ffmpeg/libavcodec/riscv/vc1dsp_rvv.S
Rémi Denis-Courmont 324eba69f7 lavc/vc1dsp: use saturating arithmetic for RVV inv_trans_dc
T-Head C908 (cycles):
vc1dsp.vc1_inv_trans_4x4_dc_c:      113.7
vc1dsp.vc1_inv_trans_4x4_dc_rvv_i32: 46.5 (before)
vc1dsp.vc1_inv_trans_4x4_dc_rvv_i32: 45.5 (after)
vc1dsp.vc1_inv_trans_4x8_dc_c:      230.7
vc1dsp.vc1_inv_trans_4x8_dc_rvv_i32: 65.7 (before)
vc1dsp.vc1_inv_trans_4x8_dc_rvv_i32: 52.5 (after)
vc1dsp.vc1_inv_trans_8x4_dc_c:      246.7
vc1dsp.vc1_inv_trans_8x4_dc_rvv_i64: 56.7 (before)
vc1dsp.vc1_inv_trans_8x4_dc_rvv_i64: 45.5 (after)
vc1dsp.vc1_inv_trans_8x8_dc_c:      419.7
vc1dsp.vc1_inv_trans_8x8_dc_rvv_i64: 81.2 (before)
vc1dsp.vc1_inv_trans_8x8_dc_rvv_i64: 53.5 (after)
2024-07-30 18:41:51 +03:00

456 lines
14 KiB
ArmAsm

/*
* Copyright (c) 2023 Institue of Software Chinese Academy of Sciences (ISCAS).
* Copyright (c) 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"
.macro inv_trans_dc rows, cols, w, mat_lmul, row_lmul
func ff_vc1_inv_trans_\cols\()x\rows\()_dc_rvv, zve64x, zba
lpad 0
lh t2, (a2)
li a4, 22 - (5 * \cols) / 4
mul t2, t2, a4
vsetivli zero, \rows, e8, m\row_lmul, ta, ma
vlse\w\().v v0, (a0), a1
addi t2, t2, 4
li a5, 22 - (5 * \rows) / 4
srai t2, t2, 3
mul t2, t2, a5
.if \cols * \rows >= 32
li t0, \cols * \rows
.endif
addi t2, t2, 64
srai t2, t2, 7
.if \rows * \cols >= 32
vsetvli zero, t0, e8, m\mat_lmul, ta, ma
.else
vsetivli zero, \rows * \cols, e8, m\mat_lmul, ta, ma
.endif
bgez t2, 1f
neg t2, t2
vssubu.vx v0, v0, t2
vsetivli zero, \rows, e8, m\row_lmul, ta, ma
vsse\w\().v v0, (a0), a1
ret
1:
vsaddu.vx v0, v0, t2
vsetivli zero, \rows, e8, m\row_lmul, ta, ma
vsse\w\().v v0, (a0), a1
ret
endfunc
.endm
inv_trans_dc 8, 8, 64, 4, f2
inv_trans_dc 4, 8, 64, 2, f4
inv_trans_dc 8, 4, 32, 2, f2
inv_trans_dc 4, 4, 32, 1, f4
.variant_cc ff_vc1_inv_trans_8_rvv
func ff_vc1_inv_trans_8_rvv, zve32x
li t4, 12
vsll.vi v14, v6, 4
li t2, 6
vsll.vi v12, v2, 4
li t5, 15
vmul.vx v8, v0, t4
li t3, 9
# t[2..5] = [6 9 12 15]
vmul.vx v10, v4, t4
vmul.vx v16, v2, t2
vmacc.vx v12, t2, v6 # t3
vnmsac.vx v14, t2, v2 # -t4
vadd.vv v24, v8, v10 # t1
vsub.vv v25, v8, v10 # t2
vadd.vv v28, v24, v12 # t5
vsub.vv v31, v24, v12 # t8
vsub.vv v29, v25, v14 # t6
vadd.vv v30, v25, v14 # t7
vsll.vi v20, v1, 4
vsll.vi v22, v7, 2
vmacc.vx v20, t5, v3
vmacc.vx v22, t3, v5
