mirror of
https://github.com/FFmpeg/FFmpeg.git
synced 2024-12-07 11:13:41 +02:00
ca664f2254
In this case, the inner loop computing the scalar product can be reduced to just one multiplication and one sum even with 128-bit vectors. The result is a lot simpler, but also brings more modest performance gains: flac_lpc_16_13_c: 15241.0 flac_lpc_16_13_rvv_i32: 11230.0 flac_lpc_16_16_c: 17884.0 flac_lpc_16_16_rvv_i32: 12125.7 flac_lpc_16_29_c: 27847.7 flac_lpc_16_29_rvv_i32: 10494.0 flac_lpc_16_32_c: 30051.5 flac_lpc_16_32_rvv_i32: 10355.0
532 lines
13 KiB
ArmAsm
532 lines
13 KiB
ArmAsm
/*
|
|
* Copyright © 2023 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"
|
|
|
|
func ff_flac_lpc16_rvv, zve32x
|
|
vsetvli zero, a2, e32, m8, ta, ma
|
|
vle32.v v8, (a1)
|
|
sub a4, a4, a2
|
|
vle32.v v16, (a0)
|
|
sh2add a0, a2, a0
|
|
vmv.s.x v0, zero
|
|
1:
|
|
vmul.vv v24, v8, v16
|
|
lw t0, (a0)
|
|
vredsum.vs v24, v24, v0
|
|
addi a4, a4, -1
|
|
vmv.x.s t1, v24
|
|
sra t1, t1, a3
|
|
add t0, t0, t1
|
|
vslide1down.vx v16, v16, t0
|
|
sw t0, (a0)
|
|
addi a0, a0, 4
|
|
bnez a4, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
#if (__riscv_xlen == 64)
|
|
func ff_flac_lpc32_rvv, zve32x
|
|
addi t2, a2, -16
|
|
ble t2, zero, ff_flac_lpc32_rvv_simple
|
|
vsetivli zero, 1, e64, m1, ta, ma
|
|
vmv.s.x v0, zero
|
|
vsetvli zero, a2, e32, m8, ta, ma
|
|
vle32.v v8, (a1)
|
|
sub a4, a4, a2
|
|
vle32.v v16, (a0)
|
|
sh2add a0, a2, a0
|
|
1:
|
|
vsetvli zero, a2, e32, m4, ta, ma
|
|
vwmul.vv v24, v8, v16
|
|
vsetvli zero, t2, e32, m4, tu, ma
|
|
vwmacc.vv v24, v12, v20
|
|
vsetvli zero, a2, e64, m8, ta, ma
|
|
vredsum.vs v24, v24, v0
|
|
lw t0, (a0)
|
|
addi a4, a4, -1
|
|
vmv.x.s t1, v24
|
|
vsetvli zero, a2, e32, m8, ta, ma
|
|
sra t1, t1, a3
|
|
add t0, t0, t1
|
|
vslide1down.vx v16, v16, t0
|
|
sw t0, (a0)
|
|
addi a0, a0, 4
|
|
bnez a4, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_lpc32_rvv_simple, zve32x
|
|
vsetivli zero, 1, e64, m1, ta, ma
|
|
vmv.s.x v0, zero
|
|
vsetvli zero, a2, e32, m4, ta, ma
|
|
vle32.v v8, (a1)
|
|
sub a4, a4, a2
|
|
vle32.v v16, (a0)
|
|
sh2add a0, a2, a0
|
|
1:
|
|
vwmul.vv v24, v8, v16
|
|
vsetvli zero, zero, e64, m8, ta, ma
|
|
vredsum.vs v24, v24, v0
|
|
lw t0, (a0)
|
|
addi a4, a4, -1
|
|
vmv.x.s t1, v24
|
|
vsetvli zero, zero, e32, m4, ta, ma
|
|
sra t1, t1, a3
|
|
add t0, t0, t1
|
|
vslide1down.vx v16, v16, t0
|
|
sw t0, (a0)
|
|
addi a0, a0, 4
|
|
bnez a4, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_indep2_16_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m8, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v8, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vsll.vx v8, v8, a4
|
|
vsetvli zero, zero, e16, m4, ta, ma
|
|
vncvt.x.x.w v16, v0
|
|
vncvt.x.x.w v20, v8
|
|
vsseg2e16.v v16, (a0)
|
|
sh2add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_indep4_16_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld t1, 16(a1)
|
|
ld t2, 24(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m4, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v4, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vle32.v v8, (t1)
|
|
sh2add t1, t0, t1
|
|
vsll.vx v4, v4, a4
|
|
vle32.v v12, (t2)
|
|
sh2add t2, t0, t2
|
|
vsll.vx v8, v8, a4
|
|
vsll.vx v12, v12, a4
|
|
vsetvli zero, zero, e16, m2, ta, ma
|
|
vncvt.x.x.w v16, v0
|
|
vncvt.x.x.w v18, v4
|
|
vncvt.x.x.w v20, v8
|
|
vncvt.x.x.w v22, v12
|
|
vsseg4e16.v v16, (a0)
