mirror of
https://github.com/FFmpeg/FFmpeg.git
synced 2024-12-23 12:43:46 +02:00
fmtconvert: drop unused functions
This commit is contained in:
parent
ee964145b5
commit
d74a8cb7e4
@ -91,7 +91,6 @@ VFP-OBJS += arm/fmtconvert_vfp.o
|
||||
# subsystems
|
||||
VFP-OBJS-$(CONFIG_FFT) += arm/fft_vfp.o
|
||||
VFP-OBJS-$(CONFIG_MDCT) += arm/mdct_vfp.o
|
||||
VFP-OBJS-$(HAVE_ARMV6) += arm/fmtconvert_vfp_armv6.o
|
||||
|
||||
# decoders/encoders
|
||||
VFP-OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_vfp.o \
|
||||
|
@ -34,11 +34,6 @@ void ff_int32_to_float_fmul_array8_vfp(FmtConvertContext *c, float *dst,
|
||||
const int32_t *src, const float *mul,
|
||||
int len);
|
||||
|
||||
void ff_float_to_int16_neon(int16_t *dst, const float *src, long len);
|
||||
void ff_float_to_int16_interleave_neon(int16_t *, const float **, long, int);
|
||||
|
||||
void ff_float_to_int16_vfp(int16_t *dst, const float *src, long len);
|
||||
|
||||
av_cold void ff_fmt_convert_init_arm(FmtConvertContext *c, AVCodecContext *avctx)
|
||||
{
|
||||
int cpu_flags = av_get_cpu_flags();
|
||||
@ -48,18 +43,9 @@ av_cold void ff_fmt_convert_init_arm(FmtConvertContext *c, AVCodecContext *avctx
|
||||
c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_vfp;
|
||||
c->int32_to_float_fmul_array8 = ff_int32_to_float_fmul_array8_vfp;
|
||||
}
|
||||
|
||||
if (have_armv6(cpu_flags)) {
|
||||
c->float_to_int16 = ff_float_to_int16_vfp;
|
||||
}
|
||||
}
|
||||
|
||||
if (have_neon(cpu_flags)) {
|
||||
c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_neon;
|
||||
|
||||
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
|
||||
c->float_to_int16 = ff_float_to_int16_neon;
|
||||
c->float_to_int16_interleave = ff_float_to_int16_interleave_neon;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
@ -22,347 +22,6 @@
|
||||
#include "config.h"
|
||||
#include "libavutil/arm/asm.S"
|
||||
|
||||
function ff_float_to_int16_neon, export=1
|
||||
subs r2, r2, #8
|
||||
vld1.64 {d0-d1}, [r1,:128]!
|
||||
vcvt.s32.f32 q8, q0, #16
|
||||
vld1.64 {d2-d3}, [r1,:128]!
|
||||
vcvt.s32.f32 q9, q1, #16
|
||||
beq 3f
|
||||
bics ip, r2, #15
|
||||
beq 2f
|
||||
1: subs ip, ip, #16
|
||||
vshrn.s32 d4, q8, #16
|
||||
vld1.64 {d0-d1}, [r1,:128]!
|
||||
vcvt.s32.f32 q0, q0, #16
|
||||
vshrn.s32 d5, q9, #16
|
||||
vld1.64 {d2-d3}, [r1,:128]!
|
||||
vcvt.s32.f32 q1, q1, #16
|
||||
vshrn.s32 d6, q0, #16
|
||||
vst1.64 {d4-d5}, [r0,:128]!
|
||||
vshrn.s32 d7, q1, #16
|
||||
vld1.64 {d16-d17},[r1,:128]!
|
||||
vcvt.s32.f32 q8, q8, #16
|
||||
vld1.64 {d18-d19},[r1,:128]!
|
||||
vcvt.s32.f32 q9, q9, #16
|
||||
vst1.64 {d6-d7}, [r0,:128]!
|
||||
bne 1b
|
||||
ands r2, r2, #15
|
||||
beq 3f
|
||||
2: vld1.64 {d0-d1}, [r1,:128]!
|
||||
vshrn.s32 d4, q8, #16
|
||||
vcvt.s32.f32 q0, q0, #16
|
||||
vld1.64 {d2-d3}, [r1,:128]!
|
||||
vshrn.s32 d5, q9, #16
|
||||
vcvt.s32.f32 q1, q1, #16
|
||||
vshrn.s32 d6, q0, #16
|
||||
vst1.64 {d4-d5}, [r0,:128]!
|
||||
vshrn.s32 d7, q1, #16
|
||||
vst1.64 {d6-d7}, [r0,:128]!
|
||||
bx lr
|
||||
3: vshrn.s32 d4, q8, #16
|
||||
vshrn.s32 d5, q9, #16
|
||||
vst1.64 {d4-d5}, [r0,:128]!
|
||||
bx lr
|
||||
endfunc
|
||||
|
||||
function ff_float_to_int16_interleave_neon, export=1
|
||||
cmp r3, #2
|
||||
itt lt
|
||||
ldrlt r1, [r1]
|
||||
blt X(ff_float_to_int16_neon)
|
||||
bne 4f
|
||||
|
||||
ldr r3, [r1]
|
||||
ldr r1, [r1, #4]
|
||||
|
||||
subs r2, r2, #8
|
||||
vld1.64 {d0-d1}, [r3,:128]!
|
||||
vcvt.s32.f32 q8, q0, #16
|
||||
vld1.64 {d2-d3}, [r3,:128]!
|
||||
vcvt.s32.f32 q9, q1, #16
|
||||
vld1.64 {d20-d21},[r1,:128]!
|
||||
vcvt.s32.f32 q10, q10, #16
|
||||
vld1.64 {d22-d23},[r1,:128]!
|
||||
vcvt.s32.f32 q11, q11, #16
|
||||
beq 3f
|
||||
bics ip, r2, #15
|
||||
beq 2f
|
||||
1: subs ip, ip, #16
|
||||
vld1.64 {d0-d1}, [r3,:128]!
|
||||
vcvt.s32.f32 q0, q0, #16
|
||||
vsri.32 q10, q8, #16
|
||||
vld1.64 {d2-d3}, [r3,:128]!
|
||||
vcvt.s32.f32 q1, q1, #16
|
||||
vld1.64 {d24-d25},[r1,:128]!
|
||||
vcvt.s32.f32 q12, q12, #16
|
||||
vld1.64 {d26-d27},[r1,:128]!
|
||||
vsri.32 q11, q9, #16
|
||||
vst1.64 {d20-d21},[r0,:128]!
|
||||
vcvt.s32.f32 q13, q13, #16
|
||||
vst1.64 {d22-d23},[r0,:128]!
|
||||
vsri.32 q12, q0, #16
|
||||
vld1.64 {d16-d17},[r3,:128]!
|
||||
vsri.32 q13, q1, #16
|
||||
vst1.64 {d24-d25},[r0,:128]!
|
||||
vcvt.s32.f32 q8, q8, #16
|
||||
vld1.64 {d18-d19},[r3,:128]!
|
||||
vcvt.s32.f32 q9, q9, #16
|
||||
vld1.64 {d20-d21},[r1,:128]!
|
||||
vcvt.s32.f32 q10, q10, #16
|
||||
vld1.64 {d22-d23},[r1,:128]!
|
||||
vcvt.s32.f32 q11, q11, #16
|
||||
vst1.64 {d26-d27},[r0,:128]!
