Merge commit '42d324694883cdf1fff1612ac70fa403692a1ad4'

* commit '42d324694883cdf1fff1612ac70fa403692a1ad4':
  floatdsp: move vector_fmul_reverse from dsputil to avfloatdsp.

Conflicts:
	libavcodec/arm/dsputil_init_vfp.c
	libavcodec/arm/dsputil_vfp.S
	libavcodec/dsputil.c
	libavcodec/ppc/float_altivec.c
	libavcodec/x86/dsputil.asm
	libavutil/x86/float_dsp.asm

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

3 files changed, 33 insertions, 0 deletions

diff --git a/libavutil/ppc/float_dsp_altivec.c b/libavutil/ppc/float_dsp_altivec.c
index fa66d5c1ca..8cee82c1c7 100644
--- a/libavutil/ppc/float_dsp_altivec.c
+++ b/libavutil/ppc/float_dsp_altivec.c
@@ -93,3 +93,32 @@ void ff_vector_fmul_add_altivec(float *dst, const float *src0,
         vec_st(t0, 0, dst + i);
     }
 }
+
+void ff_vector_fmul_reverse_altivec(float *dst, const float *src0,
+                                    const float *src1, int len)
+{
+    int i;
+    vector float d, s0, s1, h0, l0,
+                s2, s3, zero = (vector float) vec_splat_u32(0);
+
+    src1 += len-4;
+    for(i = 0; i < len - 7; i += 8) {
+        s1 = vec_ld(0, src1 - i);              // [a,b,c,d]
+        s0 = vec_ld(0, src0 + i);
+        l0 = vec_mergel(s1, s1);               // [c,c,d,d]
+        s3 = vec_ld(-16, src1 - i);
+        h0 = vec_mergeh(s1, s1);               // [a,a,b,b]
+        s2 = vec_ld(16, src0 + i);
+        s1 = vec_mergeh(vec_mergel(l0, h0),    // [d,b,d,b]
+                        vec_mergeh(l0, h0));   // [c,a,c,a]
+        // [d,c,b,a]
+        l0 = vec_mergel(s3, s3);
+        d = vec_madd(s0, s1, zero);
+        h0 = vec_mergeh(s3, s3);
+        vec_st(d, 0, dst + i);
+        s3 = vec_mergeh(vec_mergel(l0, h0),
+                        vec_mergeh(l0, h0));
+        d = vec_madd(s2, s3, zero);
+        vec_st(d, 16, dst + i);
+    }
+}
diff --git a/libavutil/ppc/float_dsp_altivec.h b/libavutil/ppc/float_dsp_altivec.h
index 2bb87cd281..b262a83548 100644
--- a/libavutil/ppc/float_dsp_altivec.h
+++ b/libavutil/ppc/float_dsp_altivec.h
@@ -32,4 +32,7 @@ extern void ff_vector_fmul_add_altivec(float *dst, const float *src0,
                                        const float *src1, const float *src2,
                                        int len);
 
+extern void ff_vector_fmul_reverse_altivec(float *dst, const float *src0,
+                                           const float *src1, int len);
+
 #endif /* AVUTIL_PPC_FLOAT_DSP_ALTIVEC_H */
diff --git a/libavutil/ppc/float_dsp_init.c b/libavutil/ppc/float_dsp_init.c
index f00cae0487..d9ca53eeec 100644
--- a/libavutil/ppc/float_dsp_init.c
+++ b/libavutil/ppc/float_dsp_init.c
@@ -33,6 +33,7 @@ void ff_float_dsp_init_ppc(AVFloatDSPContext *fdsp, int bit_exact)
 
     fdsp->vector_fmul = ff_vector_fmul_altivec;
     fdsp->vector_fmul_add = ff_vector_fmul_add_altivec;
+    fdsp->vector_fmul_reverse = ff_vector_fmul_reverse_altivec;
 
     if (!bit_exact) {
         fdsp->vector_fmul_window = ff_vector_fmul_window_altivec;