Merge remote-tracking branch 'cigaes/master'
authorMichael Niedermayer <michaelni@gmx.at>
Mon, 1 Apr 2013 13:13:02 +0000 (15:13 +0200)
committerMichael Niedermayer <michaelni@gmx.at>
Mon, 1 Apr 2013 13:13:07 +0000 (15:13 +0200)
* cigaes/master:
  lavfi/af_asetnsamples: fix EOF handling.
  lavu/opt: make sure av_opt_set_bin() handles NULL/0.
  lavd/v4l2: fully init an ioctl argument.
  lavfi: detect merge failure for unknown layouts.

Merged-by: Michael Niedermayer <michaelni@gmx.at>
Changelog
configure
libavformat/oma.c
libavformat/omadec.c
libavutil/Makefile
libavutil/opencl.c [new file with mode: 0644]
libavutil/opencl.h [new file with mode: 0644]

index 72a51ea..9f57061 100644 (file)
--- a/Changelog
+++ b/Changelog
@@ -14,6 +14,7 @@ version <next>:
 - WebP demuxing and decoding support
 - new ffmpeg options -filter_script and -filter_complex_script, which allow a
   filtergraph description to be read from a file
+- OpenCL support
 
 
 version 1.2:
index 8dc2500..f5196e1 100755 (executable)
--- a/configure
+++ b/configure
@@ -234,6 +234,7 @@ External library support:
   --enable-libxvid         enable Xvid encoding via xvidcore,
                            native MPEG-4/Xvid encoder exists [no]
   --enable-openal          enable OpenAL 1.1 capture support [no]
+  --enable-opencl          enable OpenCL code
   --enable-openssl         enable openssl [no]
   --enable-x11grab         enable X11 grabbing [no]
   --enable-zlib            enable zlib [autodetect]
@@ -1179,6 +1180,7 @@ EXTERNAL_LIBRARY_LIST="
     libxavs
     libxvid
     openal
+    opencl
     openssl
     x11grab
     zlib
@@ -3987,6 +3989,7 @@ enabled openal     && { { for al_libs in "${OPENAL_LIBS}" "-lopenal" "-lOpenAL32
                         die "ERROR: openal not found"; } &&
                       { check_cpp_condition "AL/al.h" "defined(AL_VERSION_1_1)" ||
                         die "ERROR: openal must be installed and version must be 1.1 or compatible"; }
+enabled opencl     && require2 opencl CL/cl.h clEnqueueNDRangeKernel -lOpenCL
 enabled openssl    && { check_lib openssl/ssl.h SSL_library_init -lssl -lcrypto ||
                         check_lib openssl/ssl.h SSL_library_init -lssl32 -leay32 ||
                         check_lib openssl/ssl.h SSL_library_init -lssl -lcrypto -lws2_32 -lgdi32 ||
@@ -4355,6 +4358,7 @@ echo "network support           ${network-no}"
 echo "threading support         ${thread_type-no}"
 echo "safe bitstream reader     ${safe_bitstream_reader-no}"
 echo "SDL support               ${sdl-no}"
+echo "opencl enabled            ${opencl-no}"
 echo "texi2html enabled         ${texi2html-no}"
 echo "perl enabled              ${perl-no}"
 echo "pod2man enabled           ${pod2man-no}"
index 9e4bfbd..fc926bf 100644 (file)
@@ -22,7 +22,7 @@
 #include "oma.h"
 #include "libavcodec/avcodec.h"
 
-const uint16_t ff_oma_srate_tab[8] = { 320, 441, 480, 882, 960, 0, 0, 0};
+const uint16_t ff_oma_srate_tab[8] = { 320, 441, 480, 882, 960, 0 };
 
 const AVCodecTag ff_oma_codec_tags[] = {
     { AV_CODEC_ID_ATRAC3,      OMA_CODECID_ATRAC3  },
index 9f65db0..ae35907 100644 (file)
@@ -309,7 +309,11 @@ static int oma_read_header(AVFormatContext *s)
 
     switch (buf[32]) {
         case OMA_CODECID_ATRAC3:
-            samplerate = ff_oma_srate_tab[(codec_params >> 13) & 7]*100;
+            samplerate = ff_oma_srate_tab[(codec_params >> 13) & 7] * 100;
+            if (!samplerate) {
+                av_log(s, AV_LOG_ERROR, "Unsupported sample rate\n");
+                return AVERROR_INVALIDDATA;
+            }
             if (samplerate != 44100)
                 avpriv_request_sample(s, "Sample rate %d", samplerate);
 
