fft: remove 16-bit FFT and MDCT code

No longer used by anything.
Unfortunately the old FFT_FLOAT/FFT_FIXED_32 is left as-is. It's
simply too much work for code meant to be all removed anyway.

4 files changed, 3 insertions, 510 deletions

diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index c6be814153..c4ab93aeeb 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -5,8 +5,7 @@ OBJS-$(CONFIG_AC3DSP)                  += arm/ac3dsp_init_arm.o         \
                                           arm/ac3dsp_arm.o
 OBJS-$(CONFIG_AUDIODSP)                += arm/audiodsp_init_arm.o
 OBJS-$(CONFIG_BLOCKDSP)                += arm/blockdsp_init_arm.o
-OBJS-$(CONFIG_FFT)                     += arm/fft_init_arm.o            \
-                                          arm/fft_fixed_init_arm.o
+OBJS-$(CONFIG_FFT)                     += arm/fft_init_arm.o
 OBJS-$(CONFIG_FLACDSP)                 += arm/flacdsp_init_arm.o        \
                                           arm/flacdsp_arm.o
 OBJS-$(CONFIG_FMTCONVERT)              += arm/fmtconvert_init_arm.o
@@ -108,8 +107,7 @@ NEON-OBJS-$(CONFIG_AUDIODSP)           += arm/audiodsp_init_neon.o      \
                                           arm/int_neon.o
 NEON-OBJS-$(CONFIG_BLOCKDSP)           += arm/blockdsp_init_neon.o      \
                                           arm/blockdsp_neon.o
-NEON-OBJS-$(CONFIG_FFT)                += arm/fft_neon.o                \
-                                          arm/fft_fixed_neon.o
+NEON-OBJS-$(CONFIG_FFT)                += arm/fft_neon.o
 NEON-OBJS-$(CONFIG_FMTCONVERT)         += arm/fmtconvert_neon.o
 NEON-OBJS-$(CONFIG_G722DSP)            += arm/g722dsp_neon.o
 NEON-OBJS-$(CONFIG_H264CHROMA)         += arm/h264cmc_neon.o
@@ -123,8 +121,7 @@ NEON-OBJS-$(CONFIG_HPELDSP)            += arm/hpeldsp_init_neon.o       \
 NEON-OBJS-$(CONFIG_IDCTDSP)            += arm/idctdsp_init_neon.o       \
                                           arm/idctdsp_neon.o            \
                                           arm/simple_idct_neon.o
-NEON-OBJS-$(CONFIG_MDCT)               += arm/mdct_neon.o               \
-                                          arm/mdct_fixed_neon.o
+NEON-OBJS-$(CONFIG_MDCT)               += arm/mdct_neon.o
 NEON-OBJS-$(CONFIG_MPEGVIDEO)          += arm/mpegvideo_neon.o
 NEON-OBJS-$(CONFIG_PIXBLOCKDSP)        += arm/pixblockdsp_neon.o
 NEON-OBJS-$(CONFIG_RDFT)               += arm/rdft_neon.o
diff --git a/libavcodec/arm/fft_fixed_init_arm.c b/libavcodec/arm/fft_fixed_init_arm.c
deleted file mode 100644
index 11226d65ff..0000000000
--- a/libavcodec/arm/fft_fixed_init_arm.c
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2009 Mans Rullgard <mans@mansr.com>
- *
- * 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/attributes.h"
-#include "libavutil/cpu.h"
-#include "libavutil/arm/cpu.h"
-
-#define FFT_FLOAT 0
-#include "libavcodec/fft.h"
-
-void ff_fft_fixed_calc_neon(FFTContext *s, FFTComplex *z);
-void ff_mdct_fixed_calc_neon(FFTContext *s, FFTSample *o, const FFTSample *i);
-void ff_mdct_fixed_calcw_neon(FFTContext *s, FFTDouble *o, const FFTSample *i);
-
-av_cold void ff_fft_fixed_init_arm(FFTContext *s)
-{
-    int cpu_flags = av_get_cpu_flags();
-
-    if (have_neon(cpu_flags)) {
-        s->fft_permutation = FF_FFT_PERM_SWAP_LSBS;
-#if CONFIG_FFT
-        s->fft_calc        = ff_fft_fixed_calc_neon;
-#endif
-
-#if CONFIG_MDCT
-        if (!s->inverse && s->nbits >= 3) {
-            s->mdct_permutation = FF_MDCT_PERM_INTERLEAVE;