vsll.vi v21, v3, 2
vsll.vi v23, v5, 4
vadd.vv v24, v20, v22 # t1
vnmsac.vx v21, t5, v1
vmacc.vx v23, t3, v7
vsll.vi v20, v3, 4
vsll.vi v22, v5, 2
vadd.vv v25, v21, v23 # -t2
vnmsac.vx v20, t3, v1
vmacc.vx v22, t5, v7
vsll.vi v21, v1, 2
vsll.vi v23, v7, 4
vsub.vv v26, v22, v20 # t3
vnmsac.vx v21, t3, v3
vnmsac.vx v23, t5, v5
srli t2, t1, 2
vwadd.vv v8, v28, v24
vwsub.vv v10, v29, v25
vsub.vv v27, v21, v23 # t4
vwadd.vv v12, v30, v26
vwadd.vv v14, v31, v27
beqz t2, 1f # faster than 4x add t2=zero
.irp n,31,30,29,28
vadd.vi v\n, v\n, 1
.endr
1:
vwsub.vv v16, v31, v27
vwsub.vv v18, v30, v26
vwadd.vv v20, v29, v25
vwsub.vv v22, v28, v24
vnclip.wx v0, v8, t1
vnclip.wx v1, v10, t1
vnclip.wx v2, v12, t1
vnclip.wx v3, v14, t1
vnclip.wx v4, v16, t1
vnclip.wx v5, v18, t1
vnclip.wx v6, v20, t1
vnclip.wx v7, v22, t1
jr t0
endfunc
.variant_cc ff_vc1_inv_trans_4_rvv
func ff_vc1_inv_trans_4_rvv, zve32x
li t3, 17
vmul.vx v8, v0, t3
li t4, 22
vmul.vx v10, v2, t3
li t2, 10
vmul.vx v26, v1, t4
vmul.vx v27, v3, t4
vadd.vv v24, v8, v10 # t1
vsub.vv v25, v8, v10 # t2
vmacc.vx v26, t2, v3 # t3
vnmsac.vx v27, t2, v1 # t4
vwadd.vv v8, v24, v26
vwsub.vv v10, v25, v27
vwadd.vv v12, v25, v27
vwsub.vv v14, v24, v26
vnclip.wx v0, v8, t1
vnclip.wx v1, v10, t1
vnclip.wx v2, v12, t1
vnclip.wx v3, v14, t1
jr t0
endfunc
func ff_vc1_inv_trans_8x8_rvv, zve32x
lpad 0
csrwi vxrm, 0
vsetivli zero, 8, e16, m1, ta, ma
addi a1, a0, 1 * 8 * 2
vle16.v v0, (a0)
addi a2, a0, 2 * 8 * 2
vle16.v v1, (a1)
addi a3, a0, 3 * 8 * 2
vle16.v v2, (a2)
addi a4, a0, 4 * 8 * 2
vle16.v v3, (a3)
addi a5, a0, 5 * 8 * 2
vle16.v v4, (a4)
addi a6, a0, 6 * 8 * 2
vle16.v v5, (a5)
addi a7, a0, 7 * 8 * 2
vle16.v v6, (a6)
vle16.v v7, (a7)
li t1, 3
jal t0, ff_vc1_inv_trans_8_rvv
vsseg8e16.v v0, (a0)
.irp n,0,1,2,3,4,5,6,7
vle16.v v\n, (a\n)
.endr
li t1, 7
jal t0, ff_vc1_inv_trans_8_rvv
vse16.v v0, (a0)
vse16.v v1, (a1)
vse16.v v2, (a2)
vse16.v v3, (a3)
vse16.v v4, (a4)
vse16.v v5, (a5)
vse16.v v6, (a6)
vse16.v v7, (a7)
ret
endfunc
func ff_vc1_inv_trans_8x4_rvv, zve32x
lpad 0
csrwi vxrm, 0
vsetivli zero, 4, e16, mf2, ta, ma
vlseg8e16.v v0, (a2)
li t1, 3
jal t0, ff_vc1_inv_trans_8_rvv
vsseg8e16.v v0, (a2)
addi a3, a2, 1 * 8 * 2
vsetivli zero, 8, e16, m1, ta, ma
vle16.v v0, (a2)
addi a4, a2, 2 * 8 * 2
vle16.v v1, (a3)
addi a5, a2, 3 * 8 * 2
vle16.v v2, (a4)
