dcadsp: add int8x8_fmul_int32 to dsp context

It is currently declared as a macro who is set to inlinable functions,
among which a Neon and a default C implementations.

Add a DSP parameter to each inline function, unused except by the
default C implementation which calls a function from the DSP context.

On an Arrandale CPU, gain for an inlined SSE2 function vs. a call:
- Win32: 29 to 26 cycles
- Win64: 25 to 23 cycles

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

4 files changed, 16 insertions, 7 deletions

diff --git a/libavcodec/arm/dca.h b/libavcodec/arm/dca.h
index 39ec2b6bc8..06e3ea6ba3 100644
--- a/libavcodec/arm/dca.h
+++ b/libavcodec/arm/dca.h
@@ -83,7 +83,8 @@ static inline int decode_blockcodes(int code1, int code2, int levels,
 #if HAVE_NEON_INLINE && HAVE_ASM_MOD_Y
 
 #define int8x8_fmul_int32 int8x8_fmul_int32
-static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
+static inline void int8x8_fmul_int32(av_unused DCADSPContext *dsp,
+                                     float *dst, const int8_t *src, int scale)
 {
     __asm__ ("vcvt.f32.s32 %2,  %2,  #4         \n"
              "vld1.8       {d0},     [%1,:64]   \n"
diff --git a/libavcodec/dcadec.c b/libavcodec/dcadec.c
index f9e39bcca4..b6df3b9f2f 100644
--- a/libavcodec/dcadec.c
+++ b/libavcodec/dcadec.c
@@ -1086,12 +1086,10 @@ static const uint8_t abits_sizes[7]  = { 7, 10, 12, 13, 15, 17, 19 };
 static const uint8_t abits_levels[7] = { 3,  5,  7,  9, 13, 17, 25 };
 
 #ifndef int8x8_fmul_int32
-static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
+static inline void int8x8_fmul_int32(DCADSPContext *dsp, float *dst,
+                                     const int8_t *src, int scale)
 {
-    float fscale = scale / 16.0;
-    int i;
-    for (i = 0; i < 8; i++)
-        dst[i] = src[i] * fscale;
+    dsp->int8x8_fmul_int32(dst, src, scale);
 }
 #endif
 
@@ -1219,7 +1217,7 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
                 s->debug_flag |= 0x01;
             }
 
-            int8x8_fmul_int32(subband_samples[k][l],
+            int8x8_fmul_int32(&s->dcadsp, subband_samples[k][l],
                               &high_freq_vq[hfvq][subsubframe * 8],
                               s->scale_factor[k][l][0]);
         }
diff --git a/libavcodec/dcadsp.c b/libavcodec/dcadsp.c
index 57d716e402..b98486422b 100644
--- a/libavcodec/dcadsp.c
+++ b/libavcodec/dcadsp.c
@@ -24,6 +24,14 @@
 #include "libavutil/intreadwrite.h"
 #include "dcadsp.h"
 
+static void int8x8_fmul_int32_c(float *dst, const int8_t *src, int scale)
+{
+    float fscale = scale / 16.0;
+    int i;
+    for (i = 0; i < 8; i++)
+        dst[i] = src[i] * fscale;
+}
+
 static void dca_lfe_fir_c(float *out, const float *in, const float *coefs,
                           int decifactor, float scale)
 {
@@ -78,5 +86,6 @@ av_cold void ff_dcadsp_init(DCADSPContext *s)
 {
     s->lfe_fir = dca_lfe_fir_c;
     s->qmf_32_subbands = dca_qmf_32_subbands;
+    s->int8x8_fmul_int32 = int8x8_fmul_int32_c;
     if (ARCH_ARM) ff_dcadsp_init_arm(s);
 }
diff --git a/libavcodec/dcadsp.h b/libavcodec/dcadsp.h
index ec88be71f0..0f79dd643e 100644
--- a/libavcodec/dcadsp.h
+++ b/libavcodec/dcadsp.h
@@ -31,6 +31,7 @@ typedef struct DCADSPContext {
                             int *synth_buf_offset, float synth_buf2[32],
                             const float window[512], float *samples_out,
                             float raXin[32], float scale);
+    void (*int8x8_fmul_int32)(float *dst, const int8_t *src, int scale);
 } DCADSPContext;
 
 void ff_dcadsp_init(DCADSPContext *s);