[FFmpeg-devel,2/3] avfilter/vf_yadif_cuda: CUDA accelerated deinterlacer

Submitted by Philip Langdale on Nov. 2, 2018, 3:08 a.m.

Details

Message ID 20181102030839.8674-3-philipl@overt.org
State New
Headers show

Commit Message

Philip Langdale Nov. 2, 2018, 3:08 a.m.
Signed-off-by: Philip Langdale <philipl@overt.org>
---
 Changelog                    |   1 +
 configure                    |   1 +
 doc/filters.texi             |  58 +++++
 libavfilter/Makefile         |   1 +
 libavfilter/allfilters.c     |   1 +
 libavfilter/vf_yadif_cuda.c  | 426 +++++++++++++++++++++++++++++++++++
 libavfilter/vf_yadif_cuda.cu | 296 ++++++++++++++++++++++++
 7 files changed, 784 insertions(+)
 create mode 100644 libavfilter/vf_yadif_cuda.c
 create mode 100644 libavfilter/vf_yadif_cuda.cu

Patch hide | download patch | download mbox

diff --git a/Changelog b/Changelog
index 8430da3c6a..f92886fc2e 100644
--- a/Changelog
+++ b/Changelog
@@ -44,6 +44,7 @@  version 4.1:
 - xstack filter
 - pcm vidc decoder and encoder
 - (a)graphmonitor filter
+- yadif_cuda filter
 
 
 version 4.0:
diff --git a/configure b/configure
index 2606b885b0..f3fa0cde86 100755
--- a/configure
+++ b/configure
@@ -3482,6 +3482,7 @@  zscale_filter_deps="libzimg const_nan"
 scale_vaapi_filter_deps="vaapi"
 vpp_qsv_filter_deps="libmfx"
 vpp_qsv_filter_select="qsvvpp"
+yadif_cuda_filter_deps="cuda_sdk"
 
 # examples
 avio_dir_cmd_deps="avformat avutil"
diff --git a/doc/filters.texi b/doc/filters.texi
index 4345a4931b..5d4bfd2e8e 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -17943,6 +17943,64 @@  filter").
 It accepts the following parameters:
 
 
+@table @option
+
+@item mode
+The interlacing mode to adopt. It accepts one of the following values:
+
+@table @option
+@item 0, send_frame
+Output one frame for each frame.
+@item 1, send_field
+Output one frame for each field.
+@item 2, send_frame_nospatial
+Like @code{send_frame}, but it skips the spatial interlacing check.
+@item 3, send_field_nospatial
+Like @code{send_field}, but it skips the spatial interlacing check.
+@end table
+
+The default value is @code{send_frame}.
+
+@item parity
+The picture field parity assumed for the input interlaced video. It accepts one
+of the following values:
+
+@table @option
+@item 0, tff
+Assume the top field is first.
+@item 1, bff
+Assume the bottom field is first.
+@item -1, auto
+Enable automatic detection of field parity.
+@end table
+
+The default value is @code{auto}.
+If the interlacing is unknown or the decoder does not export this information,
+top field first will be assumed.
+
+@item deint
+Specify which frames to deinterlace. Accept one of the following
+values:
+
+@table @option
+@item 0, all
+Deinterlace all frames.
+@item 1, interlaced
+Only deinterlace frames marked as interlaced.
+@end table
+
+The default value is @code{all}.
+@end table
+
+@section yadif_cuda
+
+Deinterlace the input video using the @ref{yadif} algorithm, but implemented
+in CUDA so that it can work as part of a GPU accelerated pipeline with nvdec
+and/or nvenc.
