avfilter/x86/vf_gblur: add postscale SIMD

3 files changed, 70 insertions, 9 deletions

diff --git a/libavfilter/vf_gblur.c b/libavfilter/vf_gblur.c
index 70e2a668b4..109a7a95f9 100644
--- a/libavfilter/vf_gblur.c
+++ b/libavfilter/vf_gblur.c
@@ -171,13 +171,14 @@ static int filter_postscale(AVFilterContext *ctx, void *arg, int jobnr, int nb_j
     const float min = s->flt ? -FLT_MAX : 0.f;
     const int height = td->height;
     const int width = td->width;
-    const int64_t numpixels = width * (int64_t)height;
-    const int slice_start = (numpixels *  jobnr   ) / nb_jobs;
-    const int slice_end   = (numpixels * (jobnr+1)) / nb_jobs;
+    const int awidth = FFALIGN(width, 64);
+    const int slice_start = (height *  jobnr   ) / nb_jobs;
+    const int slice_end   = (height * (jobnr+1)) / nb_jobs;
     const float postscale = s->postscale * s->postscaleV;
-    float *buffer = s->buffer + slice_start;
+    const int slice_size = slice_end - slice_start;
 
-    s->postscale_slice(buffer, slice_end - slice_start, postscale, min, max);
+    s->postscale_slice(s->buffer + slice_start * awidth,
+                       slice_size * awidth, postscale, min, max);
 
     return 0;
 }
@@ -251,7 +252,7 @@ static int config_input(AVFilterLink *inlink)
 
     s->nb_planes = av_pix_fmt_count_planes(inlink->format);
 
-    s->buffer = av_malloc_array(FFALIGN(inlink->w, 16), FFALIGN(inlink->h, 16) * sizeof(*s->buffer));
+    s->buffer = av_malloc_array(FFALIGN(inlink->w, 64), FFALIGN(inlink->h, 64) * sizeof(*s->buffer));
     if (!s->buffer)
         return AVERROR(ENOMEM);
 
diff --git a/libavfilter/x86/vf_gblur.asm b/libavfilter/x86/vf_gblur.asm
index a25b1659f5..c29ecba889 100644
--- a/libavfilter/x86/vf_gblur.asm
+++ b/libavfilter/x86/vf_gblur.asm
@@ -183,3 +183,52 @@ HORIZ_SLICE
 INIT_XMM avx2
 HORIZ_SLICE
 %endif
+
+%macro POSTSCALE_SLICE 0
+%if UNIX64
+cglobal postscale_slice, 2, 2, 4, ptr, length
+%else
+cglobal postscale_slice, 5, 5, 4, ptr, length, postscale, min, max
+%endif
+    shl lengthd, 2
+    add ptrq, lengthq
+    neg lengthq
+%if WIN64
+    SWAP 0, 2
+    SWAP 1, 3
+    SWAP 2, 4
+%endif
+%if cpuflag(avx2)
+    vbroadcastss  m0, xm0
+    vbroadcastss  m1, xm1
+    vbroadcastss  m2, xm2
+%else
+    shufps   xm0, xm0, 0
+    shufps   xm1, xm1, 0
+    shufps   xm2, xm2, 0
+%endif
+
+    .loop:
+%if cpuflag(avx2)
+    mulps         m3, m0, [ptrq + lengthq]
+%else
+    movu          m3, [ptrq + lengthq]
+    mulps         m3, m0
+%endif
+    maxps         m3, m1
+    minps         m3, m2
+    movu   [ptrq+lengthq], m3
+
+    add lengthq, mmsize
+    jl .loop
+
+    RET
+%endmacro
+
+INIT_XMM sse
+POSTSCALE_SLICE
+
+%if HAVE_AVX2_EXTERNAL
+INIT_YMM avx2
+POSTSCALE_SLICE
+%endif
diff --git a/libavfilter/x86/vf_gblur_init.c b/libavfilter/x86/vf_gblur_init.c
index e63e59fe23..d80fb46fe4 100644
--- a/libavfilter/x86/vf_gblur_init.c
+++ b/libavfilter/x86/vf_gblur_init.c
@@ -27,14 +27,25 @@
 void ff_horiz_slice_sse4(float *ptr, int width, int height, int steps, float nu, float bscale);
 void ff_horiz_slice_avx2(float *ptr, int width, int height, int steps, float nu, float bscale);
 
+void ff_postscale_slice_sse(float *ptr, int length, float postscale, float min, float max);
+void ff_postscale_slice_avx2(float *ptr, int length, float postscale, float min, float max);
+
 av_cold void ff_gblur_init_x86(GBlurContext *s)
 {
-#if ARCH_X86_64
     int cpu_flags = av_get_cpu_flags();
 
-    if (EXTERNAL_SSE4(cpu_flags))
+    if (EXTERNAL_SSE(cpu_flags)) {
+        s->postscale_slice = ff_postscale_slice_sse;
+    }
+    if (EXTERNAL_AVX2_FAST(cpu_flags)) {
+        s->postscale_slice = ff_postscale_slice_avx2;
+    }
+#if ARCH_X86_64
+    if (EXTERNAL_SSE4(cpu_flags)) {
         s->horiz_slice = ff_horiz_slice_sse4;
-    if (EXTERNAL_AVX2(cpu_flags))
+    }
+    if (EXTERNAL_AVX2(cpu_flags)) {
         s->horiz_slice = ff_horiz_slice_avx2;
+    }
 #endif
 }