@@ -339,9 +343,14 @@ static int oma_read_header(AVFormatContext *s)
         case OMA_CODECID_ATRAC3P:
             st->codec->channels = (codec_params >> 10) & 7;
             framesize = ((codec_params & 0x3FF) * 8) + 8;
-            st->codec->sample_rate = ff_oma_srate_tab[(codec_params >> 13) & 7]*100;
-            st->codec->bit_rate    = st->codec->sample_rate * framesize * 8 / 1024;
-            avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
+            samplerate = ff_oma_srate_tab[(codec_params >> 13) & 7] * 100;
+            if (!samplerate) {
+                av_log(s, AV_LOG_ERROR, "Unsupported sample rate\n");
+                return AVERROR_INVALIDDATA;
+            }
+            st->codec->sample_rate = samplerate;
+            st->codec->bit_rate    = samplerate * framesize * 8 / 1024;
+            avpriv_set_pts_info(st, 64, 1, samplerate);
             av_log(s, AV_LOG_ERROR, "Unsupported codec ATRAC3+!\n");
             break;
         case OMA_CODECID_MP3:
index 103ce5e..b520473 100644 (file)
@@ -52,6 +52,8 @@ HEADERS = adler32.h                                                     \
 
 HEADERS-$(CONFIG_LZO)                   += lzo.h
 
+HEADERS-$(CONFIG_OPENCL)                += opencl.h
+
 ARCH_HEADERS = bswap.h                                                  \
                intmath.h                                                \
                intreadwrite.h                                           \
@@ -106,6 +108,7 @@ OBJS = adler32.o                                                        \
        xtea.o                                                           \
 
 OBJS-$(CONFIG_LZO)                      += lzo.o
+OBJS-$(CONFIG_OPENCL)                   += opencl.o
 
 OBJS += $(COMPAT_OBJS:%=../compat/%)
 
