lavfi: Add OpenCL overlay filter

Input and output formats must be the same, the overlay format must be
the same as the input except possibly with an additional alpha component.

8 files changed, 472 insertions, 1 deletions

diff --git a/Changelog b/Changelog
index 74ed35cfe6..f6b70c4a34 100644
--- a/Changelog
+++ b/Changelog
@@ -17,6 +17,7 @@ version <next>:
 - Intel QSV-accelerated overlay filter
 - mcompand audio filter
 - acontrast audio filter
+- OpenCL overlay filter
 
 
 version 3.4:
diff --git a/configure b/configure
index e03d68aa92..e8c605bb97 100755
--- a/configure
+++ b/configure
@@ -3239,6 +3239,7 @@ negate_filter_deps="lut_filter"
 nnedi_filter_deps="gpl"
 ocr_filter_deps="libtesseract"
 ocv_filter_deps="libopencv"
+overlay_opencl_filter_deps="opencl"
 overlay_qsv_filter_deps="libmfx"
 overlay_qsv_filter_select="qsvvpp"
 owdenoise_filter_deps="gpl"
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index 545ef62e1a..77cc3acfcd 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -253,6 +253,8 @@ OBJS-$(CONFIG_OCV_FILTER)                    += vf_libopencv.o
 OBJS-$(CONFIG_OPENCL)                        += deshake_opencl.o unsharp_opencl.o
 OBJS-$(CONFIG_OSCILLOSCOPE_FILTER)           += vf_datascope.o
 OBJS-$(CONFIG_OVERLAY_FILTER)                += vf_overlay.o framesync.o
+OBJS-$(CONFIG_OVERLAY_OPENCL_FILTER)         += vf_overlay_opencl.o opencl.o \
+                                                opencl/overlay.o framesync.o
 OBJS-$(CONFIG_OVERLAY_QSV_FILTER)            += vf_overlay_qsv.o
 OBJS-$(CONFIG_OWDENOISE_FILTER)              += vf_owdenoise.o
 OBJS-$(CONFIG_PAD_FILTER)                    += vf_pad.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index 6d92b3ab5a..0db440d128 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -262,6 +262,7 @@ static void register_all(void)
     REGISTER_FILTER(OCV,            ocv,            vf);
     REGISTER_FILTER(OSCILLOSCOPE,   oscilloscope,   vf);
     REGISTER_FILTER(OVERLAY,        overlay,        vf);
+    REGISTER_FILTER(OVERLAY_OPENCL, overlay_opencl, vf);
     REGISTER_FILTER(OVERLAY_QSV,    overlay_qsv,    vf);
     REGISTER_FILTER(OWDENOISE,      owdenoise,      vf);
     REGISTER_FILTER(PAD,            pad,            vf);
diff --git a/libavfilter/opencl/overlay.cl b/libavfilter/opencl/overlay.cl
new file mode 100644
index 0000000000..8c783d0edc
--- /dev/null
+++ b/libavfilter/opencl/overlay.cl
@@ -0,0 +1,104 @@
+/*
+ * 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
+ */
+
+__kernel void overlay_no_alpha(__write_only image2d_t dst,
+                               __read_only  image2d_t main,
+                               __read_only  image2d_t overlay,
+                               int x_position,
+                               int y_position)
+{
+    const sampler_t sampler = (CLK_NORMALIZED_COORDS_FALSE |
+                               CLK_FILTER_NEAREST);
+
+    int2 overlay_size = get_image_dim(overlay);
+    int2 loc = (int2)(get_global_id(0), get_global_id(1));
+
+    if (loc.x <  x_position ||
+        loc.y <  y_position ||
+        loc.x >= overlay_size.x + x_position ||
+        loc.y >= overlay_size.y + y_position) {
+        float4 val = read_imagef(main, sampler, loc);
+        write_imagef(dst, loc, val);
+    } else {
+        int2 loc_overlay = (int2)(x_position, y_position);
+        float4 val       = read_imagef(overlay, sampler, loc - loc_overlay);
+        write_imagef(dst, loc, val);
+    }
+}
+
+__kernel void overlay_internal_alpha(__write_only image2d_t dst,
+                                     __read_only  image2d_t main,
+                                     __read_only  image2d_t overlay,