+
+It accepts the following parameters:
+
+
 @table @option
 
 @item mode
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index ffbcb40806..4b78b29fad 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -408,6 +408,7 @@  OBJS-$(CONFIG_WEAVE_FILTER)                  += vf_weave.o
 OBJS-$(CONFIG_XBR_FILTER)                    += vf_xbr.o
 OBJS-$(CONFIG_XSTACK_FILTER)                 += vf_stack.o framesync.o
 OBJS-$(CONFIG_YADIF_FILTER)                  += vf_yadif.o yadif_common.o
+OBJS-$(CONFIG_YADIF_CUDA_FILTER)             += vf_yadif_cuda.o vf_yadif_cuda.ptx.o yadif_common.o
 OBJS-$(CONFIG_ZMQ_FILTER)                    += f_zmq.o
 OBJS-$(CONFIG_ZOOMPAN_FILTER)                += vf_zoompan.o
 OBJS-$(CONFIG_ZSCALE_FILTER)                 += vf_zscale.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index d5a211bda5..c40c7e3a3c 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -389,6 +389,7 @@  extern AVFilter ff_vf_weave;
 extern AVFilter ff_vf_xbr;
 extern AVFilter ff_vf_xstack;
 extern AVFilter ff_vf_yadif;
+extern AVFilter ff_vf_yadif_cuda;
 extern AVFilter ff_vf_zmq;
 extern AVFilter ff_vf_zoompan;
 extern AVFilter ff_vf_zscale;
diff --git a/libavfilter/vf_yadif_cuda.c b/libavfilter/vf_yadif_cuda.c
new file mode 100644
index 0000000000..be22344d9d
--- /dev/null
+++ b/libavfilter/vf_yadif_cuda.c
@@ -0,0 +1,426 @@ 
+/*
+ * Copyright (C) 2018 Philip Langdale <philipl@overt.org>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <cuda.h>
+#include "libavutil/avassert.h"
+#include "libavutil/hwcontext_cuda.h"
+#include "internal.h"
+#include "yadif.h"
+
+extern char vf_yadif_cuda_ptx[];
+
+typedef struct DeintCUDAContext {
+    YADIFContext yadif;
+
+    AVCUDADeviceContext *hwctx;
+    AVBufferRef         *device_ref;
+    AVBufferRef         *input_frames_ref;
+    AVHWFramesContext   *input_frames;
+
+    CUcontext   cu_ctx;
+    CUstream    stream;
+    CUmodule    cu_module;
+    CUfunction  cu_func_uchar;
+    CUfunction  cu_func_uchar2;
+    CUfunction  cu_func_ushort;
+    CUfunction  cu_func_ushort2;
+} DeintCUDAContext;
+
+#define DIV_UP(a, b) ( ((a) + (b) - 1) / (b) )
+#define ALIGN_UP(a, b) (((a) + (b) - 1) & ~((b) - 1))
+#define BLOCKX 32
+#define BLOCKY 16
+
+static int check_cu(AVFilterContext *avctx, CUresult err, const char *func)
+{
+    const char *err_name;
+    const char *err_string;
+
+    av_log(avctx, AV_LOG_TRACE, "Calling %s\n", func);
+
+    if (err == CUDA_SUCCESS)
+        return 0;
+
+    cuGetErrorName(err, &err_name);
+    cuGetErrorString(err, &err_string);
+
+    av_log(avctx, AV_LOG_ERROR, "%s failed", func);
+    if (err_name && err_string)
+        av_log(avctx, AV_LOG_ERROR, " -> %s: %s", err_name, err_string);
+    av_log(avctx, AV_LOG_ERROR, "\n");
+
+    return AVERROR_EXTERNAL;
+}
+
+#define CHECK_CU(x) check_cu(ctx, (x), #x)
+
+static CUresult call_kernel(AVFilterContext *ctx, CUfunction func,
+                            CUdeviceptr prev, CUdeviceptr cur, CUdeviceptr next,
+                            CUarray_format format, int channels,
+                            int src_width,  // Width is pixels per channel
+                            int src_height, // Height is pixels per channel
+                            int src_pitch,  // Pitch is bytes
+                            CUdeviceptr dst,
+                            int dst_width,  // Width is pixels per channel
+                            int dst_height, // Height is pixels per channel
+                            int dst_pitch,  // Pitch is pixels per channel
