ffmpeg/libswscale/riscv/input_rvv.S

310 lines
8.4 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"
func ff_bgr24ToY_rvv, zve32x
lw t1, 8(a5) # BY
lw t3, 0(a5) # RY
j 1f
endfunc
func ff_rgb24ToY_rvv, zve32x
lw t1, 0(a5) # RY
lw t3, 8(a5) # BY
1:
lw t2, 4(a5) # GY
li t4, (32 << (15 - 1)) + (1 << (15 - 7))
2:
vsetvli t0, a4, e32, m8, ta, ma
vlseg3e8.v v0, (a1)
sub a4, a4, t0
vzext.vf4 v8, v0
sh1add t5, t0, t0 # t1 = 3 * t0
vzext.vf4 v16, v2
vzext.vf4 v24, v4
add a1, t5, a1
vmul.vx v8, v8, t1
vmacc.vx v8, t2, v16
vmacc.vx v8, t3, v24
vadd.vx v8, v8, t4
vsetvli zero, zero, e16, m4, ta, ma
vnsra.wi v0, v8, 15 - 6
vse16.v v0, (a0)
sh1add a0, t0, a0
bnez a4, 2b
ret
endfunc
func ff_bgr24ToUV_rvv, zve32x
lw t1, 20(a6) # BU
lw t4, 32(a6) # BV
lw t3, 12(a6) # RU
lw t6, 24(a6) # RV
j 1f
endfunc
func ff_rgb24ToUV_rvv, zve32x
lw t1, 12(a6) # RU
lw t4, 24(a6) # RV
lw t3, 20(a6) # BU
lw t6, 32(a6) # BV
1:
lw t2, 16(a6) # GU
lw t5, 28(a6) # GV
li a7, (256 << (15 - 1)) + (1 << (15 - 7))
2:
vsetvli t0, a5, e32, m8, ta, ma
vlseg3e8.v v0, (a3)
sub a5, a5, t0
vzext.vf4 v16, v0
sh1add a6, t0, t0
vzext.vf4 v24, v2
vmul.vx v8, v16, t1
add a3, a6, a3
vmul.vx v16, v16, t4
vmacc.vx v8, t2, v24
vmacc.vx v16, t5, v24
vzext.vf4 v24, v4
vadd.vx v8, v8, a7
vadd.vx v16, v16, a7
vmacc.vx v8, t3, v24
vmacc.vx v16, t6, v24
vsetvli zero, zero, e16, m4, ta, ma
vnsra.wi v0, v8, 15 - 6
vnsra.wi v4, v16, 15 - 6
vse16.v v0, (a0)
sh1add a0, t0, a0
vse16.v v4, (a1)
sh1add a1, t0, a1
bnez a5, 2b
ret
endfunc
func ff_bgr24ToUV_half_rvv, zve32x
lw t1, 20(a6) # BU
lw t4, 32(a6) # BV
lw t3, 12(a6) # RU
lw t6, 24(a6) # RV
j 1f
endfunc
func ff_rgb24ToUV_half_rvv, zve32x
lw t1, 12(a6) # RU
lw t4, 24(a6) # RV
lw t3, 20(a6) # BU
lw t6, 32(a6) # BV
1:
lw t2, 16(a6) # GU
lw t5, 28(a6) # GV
li a7, (256 << 15) + (1 << (15 - 6))
2:
vsetvli t0, a5, e8, m1, ta, ma
vlseg6e8.v v0, (a3)
sh1add a6, t0, t0
vwaddu.vv v8, v0, v3
sub a5, a5, t0
vwaddu.vv v10, v1, v4
sh1add a3, a6, a3
vwaddu.vv v12, v2, v5
vsetvli zero, zero, e32, m4, ta, ma
vzext.vf2 v20, v8
vzext.vf2 v24, v10
vzext.vf2 v28, v12
vmul.vx v0, v20, t1
vmul.vx v4, v20, t4
vmacc.vx v0, t2, v24
vmacc.vx v4, t5, v24
vmacc.vx v0, t3, v28
vmacc.vx v4, t6, v28
vadd.vx v0, v0, a7
vadd.vx v4, v4, a7
vsetvli zero, zero, e16, m2, ta, ma
vnsra.wi v0, v0, 15 - 5
vnsra.wi v2, v4, 15 - 5
vse16.v v0, (a0)
sh1add a0, t0, a0
vse16.v v2, (a1)
sh1add a1, t0, a1
bnez a5, 2b
ret
endfunc
.macro rgba_input chr0, chr1, high
func ff_\chr1\()ToY_rvv, zve32x
lw t1, 8(a5) # BY
lw t3, 0(a5) # RY
j 1f
endfunc
func ff_\chr0\()ToY_rvv, zve32x
lw t1, 0(a5) # RY
lw t3, 8(a5) # BY
1:
lw t2, 4(a5) # GY