diff --git a/libavutil/opencl.c b/libavutil/opencl.c
new file mode 100644 (file)
index 0000000..4ad39a8
--- /dev/null
@@ -0,0 +1,718 @@
+/*
+ * Copyright (C) 2012 Peng Gao <peng@multicorewareinc.com>
+ * Copyright (C) 2012 Li   Cao <li@multicorewareinc.com>
+ * Copyright (C) 2012 Wei  Gao <weigao@multicorewareinc.com>
+ *
+ * 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 "opencl.h"
+#include "avstring.h"
+#include "log.h"
+#include "avassert.h"
+
+#if HAVE_PTHREADS
+
+#include <pthread.h>
+static pthread_mutex_t atomic_opencl_lock = PTHREAD_MUTEX_INITIALIZER;
+
+#define LOCK_OPENCL pthread_mutex_lock(&atomic_opencl_lock);
+#define UNLOCK_OPENCL pthread_mutex_unlock(&atomic_opencl_lock);
+
+#elif !HAVE_THREADS
+#define LOCK_OPENCL
+#define UNLOCK_OPENCL
+#endif
+
+
+#define MAX_KERNEL_NUM 500
+#define MAX_KERNEL_CODE_NUM 200
+
+typedef struct {
+    int dev_idx;
+    int platform_idx;
+} UserSpecDevInfo;
+
+typedef struct {
+    int is_compiled;
+    const char *kernel_string;
+} KernelCode;
+
+typedef struct {
+    int init_count;
+    UserSpecDevInfo usr_spec_dev_info;
+    cl_platform_id platform_id;
+    cl_device_type device_type;
+    cl_context context;
+    cl_device_id *device_ids;
+    cl_device_id device_id;
+    cl_command_queue command_queue;
+    int program_count;
+    cl_program programs[MAX_KERNEL_CODE_NUM];
+    int kernel_code_count;
+    KernelCode kernel_code[MAX_KERNEL_CODE_NUM];
+    int kernel_count;
+    /**
+         * if set to 1, the OpenCL environment was created by the user and
+         * passed as AVOpenCLExternalEnv when initing ,0:created by opencl wrapper.
+         */
+    int is_user_created;
+} GPUEnv;
+
+typedef struct {
+    const AVClass *class;
+    int log_offset;
+    void *log_ctx;
+} OpenclUtils;
+
+static const AVClass openclutils_class = {"OPENCLUTILS", av_default_item_name,
+                                                   NULL, LIBAVUTIL_VERSION_INT,
+                                                   offsetof(OpenclUtils, log_offset),
+                                                   offsetof(OpenclUtils, log_ctx)};
+static OpenclUtils openclutils = {&openclutils_class};
+static GPUEnv gpu_env;
+
+typedef struct {
+    int err_code;
+    const char *err_str;
+} OpenclErrorMsg;
+
+static const OpenclErrorMsg opencl_err_msg[] = {
+        {CL_DEVICE_NOT_FOUND,                               "DEVICE NOT FOUND"},
+        {CL_DEVICE_NOT_AVAILABLE,                           "DEVICE NOT AVAILABLE"},
+        {CL_COMPILER_NOT_AVAILABLE,                         "COMPILER NOT AVAILABLE"},
+        {CL_MEM_OBJECT_ALLOCATION_FAILURE,                  "MEM OBJECT ALLOCATION FAILURE"},
+        {CL_OUT_OF_RESOURCES,                               "OUT OF RESOURCES"},
+        {CL_OUT_OF_HOST_MEMORY,                             "OUT OF HOST MEMORY"},
+        {CL_PROFILING_INFO_NOT_AVAILABLE,                   "PROFILING INFO NOT AVAILABLE"},
+        {CL_MEM_COPY_OVERLAP,                               "MEM COPY OVERLAP"},
+        {CL_IMAGE_FORMAT_MISMATCH,                          "IMAGE FORMAT MISMATCH"},
+        {CL_IMAGE_FORMAT_NOT_SUPPORTED,                     "IMAGE FORMAT NOT_SUPPORTED"},
+        {CL_BUILD_PROGRAM_FAILURE,                          "BUILD PROGRAM FAILURE"},
+        {CL_MAP_FAILURE,                                    "MAP FAILURE"},
+        {CL_MISALIGNED_SUB_BUFFER_OFFSET,                   "MISALIGNED SUB BUFFER OFFSET"},
+        {CL_EXEC_STATUS_ERROR_FOR_EVENTS_IN_WAIT_LIST,      "EXEC STATUS ERROR FOR EVENTS IN WAIT LIST"},
+        {CL_COMPILE_PROGRAM_FAILURE,                        "COMPILE PROGRAM FAILURE"},
+        {CL_LINKER_NOT_AVAILABLE,                           "LINKER NOT AVAILABLE"},
+        {CL_LINK_PROGRAM_FAILURE,                           "LINK PROGRAM FAILURE"},
+        {CL_DEVICE_PARTITION_FAILED,                        "DEVICE PARTITION FAILED"},
+        {CL_KERNEL_ARG_INFO_NOT_AVAILABLE,                  "KERNEL ARG INFO NOT AVAILABLE"},
+        {CL_INVALID_VALUE,                                  "INVALID VALUE"},
+        {CL_INVALID_DEVICE_TYPE,                            "INVALID DEVICE TYPE"},
+        {CL_INVALID_PLATFORM,                               "INVALID PLATFORM"},
+        {CL_INVALID_DEVICE,                                 "INVALID DEVICE"},
+        {CL_INVALID_CONTEXT,                                "INVALID CONTEXT"},
+        {CL_INVALID_QUEUE_PROPERTIES,                       "INVALID QUEUE PROPERTIES"},
+        {CL_INVALID_COMMAND_QUEUE,                          "INVALID COMMAND QUEUE"},
+        {CL_INVALID_HOST_PTR,                               "INVALID HOST PTR"},