vle16.v v3, (a5)
li t1, 7
jal t0, ff_vc1_inv_trans_4_rvv
add a3, a1, a0
vle8.v v8, (a0)
add a4, a1, a3
vle8.v v9, (a3)
add a5, a1, a4
vle8.v v10, (a4)
vle8.v v11, (a5)
vsetvli zero, zero, e8, mf2, ta, ma
vwaddu.wv v0, v0, v8
vwaddu.wv v1, v1, v9
vwaddu.wv v2, v2, v10
vwaddu.wv v3, v3, v11
vsetvli zero, zero, e16, m1, ta, ma
.irp n,0,1,2,3
vmax.vx v\n, v\n, zero
.endr
vsetvli zero, zero, e8, mf2, ta, ma
vnclipu.wi v8, v0, 0
vnclipu.wi v9, v1, 0
vse8.v v8, (a0)
vnclipu.wi v10, v2, 0
vse8.v v9, (a3)
vnclipu.wi v11, v3, 0
vse8.v v10, (a4)
vse8.v v11, (a5)
ret
endfunc
func ff_vc1_inv_trans_4x8_rvv, zve32x
lpad 0
li a3, 8 * 2
csrwi vxrm, 0
vsetivli zero, 8, e16, m1, ta, ma
vlsseg4e16.v v0, (a2), a3
li t1, 3
jal t0, ff_vc1_inv_trans_4_rvv
vssseg4e16.v v0, (a2), a3
vsetivli zero, 4, e16, mf2, ta, ma
addi t1, a2, 1 * 8 * 2
vle16.v v0, (a2)
addi t2, a2, 2 * 8 * 2
vle16.v v1, (t1)
addi t3, a2, 3 * 8 * 2
vle16.v v2, (t2)
addi t4, a2, 4 * 8 * 2
vle16.v v3, (t3)
addi t5, a2, 5 * 8 * 2
vle16.v v4, (t4)
addi t6, a2, 6 * 8 * 2
vle16.v v5, (t5)
addi t1, a2, 7 * 8 * 2
vle16.v v6, (t6)
vle16.v v7, (t1)
li t1, 7
jal t0, ff_vc1_inv_trans_8_rvv
add t0, a1, a0
vle8.v v8, (a0)
add t1, a1, t0
vle8.v v9, (t0)
add t2, a1, t1
vle8.v v10, (t1)
add t3, a1, t2
vle8.v v11, (t2)
add t4, a1, t3
vle8.v v12, (t3)
add t5, a1, t4
vle8.v v13, (t4)
add t6, a1, t5
vle8.v v14, (t5)
vle8.v v15, (t6)
vsetvli zero, zero, e8, mf4, ta, ma
vwaddu.wv v0, v0, v8
vwaddu.wv v1, v1, v9
vwaddu.wv v2, v2, v10
vwaddu.wv v3, v3, v11
vwaddu.wv v4, v4, v12
vwaddu.wv v5, v5, v13
vwaddu.wv v6, v6, v14
vwaddu.wv v7, v7, v15
vsetvli zero, zero, e16, mf2, ta, ma
.irp n,0,1,2,3,4,5,6,7
vmax.vx v\n, v\n, zero
.endr
vsetvli zero, zero, e8, mf4, ta, ma
vnclipu.wi v8, v0, 0
vnclipu.wi v9, v1, 0
vse8.v v8, (a0)
vnclipu.wi v10, v2, 0
vse8.v v9, (t0)
vnclipu.wi v11, v3, 0
vse8.v v10, (t1)
vnclipu.wi v12, v4, 0
vse8.v v11, (t2)
vnclipu.wi v13, v5, 0
vse8.v v12, (t3)
vnclipu.wi v14, v6, 0
vse8.v v13, (t4)
vnclipu.wi v15, v7, 0
vse8.v v14, (t5)
vse8.v v15, (t6)
ret
endfunc
func ff_vc1_inv_trans_4x4_rvv, zve32x
lpad 0
li a3, 8 * 2
csrwi vxrm, 0
vsetivli zero, 4, e16, mf2, ta, ma
vlsseg4e16.v v0, (a2), a3
li t1, 3
jal t0, ff_vc1_inv_trans_4_rvv
vssseg4e16.v v0, (a2), a3
addi t1, a2, 2 * 4 * 2
vle16.v v0, (a2)
addi t2, a2, 4 * 4 * 2
vle16.v v1, (t1)
addi t3, a2, 6 * 4 * 2