-            s->mdct_calc        = ff_mdct_fixed_calc_neon;
-            s->mdct_calcw       = ff_mdct_fixed_calcw_neon;
-        }
-#endif
-    }
-}
diff --git a/libavcodec/arm/fft_fixed_neon.S b/libavcodec/arm/fft_fixed_neon.S
deleted file mode 100644
index 2651607544..0000000000
--- a/libavcodec/arm/fft_fixed_neon.S
+++ /dev/null
@@ -1,261 +0,0 @@
-/*
- * Copyright (c) 2011 Mans Rullgard <mans@mansr.com>
- *
- * 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/arm/asm.S"
-
-.macro  bflies          d0,  d1,  r0,  r1
-        vrev64.32       \r0, \d1                @ t5, t6, t1, t2
-        vhsub.s16       \r1, \d1, \r0           @ t1-t5, t2-t6, t5-t1, t6-t2
-        vhadd.s16       \r0, \d1, \r0           @ t1+t5, t2+t6, t5+t1, t6+t2
-        vext.16         \r1, \r1, \r1, #1       @ t2-t6, t5-t1, t6-t2, t1-t5
-        vtrn.32         \r0, \r1                @ t1+t5, t2+t6, t2-t6, t5-t1
-                                                @ t5,    t6,    t4,    t3
-        vhsub.s16       \d1, \d0, \r0
-        vhadd.s16       \d0, \d0, \r0
-.endm
-
-.macro  transform01     q0,  q1,  d3,  c0,  c1,  r0,  w0,  w1
-        vrev32.16       \r0, \d3
-        vmull.s16       \w0, \d3, \c0
-        vmlal.s16       \w0, \r0, \c1
-        vshrn.s32       \d3, \w0, #15
-        bflies          \q0, \q1, \w0, \w1
-.endm
-
-.macro  transform2      d0,  d1,  d2,  d3,  q0,  q1,  c0,  c1,  c2,  c3, \
-                        r0,  r1,  w0,  w1
-        vrev32.16       \r0, \d1
-        vrev32.16       \r1, \d3
-        vmull.s16       \w0, \d1, \c0
-        vmlal.s16       \w0, \r0, \c1
-        vmull.s16       \w1, \d3, \c2
-        vmlal.s16       \w1, \r1, \c3
-        vshrn.s32       \d1, \w0, #15
-        vshrn.s32       \d3, \w1, #15
-        bflies          \q0, \q1, \w0, \w1
-.endm
-
-.macro  fft4            d0,  d1,  r0,  r1
-        vhsub.s16       \r0, \d0, \d1           @ t3, t4, t8, t7
-        vhsub.s16       \r1, \d1, \d0
-        vhadd.s16       \d0, \d0, \d1           @ t1, t2, t6, t5
-        vmov.i64        \d1, #0xffff00000000
-        vbit            \r0, \r1, \d1
-        vrev64.16       \r1, \r0                @ t7, t8, t4, t3
-        vtrn.32         \r0, \r1                @ t3, t4, t7, t8
-        vtrn.32         \d0, \r0                @ t1, t2, t3, t4, t6, t5, t8, t7
-        vhsub.s16       \d1, \d0, \r0           @ r2, i2, r3, i1
-        vhadd.s16       \d0, \d0, \r0           @ r0, i0, r1, i3
-.endm
-
-.macro  fft8            d0,  d1,  d2,  d3,  q0,  q1,  c0,  c1,  r0,  r1, w0, w1
-        fft4            \d0, \d1, \r0, \r1
-        vtrn.32         \d0, \d1                @ z0, z2, z1, z3
-        vhadd.s16       \r0, \d2, \d3           @ t1, t2, t3, t4
-        vhsub.s16       \d3, \d2, \d3           @ z5, z7
-        vmov            \d2, \r0
-        transform01     \q0, \q1, \d3, \c0, \c1, \r0, \w0, \w1
-.endm
-
-function fft4_neon
-        vld1.16         {d0-d1},  [r0]
-        fft4            d0,  d1,  d2,  d3
-        vst1.16         {d0-d1},  [r0]
-        bx              lr
-endfunc
-
-function fft8_neon
-        vld1.16         {d0-d3},  [r0,:128]
-        movrel          r1,  coefs
-        vld1.16         {d30},    [r1,:64]
-        vdup.16         d31, d30[0]
-        fft8            d0,  d1,  d2,  d3,  q0,  q1,  d31, d30, d20, d21, q8, q9
-        vtrn.32         d0,  d1
-        vtrn.32         d2,  d3
-        vst1.16         {d0-d3},  [r0,:128]
-        bx              lr
-endfunc
-
-function fft16_neon
-        vld1.16         {d0-d3},  [r0,:128]!