+                            int parity, int tff)
+{
+    DeintCUDAContext *s = ctx->priv;
+    CUtexObject tex_prev = 0, tex_cur = 0, tex_next = 0;
+    CUresult err;
+    int skip_spatial_check = s->yadif.mode&2;
+
+    void *args[] = { &dst, &tex_prev, &tex_cur, &tex_next,
+                     &dst_width, &dst_height, &dst_pitch,
+                     &src_width, &src_height, &parity, &tff,
+                     &skip_spatial_check };
+
+    CUDA_TEXTURE_DESC tex_desc = {
+        .filterMode = CU_TR_FILTER_MODE_POINT,
+        .flags = CU_TRSF_READ_AS_INTEGER,
+    };
+
+    CUDA_RESOURCE_DESC res_desc = {
+        .resType = CU_RESOURCE_TYPE_PITCH2D,
+        .res.pitch2D.format = format,
+        .res.pitch2D.numChannels = channels,
+        .res.pitch2D.width = src_width,
+        .res.pitch2D.height = src_height,
+        .res.pitch2D.pitchInBytes = src_pitch,
+    };
+
+    res_desc.res.pitch2D.devPtr = (CUdeviceptr)prev;
+    err = CHECK_CU(cuTexObjectCreate(&tex_prev, &res_desc, &tex_desc, NULL));
+    if (err != CUDA_SUCCESS) {
+        goto exit;
+    }
+
+    res_desc.res.pitch2D.devPtr = (CUdeviceptr)cur;
+    err = CHECK_CU(cuTexObjectCreate(&tex_cur, &res_desc, &tex_desc, NULL));
+    if (err != CUDA_SUCCESS) {
+        goto exit;
+    }
+
+    res_desc.res.pitch2D.devPtr = (CUdeviceptr)next;
+    err = CHECK_CU(cuTexObjectCreate(&tex_next, &res_desc, &tex_desc, NULL));
+    if (err != CUDA_SUCCESS) {
+        goto exit;
+    }
+
+    err = CHECK_CU(cuLaunchKernel(func,
+                                  DIV_UP(dst_width, BLOCKX), DIV_UP(dst_height, BLOCKY), 1,
+                                  BLOCKX, BLOCKY, 1,
+                                  0, s->stream, args, NULL));
+
+exit:
+    if (tex_prev)
+        CHECK_CU(cuTexObjectDestroy(tex_prev));
+    if (tex_cur)
+        CHECK_CU(cuTexObjectDestroy(tex_cur));
+    if (tex_next)
+        CHECK_CU(cuTexObjectDestroy(tex_next));
+
+    return err;
+}
+
+static void filter(AVFilterContext *ctx, AVFrame *dst,
+                   int parity, int tff)
+{
+    DeintCUDAContext *s = ctx->priv;
+    YADIFContext *y = &s->yadif;
+    CUcontext dummy;
+    CUresult err;
+    int i;
+
+    err = CHECK_CU(cuCtxPushCurrent(s->cu_ctx));
+    if (err != CUDA_SUCCESS) {
+        goto exit;
+    }
+
+    for (i = 0; i < y->csp->nb_components; i++) {
+        CUfunction func;
+        CUarray_format format;
+        int pixel_size, channels;
+        const AVComponentDescriptor *comp = &y->csp->comp[i];
+
+        if (comp->plane < i) {
+            // We process planes as a whole, so don't reprocess
+            // them for additional components
+            continue;
+        }
+
+        pixel_size = (comp->depth + comp->shift) / 8;
+        channels = comp->step / pixel_size;
+        if (pixel_size > 2 || channels > 2) {
+            av_log(ctx, AV_LOG_ERROR, "Unsupported pixel format: %s\n", y->csp->name);
+            goto exit;
+        }
+        switch (pixel_size) {
+        case 1:
+            func = channels == 1 ? s->cu_func_uchar : s->cu_func_uchar2;
+            format = CU_AD_FORMAT_UNSIGNED_INT8;
+            break;
+        case 2:
+            func = channels == 1 ? s->cu_func_ushort : s->cu_func_ushort2;
+            format = CU_AD_FORMAT_UNSIGNED_INT16;
+            break;
+        default:
+            av_log(ctx, AV_LOG_ERROR, "Unsupported pixel format: %s\n", y->csp->name);
+            goto exit;
+        }
+        av_log(ctx, AV_LOG_TRACE,
+               "Deinterlacing plane %d: pixel_size: %d channels: %d\n",
