float_dsp: add x86-optimized functions for vector_fmac_scalar()

libavutil/float_dsp.h

@@ -42,12 +42,12 @@ typedef struct AVFloatDSPContext {
      * overlap exactly or not at all.
      *
      * @param dst result vector
-     *            constraints: 16-byte aligned
+     *            constraints: 32-byte aligned
      * @param src input vector
-     *            constraints: 16-byte aligned
+     *            constraints: 32-byte aligned
      * @param mul scalar value
      * @param len length of vector
-     *            constraints: multiple of 4
+     *            constraints: multiple of 16
      */
     void (*vector_fmac_scalar)(float *dst, const float *src, float mul,
                                int len);

libavutil/x86/float_dsp.asm

@@ -19,6 +19,7 @@
 ;******************************************************************************
 
 %include "x86inc.asm"
+%include "x86util.asm"
 
 SECTION .text
 
@@ -53,3 +54,49 @@ VECTOR_FMUL
 INIT_YMM avx
 VECTOR_FMUL
 %endif
+
+;------------------------------------------------------------------------------
+; void ff_vector_fmac_scalar(float *dst, const float *src, float mul, int len)
+;------------------------------------------------------------------------------
+
+%macro VECTOR_FMAC_SCALAR 0
+%if UNIX64
+cglobal vector_fmac_scalar, 3,3,3, dst, src, len
+%else
+cglobal vector_fmac_scalar, 4,4,3, dst, src, mul, len
+%endif
+%if WIN64
+    SWAP 0, 2
+%endif
+%if ARCH_X86_32
+    VBROADCASTSS m0, mulm
+%else
+    shufps     xmm0, xmm0, 0
+%if cpuflag(avx)
+    vinsertf128  m0, m0, xmm0, 1
+%endif
+%endif
+    lea    lenq, [lend*4-2*mmsize]
+.loop
+    mulps    m1, m0, [srcq+lenq       ]
+    mulps    m2, m0, [srcq+lenq+mmsize]
+    addps    m1, m1, [dstq+lenq       ]
+    addps    m2, m2, [dstq+lenq+mmsize]
+    mova  [dstq+lenq       ], m1
+    mova  [dstq+lenq+mmsize], m2
+    sub    lenq, 2*mmsize
+    jge .loop
+%if mmsize == 32
+    vzeroupper
+    RET
+%else
+    REP_RET
+%endif
+%endmacro
+
+INIT_XMM sse
+VECTOR_FMAC_SCALAR
+%if HAVE_AVX
+INIT_YMM avx
+VECTOR_FMAC_SCALAR
+%endif

libavutil/x86/float_dsp_init.c

@@ -26,6 +26,11 @@ extern void ff_vector_fmul_sse(float *dst, const float *src0, const float *src1,
 extern void ff_vector_fmul_avx(float *dst, const float *src0, const float *src1,
                                int len);
 
+extern void ff_vector_fmac_scalar_sse(float *dst, const float *src, float mul,
+                                      int len);
+extern void ff_vector_fmac_scalar_avx(float *dst, const float *src, float mul,
+                                      int len);
+
 void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp)
 {
 #if HAVE_YASM
@@ -33,9 +38,11 @@ void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp)
 
     if (mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE) {
         fdsp->vector_fmul = ff_vector_fmul_sse;
+        fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_sse;
     }
     if (mm_flags & AV_CPU_FLAG_AVX && HAVE_AVX) {
         fdsp->vector_fmul = ff_vector_fmul_avx;
+        fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_avx;
     }
 #endif
 }