-        vld1.16         {d4-d7},  [r0,:128]
-        movrel          r1,  coefs
-        sub             r0,  r0,  #32
-        vld1.16         {d28-d31},[r1,:128]
-        vdup.16         d31, d28[0]
-        fft8            d0,  d1,  d2,  d3,  q0,  q1,  d31, d28, d20, d21, q8, q9
-        vswp            d5,  d6
-        fft4            q2,  q3,  q8,  q9
-        vswp            d5,  d6
-        vtrn.32         q0,  q1             @ z0, z4, z2, z6, z1, z5, z3, z7
-        vtrn.32         q2,  q3             @ z8, z12,z10,z14,z9, z13,z11,z15
-        vswp            d1,  d2
-        vdup.16         d31, d28[0]
-        transform01     q0,  q2,  d5,  d31, d28, d20, q8, q9
-        vdup.16         d26, d29[0]
-        vdup.16         d27, d30[0]
-        transform2      d2,  d6,  d3,  d7,  q1,  q3,  d26, d30, d27, d29, \
-                        d20, d21, q8,  q9
-        vtrn.32         q0,  q1
-        vtrn.32         q2,  q3
-        vst1.16         {d0-d3},  [r0,:128]!
-        vst1.16         {d4-d7},  [r0,:128]
-        bx              lr
-endfunc
-
-function fft_pass_neon
-        push            {r4,lr}
-        movrel          lr,  coefs + 24
-        vld1.16         {d30},    [lr,:64]
-        lsl             r12, r2,  #3
-        vmov            d31, d30
-        add             r3,  r1,  r2,  lsl #2
-        mov             lr,  #-8
-        sub             r3,  r3,  #2
-        mov             r4,  r0
-        vld1.16         {d27[]},  [r3,:16]
-        sub             r3,  r3,  #6
-        vld1.16         {q0},     [r4,:128], r12
-        vld1.16         {q1},     [r4,:128], r12
-        vld1.16         {q2},     [r4,:128], r12
-        vld1.16         {q3},     [r4,:128], r12
-        vld1.16         {d28},    [r1,:64]!
-        vld1.16         {d29},    [r3,:64], lr
-        vswp            d1,  d2
-        vswp            d5,  d6
-        vtrn.32         d0,  d1
-        vtrn.32         d4,  d5
-        vdup.16         d25, d28[1]
-        vmul.s16        d27, d27, d31
-        transform01     q0,  q2,  d5,  d25, d27, d20, q8,  q9
-        b               2f
-1:
-        mov             r4,  r0
-        vdup.16         d26, d29[0]
-        vld1.16         {q0},     [r4,:128], r12
-        vld1.16         {q1},     [r4,:128], r12
-        vld1.16         {q2},     [r4,:128], r12
-        vld1.16         {q3},     [r4,:128], r12
-        vld1.16         {d28},    [r1,:64]!