+               comp->plane, pixel_size, channels);
+        call_kernel(ctx, func,
+                    (CUdeviceptr)y->prev->data[i],
+                    (CUdeviceptr)y->cur->data[i],
+                    (CUdeviceptr)y->next->data[i],
+                    format, channels,
+                    AV_CEIL_RSHIFT(y->cur->width, i ? y->csp->log2_chroma_w : 0),
+                    AV_CEIL_RSHIFT(y->cur->height, i ? y->csp->log2_chroma_h : 0),
+                    y->cur->linesize[i],
+                    (CUdeviceptr)dst->data[i],
+                    AV_CEIL_RSHIFT(dst->width, i ? y->csp->log2_chroma_w : 0),
+                    AV_CEIL_RSHIFT(dst->height, i ? y->csp->log2_chroma_h : 0),
+                    dst->linesize[i] / comp->step,
+                    parity, tff);
+    }
+
+    err = CHECK_CU(cuStreamSynchronize(s->stream));
+    if (err != CUDA_SUCCESS) {
+        goto exit;
+    }
+
+exit:
+    CHECK_CU(cuCtxPopCurrent(&dummy));
+    return;
+}
+
+static av_cold void deint_cuda_uninit(AVFilterContext *ctx)
+{
+    CUcontext dummy;
+    DeintCUDAContext *s = ctx->priv;
+    YADIFContext *y = &s->yadif;
+
+    if (s->cu_module) {
+        CHECK_CU(cuCtxPushCurrent(s->cu_ctx));
+        CHECK_CU(cuModuleUnload(s->cu_module));
+        CHECK_CU(cuCtxPopCurrent(&dummy));
+    }
+
+    av_frame_free(&y->prev);
+    av_frame_free(&y->cur);
+    av_frame_free(&y->next);
+
+    av_buffer_unref(&s->device_ref);
+    s->hwctx = NULL;
+    av_buffer_unref(&s->input_frames_ref);
+    s->input_frames = NULL;
+}
+
+static int deint_cuda_query_formats(AVFilterContext *ctx)
+{
+    enum AVPixelFormat pix_fmts[] = {
+        AV_PIX_FMT_CUDA, AV_PIX_FMT_NONE,
+    };
+    int ret;
+
+    if ((ret = ff_formats_ref(ff_make_format_list(pix_fmts),
+                              &ctx->inputs[0]->out_formats)) < 0)
+        return ret;
+    if ((ret = ff_formats_ref(ff_make_format_list(pix_fmts),
+                              &ctx->outputs[0]->in_formats)) < 0)
+        return ret;
+
+    return 0;
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+    AVFilterContext *ctx = inlink->dst;
+    DeintCUDAContext *s  = ctx->priv;
+
+    if (!inlink->hw_frames_ctx) {
+        av_log(ctx, AV_LOG_ERROR, "A hardware frames reference is "
+               "required to associate the processing device.\n");
+        return AVERROR(EINVAL);
+    }
+
+    s->input_frames_ref = av_buffer_ref(inlink->hw_frames_ctx);
+    if (!s->input_frames_ref) {
+        av_log(ctx, AV_LOG_ERROR, "A input frames reference create "
+               "failed.\n");
+        return AVERROR(ENOMEM);
+    }
+    s->input_frames = (AVHWFramesContext*)s->input_frames_ref->data;
+
+    return 0;
+}
+
+static int config_output(AVFilterLink *link)
+{
+    AVHWFramesContext *output_frames;
+    AVFilterContext *ctx = link->src;
+    DeintCUDAContext *s = ctx->priv;
+    YADIFContext *y = &s->yadif;
+    int ret = 0;
+    CUcontext dummy;
+    CUresult err;
+
+    av_assert0(s->input_frames);
+    s->device_ref = av_buffer_ref(s->input_frames->device_ref);
+    if (!s->device_ref) {
+        av_log(ctx, AV_LOG_ERROR, "A device reference create "
+               "failed.\n");
+        return AVERROR(ENOMEM);
+    }
+    s->hwctx = ((AVHWDeviceContext*)s->device_ref->data)->hwctx;
+    s->cu_ctx = s->hwctx->cuda_ctx;
+    s->stream = s->hwctx->stream;
+
+    link->hw_frames_ctx = av_hwframe_ctx_alloc(s->device_ref);
+    if (!link->hw_frames_ctx) {
+        av_log(ctx, AV_LOG_ERROR, "Failed to create HW frame context "
