aarch64: NEON float (i)MDCT

Approximately as fast as the ARM NEON version on Apple's A7.

libavcodec/aarch64/Makefile

@@ -18,3 +18,4 @@ NEON-OBJS-$(CONFIG_H264DSP)             += aarch64/h264dsp_neon.o              \
 NEON-OBJS-$(CONFIG_H264QPEL)            += aarch64/h264qpel_neon.o             \
                                            aarch64/hpeldsp_neon.o
 NEON-OBJS-$(CONFIG_HPELDSP)             += aarch64/hpeldsp_neon.o
+NEON-OBJS-$(CONFIG_MDCT)                += aarch64/mdct_neon.o

libavcodec/aarch64/fft_init_aarch64.c

@@ -26,6 +26,10 @@
 void ff_fft_permute_neon(FFTContext *s, FFTComplex *z);
 void ff_fft_calc_neon(FFTContext *s, FFTComplex *z);
 
+void ff_imdct_calc_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
+void ff_imdct_half_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
+void ff_mdct_calc_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
+
 av_cold void ff_fft_init_aarch64(FFTContext *s)
 {
     int cpu_flags = av_get_cpu_flags();
@@ -33,5 +37,11 @@ av_cold void ff_fft_init_aarch64(FFTContext *s)
     if (have_neon(cpu_flags)) {
         s->fft_permute  = ff_fft_permute_neon;
         s->fft_calc     = ff_fft_calc_neon;
+#if CONFIG_MDCT
+        s->imdct_calc   = ff_imdct_calc_neon;
+        s->imdct_half   = ff_imdct_half_neon;
+        s->mdct_calc    = ff_mdct_calc_neon;
+        s->mdct_permutation = FF_MDCT_PERM_INTERLEAVE;
+#endif
     }
 }