|
||||
bne 1b
|
||||
ands r2, r2, #15
|
||||
beq 3f
|
||||
2: vsri.32 q10, q8, #16
|
||||
vld1.64 {d0-d1}, [r3,:128]!
|
||||
vcvt.s32.f32 q0, q0, #16
|
||||
vld1.64 {d2-d3}, [r3,:128]!
|
||||
vcvt.s32.f32 q1, q1, #16
|
||||
vld1.64 {d24-d25},[r1,:128]!
|
||||
vcvt.s32.f32 q12, q12, #16
|
||||
vsri.32 q11, q9, #16
|
||||
vld1.64 {d26-d27},[r1,:128]!
|
||||
vcvt.s32.f32 q13, q13, #16
|
||||
vst1.64 {d20-d21},[r0,:128]!
|
||||
vsri.32 q12, q0, #16
|
||||
vst1.64 {d22-d23},[r0,:128]!
|
||||
vsri.32 q13, q1, #16
|
||||
vst1.64 {d24-d27},[r0,:128]!
|
||||
bx lr
|
||||
3: vsri.32 q10, q8, #16
|
||||
vsri.32 q11, q9, #16
|
||||
vst1.64 {d20-d23},[r0,:128]!
|
||||
bx lr
|
||||
|
||||
4: push {r4-r8,lr}
|
||||
cmp r3, #4
|
||||
lsl ip, r3, #1
|
||||
blt 4f
|
||||
|
||||
@ 4 channels
|
||||
5: ldmia r1!, {r4-r7}
|
||||
mov lr, r2
|
||||
mov r8, r0
|
||||
vld1.64 {d16-d17},[r4,:128]!
|
||||
vcvt.s32.f32 q8, q8, #16
|
||||
vld1.64 {d18-d19},[r5,:128]!
|
||||
vcvt.s32.f32 q9, q9, #16
|
||||
vld1.64 {d20-d21},[r6,:128]!
|
||||
vcvt.s32.f32 q10, q10, #16
|
||||
vld1.64 {d22-d23},[r7,:128]!
|
||||
vcvt.s32.f32 q11, q11, #16
|
||||
6: subs lr, lr, #8
|
||||
vld1.64 {d0-d1}, [r4,:128]!
|
||||
vcvt.s32.f32 q0, q0, #16
|
||||
vsri.32 q9, q8, #16
|
||||
vld1.64 {d2-d3}, [r5,:128]!
|
||||
vcvt.s32.f32 q1, q1, #16
|
||||
vsri.32 q11, q10, #16
|
||||
vld1.64 {d4-d5}, [r6,:128]!
|
||||
vcvt.s32.f32 q2, q2, #16
|
||||
vzip.32 d18, d22
|
||||
vld1.64 {d6-d7}, [r7,:128]!
|
||||
vcvt.s32.f32 q3, q3, #16
|
||||
vzip.32 d19, d23
|
||||
vst1.64 {d18}, [r8], ip
|
||||
vsri.32 q1, q0, #16
|
||||
vst1.64 {d22}, [r8], ip
|
||||
vsri.32 q3, q2, #16
|
||||
vst1.64 {d19}, [r8], ip
|
||||
vzip.32 d2, d6
|
||||
vst1.64 {d23}, [r8], ip
|
||||
vzip.32 d3, d7
|
||||
beq 7f
|
||||
vld1.64 {d16-d17},[r4,:128]!
|
||||
vcvt.s32.f32 q8, q8, #16
|
||||
vst1.64 {d2}, [r8], ip
|
||||
vld1.64 {d18-d19},[r5,:128]!
|
||||
vcvt.s32.f32 q9, q9, #16
|
||||
vst1.64 {d6}, [r8], ip
|
||||
vld1.64 {d20-d21},[r6,:128]!
|
||||
vcvt.s32.f32 q10, q10, #16
|
||||
vst1.64 {d3}, [r8], ip
|
||||
vld1.64 {d22-d23},[r7,:128]!
|
||||
vcvt.s32.f32 q11, q11, #16
|
||||
vst1.64 {d7}, [r8], ip
|
||||
b 6b
|
||||
7: vst1.64 {d2}, [r8], ip
|
||||
vst1.64 {d6}, [r8], ip
|
||||
vst1.64 {d3}, [r8], ip
|
||||
vst1.64 {d7}, [r8], ip
|
||||
subs r3, r3, #4
|
||||
it eq
|
||||
popeq {r4-r8,pc}
|
||||
cmp r3, #4
|
||||
add r0, r0, #8
|
||||
bge 5b
|
||||
|
||||
@ 2 channels
|
||||
4: cmp r3, #2
|
||||
blt 4f
|
||||
ldmia r1!, {r4-r5}
|
||||
mov lr, r2
|
||||
mov r8, r0
|
||||
tst lr, #8
|
||||
vld1.64 {d16-d17},[r4,:128]!
|
||||
vcvt.s32.f32 q8, q8, #16
|
||||
vld1.64 {d18-d19},[r5,:128]!
|
||||
vcvt.s32.f32 q9, q9, #16
|
||||
vld1.64 {d20-d21},[r4,:128]!
|
||||
vcvt.s32.f32 q10, q10, #16
|
||||
vld1.64 {d22-d23},[r5,:128]!
|
||||
vcvt.s32.f32 q11, q11, #16
|
||||
beq 6f
|
||||
subs lr, lr, #8
|
||||
beq 7f
|
||||
vsri.32 d18, d16, #16
|
||||
vsri.32 d19, d17, #16
|
||||
vld1.64 {d16-d17},[r4,:128]!
|
||||
vcvt.s32.f32 q8, q8, #16
|
||||
vst1.32 {d18[0]}, [r8], ip
|
||||
vsri.32 d22, d20, #16
|
||||
vst1.32 {d18[1]}, [r8], ip
|
||||
vsri.32 d23, d21, #16
|
||||
vst1.32 {d19[0]}, [r8], ip
|
||||
vst1.32 {d19[1]}, [r8], ip
|
||||
vld1.64 {d18-d19},[r5,:128]!
|
||||
vcvt.s32.f32 q9, q9, #16
|
||||
vst1.32 {d22[0]}, [r8], ip
|
||||
vst1.32 {d22[1]}, [r8], ip
|
||||
vld1.64 {d20-d21},[r4,:128]!
|
||||
vcvt.s32.f32 q10, q10, #16
|
||||
vst1.32 {d23[0]}, [r8], ip
|
||||
vst1.32 {d23[1]}, [r8], ip
|
||||
vld1.64 {d22-d23},[r5,:128]!
|
||||
vcvt.s32.f32 q11, q11, #16
|
||||
6: subs lr, lr, #16
|
||||
vld1.64 {d0-d1}, [r4,:128]!
|
||||
vcvt.s32.f32 q0, q0, #16
|
||||
vsri.32 d18, d16, #16
|
||||
vld1.64 {d2-d3}, [r5,:128]!
|
||||
vcvt.s32.f32 q1, q1, #16
|
||||
vsri.32 d19, d17, #16
|
||||
vld1.64 {d4-d5}, [r4,:128]!
|
||||
vcvt.s32.f32 q2, q2, #16
|
||||
vld1.64 {d6-d7}, [r5,:128]!
