swr: split out DSP functions.

DSP bits of swri_resample go into their own mini-DSP functions; DSP
init goes from a per-call branch in multiple_resample to a proper
DSP init routine; x86 bits go into x86/; swri_resample() moves out of
resample_template.c into resample.c because it's independent of DSP
code or sample type; multiple_resample() is simplified.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

1 files changed, 49 insertions, 100 deletions

diff --git a/libswresample/resample.c b/libswresample/resample.c
index 5b7a0f9f23..9a67e6340c 100644
--- a/libswresample/resample.c
+++ b/libswresample/resample.c
@@ -25,32 +25,8 @@
  * @author Michael Niedermayer <michaelni@gmx.at>
  */
 
-#include "libavutil/log.h"
 #include "libavutil/avassert.h"
-#include "swresample_internal.h"
-
-
-typedef struct ResampleContext {
-    const AVClass *av_class;
-    uint8_t *filter_bank;
-    int filter_length;
-    int filter_alloc;
-    int ideal_dst_incr;
-    int dst_incr;
-    int index;
-    int frac;
-    int src_incr;
-    int compensation_distance;
-    int phase_shift;
-    int phase_mask;
-    int linear;
-    enum SwrFilterType filter_type;
-    int kaiser_beta;
-    double factor;
-    enum AVSampleFormat format;
-    int felem_size;
-    int filter_shift;
-} ResampleContext;
+#include "resample.h"
 
 /**
  * 0th order modified bessel function of the first kind.
@@ -197,7 +173,8 @@ static int build_filter(ResampleContext *c, void *filter, double factor, int tap
 
 static ResampleContext *resample_init(ResampleContext *c, int out_rate, int in_rate, int filter_size, int phase_shift, int linear,
                                     double cutoff0, enum AVSampleFormat format, enum SwrFilterType filter_type, int kaiser_beta,
-                                    double precision, int cheby){
+                                    double precision, int cheby)
+{
     double cutoff = cutoff0? cutoff0 : 0.97;
     double factor= FFMIN(out_rate * cutoff / in_rate, 1.0);
     int phase_count= 1<<phase_shift;
@@ -259,6 +236,8 @@ static ResampleContext *resample_init(ResampleContext *c, int out_rate, int in_r
     c->index= -phase_count*((c->filter_length-1)/2);
     c->frac= 0;
 
+    swresample_dsp_init(c);
+
     return c;
 error:
     av_freep(&c->filter_bank);
@@ -282,59 +261,53 @@ static int set_compensation(ResampleContext *c, int sample_delta, int compensati
     return 0;
 }
 
-#define TEMPLATE_RESAMPLE_S16
-#include "resample_template.c"
-#undef TEMPLATE_RESAMPLE_S16
-
-#define TEMPLATE_RESAMPLE_S32
-#include "resample_template.c"
-#undef TEMPLATE_RESAMPLE_S32
-
-#define TEMPLATE_RESAMPLE_FLT
-#include "resample_template.c"
-#undef TEMPLATE_RESAMPLE_FLT
-
-#define TEMPLATE_RESAMPLE_DBL
-#include "resample_template.c"
-#undef TEMPLATE_RESAMPLE_DBL
-
-// XXX FIXME the whole C loop should be written in asm so this x86 specific code here isnt needed
-#if HAVE_MMXEXT_INLINE
-
-#include "x86/resample_mmx.h"
-
-#define TEMPLATE_RESAMPLE_S16_MMX2
-#include "resample_template.c"
-#undef TEMPLATE_RESAMPLE_S16_MMX2
-
-#if HAVE_SSE_INLINE
-#define TEMPLATE_RESAMPLE_FLT_SSE
-#include "resample_template.c"
-#undef TEMPLATE_RESAMPLE_FLT_SSE
-#endif
-
-#if HAVE_SSE2_INLINE
-#define TEMPLATE_RESAMPLE_S16_SSE2
-#include "resample_template.c"
-#undef TEMPLATE_RESAMPLE_S16_SSE2
-
-#define TEMPLATE_RESAMPLE_DBL_SSE2
-#include "resample_template.c"
-#undef TEMPLATE_RESAMPLE_DBL_SSE2
-#endif
-
-#if HAVE_AVX_INLINE
-#define TEMPLATE_RESAMPLE_FLT_AVX
-#include "resample_template.c"
-#undef TEMPLATE_RESAMPLE_FLT_AVX
-#endif
+static int swri_resample(ResampleContext *c,
+                         uint8_t *dst, const uint8_t *src, int *consumed,
+                         int src_size, int dst_size, int update_ctx)
+{
+    int fn_idx = c->format - AV_SAMPLE_FMT_S16P;
+
+    if (c->filter_length == 1 && c->phase_shift == 0) {
+        int index= c->index;
+        int frac= c->frac;
+        int dst_incr_frac= c->dst_incr % c->src_incr;
+        int dst_incr=      c->dst_incr / c->src_incr;
+        int64_t index2= (1LL<<32)*c->frac/c->src_incr + (1LL<<32)*index;
+        int64_t incr= (1LL<<32) * c->dst_incr / c->src_incr;
+        int new_size = (src_size * (int64_t)c->src_incr - frac + c->dst_incr - 1) / c->dst_incr;
+
+        dst_size= FFMIN(dst_size, new_size);
+        c->dsp.resample_one[fn_idx](dst, src, dst_size, index2, incr);
+
+        index += dst_size * dst_incr;
+        index += (frac + dst_size * (int64_t)dst_incr_frac) / c->src_incr;
+        av_assert2(index >= 0);
+        *consumed= index;
+        if (update_ctx) {
+            c->frac   = (frac + dst_size * (int64_t)dst_incr_frac) % c->src_incr;
+            c->index = 0;
+        }
+    } else {
+        int64_t end_index = (1LL + src_size - c->filter_length) << c->phase_shift;
+        int64_t delta_frac = (end_index - c->index) * c->src_incr - c->frac;
+        int delta_n = (delta_frac + c->dst_incr - 1) / c->dst_incr;
+
+        dst_size = FFMIN(dst_size, delta_n);
+        if (!c->linear) {
+            *consumed = c->dsp.resample_common[fn_idx](c, dst, src, dst_size, update_ctx);
+        } else {
+            *consumed = c->dsp.resample_linear[fn_idx](c, dst, src, dst_size, update_ctx);
+        }
+    }
 