+               "for output.\n");
+        ret = AVERROR(ENOMEM);
+        goto exit;
+    }
+
+    output_frames = (AVHWFramesContext*)link->hw_frames_ctx->data;
+
+    output_frames->format    = AV_PIX_FMT_CUDA;
+    output_frames->sw_format = s->input_frames->sw_format;
+    output_frames->width     = ctx->inputs[0]->w;
+    output_frames->height    = ctx->inputs[0]->h;
+
+    output_frames->initial_pool_size = 4;
+
+    ret = ff_filter_init_hw_frames(ctx, link, 10);
+    if (ret < 0)
+        goto exit;
+
+    ret = av_hwframe_ctx_init(link->hw_frames_ctx);
+    if (ret < 0) {
+        av_log(ctx, AV_LOG_ERROR, "Failed to initialise CUDA frame "
+               "context for output: %d\n", ret);
+        goto exit;
+    }
+
+    link->time_base.num = ctx->inputs[0]->time_base.num;
+    link->time_base.den = ctx->inputs[0]->time_base.den * 2;
+    link->w             = ctx->inputs[0]->w;
+    link->h             = ctx->inputs[0]->h;
+
+    if(y->mode & 1)
+        link->frame_rate = av_mul_q(ctx->inputs[0]->frame_rate,
+                                    (AVRational){2, 1});
+
+    if (link->w < 3 || link->h < 3) {
+        av_log(ctx, AV_LOG_ERROR, "Video of less than 3 columns or lines is not supported\n");
+        ret = AVERROR(EINVAL);
+        goto exit;
+    }
+
+    y->csp = av_pix_fmt_desc_get(output_frames->sw_format);
+    y->filter = filter;
+
+    err = CHECK_CU(cuCtxPushCurrent(s->cu_ctx));
+    if (err != CUDA_SUCCESS) {
+        ret = AVERROR_EXTERNAL;
+        goto exit;
+    }
+
+    err = CHECK_CU(cuModuleLoadData(&s->cu_module, vf_yadif_cuda_ptx));
+    if (err != CUDA_SUCCESS) {
+        ret = AVERROR_INVALIDDATA;
+        goto exit;
+    }
+
+    err = CHECK_CU(cuModuleGetFunction(&s->cu_func_uchar, s->cu_module, "yadif_uchar"));
+    if (err != CUDA_SUCCESS) {
+        ret = AVERROR_INVALIDDATA;
+        goto exit;
+    }
+
+    err = CHECK_CU(cuModuleGetFunction(&s->cu_func_uchar2, s->cu_module, "yadif_uchar2"));
+    if (err != CUDA_SUCCESS) {
+        ret = AVERROR_INVALIDDATA;
+        goto exit;
+    }
+
+    err= CHECK_CU(cuModuleGetFunction(&s->cu_func_ushort, s->cu_module, "yadif_ushort"));
+    if (err != CUDA_SUCCESS) {
+        ret = AVERROR_INVALIDDATA;
+        goto exit;
+    }
+
+    err = CHECK_CU(cuModuleGetFunction(&s->cu_func_ushort2, s->cu_module, "yadif_ushort2"));
+    if (err != CUDA_SUCCESS) {
+        ret = AVERROR_INVALIDDATA;
+        goto exit;
+    }
+
+exit:
+    CHECK_CU(cuCtxPopCurrent(&dummy));
+
+    return ret;
+}
+
+static const AVClass yadif_cuda_class = {
+    .class_name = "yadif_cuda",
+    .item_name  = av_default_item_name,
+    .option     = ff_yadif_options,
+    .version    = LIBAVUTIL_VERSION_INT,
+    .category   = AV_CLASS_CATEGORY_FILTER,
+};
+
+static const AVFilterPad deint_cuda_inputs[] = {
+    {
+        .name          = "default",
+        .type          = AVMEDIA_TYPE_VIDEO,
+        .filter_frame  = ff_yadif_filter_frame,
+        .config_props  = config_input,
+    },
+    { NULL }
+};
+
+static const AVFilterPad deint_cuda_outputs[] = {
+    {
+        .name          = "default",
+        .type          = AVMEDIA_TYPE_VIDEO,
+        .request_frame = ff_yadif_request_frame,
+        .config_props  = config_output,
+    },
+    { NULL }
+};
+
+AVFilter ff_vf_yadif_cuda = {
+    .name           = "yadif_cuda",
+    .description    = NULL_IF_CONFIG_SMALL("Deinterlace CUDA frames"),
+    .priv_size      = sizeof(DeintCUDAContext),
+    .priv_class     = &yadif_cuda_class,
+    .uninit         = deint_cuda_uninit,