|
||||
vcvt.s32.f32 q3, q3, #16
|
||||
vst1.32 {d18[0]}, [r8], ip
|
||||
vsri.32 d22, d20, #16
|
||||
vst1.32 {d18[1]}, [r8], ip
|
||||
vsri.32 d23, d21, #16
|
||||
vst1.32 {d19[0]}, [r8], ip
|
||||
vsri.32 d2, d0, #16
|
||||
vst1.32 {d19[1]}, [r8], ip
|
||||
vsri.32 d3, d1, #16
|
||||
vst1.32 {d22[0]}, [r8], ip
|
||||
vsri.32 d6, d4, #16
|
||||
vst1.32 {d22[1]}, [r8], ip
|
||||
vsri.32 d7, d5, #16
|
||||
vst1.32 {d23[0]}, [r8], ip
|
||||
vst1.32 {d23[1]}, [r8], ip
|
||||
beq 6f
|
||||
vld1.64 {d16-d17},[r4,:128]!
|
||||
vcvt.s32.f32 q8, q8, #16
|
||||
vst1.32 {d2[0]}, [r8], ip
|
||||
vst1.32 {d2[1]}, [r8], ip
|
||||
vld1.64 {d18-d19},[r5,:128]!
|
||||
vcvt.s32.f32 q9, q9, #16
|
||||
vst1.32 {d3[0]}, [r8], ip
|
||||
vst1.32 {d3[1]}, [r8], ip
|
||||
vld1.64 {d20-d21},[r4,:128]!
|
||||
vcvt.s32.f32 q10, q10, #16
|
||||
vst1.32 {d6[0]}, [r8], ip
|
||||
vst1.32 {d6[1]}, [r8], ip
|
||||
vld1.64 {d22-d23},[r5,:128]!
|
||||
vcvt.s32.f32 q11, q11, #16
|
||||
vst1.32 {d7[0]}, [r8], ip
|
||||
vst1.32 {d7[1]}, [r8], ip
|
||||
bgt 6b
|
||||
6: vst1.32 {d2[0]}, [r8], ip
|
||||
vst1.32 {d2[1]}, [r8], ip
|
||||
vst1.32 {d3[0]}, [r8], ip
|
||||
vst1.32 {d3[1]}, [r8], ip
|
||||
vst1.32 {d6[0]}, [r8], ip
|
||||
vst1.32 {d6[1]}, [r8], ip
|
||||
vst1.32 {d7[0]}, [r8], ip
|
||||
vst1.32 {d7[1]}, [r8], ip
|
||||
b 8f
|
||||
7: vsri.32 d18, d16, #16
|
||||
vsri.32 d19, d17, #16
|
||||
vst1.32 {d18[0]}, [r8], ip
|
||||
vsri.32 d22, d20, #16
|
||||
vst1.32 {d18[1]}, [r8], ip
|
||||
vsri.32 d23, d21, #16
|
||||
vst1.32 {d19[0]}, [r8], ip
|
||||
vst1.32 {d19[1]}, [r8], ip
|
||||
vst1.32 {d22[0]}, [r8], ip
|
||||
vst1.32 {d22[1]}, [r8], ip
|
||||
vst1.32 {d23[0]}, [r8], ip
|
||||
vst1.32 {d23[1]}, [r8], ip
|
||||
8: subs r3, r3, #2
|
||||
add r0, r0, #4
|
||||
it eq
|
||||
popeq {r4-r8,pc}
|
||||
|
||||
@ 1 channel
|
||||
4: ldr r4, [r1],#4
|
||||
tst r2, #8
|
||||
mov lr, r2
|
||||
mov r5, r0
|
||||
vld1.64 {d0-d1}, [r4,:128]!
|
||||
vcvt.s32.f32 q0, q0, #16
|
||||
vld1.64 {d2-d3}, [r4,:128]!
|
||||
vcvt.s32.f32 q1, q1, #16
|
||||
bne 8f
|
||||
6: subs lr, lr, #16
|
||||
vld1.64 {d4-d5}, [r4,:128]!
|
||||
vcvt.s32.f32 q2, q2, #16
|
||||
vld1.64 {d6-d7}, [r4,:128]!
|
||||
vcvt.s32.f32 q3, q3, #16
|
||||
vst1.16 {d0[1]}, [r5,:16], ip
|
||||
vst1.16 {d0[3]}, [r5,:16], ip
|
||||
vst1.16 {d1[1]}, [r5,:16], ip
|
||||
vst1.16 {d1[3]}, [r5,:16], ip
|
||||
vst1.16 {d2[1]}, [r5,:16], ip
|
||||
vst1.16 {d2[3]}, [r5,:16], ip
|
||||
vst1.16 {d3[1]}, [r5,:16], ip
|
||||
vst1.16 {d3[3]}, [r5,:16], ip
|
||||
beq 7f
|
||||
vld1.64 {d0-d1}, [r4,:128]!
|
||||
vcvt.s32.f32 q0, q0, #16
|
||||
vld1.64 {d2-d3}, [r4,:128]!
|
||||
vcvt.s32.f32 q1, q1, #16
|
||||
7: vst1.16 {d4[1]}, [r5,:16], ip
|
||||
vst1.16 {d4[3]}, [r5,:16], ip
|
||||
vst1.16 {d5[1]}, [r5,:16], ip
|
||||
vst1.16 {d5[3]}, [r5,:16], ip
|
||||
vst1.16 {d6[1]}, [r5,:16], ip
|
||||
vst1.16 {d6[3]}, [r5,:16], ip
|
||||
vst1.16 {d7[1]}, [r5,:16], ip
|
||||
vst1.16 {d7[3]}, [r5,:16], ip
|
||||
bgt 6b
|
||||
pop {r4-r8,pc}
|
||||
8: subs lr, lr, #8
|
||||
vst1.16 {d0[1]}, [r5,:16], ip
|
||||
vst1.16 {d0[3]}, [r5,:16], ip
|
||||
vst1.16 {d1[1]}, [r5,:16], ip
|
||||
vst1.16 {d1[3]}, [r5,:16], ip
|
||||
vst1.16 {d2[1]}, [r5,:16], ip
|
||||
vst1.16 {d2[3]}, [r5,:16], ip
|
||||
vst1.16 {d3[1]}, [r5,:16], ip
|
||||
vst1.16 {d3[3]}, [r5,:16], ip
|
||||
it eq
|
||||
popeq {r4-r8,pc}
|
||||
vld1.64 {d0-d1}, [r4,:128]!
|
||||
vcvt.s32.f32 q0, q0, #16
|
||||
vld1.64 {d2-d3}, [r4,:128]!
|
||||
vcvt.s32.f32 q1, q1, #16
|
||||
b 6b
|
||||
endfunc
|
||||
|
||||
function ff_int32_to_float_fmul_scalar_neon, export=1
|
||||
VFP vdup.32 q0, d0[0]
|
||||
VFP len .req r2
|
||||
|
@ -1,78 +0,0 @@
|
||||
/*
|
||||
* Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net>
|
||||
*
|
||||
* This file is part of Libav.
|
||||
*
|
||||
* Libav 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.
|
||||
*
|
||||
* Libav 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 Libav; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#include "config.h"
|
||||
#include "libavutil/arm/asm.S"
|
||||
|
||||
/**
|
||||
* ARM VFP optimized float to int16 conversion.