-#endif // HAVE_MMXEXT_INLINE
+    return dst_size;
+}
 
 static int multiple_resample(ResampleContext *c, AudioData *dst, int dst_size, AudioData *src, int src_size, int *consumed){
     int i, ret= -1;
     int av_unused mm_flags = av_get_cpu_flags();
-    int need_emms= 0;
+    int need_emms = c->format == AV_SAMPLE_FMT_S16P && ARCH_X86_32 &&
+                    (mm_flags & (AV_CPU_FLAG_MMX2 | AV_CPU_FLAG_SSE2)) == AV_CPU_FLAG_MMX2;
     int64_t max_src_size = (INT64_MAX >> (c->phase_shift+1)) / c->src_incr;
 
     if (c->compensation_distance)
@@ -342,32 +315,8 @@ static int multiple_resample(ResampleContext *c, AudioData *dst, int dst_size, A
     src_size = FFMIN(src_size, max_src_size);
 
     for(i=0; i<dst->ch_count; i++){
-#if HAVE_MMXEXT_INLINE
-#if HAVE_SSE2_INLINE
-             if(c->format == AV_SAMPLE_FMT_S16P && (mm_flags&AV_CPU_FLAG_SSE2)) ret= swri_resample_int16_sse2 (c, (int16_t*)dst->ch[i], (const int16_t*)src->ch[i], consumed, src_size, dst_size, i+1==dst->ch_count);
-        else
-#endif
-             if(c->format == AV_SAMPLE_FMT_S16P && (mm_flags&AV_CPU_FLAG_MMX2 )){
-                 ret= swri_resample_int16_mmx2 (c, (int16_t*)dst->ch[i], (const int16_t*)src->ch[i], consumed, src_size, dst_size, i+1==dst->ch_count);
-                 need_emms= 1;
-             } else
-#endif
-             if(c->format == AV_SAMPLE_FMT_S16P) ret= swri_resample_int16(c, (int16_t*)dst->ch[i], (const int16_t*)src->ch[i], consumed, src_size, dst_size, i+1==dst->ch_count);
-        else if(c->format == AV_SAMPLE_FMT_S32P) ret= swri_resample_int32(c, (int32_t*)dst->ch[i], (const int32_t*)src->ch[i], consumed, src_size, dst_size, i+1==dst->ch_count);
-#if HAVE_AVX_INLINE
-        else if(c->format == AV_SAMPLE_FMT_FLTP && (mm_flags&AV_CPU_FLAG_AVX))
-                                                 ret= swri_resample_float_avx (c, (float*)dst->ch[i], (const float*)src->ch[i], consumed, src_size, dst_size, i+1==dst->ch_count);
-#endif
-#if HAVE_SSE_INLINE
-        else if(c->format == AV_SAMPLE_FMT_FLTP && (mm_flags&AV_CPU_FLAG_SSE))
-                                                 ret= swri_resample_float_sse (c, (float*)dst->ch[i], (const float*)src->ch[i], consumed, src_size, dst_size, i+1==dst->ch_count);
-#endif
-        else if(c->format == AV_SAMPLE_FMT_FLTP) ret= swri_resample_float(c, (float  *)dst->ch[i], (const float  *)src->ch[i], consumed, src_size, dst_size, i+1==dst->ch_count);
-#if HAVE_SSE2_INLINE
-        else if(c->format == AV_SAMPLE_FMT_DBLP && (mm_flags&AV_CPU_FLAG_SSE2))
-                                                 ret= swri_resample_double_sse2(c,(double *)dst->ch[i], (const double *)src->ch[i], consumed, src_size, dst_size, i+1==dst->ch_count);
-#endif
-        else if(c->format == AV_SAMPLE_FMT_DBLP) ret= swri_resample_double(c,(double *)dst->ch[i], (const double *)src->ch[i], consumed, src_size, dst_size, i+1==dst->ch_count);
+        ret= swri_resample(c, dst->ch[i], src->ch[i],
+                           consumed, src_size, dst_size, i+1==dst->ch_count);
     }
     if(need_emms)
         emms_c();