avfilter: add hflip x86 SIMD

Signed-off-by: Paul B Mahol <onemda@gmail.com>

1 files changed, 86 insertions, 47 deletions

diff --git a/libavfilter/vf_hflip.c b/libavfilter/vf_hflip.c
index cf20c193f7..957ddd9900 100644
--- a/libavfilter/vf_hflip.c
+++ b/libavfilter/vf_hflip.c
@@ -29,6 +29,7 @@
 #include "libavutil/opt.h"
 #include "avfilter.h"
 #include "formats.h"
+#include "hflip.h"
 #include "internal.h"
 #include "video.h"
 #include "libavutil/pixdesc.h"
@@ -36,13 +37,6 @@
 #include "libavutil/intreadwrite.h"
 #include "libavutil/imgutils.h"
 
-typedef struct FlipContext {
-    const AVClass *class;
-    int max_step[4];    ///< max pixel step for each plane, expressed as a number of bytes
-    int planewidth[4];  ///< width of each plane
-    int planeheight[4]; ///< height of each plane
-} FlipContext;
-
 static const AVOption hflip_options[] = {
     { NULL }
 };
@@ -67,12 +61,77 @@ static int query_formats(AVFilterContext *ctx)
     return ff_set_common_formats(ctx, pix_fmts);
 }
 
+static void hflip_byte_c(const uint8_t *src, uint8_t *dst, int w)
+{
+    int j;
+
+    for (j = 0; j < w; j++)
+        dst[j] = src[-j];
+}
+
+static void hflip_short_c(const uint8_t *ssrc, uint8_t *ddst, int w)
+{
+    const uint16_t *src = (const uint16_t *)ssrc;
+    uint16_t *dst = (uint16_t *)ddst;
+    int j;
+
+    for (j = 0; j < w; j++)
+        dst[j] = src[-j];
+}
+
+static void hflip_dword_c(const uint8_t *ssrc, uint8_t *ddst, int w)
+{
+    const uint32_t *src = (const uint32_t *)ssrc;
+    uint32_t *dst = (uint32_t *)ddst;
+    int j;
+
+    for (j = 0; j < w; j++)
+        dst[j] = src[-j];
+}
+
+static void hflip_b24_c(const uint8_t *src, uint8_t *dst, int w)
+{
+    const uint8_t *in  = src;
+    uint8_t *out = dst;
+    int j;
+
+    for (j = 0; j < w; j++, out += 3, in -= 3) {
+        int32_t v = AV_RB24(in);
+
+        AV_WB24(out, v);
+    }
+}
+
+static void hflip_b48_c(const uint8_t *src, uint8_t *dst, int w)
+{
+    const uint8_t *in  = src;
+    uint8_t *out = dst;
+    int j;
+
+    for (j = 0; j < w; j++, out += 6, in -= 6) {
+        int64_t v = AV_RB48(in);
+
+        AV_WB48(out, v);
+    }
+}
+
+static void hflip_qword_c(const uint8_t *ssrc, uint8_t *ddst, int w)
+{
+    const uint64_t *src = (const uint64_t *)ssrc;
+    uint64_t *dst = (uint64_t *)ddst;
+    int j;
+
+    for (j = 0; j < w; j++)
+        dst[j] = src[-j];
+}
+
 static int config_props(AVFilterLink *inlink)
 {
     FlipContext *s = inlink->dst->priv;
     const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
     const int hsub = pix_desc->log2_chroma_w;
     const int vsub = pix_desc->log2_chroma_h;
+    int nb_planes, i;
 
     av_image_fill_max_pixsteps(s->max_step, NULL, pix_desc);
     s->planewidth[0]  = s->planewidth[3]  = inlink->w;
@@ -80,6 +139,24 @@ static int config_props(AVFilterLink *inlink)
     s->planeheight[0] = s->planeheight[3] = inlink->h;
     s->planeheight[1] = s->planeheight[2] = AV_CEIL_RSHIFT(inlink->h, vsub);
 
+    nb_planes = av_pix_fmt_count_planes(inlink->format);
+
+    for (i = 0; i < nb_planes; i++) {
+        switch (s->max_step[i]) {
+        case 1: s->flip_line[i] = hflip_byte_c;  break;
+        case 2: s->flip_line[i] = hflip_short_c; break;
+        case 3: s->flip_line[i] = hflip_b24_c;   break;
+        case 4: s->flip_line[i] = hflip_dword_c; break;
+        case 6: s->flip_line[i] = hflip_b48_c;   break;
+        case 8: s->flip_line[i] = hflip_qword_c; break;
+        default:
+            return AVERROR_BUG;
+        }
+    }
+
+    if (ARCH_X86)
+        ff_hflip_init_x86(s, s->max_step, nb_planes);
+
     return 0;
 }
 
@@ -94,7 +171,7 @@ static int filter_slice(AVFilterContext *ctx, void *arg, int job, int nb_jobs)
     AVFrame *in = td->in;
     AVFrame *out = td->out;
     uint8_t *inrow, *outrow;
-    int i, j, plane, step;
+    int i, plane, step;
 
     for (plane = 0; plane < 4 && in->data[plane] && in->linesize[plane]; plane++) {
         const int width  = s->planewidth[plane];
@@ -107,45 +184,7 @@ static int filter_slice(AVFilterContext *ctx, void *arg, int job, int nb_jobs)
         outrow = out->data[plane] + start * out->linesize[plane];
         inrow  = in ->data[plane] + start * in->linesize[plane] + (width - 1) * step;
         for (i = start; i < end; i++) {
-            switch (step) {
-            case 1:
-                for (j = 0; j < width; j++)
-                    outrow[j] = inrow[-j];
-            break;
-
-            case 2:
-            {
-                uint16_t *outrow16 = (uint16_t *)outrow;
-                uint16_t * inrow16 = (uint16_t *) inrow;
-                for (j = 0; j < width; j++)
-                    outrow16[j] = inrow16[-j];
-            }
-            break;
-
-            case 3:
-            {
-                uint8_t *in  =  inrow;
-                uint8_t *out = outrow;
-                for (j = 0; j < width; j++, out += 3, in -= 3) {
-                    int32_t v = AV_RB24(in);
-                    AV_WB24(out, v);
-                }
-            }
-            break;
-
-            case 4:
-            {
-                uint32_t *outrow32 = (uint32_t *)outrow;
-                uint32_t * inrow32 = (uint32_t *) inrow;
-                for (j = 0; j < width; j++)
-                    outrow32[j] = inrow32[-j];
-            }
-            break;
-
-            default:
-                for (j = 0; j < width; j++)
-                    memcpy(outrow + j*step, inrow - j*step, step);
-            }
+            s->flip_line[plane](inrow, outrow, width);
 
             inrow  += in ->linesize[plane];
             outrow += out->linesize[plane];