+    .query_formats  = deint_cuda_query_formats,
+    .inputs         = deint_cuda_inputs,
+    .outputs        = deint_cuda_outputs,
+    .flags          = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
+    .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE,
+};
diff --git a/libavfilter/vf_yadif_cuda.cu b/libavfilter/vf_yadif_cuda.cu
new file mode 100644
index 0000000000..65a902c66b
--- /dev/null
+++ b/libavfilter/vf_yadif_cuda.cu
@@ -0,0 +1,296 @@ 
+/*
+ * Copyright (C) 2018 Philip Langdale <philipl@overt.org>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+template<typename T>
+__inline__ __device__ T spatial_predictor(T a, T b, T c, T d, T e, T f, T g,
+                                          T h, T i, T j, T k, T l, T m, T n)
+{
+    int spatial_pred = (d + k)/2;
+    int spatial_score = abs(c - j) + abs(d - k) + abs(e - l);
+
+    int score = abs(b - k) + abs(c - l) + abs(d - m);
+    if (score < spatial_score) {
+        spatial_pred = (c + l)/2;
+        spatial_score = score;
+        score = abs(a - l) + abs(b - m) + abs(c - n);
+        if (score < spatial_score) {
+          spatial_pred = (b + m)/2;
+          spatial_score = score;
+        }
+    }
+    score = abs(d - i) + abs(e - j) + abs(f - k);
+    if (score < spatial_score) {
+        spatial_pred = (e + j)/2;
+        spatial_score = score;
+        score = abs(e - h) + abs(f - i) + abs(g - j);
+        if (score < spatial_score) {
+          spatial_pred = (f + i)/2;
+          spatial_score = score;
+        }
+    }
+    return spatial_pred;
+}
+
+__inline__ __device__ int max3(int a, int b, int c)
+{
+    int x = max(a, b);
+    return max(x, c);
+}
+
+__inline__ __device__ int min3(int a, int b, int c)
+{
+    int x = min(a, b);
+    return min(x, c);
+}
+
+template<typename T>
+__inline__ __device__ T temporal_predictor(T A, T B, T C, T D, T E, T F,
+                                           T G, T H, T I, T J, T K, T L,
+                                           T spatial_pred, bool skip_check)
+{
+    int p0 = (C + H) / 2;
+    int p1 = F;
+    int p2 = (D + I) / 2;
+    int p3 = G;
+    int p4 = (E + J) / 2;
+
+    int tdiff0 = abs(D - I);
+    int tdiff1 = (abs(A - F) + abs(B - G)) / 2;
+    int tdiff2 = (abs(K - F) + abs(G - L)) / 2;
+
+    int diff = max3(tdiff0, tdiff1, tdiff2);
+
+    if (!skip_check) {
+      int maxi = max3(p2 - p3, p2 - p1, min(p0 - p1, p4 - p3));
+      int mini = min3(p2 - p3, p2 - p1, max(p0 - p1, p4 - p3));
+      diff = max3(diff, mini, -maxi);
+    }
+
+    if (spatial_pred > p2 + diff) {
+      spatial_pred = p2 + diff;
+    }
+    if (spatial_pred < p2 - diff) {
+      spatial_pred = p2 - diff;
+    }
+
+    return spatial_pred;
+}
+
+template<typename T>
+__inline__ __device__ void yadif_single(T *dst,
+                                        cudaTextureObject_t prev,
+                                        cudaTextureObject_t cur,
+                                        cudaTextureObject_t next,
+                                        int dst_width, int dst_height, int dst_pitch,
+                                        int src_width, int src_height,
+                                        int parity, int tff, bool skip_spatial_check)
+{
+    // Identify location
+    int xo = blockIdx.x * blockDim.x + threadIdx.x;
+    int yo = blockIdx.y * blockDim.y + threadIdx.y;
+
+    if (xo >= dst_width || yo >= dst_height) {
+        return;
+    }
+
+    // Don't modify the primary field
+    if (yo % 2 == parity) {