+                                     int x_position,
+                                     int y_position)
+{
+    const sampler_t sampler = (CLK_NORMALIZED_COORDS_FALSE |
+                               CLK_FILTER_NEAREST);
+
+    int2 overlay_size = get_image_dim(overlay);
+    int2 loc = (int2)(get_global_id(0), get_global_id(1));
+
+    if (loc.x <  x_position ||
+        loc.y <  y_position ||
+        loc.x >= overlay_size.x + x_position ||
+        loc.y >= overlay_size.y + y_position) {
+        float4 val = read_imagef(main, sampler, loc);
+        write_imagef(dst, loc, val);
+    } else {
+        int2 loc_overlay  = (int2)(x_position, y_position);
+        float4 in_main    = read_imagef(main,    sampler, loc);
+        float4 in_overlay = read_imagef(overlay, sampler, loc - loc_overlay);
+        float4 val        = in_overlay * in_overlay.w + in_main * (1.0f - in_overlay.w);
+        write_imagef(dst, loc, val);
+    }
+}
+
+__kernel void overlay_external_alpha(__write_only image2d_t dst,
+                                     __read_only  image2d_t main,
+                                     __read_only  image2d_t overlay,
+                                     __read_only  image2d_t alpha,
+                                     int x_position,
+                                     int y_position,
+                                     int alpha_adj_x,
+                                     int alpha_adj_y)
+{
+    const sampler_t sampler = (CLK_NORMALIZED_COORDS_FALSE |
+                               CLK_FILTER_NEAREST);
+
+    int2 overlay_size = get_image_dim(overlay);
+    int2 loc = (int2)(get_global_id(0), get_global_id(1));
+
+    if (loc.x <  x_position ||
+        loc.y <  y_position ||
+        loc.x >= overlay_size.x + x_position ||
+        loc.y >= overlay_size.y + y_position) {
+        float4 val = read_imagef(main, sampler, loc);
+        write_imagef(dst, loc, val);
+    } else {
+        int2 loc_overlay  = (int2)(x_position, y_position);
+        float4 in_main    = read_imagef(main,    sampler, loc);
+        float4 in_overlay = read_imagef(overlay, sampler, loc - loc_overlay);
+
+        int2 loc_alpha    = (int2)(loc.x * alpha_adj_x,
+                                   loc.y * alpha_adj_y) - loc_overlay;
+        float4 in_alpha   = read_imagef(alpha,   sampler, loc_alpha);
+
+        float4 val = in_overlay * in_alpha.x + in_main * (1.0f - in_alpha.x);
+        write_imagef(dst, loc, val);
+    }
+}
diff --git a/libavfilter/opencl_source.h b/libavfilter/opencl_source.h
index 8674a03a94..e7af58bcfa 100644
--- a/libavfilter/opencl_source.h
+++ b/libavfilter/opencl_source.h
@@ -19,4 +19,6 @@
 #ifndef AVFILTER_OPENCL_SOURCE_H
 #define AVFILTER_OPENCL_SOURCE_H
 
+extern const char *ff_opencl_source_overlay;
+
 #endif /* AVFILTER_OPENCL_SOURCE_H */
diff --git a/libavfilter/version.h b/libavfilter/version.h
index d8484e4263..25087957eb 100644
--- a/libavfilter/version.h
+++ b/libavfilter/version.h
@@ -31,7 +31,7 @@
 
 #define LIBAVFILTER_VERSION_MAJOR   7
 #define LIBAVFILTER_VERSION_MINOR   2
-#define LIBAVFILTER_VERSION_MICRO 100
+#define LIBAVFILTER_VERSION_MICRO 101
 
 #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \
                                                LIBAVFILTER_VERSION_MINOR, \
diff --git a/libavfilter/vf_overlay_opencl.c b/libavfilter/vf_overlay_opencl.c
new file mode 100644
index 0000000000..46ce42df84
--- /dev/null
+++ b/libavfilter/vf_overlay_opencl.c
@@ -0,0 +1,360 @@
+/*
+ * 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 "libavutil/avassert.h"
+#include "libavutil/buffer.h"
+#include "libavutil/common.h"