-        vld1.16         {d29},    [r3,:64], lr
-        vswp            d1,  d2
-        vswp            d5,  d6
-        vtrn.32         d0,  d1
-        vtrn.32         d4,  d5
-        vdup.16         d24, d28[0]
-        vdup.16         d25, d28[1]
-        vdup.16         d27, d29[3]
-        vmul.s16        q13, q13, q15
-        transform2      d0,  d4,  d1,  d5,  q0,  q2,  d24, d26, d25, d27, \
-                        d16, d17, q9,  q10
-2:
-        vtrn.32         d2,  d3
-        vtrn.32         d6,  d7
-        vdup.16         d24, d28[2]
-        vdup.16         d26, d29[2]
-        vdup.16         d25, d28[3]
-        vdup.16         d27, d29[1]
-        vmul.s16        q13, q13, q15
-        transform2      d2,  d6,  d3,  d7,  q1,  q3,  d24, d26, d25, d27, \
-                        d16, d17, q9,  q10
-        vtrn.32         d0,  d1
-        vtrn.32         d2,  d3
-        vtrn.32         d4,  d5
-        vtrn.32         d6,  d7
-        vswp            d1,  d2
-        vswp            d5,  d6
-        mov             r4,  r0
-        vst1.16         {q0},     [r4,:128], r12
-        vst1.16         {q1},     [r4,:128], r12
-        vst1.16         {q2},     [r4,:128], r12
-        vst1.16         {q3},     [r4,:128], r12
-        add             r0,  r0,  #16
-        subs            r2,  r2,  #2
-        bgt             1b
-        pop             {r4,pc}
-endfunc
-
-#define F_SQRT1_2   23170
-#define F_COS_16_1  30274
-#define F_COS_16_3  12540
-
-const   coefs, align=4
-        .short          F_SQRT1_2, -F_SQRT1_2, -F_SQRT1_2,  F_SQRT1_2
-        .short          F_COS_16_1,-F_COS_16_1,-F_COS_16_1, F_COS_16_1
-        .short          F_COS_16_3,-F_COS_16_3,-F_COS_16_3, F_COS_16_3
-        .short          1,         -1,         -1,          1
-endconst
-
-.macro  def_fft n, n2, n4
-function fft\n\()_neon
-        push            {r4, lr}
-        mov             r4,  r0
-        bl              fft\n2\()_neon
-        add             r0,  r4,  #\n4*2*4
-        bl              fft\n4\()_neon
-        add             r0,  r4,  #\n4*3*4
-        bl              fft\n4\()_neon
-        mov             r0,  r4
-        pop             {r4, lr}
-        movrelx         r1,  X(ff_cos_\n\()_fixed)
-        mov             r2,  #\n4/2
-        b               fft_pass_neon
-endfunc
-.endm
-
-        def_fft    32,    16,     8
-        def_fft    64,    32,    16
-        def_fft   128,    64,    32
-        def_fft   256,   128,    64
-        def_fft   512,   256,   128
-        def_fft  1024,   512,   256
-        def_fft  2048,  1024,   512
-        def_fft  4096,  2048,  1024
-        def_fft  8192,  4096,  2048
-        def_fft 16384,  8192,  4096
-        def_fft 32768, 16384,  8192
-        def_fft 65536, 32768, 16384
-
-function ff_fft_fixed_calc_neon, export=1
-        ldr             r2,  [r0]
-        sub             r2,  r2,  #2
-        movrel          r3,  fft_fixed_tab_neon
-        ldr             r3,  [r3, r2, lsl #2]
-        mov             r0,  r1
-        bx              r3
-endfunc
-
-const   fft_fixed_tab_neon, relocate=1
-        .word fft4_neon
-        .word fft8_neon
-        .word fft16_neon
-        .word fft32_neon
-        .word fft64_neon
-        .word fft128_neon
-        .word fft256_neon
-        .word fft512_neon
-        .word fft1024_neon
-        .word fft2048_neon
-        .word fft4096_neon
-        .word fft8192_neon
-        .word fft16384_neon
-        .word fft32768_neon
-        .word fft65536_neon
-endconst
diff --git a/libavcodec/arm/mdct_fixed_neon.S b/libavcodec/arm/mdct_fixed_neon.S
deleted file mode 100644
index 365c5e7faf..0000000000
--- a/libavcodec/arm/mdct_fixed_neon.S
+++ /dev/null
@@ -1,193 +0,0 @@
-/*
- * Copyright (c) 2011 Mans Rullgard <mans@mansr.com>
- *
- * 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/arm/asm.S"
-
-.macro  prerot          dst, rt
-        lsr             r3,  r6,  #2            @ n4
-        add             \rt, r4,  r6,  lsr #1   @ revtab + n4
-        add             r9,  r3,  r3,  lsl #1   @ n3
-        add             r8,  r7,  r6            @ tcos + n4
-        add             r3,  r2,  r6,  lsr #1   @ in + n4
-        add             r9,  r2,  r9,  lsl #1   @ in + n3
-        sub             r8,  r8,  #16
-        sub             r10, r3,  #16
-        sub             r11, r9,  #16
-        mov             r12, #-16
-1:
-        vld2.16         {d0,d1},  [r9, :128]!