+      dst[yo*dst_pitch+xo] = tex2D<T>(cur, xo, yo);
+      return;
+    }
+
+    // Calculate spatial prediction
+    T a = tex2D<T>(cur, xo - 3, yo - 1);
+    T b = tex2D<T>(cur, xo - 2, yo - 1);
+    T c = tex2D<T>(cur, xo - 1, yo - 1);
+    T d = tex2D<T>(cur, xo - 0, yo - 1);
+    T e = tex2D<T>(cur, xo + 1, yo - 1);
+    T f = tex2D<T>(cur, xo + 2, yo - 1);
+    T g = tex2D<T>(cur, xo + 3, yo - 1);
+
+    T h = tex2D<T>(cur, xo - 3, yo + 1);
+    T i = tex2D<T>(cur, xo - 2, yo + 1);
+    T j = tex2D<T>(cur, xo - 1, yo + 1);
+    T k = tex2D<T>(cur, xo - 0, yo + 1);
+    T l = tex2D<T>(cur, xo + 1, yo + 1);
+    T m = tex2D<T>(cur, xo + 2, yo + 1);
+    T n = tex2D<T>(cur, xo + 3, yo + 1);
+
+    T spatial_pred =
+        spatial_predictor(a, b, c, d, e, f, g, h, i, j, k, l, m, n);
+
+    // Calculate temporal prediction
+    int is_second_field = !(parity ^ tff);
+
+    cudaTextureObject_t prev2 = prev;
+    cudaTextureObject_t prev1 = is_second_field ? cur : prev;
+    cudaTextureObject_t next1 = is_second_field ? next : cur;
+    cudaTextureObject_t next2 = next;
+
+    T A = tex2D<T>(prev2, xo,  yo - 1);
+    T B = tex2D<T>(prev2, xo,  yo + 1);
+    T C = tex2D<T>(prev1, xo,  yo - 2);
+    T D = tex2D<T>(prev1, xo,  yo + 0);
+    T E = tex2D<T>(prev1, xo,  yo + 2);
+    T F = tex2D<T>(cur,   xo,  yo - 1);
+    T G = tex2D<T>(cur,   xo,  yo + 1);
+    T H = tex2D<T>(next1, xo,  yo - 2);
+    T I = tex2D<T>(next1, xo,  yo + 0);
+    T J = tex2D<T>(next1, xo,  yo + 2);
+    T K = tex2D<T>(next2, xo,  yo - 1);
+    T L = tex2D<T>(next2, xo,  yo + 1);
+
+    spatial_pred = temporal_predictor(A, B, C, D, E, F, G, H, I, J, K, L,
+                                      spatial_pred, skip_spatial_check);
+
+    dst[yo*dst_pitch+xo] = spatial_pred;
+}
+
+template <typename T>
+__inline__ __device__ void yadif_double(T *dst,
+                                        cudaTextureObject_t prev,
+                                        cudaTextureObject_t cur,
+                                        cudaTextureObject_t next,
+                                        int dst_width, int dst_height, int dst_pitch,
+                                        int src_width, int src_height,
+                                        int parity, int tff, bool skip_spatial_check)
+{
+    int xo = blockIdx.x * blockDim.x + threadIdx.x;
+    int yo = blockIdx.y * blockDim.y + threadIdx.y;
+
+    if (xo >= dst_width || yo >= dst_height) {
+        return;
+    }
+
+    if (yo % 2 == parity) {
+      // Don't modify the primary field
+      dst[yo*dst_pitch+xo] = tex2D<T>(cur, xo, yo);
+      return;
+    }
+
+    T a = tex2D<T>(cur, xo - 3, yo - 1);
+    T b = tex2D<T>(cur, xo - 2, yo - 1);
+    T c = tex2D<T>(cur, xo - 1, yo - 1);
+    T d = tex2D<T>(cur, xo - 0, yo - 1);
+    T e = tex2D<T>(cur, xo + 1, yo - 1);
+    T f = tex2D<T>(cur, xo + 2, yo - 1);
+    T g = tex2D<T>(cur, xo + 3, yo - 1);
+
+    T h = tex2D<T>(cur, xo - 3, yo + 1);
+    T i = tex2D<T>(cur, xo - 2, yo + 1);
+    T j = tex2D<T>(cur, xo - 1, yo + 1);
+    T k = tex2D<T>(cur, xo - 0, yo + 1);
+    T l = tex2D<T>(cur, xo + 1, yo + 1);
+    T m = tex2D<T>(cur, xo + 2, yo + 1);
+    T n = tex2D<T>(cur, xo + 3, yo + 1);
+
+    T spatial_pred = {
+        spatial_predictor(a.x, b.x, c.x, d.x, e.x, f.x, g.x, h.x, i.x, j.x, k.x, l.x, m.x, n.x),
+        spatial_predictor(a.y, b.y, c.y, d.y, e.y, f.y, g.y, h.y, i.y, j.y, k.y, l.y, m.y, n.y) };
+
+    // Calculate temporal prediction
+    int is_second_field = !(parity ^ tff);
+
+    cudaTextureObject_t prev2 = prev;