li t4, (32 << (15 - 1)) + (1 << (15 - 7))
li t5, 0xff
2:
vsetvli t0, a4, e32, m8, ta, ma
vle32.v v0, (a1)
sub a4, a4, t0
.if \high
vsrl.vi v8, v0, 24
.else
vand.vx v8, v0, t5
.endif
sh2add a1, t0, a1
vsrl.vi v16, v0, 8 * (1 + \high)
vmul.vx v24, v8, t1
vand.vx v16, v16, t5
vsrl.vi v8, v0, 8 * (2 - \high)
vmacc.vx v24, t2, v16
vand.vx v8, v8, t5
vadd.vx v24, v24, t4
vmacc.vx v24, t3, v8
vsetvli zero, zero, e16, m4, ta, ma
vnsra.wi v0, v24, 15 - 6
vse16.v v0, (a0)
sh1add a0, t0, a0
bnez a4, 2b
ret
endfunc
func ff_\chr1\()ToUV_rvv, zve32x
lw t1, 20(a6) # BU
lw t4, 32(a6) # BV
lw t3, 12(a6) # RU
lw t6, 24(a6) # RV
j 1f
endfunc
func ff_\chr0\()ToUV_rvv, zve32x
lw t1, 12(a6) # RU
lw t4, 24(a6) # RV
lw t3, 20(a6) # BU
lw t6, 32(a6) # BV
1:
lw t2, 16(a6) # GU
lw t5, 28(a6) # GV
li a6, 0xff
li a7, (256 << (15 - 1)) + (1 << (15 - 7))
2:
vsetvli t0, a5, e32, m8, ta, ma
vle32.v v0, (a3)
sub a5, a5, t0
.if \high
vsrl.vi v24, v0, 24
.else
vand.vx v24, v0, a6
.endif
sh2add a3, t0, a3
vsrl.vi v8, v0, 8 * (1 + \high)
vmul.vx v16, v24, t1
vand.vx v8, v8, a6
vmul.vx v24, v24, t4
vmacc.vx v16, t2, v8
vsrl.vi v0, v0, 8 * (2 - \high)
vmacc.vx v24, t5, v8
vand.vx v0, v0, a6
vadd.vx v16, v16, a7
vadd.vx v24, v24, a7
vmacc.vx v16, t3, v0
vmacc.vx v24, t6, v0
vsetvli zero, zero, e16, m4, ta, ma
vnsra.wi v0, v16, 15 - 6
vnsra.wi v4, v24, 15 - 6
vse16.v v0, (a0)
sh1add a0, t0, a0
vse16.v v4, (a1)
sh1add a1, t0, a1
bnez a5, 2b
ret
endfunc
func ff_\chr1\()ToUV_half_rvv, zve32x
lw t1, 20(a6) # BU
lw t4, 32(a6) # BV
lw t3, 12(a6) # RU
lw t6, 24(a6) # RV
j 1f
endfunc
func ff_\chr0\()ToUV_half_rvv, zve32x
lw t1, 12(a6) # RU
lw t4, 24(a6) # RV
lw t3, 20(a6) # BU
lw t6, 32(a6) # BV
1:
lw t2, 16(a6) # GU
lw t5, 28(a6) # GV
li a6, 0xff
li a7, (256 << 15) + (1 << (15 - 6))
2:
vsetvli t0, a5, e32, m4, ta, ma
vlseg2e32.v v0, (a3)
sub a5, a5, t0
.if \high
vsrl.vi v8, v0, 24
vsrl.vi v12, v4, 24
.else
vand.vx v8, v0, a6
vand.vx v12, v4, a6
.endif
sh3add a3, t0, a3
vsrl.vi v16, v0, 8 * (1 + \high)
vsrl.vi v20, v4, 8 * (1 + \high)
vsrl.vi v24, v0, 8 * (2 - \high)
vsrl.vi v28, v4, 8 * (2 - \high)
vand.vx v16, v16, a6
vand.vx v20, v20, a6
vand.vx v24, v24, a6
vand.vx v28, v28, a6
vadd.vv v8, v8, v12
vadd.vv v16, v16, v20
vadd.vv v24, v24, v28
vmul.vx v0, v8, t1
vmul.vx v4, v8, t4
vmacc.vx v0, t2, v16
vmacc.vx v4, t5, v16
vmacc.vx v0, t3, v24
vmacc.vx v4, t6, v24
vadd.vx v0, v0, a7
vadd.vx v4, v4, a7
vsetvli zero, zero, e16, m2, ta, ma
vnsra.wi v0, v0, 15 - 5
vnsra.wi v2, v4, 15 - 5
vse16.v v0, (a0)
sh1add a0, t0, a0
vse16.v v2, (a1)
sh1add a1, t0, a1
bnez a5, 2b
ret
endfunc
.endm
rgba_input rgba32, bgra32, 0
rgba_input abgr32, argb32, 1