+#include "libavutil/hwcontext.h"
+#include "libavutil/hwcontext_opencl.h"
+#include "libavutil/log.h"
+#include "libavutil/mathematics.h"
+#include "libavutil/mem.h"
+#include "libavutil/pixdesc.h"
+#include "libavutil/opt.h"
+
+#include "avfilter.h"
+#include "framesync.h"
+#include "internal.h"
+#include "opencl.h"
+#include "opencl_source.h"
+#include "video.h"
+
+typedef struct OverlayOpenCLContext {
+    OpenCLFilterContext ocf;
+
+    int              initialised;
+    cl_kernel        kernel;
+    cl_command_queue command_queue;
+
+    FFFrameSync      fs;
+
+    int              nb_planes;
+    int              x_subsample;
+    int              y_subsample;
+    int              alpha_separate;
+
+    int              x_position;
+    int              y_position;
+} OverlayOpenCLContext;
+
+static int overlay_opencl_load(AVFilterContext *avctx,
+                               enum AVPixelFormat main_format,
+                               enum AVPixelFormat overlay_format)
+{
+    OverlayOpenCLContext *ctx = avctx->priv;
+    cl_int cle;
+    const char *source = ff_opencl_source_overlay;
+    const char *kernel;
+    const AVPixFmtDescriptor *main_desc, *overlay_desc;
+    int err, i, main_planes, overlay_planes;
+
+    main_desc    = av_pix_fmt_desc_get(main_format);
+    overlay_desc = av_pix_fmt_desc_get(overlay_format);
+
+    main_planes = overlay_planes = 0;
+    for (i = 0; i < main_desc->nb_components; i++)
+        main_planes = FFMAX(main_planes,
+                            main_desc->comp[i].plane + 1);
+    for (i = 0; i < overlay_desc->nb_components; i++)
+        overlay_planes = FFMAX(overlay_planes,
+                               overlay_desc->comp[i].plane + 1);
+
+    ctx->nb_planes = main_planes;
+    ctx->x_subsample = 1 << main_desc->log2_chroma_w;
+    ctx->y_subsample = 1 << main_desc->log2_chroma_h;
+
+    if (ctx->x_position % ctx->x_subsample ||
+        ctx->y_position % ctx->y_subsample) {
+        av_log(avctx, AV_LOG_WARNING, "Warning: overlay position (%d, %d) "
+               "does not match subsampling (%d, %d).\n",
+               ctx->x_position, ctx->y_position,
+               ctx->x_subsample, ctx->y_subsample);
+    }
+
+    if (main_planes == overlay_planes) {
+        if (main_desc->nb_components == overlay_desc->nb_components)
+            kernel = "overlay_no_alpha";
+        else
+            kernel = "overlay_internal_alpha";
+        ctx->alpha_separate = 0;
+    } else {
+        kernel = "overlay_external_alpha";
+        ctx->alpha_separate = 1;
+    }
+
+    av_log(avctx, AV_LOG_DEBUG, "Using kernel %s.\n", kernel);
+
+    err = ff_opencl_filter_load_program(avctx, &source, 1);
+    if (err < 0)
+        goto fail;
+
+    ctx->command_queue = clCreateCommandQueue(ctx->ocf.hwctx->context,
+                                              ctx->ocf.hwctx->device_id,
+                                              0, &cle);
+    if (!ctx->command_queue) {
+        av_log(avctx, AV_LOG_ERROR, "Failed to create OpenCL "
+               "command queue: %d.\n", cle);
+        err = AVERROR(EIO);
+        goto fail;
+    }
+
+    ctx->kernel = clCreateKernel(ctx->ocf.program, kernel, &cle);
+    if (!ctx->kernel) {
+        av_log(avctx, AV_LOG_ERROR, "Failed to create kernel: %d.\n", cle);
+        err = AVERROR(EIO);
+        goto fail;
+    }
+
+    ctx->initialised = 1;
+    return 0;
+
+fail:
+    if (ctx->command_queue)
+        clReleaseCommandQueue(ctx->command_queue);
+    if (ctx->kernel)
+        clReleaseKernel(ctx->kernel);
+    return err;
+}
+
+static int overlay_opencl_blend(FFFrameSync *fs)
+{
+    AVFilterContext    *avctx = fs->parent;