+    cudaTextureObject_t prev1 = is_second_field ? cur : prev;
+    cudaTextureObject_t next1 = is_second_field ? next : cur;
+    cudaTextureObject_t next2 = next;
+
+    T A = tex2D<T>(prev2, xo,  yo - 1);
+    T B = tex2D<T>(prev2, xo,  yo + 1);
+    T C = tex2D<T>(prev1, xo,  yo - 2);
+    T D = tex2D<T>(prev1, xo,  yo + 0);
+    T E = tex2D<T>(prev1, xo,  yo + 2);
+    T F = tex2D<T>(cur,   xo,  yo - 1);
+    T G = tex2D<T>(cur,   xo,  yo + 1);
+    T H = tex2D<T>(next1, xo,  yo - 2);
+    T I = tex2D<T>(next1, xo,  yo + 0);
+    T J = tex2D<T>(next1, xo,  yo + 2);
+    T K = tex2D<T>(next2, xo,  yo - 1);
+    T L = tex2D<T>(next2, xo,  yo + 1);
+
+    spatial_pred = {
+        temporal_predictor(A.x, B.x, C.x, D.x, E.x, F.x, G.x, H.x, I.x, J.x, K.x, L.x,
+                           spatial_pred.x, skip_spatial_check),
+        temporal_predictor(A.y, B.y, C.y, D.y, E.y, F.y, G.y, H.y, I.y, J.y, K.y, L.y,
+                           spatial_pred.y, skip_spatial_check) };
+
+    dst[yo*dst_pitch+xo] = spatial_pred;
+}
+
+extern "C" {
+
+__global__ void yadif_uchar(unsigned char *dst,
+                            cudaTextureObject_t prev,
+                            cudaTextureObject_t cur,
+                            cudaTextureObject_t next,
+                            int dst_width, int dst_height, int dst_pitch,
+                            int src_width, int src_height,
+                            int parity, int tff, bool skip_spatial_check)
+{
+    yadif_single(dst, prev, cur, next,
+                 dst_width, dst_height, dst_pitch,
+                 src_width, src_height,
+                 parity, tff, skip_spatial_check);
+}
+
+__global__ void yadif_ushort(unsigned short *dst,
+                            cudaTextureObject_t prev,
+                            cudaTextureObject_t cur,
+                            cudaTextureObject_t next,
+                            int dst_width, int dst_height, int dst_pitch,
+                            int src_width, int src_height,
+                            int parity, int tff, bool skip_spatial_check)
+{
+    yadif_single(dst, prev, cur, next,
+                 dst_width, dst_height, dst_pitch,
+                 src_width, src_height,
+                 parity, tff, skip_spatial_check);
+}
+
+__global__ void yadif_uchar2(uchar2 *dst,
+                            cudaTextureObject_t prev,
+                            cudaTextureObject_t cur,
+                            cudaTextureObject_t next,
+                            int dst_width, int dst_height, int dst_pitch,
+                            int src_width, int src_height,
+                            int parity, int tff, bool skip_spatial_check)
+{
+    yadif_double(dst, prev, cur, next,
+                 dst_width, dst_height, dst_pitch,
+                 src_width, src_height,
+                 parity, tff, skip_spatial_check);
+}
+
+__global__ void yadif_ushort2(ushort2 *dst,
+                            cudaTextureObject_t prev,
+                            cudaTextureObject_t cur,
+                            cudaTextureObject_t next,
+                            int dst_width, int dst_height, int dst_pitch,
+                            int src_width, int src_height,
+                            int parity, int tff, bool skip_spatial_check)
+{
+    yadif_double(dst, prev, cur, next,
+                 dst_width, dst_height, dst_pitch,
+                 src_width, src_height,
+                 parity, tff, skip_spatial_check);
+}
+
+} /* extern "C" */