vle16.v v2, (t2)
vle16.v v3, (t3)
li t1, 7
jal t0, ff_vc1_inv_trans_4_rvv
add t1, a1, a0
vle8.v v8, (a0)
add t2, a1, t1
vle8.v v9, (t1)
add t3, a1, t2
vle8.v v10, (t2)
vle8.v v11, (t3)
vsetvli zero, zero, e8, mf4, ta, ma
vwaddu.wv v0, v0, v8
vwaddu.wv v1, v1, v9
vwaddu.wv v2, v2, v10
vwaddu.wv v3, v3, v11
vsetvli zero, zero, e16, mf2, ta, ma
.irp n,0,1,2,3
vmax.vx v\n, v\n, zero
.endr
vsetvli zero, zero, e8, mf4, ta, ma
vnclipu.wi v8, v0, 0
vnclipu.wi v9, v1, 0
vse8.v v8, (a0)
vnclipu.wi v10, v2, 0
vse8.v v9, (t1)
vnclipu.wi v11, v3, 0
vse8.v v10, (t2)
vse8.v v11, (t3)
ret
endfunc
.macro mspel_op op pos n1 n2
add t1, \pos, a2
v\op\()e8.v v\n1, (\pos)
sh1add \pos, a2, \pos
v\op\()e8.v v\n2, (t1)
.endm
.macro mspel_op_all op pos a1 a2 a3 a4 a5 a6 a7 a8 a9 a10 a11 a12 a13 a14 a15 a16
mspel_op \op \pos \a1 \a2
mspel_op \op \pos \a3 \a4
mspel_op \op \pos \a5 \a6
mspel_op \op \pos \a7 \a8
mspel_op \op \pos \a9 \a10
mspel_op \op \pos \a11 \a12
mspel_op \op \pos \a13 \a14
mspel_op \op \pos \a15 \a16
.endm
func ff_avg_pixels16x16_rvv, zve32x
lpad 0
li t0, 16
vsetivli zero, 16, e8, m1, ta, ma
j 1f
endfunc
func ff_avg_pixels8x8_rvv, zve32x
lpad 0
li t0, 8
vsetivli zero, 8, e8, mf2, ta, ma
1:
csrwi vxrm, 0
2:
vle8.v v16, (a1)
addi t0, t0, -1
vle8.v v8, (a0)
add a1, a1, a2
vaaddu.vv v16, v16, v8
vse8.v v16, (a0)
add a0, a0, a2
bnez t0, 2b
ret
endfunc
func ff_vc1_unescape_buffer_rvv, zve32x
lpad 0
vsetivli zero, 2, e8, m1, ta, ma
vmv.v.i v8, -1
li t4, 1
vmv.v.i v12, -1
li t3, -1
mv t5, a2
blez a1, 3f
1:
vsetvli t0, a1, e8, m4, ta, ma
vle8.v v16, (a0)
vslideup.vi v8, v16, 2
addi t0, t0, -1 # we cannot fully process the last element
vslideup.vi v12, v16, 1
vslide1down.vx v20, v16, t3
vsetvli zero, t0, e8, m4, ta, ma
vmseq.vi v0, v8, 0
vmseq.vi v1, v12, 0
vmseq.vi v2, v16, 3
vmand.mm v0, v0, v1
vmsltu.vi v3, v20, 4
vmand.mm v0, v0, v2
vmand.mm v0, v0, v3
vfirst.m t2, v0
bgez t2, 4f # found an escape byte?
vse8.v v16, (a2)
addi t2, t0, -2
add a2, a2, t0
2:
vslidedown.vx v8, v16, t2
sub a1, a1, t0
vslidedown.vi v12, v8, 1
add a0, a0, t0
bgtu a1, t4, 1b // size > 1
lb t0, (a0)
sb t0, (a2) # copy last byte (cannot be escaped)
addi a2, a2, 1
3:
sub a0, a2, t5
ret
4:
vsetvli zero, t2, e8, m4, ta, ma
vse8.v v16, (a2)
addi t0, t2, 1
add a2, a2, t2
addi t2, t2, -1
vsetvli zero, t0, e8, m4, ta, ma
j 2b
endfunc