+    AVFilterLink     *outlink = avctx->outputs[0];
+    OverlayOpenCLContext *ctx = avctx->priv;
+    AVFrame *input_main, *input_overlay;
+    AVFrame *output;
+    cl_mem mem;
+    cl_int cle, x, y;
+    size_t global_work[2];
+    int kernel_arg = 0;
+    int err, plane;
+
+    err = ff_framesync_get_frame(fs, 0, &input_main, 0);
+    if (err < 0)
+        return err;
+    err = ff_framesync_get_frame(fs, 1, &input_overlay, 0);
+    if (err < 0)
+        return err;
+
+    if (!ctx->initialised) {
+        AVHWFramesContext *main_fc =
+            (AVHWFramesContext*)input_main->hw_frames_ctx->data;
+        AVHWFramesContext *overlay_fc =
+            (AVHWFramesContext*)input_overlay->hw_frames_ctx->data;
+
+        err = overlay_opencl_load(avctx, main_fc->sw_format,
+                                  overlay_fc->sw_format);
+        if (err < 0)
+            return err;
+    }
+
+    output = ff_get_video_buffer(outlink, outlink->w, outlink->h);
+    if (!output) {
+        err = AVERROR(ENOMEM);
+        goto fail;
+    }
+
+    for (plane = 0; plane < ctx->nb_planes; plane++) {
+        kernel_arg = 0;
+
+        mem = (cl_mem)output->data[plane];
+        cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_mem), &mem);
+        if (cle != CL_SUCCESS)
+            goto fail_kernel_arg;
+
+        mem = (cl_mem)input_main->data[plane];
+        cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_mem), &mem);
+        if (cle != CL_SUCCESS)
+            goto fail_kernel_arg;
+
+        mem = (cl_mem)input_overlay->data[plane];
+        cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_mem), &mem);
+        if (cle != CL_SUCCESS)
+            goto fail_kernel_arg;
+
+        if (ctx->alpha_separate) {
+            mem = (cl_mem)input_overlay->data[ctx->nb_planes];
+            cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_mem), &mem);
+            if (cle != CL_SUCCESS)
+                goto fail_kernel_arg;
+        }
+
+        x = ctx->x_position / (plane == 0 ? 1 : ctx->x_subsample);
+        y = ctx->y_position / (plane == 0 ? 1 : ctx->y_subsample);
+
+        cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_int), &x);
+        if (cle != CL_SUCCESS)
+            goto fail_kernel_arg;
+        cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_int), &y);
+        if (cle != CL_SUCCESS)
+            goto fail_kernel_arg;
+
+        if (ctx->alpha_separate) {
+            cl_int alpha_adj_x = plane == 0 ? 1 : ctx->x_subsample;
+            cl_int alpha_adj_y = plane == 0 ? 1 : ctx->y_subsample;
+
+            cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_int), &alpha_adj_x);
+            if (cle != CL_SUCCESS)
+                goto fail_kernel_arg;
+            cle = clSetKernelArg(ctx->kernel, kernel_arg++, sizeof(cl_int), &alpha_adj_y);
+            if (cle != CL_SUCCESS)
+                goto fail_kernel_arg;
+        }
+
+        global_work[0] = output->width;
+        global_work[1] = output->height;
+
+        cle = clEnqueueNDRangeKernel(ctx->command_queue, ctx->kernel, 2, NULL,
+                                     global_work, NULL, 0, NULL, NULL);
+        if (cle != CL_SUCCESS) {
+            av_log(avctx, AV_LOG_ERROR, "Failed to enqueue "
+                   "overlay kernel for plane %d: %d.\n", cle, plane);
+            err = AVERROR(EIO);
+            goto fail;
+        }
+    }
+
+    cle = clFinish(ctx->command_queue);
+    if (cle != CL_SUCCESS) {
+        av_log(avctx, AV_LOG_ERROR, "Failed to finish "
+               "command queue: %d.\n", cle);
+        err = AVERROR(EIO);
+        goto fail;
+    }