|
||||
* Assume that len is a positive number and is multiple of 8, destination
|
||||
* buffer is at least 4 bytes aligned (8 bytes alignment is better for
|
||||
* performance), little-endian byte sex.
|
||||
*/
|
||||
@ void ff_float_to_int16_vfp(int16_t *dst, const float *src, int len)
|
||||
function ff_float_to_int16_vfp, export=1
|
||||
push {r4-r8,lr}
|
||||
vpush {d8-d11}
|
||||
vldmia r1!, {s16-s23}
|
||||
vcvt.s32.f32 s0, s16
|
||||
vcvt.s32.f32 s1, s17
|
||||
vcvt.s32.f32 s2, s18
|
||||
vcvt.s32.f32 s3, s19
|
||||
vcvt.s32.f32 s4, s20
|
||||
vcvt.s32.f32 s5, s21
|
||||
vcvt.s32.f32 s6, s22
|
||||
vcvt.s32.f32 s7, s23
|
||||
1:
|
||||
subs r2, r2, #8
|
||||
vmov r3, r4, s0, s1
|
||||
vmov r5, r6, s2, s3
|
||||
vmov r7, r8, s4, s5
|
||||
vmov ip, lr, s6, s7
|
||||
it gt
|
||||
vldmiagt r1!, {s16-s23}
|
||||
ssat r4, #16, r4
|
||||
ssat r3, #16, r3
|
||||
ssat r6, #16, r6
|
||||
ssat r5, #16, r5
|
||||
pkhbt r3, r3, r4, lsl #16
|
||||
pkhbt r4, r5, r6, lsl #16
|
||||
itttt gt
|
||||
vcvtgt.s32.f32 s0, s16
|
||||
vcvtgt.s32.f32 s1, s17
|
||||
vcvtgt.s32.f32 s2, s18
|
||||
vcvtgt.s32.f32 s3, s19
|
||||
itttt gt
|
||||
vcvtgt.s32.f32 s4, s20
|
||||
vcvtgt.s32.f32 s5, s21
|
||||
vcvtgt.s32.f32 s6, s22
|
||||
vcvtgt.s32.f32 s7, s23
|
||||
ssat r8, #16, r8
|
||||
ssat r7, #16, r7
|
||||
ssat lr, #16, lr
|
||||
ssat ip, #16, ip
|
||||
pkhbt r5, r7, r8, lsl #16
|
||||
pkhbt r6, ip, lr, lsl #16
|
||||
stmia r0!, {r3-r6}
|
||||
bgt 1b
|
||||
|
||||
vpop {d8-d11}
|
||||
pop {r4-r8,pc}
|
||||
endfunc
|
@ -41,59 +41,10 @@ static void int32_to_float_fmul_array8_c(FmtConvertContext *c, float *dst,
|
||||
c->int32_to_float_fmul_scalar(&dst[i], &src[i], *mul++, 8);
|
||||
}
|
||||
|
||||
static av_always_inline int float_to_int16_one(const float *src){
|
||||
return av_clip_int16(lrintf(*src));
|
||||
}
|
||||
|
||||
static void float_to_int16_c(int16_t *dst, const float *src, long len)
|
||||
{
|
||||
int i;
|
||||
for(i=0; i<len; i++)
|
||||
dst[i] = float_to_int16_one(src+i);
|
||||
}
|
||||
|
||||
static void float_to_int16_interleave_c(int16_t *dst, const float **src,
|
||||
long len, int channels)
|
||||
{
|
||||
int i,j,c;
|
||||
if(channels==2){
|
||||
for(i=0; i<len; i++){
|
||||
dst[2*i] = float_to_int16_one(src[0]+i);
|
||||
dst[2*i+1] = float_to_int16_one(src[1]+i);
|
||||
}
|
||||
}else{
|
||||
for(c=0; c<channels; c++)
|
||||
for(i=0, j=c; i<len; i++, j+=channels)
|
||||
dst[j] = float_to_int16_one(src[c]+i);
|
||||
}
|
||||
}
|
||||
|
||||
void ff_float_interleave_c(float *dst, const float **src, unsigned int len,
|
||||
int channels)
|
||||
{
|
||||
int j, c;
|
||||
unsigned int i;
|
||||
if (channels == 2) {
|
||||
for (i = 0; i < len; i++) {
|
||||
dst[2*i] = src[0][i];
|
||||
dst[2*i+1] = src[1][i];
|
||||
}
|
||||
} else if (channels == 1 && len < INT_MAX / sizeof(float)) {
|
||||
memcpy(dst, src[0], len * sizeof(float));
|
||||
} else {
|
||||
for (c = 0; c < channels; c++)
|
||||
for (i = 0, j = c; i < len; i++, j += channels)
|
||||
dst[j] = src[c][i];
|
||||
}
|
||||
}
|
||||
|
||||
av_cold void ff_fmt_convert_init(FmtConvertContext *c, AVCodecContext *avctx)
|
||||
{
|
||||
c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_c;
|
||||
c->int32_to_float_fmul_array8 = int32_to_float_fmul_array8_c;
|
||||
c->float_to_int16 = float_to_int16_c;
|
||||
c->float_to_int16_interleave = float_to_int16_interleave_c;
|
||||
c->float_interleave = ff_float_interleave_c;
|
||||
|
||||
if (ARCH_ARM) ff_fmt_convert_init_arm(c, avctx);
|
||||
if (ARCH_PPC) ff_fmt_convert_init_ppc(c, avctx);
|
||||
|
@ -54,56 +54,8 @@ typedef struct FmtConvertContext {
|
||||
float *dst, const int32_t *src,
|
||||
const float *mul, int len);
|
||||
|
||||
/**
|
||||
* Convert an array of float to an array of int16_t.
|
||||
*
|
||||
* Convert floats from in the range [-32768.0,32767.0] to ints
|
||||
* without rescaling
|
||||
*
|
||||
* @param dst destination array of int16_t.
|
||||
* constraints: 16-byte aligned
|
||||
* @param src source array of float.
|
||||
* constraints: 16-byte aligned
|
||||
* @param len number of elements to convert.
|
||||
* constraints: multiple of 8
|
||||
*/
|
||||
void (*float_to_int16)(int16_t *dst, const float *src, long len);
|
||||
|
||||
/**
|
||||
* Convert multiple arrays of float to an interleaved array of int16_t.
|
||||
*
|
||||
* Convert floats from in the range [-32768.0,32767.0] to ints
|
||||
* without rescaling
|
||||
*
|
||||
* @param dst destination array of interleaved int16_t.
|
||||
* constraints: 16-byte aligned
|
||||
* @param src source array of float arrays, one for each channel.
|
||||
* constraints: 16-byte aligned
|
||||
* @param len number of elements to convert.
|
||||
* constraints: multiple of 8
|
||||
* @param channels number of channels
|
||||
*/
|
||||
void (*float_to_int16_interleave)(int16_t *dst, const float **src,
|
||||
long len, int channels);
|
||||
|
||||
/**
|
||||
* Convert multiple arrays of float to an array of interleaved float.
|
||||
*
|
||||
* @param dst destination array of interleaved float.
|
||||
* constraints: 16-byte aligned
|
||||
* @param src source array of float arrays, one for each channel.
|
||||
* constraints: 16-byte aligned
|
||||
* @param len number of elements to convert.