-        vld2.16         {d2,d3},  [r11,:128], r12
-        vld2.16         {d4,d5},  [r3, :128]!
-        vld2.16         {d6,d7},  [r10,:128], r12
-        vld2.16         {d16,d17},[r7, :128]!   @ cos, sin
-        vld2.16         {d18,d19},[r8, :128], r12
-        vrev64.16       q1,  q1
-        vrev64.16       q3,  q3
-        vrev64.16       q9,  q9
-        vneg.s16        d0,  d0
-        vneg.s16        d2,  d2
-        vneg.s16        d16, d16
-        vneg.s16        d18, d18
-        vhsub.s16       d0,  d0,  d3            @ re
-        vhsub.s16       d4,  d7,  d4            @ im
-        vhsub.s16       d6,  d6,  d5
-        vhsub.s16       d2,  d2,  d1
-        vmull.s16       q10, d0,  d16
-        vmlsl.s16       q10, d4,  d17
-        vmull.s16       q11, d0,  d17
-        vmlal.s16       q11, d4,  d16
-        vmull.s16       q12, d6,  d18
-        vmlsl.s16       q12, d2,  d19
-        vmull.s16       q13, d6,  d19
-        vmlal.s16       q13, d2,  d18
-        vshrn.s32       d0,  q10, #15
-        vshrn.s32       d1,  q11, #15
-        vshrn.s32       d2,  q12, #15
-        vshrn.s32       d3,  q13, #15
-        vzip.16         d0,  d1
-        vzip.16         d2,  d3
-        ldrh            lr,  [r4], #2
-        ldrh            r2,  [\rt, #-2]!
-        add             lr,  \dst, lr,  lsl #2
-        add             r2,  \dst, r2,  lsl #2
-        vst1.32         {d0[0]},  [lr,:32]
-        vst1.32         {d2[0]},  [r2,:32]
-        ldrh            lr,  [r4], #2
-        ldrh            r2,  [\rt, #-2]!
-        add             lr,  \dst, lr,  lsl #2
-        add             r2,  \dst, r2,  lsl #2
-        vst1.32         {d0[1]},  [lr,:32]
-        vst1.32         {d2[1]},  [r2,:32]
-        ldrh            lr,  [r4], #2
-        ldrh            r2,  [\rt, #-2]!
-        add             lr,  \dst, lr,  lsl #2
-        add             r2,  \dst, r2,  lsl #2
-        vst1.32         {d1[0]},  [lr,:32]
-        vst1.32         {d3[0]},  [r2,:32]
-        ldrh            lr,  [r4], #2
-        ldrh            r2,  [\rt, #-2]!