+        {CL_INVALID_MEM_OBJECT,                             "INVALID MEM OBJECT"},
+        {CL_INVALID_IMAGE_FORMAT_DESCRIPTOR,                "INVALID IMAGE FORMAT DESCRIPTOR"},
+        {CL_INVALID_IMAGE_SIZE,                             "INVALID IMAGE SIZE"},
+        {CL_INVALID_SAMPLER,                                "INVALID SAMPLER"},
+        {CL_INVALID_BINARY,                                 "INVALID BINARY"},
+        {CL_INVALID_BUILD_OPTIONS,                          "INVALID BUILD OPTIONS"},
+        {CL_INVALID_PROGRAM,                                "INVALID PROGRAM"},
+        {CL_INVALID_PROGRAM_EXECUTABLE,                     "INVALID PROGRAM EXECUTABLE"},
+        {CL_INVALID_KERNEL_NAME,                            "INVALID KERNEL NAME"},
+        {CL_INVALID_KERNEL_DEFINITION,                      "INVALID KERNEL DEFINITION"},
+        {CL_INVALID_KERNEL,                                 "INVALID KERNEL"},
+        {CL_INVALID_ARG_INDEX,                              "INVALID ARG INDEX"},
+        {CL_INVALID_ARG_VALUE,                              "INVALID ARG VALUE"},
+        {CL_INVALID_ARG_SIZE,                               "INVALID ARG_SIZE"},
+        {CL_INVALID_KERNEL_ARGS,                            "INVALID KERNEL ARGS"},
+        {CL_INVALID_WORK_DIMENSION,                         "INVALID WORK DIMENSION"},
+        {CL_INVALID_WORK_GROUP_SIZE,                        "INVALID WORK GROUP SIZE"},
+        {CL_INVALID_WORK_ITEM_SIZE,                         "INVALID WORK ITEM SIZE"},
+        {CL_INVALID_GLOBAL_OFFSET,                          "INVALID GLOBAL OFFSET"},
+        {CL_INVALID_EVENT_WAIT_LIST,                        "INVALID EVENT WAIT LIST"},
+        {CL_INVALID_EVENT,                                  "INVALID EVENT"},
+        {CL_INVALID_OPERATION,                              "INVALID OPERATION"},
+        {CL_INVALID_GL_OBJECT,                              "INVALID GL OBJECT"},
+        {CL_INVALID_BUFFER_SIZE,                            "INVALID BUFFER SIZE"},
+        {CL_INVALID_MIP_LEVEL,                              "INVALID MIP LEVEL"},
+        {CL_INVALID_GLOBAL_WORK_SIZE,                       "INVALID GLOBAL WORK SIZE"},
+        {CL_INVALID_PROPERTY,                               "INVALID PROPERTY"},
+        {CL_INVALID_IMAGE_DESCRIPTOR,                       "INVALID IMAGE DESCRIPTOR"},
+        {CL_INVALID_COMPILER_OPTIONS,                       "INVALID COMPILER OPTIONS"},
+        {CL_INVALID_LINKER_OPTIONS,                         "INVALID LINKER OPTIONS"},
+        {CL_INVALID_DEVICE_PARTITION_COUNT,                 "INVALID DEVICE PARTITION COUNT"},
+};
+
+static const char *opencl_errstr(cl_int status)
+{
+    int i;
+    for (i = 0; i < sizeof(opencl_err_msg); i++) {
+        if (opencl_err_msg[i].err_code == status)
+            return opencl_err_msg[i].err_str;
+    }
+    return "unknown error";
+}
+
+AVOpenCLExternalEnv *av_opencl_alloc_external_env(void)
+{
+    AVOpenCLExternalEnv *ext = av_mallocz(sizeof(AVOpenCLExternalEnv));
+    if (!ext) {
+        av_log(&openclutils, AV_LOG_ERROR,
+         "Could not malloc external opencl environment data space\n");
+    }
+    return ext;
+}
+
+void av_opencl_free_external_env(AVOpenCLExternalEnv **ext_opencl_env)
+{
+    av_freep(ext_opencl_env);
+}
+
+int av_opencl_register_kernel_code(const char *kernel_code)
+{
+    int i, ret = 0;
+    LOCK_OPENCL;
+    if (gpu_env.kernel_code_count >= MAX_KERNEL_CODE_NUM) {
+        av_log(&openclutils, AV_LOG_ERROR,
+         "Could not register kernel code, maximum number of registered kernel code %d already reached\n",
+         MAX_KERNEL_CODE_NUM);
+        ret = AVERROR(EINVAL);
+        goto end;
+    }
+    for (i = 0; i < gpu_env.kernel_code_count; i++) {
+        if (gpu_env.kernel_code[i].kernel_string == kernel_code) {
+            av_log(&openclutils, AV_LOG_WARNING, "Same kernel code has been registered\n");
+            goto end;
+        }
+    }
+    gpu_env.kernel_code[gpu_env.kernel_code_count].kernel_string = kernel_code;
+    gpu_env.kernel_code[gpu_env.kernel_code_count].is_compiled = 0;
+    gpu_env.kernel_code_count++;
+end:
+    UNLOCK_OPENCL;
+    return ret;
+}
+
+int av_opencl_create_kernel(AVOpenCLKernelEnv *env, const char *kernel_name)
+{
+    cl_int status;
+    int i, ret = 0;
+    LOCK_OPENCL;
+    if (strlen(kernel_name) + 1 > AV_OPENCL_MAX_KERNEL_NAME_SIZE) {
+        av_log(&openclutils, AV_LOG_ERROR, "Created kernel name %s is too long\n", kernel_name);
+        ret = AVERROR(EINVAL);
+        goto end;
+    }
+    if (!env->kernel) {
+        if (gpu_env.kernel_count >= MAX_KERNEL_NUM) {