|
||||
* constraints: multiple of 8
|
||||
* @param channels number of channels
|
||||
*/
|
||||
void (*float_interleave)(float *dst, const float **src, unsigned int len,
|
||||
int channels);
|
||||
} FmtConvertContext;
|
||||
|
||||
void ff_float_interleave_c(float *dst, const float **src, unsigned int len,
|
||||
int channels);
|
||||
|
||||
void ff_fmt_convert_init(FmtConvertContext *c, AVCodecContext *avctx);
|
||||
|
||||
void ff_fmt_convert_init_arm(FmtConvertContext *c, AVCodecContext *avctx);
|
||||
|
@ -52,113 +52,6 @@ static void int32_to_float_fmul_scalar_altivec(float *dst, const int32_t *src,
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
static vector signed short float_to_int16_one_altivec(const float *src)
|
||||
{
|
||||
vector float s0 = vec_ld(0, src);
|
||||
vector float s1 = vec_ld(16, src);
|
||||
vector signed int t0 = vec_cts(s0, 0);
|
||||
vector signed int t1 = vec_cts(s1, 0);
|
||||
return vec_packs(t0,t1);
|
||||
}
|
||||
|
||||
static void float_to_int16_altivec(int16_t *dst, const float *src, long len)
|
||||
{
|
||||
int i;
|
||||
vector signed short d0, d1, d;
|
||||
vector unsigned char align;
|
||||
if (((long)dst) & 15) { //FIXME
|
||||
for (i = 0; i < len - 7; i += 8) {
|
||||
d0 = vec_ld(0, dst+i);
|
||||
d = float_to_int16_one_altivec(src + i);
|
||||
d1 = vec_ld(15, dst+i);
|
||||
d1 = vec_perm(d1, d0, vec_lvsl(0, dst + i));
|
||||
align = vec_lvsr(0, dst + i);
|
||||
d0 = vec_perm(d1, d, align);
|
||||
d1 = vec_perm(d, d1, align);
|
||||
vec_st(d0, 0, dst + i);
|
||||
vec_st(d1, 15, dst + i);
|
||||
}
|
||||
} else {
|
||||
for (i = 0; i < len - 7; i += 8) {
|
||||
d = float_to_int16_one_altivec(src + i);
|
||||
vec_st(d, 0, dst + i);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
#define VSTE_INC(dst, v, elem, inc) do { \
|
||||
vector signed short s = vec_splat(v, elem); \
|
||||
vec_ste(s, 0, dst); \
|
||||
dst += inc; \
|
||||
} while (0)
|
||||
|
||||
static void float_to_int16_stride_altivec(int16_t *dst, const float *src,
|
||||
long len, int stride)
|
||||
{
|
||||
int i;
|
||||
vector signed short d;
|
||||
|
||||
for (i = 0; i < len - 7; i += 8) {
|
||||
d = float_to_int16_one_altivec(src + i);
|
||||
VSTE_INC(dst, d, 0, stride);
|
||||
VSTE_INC(dst, d, 1, stride);
|
||||
VSTE_INC(dst, d, 2, stride);
|
||||
VSTE_INC(dst, d, 3, stride);
|
||||
VSTE_INC(dst, d, 4, stride);
|
||||
VSTE_INC(dst, d, 5, stride);
|
||||
VSTE_INC(dst, d, 6, stride);
|
||||
VSTE_INC(dst, d, 7, stride);
|
||||
}
|
||||
}
|
||||
|
||||
static void float_to_int16_interleave_altivec(int16_t *dst, const float **src,
|
||||
long len, int channels)
|
||||
{
|
||||
int i;
|
||||
vector signed short d0, d1, d2, c0, c1, t0, t1;
|
||||
vector unsigned char align;
|
||||
|
||||
if (channels == 1)
|
||||
float_to_int16_altivec(dst, src[0], len);
|
||||
else {
|
||||
if (channels == 2) {
|
||||
if (((long)dst) & 15) {
|
||||
for (i = 0; i < len - 7; i += 8) {
|
||||
d0 = vec_ld(0, dst + i);
|
||||
t0 = float_to_int16_one_altivec(src[0] + i);
|
||||
d1 = vec_ld(31, dst + i);
|
||||
t1 = float_to_int16_one_altivec(src[1] + i);
|
||||
c0 = vec_mergeh(t0, t1);
|
||||
c1 = vec_mergel(t0, t1);
|
||||
d2 = vec_perm(d1, d0, vec_lvsl(0, dst + i));
|
||||
align = vec_lvsr(0, dst + i);
|
||||
d0 = vec_perm(d2, c0, align);
|
||||
d1 = vec_perm(c0, c1, align);
|
||||
vec_st(d0, 0, dst + i);
|
||||
d0 = vec_perm(c1, d2, align);
|
||||
vec_st(d1, 15, dst + i);
|
||||
vec_st(d0, 31, dst + i);
|
||||
dst += 8;
|
||||
}
|
||||
} else {
|
||||
for (i = 0; i < len - 7; i += 8) {
|
||||
t0 = float_to_int16_one_altivec(src[0] + i);
|
||||
t1 = float_to_int16_one_altivec(src[1] + i);
|
||||
d0 = vec_mergeh(t0, t1);
|
||||
d1 = vec_mergel(t0, t1);
|
||||
vec_st(d0, 0, dst + i);
|
||||
vec_st(d1, 16, dst + i);
|
||||
dst += 8;
|
||||
}
|
||||
}
|
||||
} else {
|
||||
for (i = 0; i < channels; i++)
|
||||
float_to_int16_stride_altivec(dst + i, src[i], len, channels);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
#endif /* HAVE_ALTIVEC */
|
||||
|
||||
av_cold void ff_fmt_convert_init_ppc(FmtConvertContext *c,
|
||||
@ -169,9 +62,5 @@ av_cold void ff_fmt_convert_init_ppc(FmtConvertContext *c,
|
||||
return;
|
||||
|
||||
c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_altivec;
|
||||
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
|
||||
c->float_to_int16 = float_to_int16_altivec;
|
||||
c->float_to_int16_interleave = float_to_int16_interleave_altivec;
|
||||
}
|
||||
#endif /* HAVE_ALTIVEC */
|
||||
}
|
||||
|
@ -23,14 +23,6 @@
|
||||
|
||||
SECTION_TEXT
|
||||
|
||||
%macro CVTPS2PI 2
|
||||
%if cpuflag(sse)
|
||||
cvtps2pi %1, %2
|
||||
%elif cpuflag(3dnow)
|
||||
pf2id %1, %2
|
||||
%endif
|
||||
%endmacro
|
||||
|
||||
;------------------------------------------------------------------------------
|
||||
; void ff_int32_to_float_fmul_scalar(float *dst, const int32_t *src, float mul,
|
||||
; int len);
|
||||
@ -76,357 +68,3 @@ INIT_XMM sse
|
||||
INT32_TO_FLOAT_FMUL_SCALAR 5
|
||||
INIT_XMM sse2
|
||||
INT32_TO_FLOAT_FMUL_SCALAR 3