+
+    err = av_frame_copy_props(output, input_main);
+
+    av_log(avctx, AV_LOG_DEBUG, "Filter output: %s, %ux%u (%"PRId64").\n",
+           av_get_pix_fmt_name(output->format),
+           output->width, output->height, output->pts);
+
+    return ff_filter_frame(outlink, output);
+
+fail_kernel_arg:
+    av_log(avctx, AV_LOG_ERROR, "Failed to set kernel arg %d: %d.\n",
+           kernel_arg, cle);
+    err = AVERROR(EIO);
+fail:
+    return err;
+}
+
+static int overlay_opencl_config_output(AVFilterLink *outlink)
+{
+    AVFilterContext *avctx = outlink->src;
+    OverlayOpenCLContext *ctx = avctx->priv;
+    int err;
+
+    err = ff_opencl_filter_config_output(outlink);
+    if (err < 0)
+        return err;
+
+    err = ff_framesync_init_dualinput(&ctx->fs, avctx);
+    if (err < 0)
+        return err;
+
+    return ff_framesync_configure(&ctx->fs);
+}
+
+static av_cold int overlay_opencl_init(AVFilterContext *avctx)
+{
+    OverlayOpenCLContext *ctx = avctx->priv;
+
+    ctx->fs.on_event = &overlay_opencl_blend;
+
+    return ff_opencl_filter_init(avctx);
+}
+
+static int overlay_opencl_activate(AVFilterContext *avctx)
+{
+    OverlayOpenCLContext *ctx = avctx->priv;
+
+    return ff_framesync_activate(&ctx->fs);
+}
+
+static av_cold void overlay_opencl_uninit(AVFilterContext *avctx)
+{
+    OverlayOpenCLContext *ctx = avctx->priv;
+    cl_int cle;
+
+    if (ctx->kernel) {
+        cle = clReleaseKernel(ctx->kernel);
+        if (cle != CL_SUCCESS)
+            av_log(avctx, AV_LOG_ERROR, "Failed to release "
+                   "kernel: %d.\n", cle);
+    }
+
+    if (ctx->command_queue) {
+        cle = clReleaseCommandQueue(ctx->command_queue);
+        if (cle != CL_SUCCESS)
+            av_log(avctx, AV_LOG_ERROR, "Failed to release "
+                   "command queue: %d.\n", cle);
+    }
+
+    ff_opencl_filter_uninit(avctx);
+
+    ff_framesync_uninit(&ctx->fs);
+}
+
+#define OFFSET(x) offsetof(OverlayOpenCLContext, x)
+#define FLAGS (AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_VIDEO_PARAM)
+static const AVOption overlay_opencl_options[] = {
+    { "x", "Overlay x position",
+      OFFSET(x_position), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, .flags = FLAGS },
+    { "y", "Overlay y position",
+      OFFSET(y_position), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, .flags = FLAGS },
+    { NULL },
+};
+
+AVFILTER_DEFINE_CLASS(overlay_opencl);
+
+static const AVFilterPad overlay_opencl_inputs[] = {
+    {
+        .name         = "main",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .config_props = &ff_opencl_filter_config_input,
+    },
+    {
+        .name         = "overlay",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .config_props = &ff_opencl_filter_config_input,
+    },
+    { NULL }
+};
+
+static const AVFilterPad overlay_opencl_outputs[] = {
+    {
+        .name          = "default",
+        .type          = AVMEDIA_TYPE_VIDEO,
+        .config_props  = &overlay_opencl_config_output,
+    },
+    { NULL }
+};
+
+AVFilter ff_vf_overlay_opencl = {
+    .name            = "overlay_opencl",
+    .description     = NULL_IF_CONFIG_SMALL("Overlay one video on top of another"),
+    .priv_size       = sizeof(OverlayOpenCLContext),
+    .priv_class      = &overlay_opencl_class,
+    .init            = &overlay_opencl_init,
+    .uninit          = &overlay_opencl_uninit,
+    .query_formats   = &ff_opencl_filter_query_formats,
+    .activate        = &overlay_opencl_activate,
+    .inputs          = overlay_opencl_inputs,
+    .outputs         = overlay_opencl_outputs,
+    .flags_internal  = FF_FILTER_FLAG_HWFRAME_AWARE,
+};