
|
vmv.v.x v16, zero
|
|
vmv.s.x v1, zero
|
|
1:
|
|
vsetvli t0, t1, e64, m2, ta, ma
|
|
vlsseg2e64.v v8, (a0), a1
|
|
slli t2, t0, 3
|
|
vsetvli zero, t2, e8, m2, ta, ma
|
|
vwmulu.vv v24, v8, v8
|
|
sub t1, t1, t0
|
|
vwmulu.vv v28, v10, v10
|
|
mul t3, a1, t0
|
|
vsetvli zero, zero, e16, m4, tu, ma
|
|
vwaddu.wv v16, v16, v24
|
|
add a0, t3, a0
|
|
vwaddu.wv v16, v16, v28
|
|
bnez t1, 1b
|
|
|
|
vsetvli zero, zero, e32, m8, ta, ma
|
|
vredsum.vs v1, v16, v1
|
|
vmv.x.s a0, v1
|
|
ret
|
|
endfunc
|