|
||||
|
||||
|
||||
;------------------------------------------------------------------------------
|
||||
; void ff_float_to_int16(int16_t *dst, const float *src, long len);
|
||||
;------------------------------------------------------------------------------
|
||||
%macro FLOAT_TO_INT16 1
|
||||
cglobal float_to_int16, 3, 3, %1, dst, src, len
|
||||
add lenq, lenq
|
||||
lea srcq, [srcq+2*lenq]
|
||||
add dstq, lenq
|
||||
neg lenq
|
||||
.loop:
|
||||
%if cpuflag(sse2)
|
||||
cvtps2dq m0, [srcq+2*lenq ]
|
||||
cvtps2dq m1, [srcq+2*lenq+16]
|
||||
packssdw m0, m1
|
||||
mova [dstq+lenq], m0
|
||||
%else
|
||||
CVTPS2PI m0, [srcq+2*lenq ]
|
||||
CVTPS2PI m1, [srcq+2*lenq+ 8]
|
||||
CVTPS2PI m2, [srcq+2*lenq+16]
|
||||
CVTPS2PI m3, [srcq+2*lenq+24]
|
||||
packssdw m0, m1
|
||||
packssdw m2, m3
|
||||
mova [dstq+lenq ], m0
|
||||
mova [dstq+lenq+8], m2
|
||||
%endif
|
||||
add lenq, 16
|
||||
js .loop
|
||||
%if mmsize == 8
|
||||
emms
|
||||
%endif
|
||||
REP_RET
|
||||
%endmacro
|
||||
|
||||
INIT_XMM sse2
|
||||
FLOAT_TO_INT16 2
|
||||
INIT_MMX sse
|
||||
FLOAT_TO_INT16 0
|
||||
INIT_MMX 3dnow
|
||||
FLOAT_TO_INT16 0
|
||||
|
||||
;------------------------------------------------------------------------------
|
||||
; void ff_float_to_int16_step(int16_t *dst, const float *src, long len, long step);
|
||||
;------------------------------------------------------------------------------
|
||||
%macro FLOAT_TO_INT16_STEP 1
|
||||
cglobal float_to_int16_step, 4, 7, %1, dst, src, len, step, step3, v1, v2
|
||||
add lenq, lenq
|
||||
lea srcq, [srcq+2*lenq]
|
||||
lea step3q, [stepq*3]
|
||||
neg lenq
|
||||
.loop:
|
||||
%if cpuflag(sse2)
|
||||
cvtps2dq m0, [srcq+2*lenq ]
|
||||
cvtps2dq m1, [srcq+2*lenq+16]
|
||||
packssdw m0, m1
|
||||
movd v1d, m0
|
||||
psrldq m0, 4
|
||||
movd v2d, m0
|
||||
psrldq m0, 4
|
||||
mov [dstq], v1w
|
||||
mov [dstq+stepq*4], v2w
|
||||
shr v1d, 16
|
||||
shr v2d, 16
|
||||
mov [dstq+stepq*2], v1w
|
||||
mov [dstq+step3q*2], v2w
|
||||
lea dstq, [dstq+stepq*8]
|
||||
movd v1d, m0
|
||||
psrldq m0, 4
|
||||
movd v2d, m0
|
||||
mov [dstq], v1w
|
||||
mov [dstq+stepq*4], v2w
|
||||
shr v1d, 16
|
||||
shr v2d, 16
|
||||
mov [dstq+stepq*2], v1w
|
||||
mov [dstq+step3q*2], v2w
|
||||
lea dstq, [dstq+stepq*8]
|
||||
%else
|
||||
CVTPS2PI m0, [srcq+2*lenq ]
|
||||
CVTPS2PI m1, [srcq+2*lenq+ 8]
|
||||
CVTPS2PI m2, [srcq+2*lenq+16]
|
||||
CVTPS2PI m3, [srcq+2*lenq+24]
|
||||
packssdw m0, m1
|
||||
packssdw m2, m3
|
||||
movd v1d, m0
|
||||
psrlq m0, 32
|
||||
movd v2d, m0
|
||||
mov [dstq], v1w
|
||||
mov [dstq+stepq*4], v2w
|
||||
shr v1d, 16
|
||||
shr v2d, 16
|
||||
mov [dstq+stepq*2], v1w
|
||||
mov [dstq+step3q*2], v2w
|
||||
lea dstq, [dstq+stepq*8]
|
||||
movd v1d, m2
|
||||
psrlq m2, 32
|
||||
movd v2d, m2
|
||||
mov [dstq], v1w
|
||||
mov [dstq+stepq*4], v2w
|
||||
shr v1d, 16
|
||||
shr v2d, 16
|
||||
mov [dstq+stepq*2], v1w
|
||||
mov [dstq+step3q*2], v2w
|
||||
lea dstq, [dstq+stepq*8]
|
||||
%endif
|
||||
add lenq, 16
|
||||
js .loop
|
||||
%if mmsize == 8
|
||||
emms
|
||||
%endif
|
||||
REP_RET
|
||||
%endmacro
|
||||
|
||||
INIT_XMM sse2
|
||||
FLOAT_TO_INT16_STEP 2
|
||||
INIT_MMX sse
|
||||
FLOAT_TO_INT16_STEP 0
|
||||
INIT_MMX 3dnow
|
||||
FLOAT_TO_INT16_STEP 0
|
||||
|
||||
;-------------------------------------------------------------------------------
|
||||
; void ff_float_to_int16_interleave2(int16_t *dst, const float **src, long len);
|
||||
;-------------------------------------------------------------------------------
|
||||
%macro FLOAT_TO_INT16_INTERLEAVE2 0
|
||||
cglobal float_to_int16_interleave2, 3, 4, 2, dst, src0, src1, len
|
||||
lea lenq, [4*r2q]
|
||||
mov src1q, [src0q+gprsize]
|
||||
mov src0q, [src0q]
|
||||
add dstq, lenq
|
||||
add src0q, lenq
|
||||
add src1q, lenq
|
||||
neg lenq
|
||||
.loop:
|
||||
%if cpuflag(sse2)
|
||||
cvtps2dq m0, [src0q+lenq]
|
||||
cvtps2dq m1, [src1q+lenq]
|
||||
packssdw m0, m1
|
||||
movhlps m1, m0
|
||||
punpcklwd m0, m1
|
||||
mova [dstq+lenq], m0
|
||||
%else
|
||||
CVTPS2PI m0, [src0q+lenq ]
|
||||
CVTPS2PI m1, [src0q+lenq+8]
|
||||
CVTPS2PI m2, [src1q+lenq ]
|
||||
CVTPS2PI m3, [src1q+lenq+8]
|
||||
packssdw m0, m1
|
||||
packssdw m2, m3
|
||||
mova m1, m0
|
||||
punpcklwd m0, m2
|
||||
punpckhwd m1, m2
|
||||
mova [dstq+lenq ], m0
|
||||
mova [dstq+lenq+8], m1
|
||||
%endif
|
||||
add lenq, 16
|
||||
js .loop
|
||||
%if mmsize == 8
|
||||
emms
|
||||
%endif
|
||||
REP_RET
|
||||
%endmacro
|
||||
|
||||
INIT_MMX 3dnow
|
||||
FLOAT_TO_INT16_INTERLEAVE2
|
||||
INIT_MMX sse
|
||||
FLOAT_TO_INT16_INTERLEAVE2
|
||||
INIT_XMM sse2
|
||||
FLOAT_TO_INT16_INTERLEAVE2
|
||||
|
||||
;-----------------------------------------------------------------------------
|
||||
; void ff_float_to_int16_interleave6(int16_t *dst, const float **src, int len)
|
||||
;-----------------------------------------------------------------------------
|
||||
%macro FLOAT_TO_INT16_INTERLEAVE6 0
|
||||
cglobal float_to_int16_interleave6, 2, 8, 0, dst, src, src1, src2, src3, src4, src5, len