-        add             lr,  \dst, lr,  lsl #2
-        add             r2,  \dst, r2,  lsl #2
-        vst1.32         {d1[1]},  [lr,:32]
-        vst1.32         {d3[1]},  [r2,:32]
-        subs            r6,  r6,  #32
-        bgt             1b
-.endm
-
-function ff_mdct_fixed_calc_neon, export=1
-        push            {r1,r4-r11,lr}
-
-        ldr             r4,  [r0, #8]           @ revtab
-        ldr             r6,  [r0, #16]          @ mdct_size; n
-        ldr             r7,  [r0, #24]          @ tcos
-
-        prerot          r1,  r5
-
-        mov             r4,  r0
-        bl              X(ff_fft_fixed_calc_neon)
-
-        pop             {r5}
-        mov             r12, #-16
-        ldr             r6,  [r4, #16]          @ mdct_size; n
-        ldr             r7,  [r4, #24]          @ tcos
-        add             r5,  r5,  r6,  lsr #1
-        add             r7,  r7,  r6,  lsr #1
-        sub             r1,  r5,  #16
-        sub             r2,  r7,  #16
-1:
-        vld2.16         {d4,d5},  [r7,:128]!
-        vld2.16         {d6,d7},  [r2,:128], r12
-        vld2.16         {d0,d1},  [r5,:128]
-        vld2.16         {d2,d3},  [r1,:128]
-        vrev64.16       q3,  q3
-        vrev64.16       q1,  q1
-        vneg.s16        q3,  q3
-        vneg.s16        q2,  q2
-        vmull.s16       q11, d2,  d6
-        vmlal.s16       q11, d3,  d7
-        vmull.s16       q8,  d0,  d5
-        vmlsl.s16       q8,  d1,  d4
-        vmull.s16       q9,  d0,  d4
-        vmlal.s16       q9,  d1,  d5
-        vmull.s16       q10, d2,  d7
-        vmlsl.s16       q10, d3,  d6
-        vshrn.s32       d0,  q11, #15
-        vshrn.s32       d1,  q8,  #15
-        vshrn.s32       d2,  q9,  #15
-        vshrn.s32       d3,  q10, #15
-        vrev64.16       q0,  q0
-        vst2.16         {d2,d3},  [r5,:128]!
-        vst2.16         {d0,d1},  [r1,:128], r12
-        subs            r6,  r6,  #32
-        bgt             1b
-
-        pop             {r4-r11,pc}
-endfunc
-
-function ff_mdct_fixed_calcw_neon, export=1
-        push            {r1,r4-r11,lr}
-
-        ldrd            r4,  r5,  [r0, #8]      @ revtab, tmp_buf
-        ldr             r6,  [r0, #16]          @ mdct_size; n
-        ldr             r7,  [r0, #24]          @ tcos
-
-        prerot          r5,  r1
-
-        mov             r4,  r0
-        mov             r1,  r5
-        bl              X(ff_fft_fixed_calc_neon)
-
-        pop             {r7}
-        mov             r12, #-16
-        ldr             r6,  [r4, #16]          @ mdct_size; n
-        ldr             r9,  [r4, #24]          @ tcos
-        add             r5,  r5,  r6,  lsr #1
-        add             r7,  r7,  r6
-        add             r9,  r9,  r6,  lsr #1
-        sub             r3,  r5,  #16
-        sub             r1,  r7,  #16
-        sub             r2,  r9,  #16
-1:
-        vld2.16         {d4,d5},  [r9,:128]!
-        vld2.16         {d6,d7},  [r2,:128], r12
-        vld2.16         {d0,d1},  [r5,:128]!
-        vld2.16         {d2,d3},  [r3,:128], r12
-        vrev64.16       q3,  q3
-        vrev64.16       q1,  q1
-        vneg.s16        q3,  q3
-        vneg.s16        q2,  q2
-        vmull.s16       q8,  d2,  d6
-        vmlal.s16       q8,  d3,  d7
-        vmull.s16       q9,  d0,  d5
-        vmlsl.s16       q9,  d1,  d4
-        vmull.s16       q10, d0,  d4
-        vmlal.s16       q10, d1,  d5
-        vmull.s16       q11, d2,  d7
-        vmlsl.s16       q11, d3,  d6
-        vrev64.32       q8,  q8
-        vrev64.32       q9,  q9
-        vst2.32         {q10,q11},[r7,:128]!
-        vst2.32         {d16,d18},[r1,:128], r12
-        vst2.32         {d17,d19},[r1,:128], r12
-        subs            r6,  r6,  #32
-        bgt             1b
-
-        pop             {r4-r11,pc}
-endfunc