|
|
sh3add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_indep6_16_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld t1, 16(a1)
|
|
ld t2, 24(a1)
|
|
ld t3, 32(a1)
|
|
ld t4, 40(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m2, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v2, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vle32.v v4, (t1)
|
|
sh2add t1, t0, t1
|
|
vsll.vx v2, v2, a4
|
|
vle32.v v6, (t2)
|
|
sh2add t2, t0, t2
|
|
vsll.vx v4, v4, a4
|
|
vle32.v v8, (t3)
|
|
sh2add t3, t0, t3
|
|
vsll.vx v6, v6, a4
|
|
vle32.v v10, (t4)
|
|
sh2add t4, t0, t4
|
|
vsll.vx v8, v8, a4
|
|
slli t0, t0, 2
|
|
vsll.vx v10, v10, a4
|
|
sh1add t0, t0, t0 // t0 *= 3
|
|
vsetvli zero, zero, e16, m1, ta, ma
|
|
vncvt.x.x.w v16, v0
|
|
vncvt.x.x.w v17, v2
|
|
vncvt.x.x.w v18, v4
|
|
vncvt.x.x.w v19, v6
|
|
vncvt.x.x.w v20, v8
|
|
vncvt.x.x.w v21, v10
|
|
vsseg6e16.v v16, (a0)
|
|
add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_indep8_16_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld t1, 16(a1)
|
|
ld t2, 24(a1)
|
|
ld t3, 32(a1)
|
|
ld t4, 40(a1)
|
|
ld t5, 48(a1)
|
|
ld t6, 56(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m2, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v2, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
vle32.v v4, (t1)
|
|
sh2add a2, t0, a2
|
|
vsll.vx v2, v2, a4
|
|
sh2add t1, t0, t1
|
|
vle32.v v6, (t2)
|
|
vsll.vx v4, v4, a4
|
|
sh2add t2, t0, t2
|
|
vle32.v v8, (t3)
|
|
sh2add t3, t0, t3
|
|
vsll.vx v6, v6, a4
|
|
vle32.v v10, (t4)
|
|
sh2add t4, t0, t4
|
|
vsll.vx v8, v8, a4
|
|
vle32.v v12, (t5)
|
|
sh2add t5, t0, t5
|
|
vsll.vx v10, v10, a4
|
|
vle32.v v14, (t6)
|
|
sh2add t6, t0, t6
|
|
vsll.vx v12, v12, a4
|
|
slli t0, t0, 4
|
|
vsll.vx v14, v14, a4
|
|
vsetvli zero, zero, e16, m1, ta, ma
|
|
vncvt.x.x.w v16, v0
|
|
vncvt.x.x.w v17, v2
|
|
vncvt.x.x.w v18, v4
|
|
vncvt.x.x.w v19, v6
|
|
vncvt.x.x.w v20, v8
|
|
vncvt.x.x.w v21, v10
|
|
vncvt.x.x.w v22, v12
|
|
vncvt.x.x.w v23, v14
|
|
vsseg8e16.v v16, (a0)
|
|
add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
|
|
|
|
func ff_flac_decorrelate_ls_16_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m8, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v8, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vsll.vx v8, v8, a4
|
|
vsub.vv v8, v0, v8
|
|
vsetvli zero, zero, e16, m4, ta, ma
|
|
vncvt.x.x.w v16, v0
|
|
vncvt.x.x.w v20, v8
|
|
vsseg2e16.v v16, (a0)
|
|
sh2add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_rs_16_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m8, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v8, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vsll.vx v8, v8, a4
|
|
vadd.vv v0, v0, v8
|
|
vsetvli zero, zero, e16, m4, ta, ma
|
|
vncvt.x.x.w v16, v0
|
|
vncvt.x.x.w v20, v8
|
|
vsseg2e16.v v16, (a0)
|
|
sh2add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_ms_16_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m8, ta, ma
|
|
vle32.v v8, (a2)
|
|
sub a3, a3, t0
|
|
vle32.v v0, (a1)
|
|
sh2add a1, t0, a1
|
|
vsra.vi v16, v8, 1
|
|
sh2add a2, t0, a2
|
|
vsub.vv v24, v0, v16
|
|
vadd.vv v16, v24, v8
|
|
vsll.vx v8, v24, a4
|
|
vsll.vx v0, v16, a4
|
|
vsetvli zero, zero, e16, m4, ta, ma
|
|
vncvt.x.x.w v0, v0
|
|
vncvt.x.x.w v4, v8
|
|
vsseg2e16.v v0, (a0)
|
|