|
||||
%if ARCH_X86_64
|
||||
mov lend, r2d
|
||||
%else
|
||||
%define lend dword r2m
|
||||
%endif
|
||||
mov src1q, [srcq+1*gprsize]
|
||||
mov src2q, [srcq+2*gprsize]
|
||||
mov src3q, [srcq+3*gprsize]
|
||||
mov src4q, [srcq+4*gprsize]
|
||||
mov src5q, [srcq+5*gprsize]
|
||||
mov srcq, [srcq]
|
||||
sub src1q, srcq
|
||||
sub src2q, srcq
|
||||
sub src3q, srcq
|
||||
sub src4q, srcq
|
||||
sub src5q, srcq
|
||||
.loop:
|
||||
CVTPS2PI mm0, [srcq]
|
||||
CVTPS2PI mm1, [srcq+src1q]
|
||||
CVTPS2PI mm2, [srcq+src2q]
|
||||
CVTPS2PI mm3, [srcq+src3q]
|
||||
CVTPS2PI mm4, [srcq+src4q]
|
||||
CVTPS2PI mm5, [srcq+src5q]
|
||||
packssdw mm0, mm3
|
||||
packssdw mm1, mm4
|
||||
packssdw mm2, mm5
|
||||
PSWAPD mm3, mm0
|
||||
punpcklwd mm0, mm1
|
||||
punpckhwd mm1, mm2
|
||||
punpcklwd mm2, mm3
|
||||
PSWAPD mm3, mm0
|
||||
punpckldq mm0, mm2
|
||||
punpckhdq mm2, mm1
|
||||
punpckldq mm1, mm3
|
||||
movq [dstq ], mm0
|
||||
movq [dstq+16], mm2
|
||||
movq [dstq+ 8], mm1
|
||||
add srcq, 8
|
||||
add dstq, 24
|
||||
sub lend, 2
|
||||
jg .loop
|
||||
emms
|
||||
RET
|
||||
%endmacro ; FLOAT_TO_INT16_INTERLEAVE6
|
||||
|
||||
INIT_MMX sse
|
||||
FLOAT_TO_INT16_INTERLEAVE6
|
||||
INIT_MMX 3dnow
|
||||
FLOAT_TO_INT16_INTERLEAVE6
|
||||
INIT_MMX 3dnowext
|
||||
FLOAT_TO_INT16_INTERLEAVE6
|
||||
|
||||
;-----------------------------------------------------------------------------
|
||||
; void ff_float_interleave6(float *dst, const float **src, unsigned int len);
|
||||
;-----------------------------------------------------------------------------
|
||||
|
||||
%macro FLOAT_INTERLEAVE6 1
|
||||
cglobal float_interleave6, 2, 8, %1, dst, src, src1, src2, src3, src4, src5, len
|
||||
%if ARCH_X86_64
|
||||
mov lend, r2d
|
||||
%else
|
||||
%define lend dword r2m
|
||||
%endif
|
||||
mov src1q, [srcq+1*gprsize]
|
||||
mov src2q, [srcq+2*gprsize]
|
||||
mov src3q, [srcq+3*gprsize]
|
||||
mov src4q, [srcq+4*gprsize]
|
||||
mov src5q, [srcq+5*gprsize]
|
||||
mov srcq, [srcq]
|
||||
sub src1q, srcq
|
||||
sub src2q, srcq
|
||||
sub src3q, srcq
|
||||
sub src4q, srcq
|
||||
sub src5q, srcq
|
||||
.loop:
|
||||
%if cpuflag(sse)
|
||||
movaps m0, [srcq]
|
||||
movaps m1, [srcq+src1q]
|
||||
movaps m2, [srcq+src2q]
|
||||
movaps m3, [srcq+src3q]
|
||||
movaps m4, [srcq+src4q]
|
||||
movaps m5, [srcq+src5q]
|
||||
|
||||
SBUTTERFLYPS 0, 1, 6
|
||||
SBUTTERFLYPS 2, 3, 6
|
||||
SBUTTERFLYPS 4, 5, 6
|
||||
|
||||
movaps m6, m4
|
||||
shufps m4, m0, 0xe4
|
||||
movlhps m0, m2
|
||||
movhlps m6, m2
|
||||
movaps [dstq ], m0
|
||||
movaps [dstq+16], m4
|
||||
movaps [dstq+32], m6
|
||||
|
||||
movaps m6, m5
|
||||
shufps m5, m1, 0xe4
|
||||
movlhps m1, m3
|
||||
movhlps m6, m3
|
||||
movaps [dstq+48], m1
|
||||
movaps [dstq+64], m5
|
||||
movaps [dstq+80], m6
|
||||
%else ; mmx
|
||||
movq m0, [srcq]
|
||||
movq m1, [srcq+src1q]
|
||||
movq m2, [srcq+src2q]
|
||||
movq m3, [srcq+src3q]
|
||||
movq m4, [srcq+src4q]
|
||||
movq m5, [srcq+src5q]
|
||||
|
||||
SBUTTERFLY dq, 0, 1, 6
|
||||
SBUTTERFLY dq, 2, 3, 6
|
||||
SBUTTERFLY dq, 4, 5, 6
|
||||
movq [dstq ], m0
|
||||
movq [dstq+ 8], m2
|
||||
movq [dstq+16], m4
|
||||
movq [dstq+24], m1
|
||||
movq [dstq+32], m3
|
||||
movq [dstq+40], m5
|
||||
%endif
|
||||
add srcq, mmsize
|
||||
add dstq, mmsize*6
|
||||
sub lend, mmsize/4
|
||||
jg .loop
|
||||
%if mmsize == 8
|
||||
emms
|
||||
%endif
|
||||
REP_RET
|
||||
%endmacro
|
||||
|
||||
INIT_MMX mmx
|
||||
FLOAT_INTERLEAVE6 0
|
||||
INIT_XMM sse
|
||||
FLOAT_INTERLEAVE6 7
|
||||
|
||||
;-----------------------------------------------------------------------------
|
||||
; void ff_float_interleave2(float *dst, const float **src, unsigned int len);
|
||||
;-----------------------------------------------------------------------------
|
||||
|
||||
%macro FLOAT_INTERLEAVE2 1
|
||||
cglobal float_interleave2, 3, 4, %1, dst, src, len, src1
|
||||
mov src1q, [srcq+gprsize]
|
||||
mov srcq, [srcq ]
|
||||
sub src1q, srcq
|
||||
.loop:
|
||||
mova m0, [srcq ]
|
||||
mova m1, [srcq+src1q ]
|
||||
mova m3, [srcq +mmsize]
|
||||
mova m4, [srcq+src1q+mmsize]
|
||||
|
||||
mova m2, m0
|
||||
PUNPCKLDQ m0, m1
|
||||
PUNPCKHDQ m2, m1
|
||||
|
||||
mova m1, m3
|
||||
PUNPCKLDQ m3, m4
|
||||
PUNPCKHDQ m1, m4
|
||||
|
||||
mova [dstq ], m0
|
||||
mova [dstq+1*mmsize], m2
|
||||
mova [dstq+2*mmsize], m3
|
||||
mova [dstq+3*mmsize], m1
|
||||
|
||||
add srcq, mmsize*2
|
||||
add dstq, mmsize*4
|
||||
sub lend, mmsize/2
|
||||
jg .loop
|
||||
%if mmsize == 8
|
||||
emms
|
||||
%endif
|
||||
REP_RET
|
||||
%endmacro
|
||||
|
||||
INIT_MMX mmx
|
||||
%define PUNPCKLDQ punpckldq
|
||||
%define PUNPCKHDQ punpckhdq
|
||||
FLOAT_INTERLEAVE2 0
|
||||
INIT_XMM sse
|
||||
%define PUNPCKLDQ unpcklps
|
||||
%define PUNPCKHDQ unpckhps
|
||||
FLOAT_INTERLEAVE2 5
|
||||
|
@ -33,84 +33,6 @@
|
||||
void ff_int32_to_float_fmul_scalar_sse (float *dst, const int32_t *src, float mul, int len);
|
||||
void ff_int32_to_float_fmul_scalar_sse2(float *dst, const int32_t *src, float mul, int len);
