dcadsp: split lfe_dir cases

The x86 runs short on registers because numerous elements are not static.
In addition, splitting them allows more optimized code, at least for x86.

Arm asm changes by Janne Grunau.

Signed-off-by: Janne Grunau <janne-libav@jannau.net>

1 files changed, 16 insertions, 7 deletions

diff --git a/libavcodec/arm/dcadsp_init_arm.c b/libavcodec/arm/dcadsp_init_arm.c
index d49a1765f6..2ea12895de 100644
--- a/libavcodec/arm/dcadsp_init_arm.c
+++ b/libavcodec/arm/dcadsp_init_arm.c
@@ -24,16 +24,22 @@
 #include "libavutil/attributes.h"
 #include "libavcodec/dcadsp.h"
 
-void ff_dca_lfe_fir_vfp(float *out, const float *in, const float *coefs,
-                        int decifactor, float scale);
+void ff_dca_lfe_fir0_neon(float *out, const float *in, const float *coefs,
+                          float scale);
+void ff_dca_lfe_fir1_neon(float *out, const float *in, const float *coefs,
+                          float scale);
+
+void ff_dca_lfe_fir32_vfp(float *out, const float *in, const float *coefs,
+                          float scale);
+void ff_dca_lfe_fir64_vfp(float *out, const float *in, const float *coefs,
+                          float scale);
+
 void ff_dca_qmf_32_subbands_vfp(float samples_in[32][8], int sb_act,
                                 SynthFilterContext *synth, FFTContext *imdct,
                                 float synth_buf_ptr[512],
                                 int *synth_buf_offset, float synth_buf2[32],
                                 const float window[512], float *samples_out,
                                 float raXin[32], float scale);
-void ff_dca_lfe_fir_neon(float *out, const float *in, const float *coefs,
-                         int decifactor, float scale);
 
 void ff_synth_filter_float_vfp(FFTContext *imdct,
                                float *synth_buf_ptr, int *synth_buf_offset,
@@ -52,11 +58,14 @@ av_cold void ff_dcadsp_init_arm(DCADSPContext *s)
     int cpu_flags = av_get_cpu_flags();
 
     if (have_vfp(cpu_flags) && !have_vfpv3(cpu_flags)) {
-        s->lfe_fir = ff_dca_lfe_fir_vfp;
+        s->lfe_fir[0]      = ff_dca_lfe_fir32_vfp;
+        s->lfe_fir[1]      = ff_dca_lfe_fir64_vfp;
         s->qmf_32_subbands = ff_dca_qmf_32_subbands_vfp;
     }
-    if (have_neon(cpu_flags))
-        s->lfe_fir = ff_dca_lfe_fir_neon;
+    if (have_neon(cpu_flags)) {
+        s->lfe_fir[0] = ff_dca_lfe_fir0_neon;
+        s->lfe_fir[1] = ff_dca_lfe_fir1_neon;
+    }
 }
 
 av_cold void ff_synth_filter_init_arm(SynthFilterContext *s)