sh2add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_indep2_32_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m4, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v4, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vsll.vx v4, v4, a4
|
|
vsseg2e32.v v0, (a0)
|
|
sh3add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_indep4_32_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld t1, 16(a1)
|
|
ld t2, 24(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m2, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v2, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vle32.v v4, (t1)
|
|
sh2add t1, t0, t1
|
|
vsll.vx v2, v2, a4
|
|
vle32.v v6, (t2)
|
|
sh2add t2, t0, t2
|
|
vsll.vx v4, v4, a4
|
|
slli t0, t0, 4
|
|
vsll.vx v6, v6, a4
|
|
vsseg4e32.v v0, (a0)
|
|
add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_indep6_32_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld t1, 16(a1)
|
|
ld t2, 24(a1)
|
|
ld t3, 32(a1)
|
|
ld t4, 40(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m1, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v1, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vle32.v v2, (t1)
|
|
sh2add t1, t0, t1
|
|
vsll.vx v1, v1, a4
|
|
vle32.v v3, (t2)
|
|
sh2add t2, t0, t2
|
|
vsll.vx v2, v2, a4
|
|
vle32.v v4, (t3)
|
|
sh2add t3, t0, t3
|
|
vsll.vx v3, v3, a4
|
|
vle32.v v5, (t4)
|
|
sh2add t4, t0, t4
|
|
vsll.vx v4, v4, a4
|
|
slli t0, t0, 3
|
|
vsll.vx v5, v5, a4
|
|
sh1add t0, t0, t0 // t0 *= 3
|
|
vsseg6e32.v v0, (a0)
|
|
add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_indep8_32_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld t1, 16(a1)
|
|
ld t2, 24(a1)
|
|
ld t3, 32(a1)
|
|
ld t4, 40(a1)
|
|
ld t5, 48(a1)
|
|
ld t6, 56(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m1, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v1, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
vle32.v v2, (t1)
|
|
sh2add a2, t0, a2
|
|
vsll.vx v1, v1, a4
|
|
sh2add t1, t0, t1
|
|
vle32.v v3, (t2)
|
|
vsll.vx v2, v2, a4
|
|
sh2add t2, t0, t2
|
|
vle32.v v4, (t3)
|
|
sh2add t3, t0, t3
|
|
vsll.vx v3, v3, a4
|
|
vle32.v v5, (t4)
|
|
sh2add t4, t0, t4
|
|
vsll.vx v4, v4, a4
|
|
vle32.v v6, (t5)
|
|
sh2add t5, t0, t5
|
|
vsll.vx v5, v5, a4
|
|
vle32.v v7, (t6)
|
|
sh2add t6, t0, t6
|
|
vsll.vx v6, v6, a4
|
|
slli t0, t0, 5
|
|
vsll.vx v7, v7, a4
|
|
vsseg8e32.v v0, (a0)
|
|
add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_ls_32_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m4, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v4, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vsll.vx v4, v4, a4
|
|
vsub.vv v4, v0, v4
|
|
vsseg2e32.v v0, (a0)
|
|
sh3add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_rs_32_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m4, ta, ma
|
|
vle32.v v0, (a1)
|
|
sub a3, a3, t0
|
|
vle32.v v4, (a2)
|
|
sh2add a1, t0, a1
|
|
vsll.vx v0, v0, a4
|
|
sh2add a2, t0, a2
|
|
vsll.vx v4, v4, a4
|
|
vadd.vv v0, v0, v4
|
|
vsseg2e32.v v0, (a0)
|
|
sh3add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
|
|
func ff_flac_decorrelate_ms_32_rvv, zve32x
|
|
ld a0, (a0)
|
|
ld a2, 8(a1)
|
|
ld a1, (a1)
|
|
1:
|
|
vsetvli t0, a3, e32, m4, ta, ma
|
|
vle32.v v4, (a2)
|
|
sub a3, a3, t0
|
|
vle32.v v0, (a1)
|
|
sh2add a1, t0, a1
|
|
vsra.vi v8, v4, 1
|
|
sh2add a2, t0, a2
|
|
vsub.vv v12, v0, v8
|
|
vadd.vv v8, v12, v4
|
|
vsll.vx v4, v12, a4
|
|
vsll.vx v0, v8, a4
|
|
vsseg2e32.v v0, (a0)
|
|
sh3add a0, t0, a0
|
|
bnez a3, 1b
|
|
|
|
ret
|
|
endfunc
|
|
#endif
|