|
||||
|
||||
void ff_float_to_int16_3dnow(int16_t *dst, const float *src, long len);
|
||||
void ff_float_to_int16_sse (int16_t *dst, const float *src, long len);
|
||||
void ff_float_to_int16_sse2 (int16_t *dst, const float *src, long len);
|
||||
|
||||
void ff_float_to_int16_step_3dnow(int16_t *dst, const float *src, long len, long step);
|
||||
void ff_float_to_int16_step_sse (int16_t *dst, const float *src, long len, long step);
|
||||
void ff_float_to_int16_step_sse2 (int16_t *dst, const float *src, long len, long step);
|
||||
|
||||
void ff_float_to_int16_interleave2_3dnow(int16_t *dst, const float **src, long len);
|
||||
void ff_float_to_int16_interleave2_sse (int16_t *dst, const float **src, long len);
|
||||
void ff_float_to_int16_interleave2_sse2 (int16_t *dst, const float **src, long len);
|
||||
|
||||
void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len);
|
||||
void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int len);
|
||||
void ff_float_to_int16_interleave6_3dnowext(int16_t *dst, const float **src, int len);
|
||||
|
||||
#define ff_float_to_int16_interleave6_sse2 ff_float_to_int16_interleave6_sse
|
||||
|
||||
#define FLOAT_TO_INT16_INTERLEAVE(cpu) \
|
||||
/* gcc pessimizes register allocation if this is in the same function as float_to_int16_interleave_sse2*/\
|
||||
static av_noinline void float_to_int16_interleave_misc_##cpu(int16_t *dst, const float **src, long len, int channels){\
|
||||
int c;\
|
||||
for(c=0; c<channels; c++){\
|
||||
ff_float_to_int16_step_##cpu(dst+c, src[c], len, channels);\
|
||||
}\
|
||||
}\
|
||||
\
|
||||
static void float_to_int16_interleave_##cpu(int16_t *dst, const float **src, long len, int channels){\
|
||||
if(channels==1)\
|
||||
ff_float_to_int16_##cpu(dst, src[0], len);\
|
||||
else if(channels==2){\
|
||||
ff_float_to_int16_interleave2_##cpu(dst, src, len);\
|
||||
}else if(channels==6){\
|
||||
ff_float_to_int16_interleave6_##cpu(dst, src, len);\
|
||||
}else\
|
||||
float_to_int16_interleave_misc_##cpu(dst, src, len, channels);\
|
||||
}
|
||||
|
||||
FLOAT_TO_INT16_INTERLEAVE(3dnow)
|
||||
FLOAT_TO_INT16_INTERLEAVE(sse)
|
||||
FLOAT_TO_INT16_INTERLEAVE(sse2)
|
||||
|
||||
static void float_to_int16_interleave_3dnowext(int16_t *dst, const float **src,
|
||||
long len, int channels)
|
||||
{
|
||||
if(channels==6)
|
||||
ff_float_to_int16_interleave6_3dnowext(dst, src, len);
|
||||
else
|
||||
float_to_int16_interleave_3dnow(dst, src, len, channels);
|
||||
}
|
||||
|
||||
void ff_float_interleave2_mmx(float *dst, const float **src, unsigned int len);
|
||||
void ff_float_interleave2_sse(float *dst, const float **src, unsigned int len);
|
||||
|
||||
void ff_float_interleave6_mmx(float *dst, const float **src, unsigned int len);
|
||||
void ff_float_interleave6_sse(float *dst, const float **src, unsigned int len);
|
||||
|
||||
static void float_interleave_mmx(float *dst, const float **src,
|
||||
unsigned int len, int channels)
|
||||
{
|
||||
if (channels == 2) {
|
||||
ff_float_interleave2_mmx(dst, src, len);
|
||||
} else if (channels == 6)
|
||||
ff_float_interleave6_mmx(dst, src, len);
|
||||
else
|
||||
ff_float_interleave_c(dst, src, len, channels);
|
||||
}
|
||||
|
||||
static void float_interleave_sse(float *dst, const float **src,
|
||||
unsigned int len, int channels)
|
||||
{
|
||||
if (channels == 2) {
|
||||
ff_float_interleave2_sse(dst, src, len);
|
||||
} else if (channels == 6)
|
||||
ff_float_interleave6_sse(dst, src, len);
|
||||
else
|
||||
ff_float_interleave_c(dst, src, len, channels);
|
||||
}
|
||||
#endif /* HAVE_YASM */
|
||||
|
||||
av_cold void ff_fmt_convert_init_x86(FmtConvertContext *c, AVCodecContext *avctx)
|
||||
@ -118,30 +40,11 @@ av_cold void ff_fmt_convert_init_x86(FmtConvertContext *c, AVCodecContext *avctx
|
||||
#if HAVE_YASM
|
||||
int cpu_flags = av_get_cpu_flags();
|
||||
|
||||
if (EXTERNAL_MMX(cpu_flags)) {
|
||||
c->float_interleave = float_interleave_mmx;
|
||||
}
|
||||
if (EXTERNAL_AMD3DNOW(cpu_flags)) {
|
||||
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
|
||||
c->float_to_int16 = ff_float_to_int16_3dnow;
|
||||
c->float_to_int16_interleave = float_to_int16_interleave_3dnow;
|
||||
}
|
||||
}
|
||||
if (EXTERNAL_AMD3DNOWEXT(cpu_flags)) {
|
||||
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
|
||||
c->float_to_int16_interleave = float_to_int16_interleave_3dnowext;
|
||||
}
|
||||
}
|
||||
if (EXTERNAL_SSE(cpu_flags)) {
|
||||
c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse;
|
||||
c->float_to_int16 = ff_float_to_int16_sse;
|
||||
c->float_to_int16_interleave = float_to_int16_interleave_sse;
|
||||
c->float_interleave = float_interleave_sse;
|
||||
}
|
||||
if (EXTERNAL_SSE2(cpu_flags)) {
|
||||
c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse2;
|
||||
c->float_to_int16 = ff_float_to_int16_sse2;
|
||||
c->float_to_int16_interleave = float_to_int16_interleave_sse2;
|
||||
}
|
||||
#endif /* HAVE_YASM */
|
||||
}
|
||||
|
Loading…
Reference in New Issue
Block a user