+            av_log(&openclutils, AV_LOG_ERROR,
+            "Could not create kernel with name '%s', maximum number of kernels %d already reached\n",
+                kernel_name, MAX_KERNEL_NUM);
+            ret = AVERROR(EINVAL);
+            goto end;
+        }
+        for (i = 0; i < gpu_env.program_count; i++) {
+            env->kernel = clCreateKernel(gpu_env.programs[i], kernel_name, &status);
+            if (status == CL_SUCCESS)
+                break;
+        }
+        if (status != CL_SUCCESS) {
+            av_log(&openclutils, AV_LOG_ERROR, "Could not create OpenCL kernel: %s\n", opencl_errstr(status));
+            ret = AVERROR_EXTERNAL;
+            goto end;
+        }
+        gpu_env.kernel_count++;
+        env->command_queue = gpu_env.command_queue;
+        av_strlcpy(env->kernel_name, kernel_name, sizeof(env->kernel_name));
+    }
+end:
+    UNLOCK_OPENCL;
+    return ret;
+}
+
+void av_opencl_release_kernel(AVOpenCLKernelEnv *env)
+{
+    cl_int status;
+    LOCK_OPENCL
+    if (!env->kernel)
+        goto end;
+    status = clReleaseKernel(env->kernel);
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not release kernel: %s\n",
+              opencl_errstr(status));
+    }
+    env->kernel = NULL;
+    env->command_queue = NULL;
+    env->kernel_name[0] = 0;
+    gpu_env.kernel_count--;
+end:
+    UNLOCK_OPENCL
+}
+
+static int init_opencl_env(GPUEnv *gpu_env, AVOpenCLExternalEnv *ext_opencl_env)
+{
+    size_t device_length;
+    cl_int status;
+    cl_uint num_platforms, num_devices;
+    cl_platform_id *platform_ids = NULL;
+    cl_context_properties cps[3];
+    char platform_name[100];
+    int i, ret = 0;
+    cl_device_type device_type[] = {CL_DEVICE_TYPE_GPU, CL_DEVICE_TYPE_CPU, CL_DEVICE_TYPE_DEFAULT};
+    if (ext_opencl_env) {
+        if (gpu_env->is_user_created)
+            return 0;
+        gpu_env->platform_id     = ext_opencl_env->platform_id;
+        gpu_env->is_user_created = 1;
+        gpu_env->command_queue   = ext_opencl_env->command_queue;
+        gpu_env->context         = ext_opencl_env->context;
+        gpu_env->device_ids      = ext_opencl_env->device_ids;
+        gpu_env->device_id       = ext_opencl_env->device_id;
+        gpu_env->device_type     = ext_opencl_env->device_type;
+    } else {
+        if (!gpu_env->is_user_created) {
+            status = clGetPlatformIDs(0, NULL, &num_platforms);
+            if (status != CL_SUCCESS) {
+                av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL platform ids: %s\n", opencl_errstr(status));
+                return AVERROR_EXTERNAL;
+            }
+            if (gpu_env->usr_spec_dev_info.platform_idx >= 0) {
+                if (num_platforms < gpu_env->usr_spec_dev_info.platform_idx + 1) {
+                    av_log(&openclutils, AV_LOG_ERROR, "User set platform index not exist\n");
+                    return AVERROR(EINVAL);
+                }
+            }
+            if (num_platforms > 0) {
+                platform_ids = av_mallocz(num_platforms * sizeof(cl_platform_id));
+                if (!platform_ids) {
+                    ret = AVERROR(ENOMEM);
+                    goto end;
+                }
+                status = clGetPlatformIDs(num_platforms, platform_ids, NULL);
+                if (status != CL_SUCCESS) {
+                    av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL platform ids: %s\n", opencl_errstr(status));
+                    ret = AVERROR_EXTERNAL;
+                    goto end;
+                }
+                i = 0;
+                if (gpu_env->usr_spec_dev_info.platform_idx >= 0) {
+                    i = gpu_env->usr_spec_dev_info.platform_idx;
+                }
+                while (i < num_platforms) {
+                    status = clGetPlatformInfo(platform_ids[i], CL_PLATFORM_VENDOR,
+                                               sizeof(platform_name), platform_name,
+                                               NULL);
+
+                    if (status != CL_SUCCESS) {
+                        av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL platform info: %s\n", opencl_errstr(status));
+                        ret = AVERROR_EXTERNAL;
+                        goto end;
+                    }
+                    gpu_env->platform_id = platform_ids[i];
+                    status = clGetDeviceIDs(gpu_env->platform_id, CL_DEVICE_TYPE_GPU,
+                                            0, NULL, &num_devices);
+                    if (status != CL_SUCCESS) {
+                        av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL device number:%s\n", opencl_errstr(status));
+                        ret = AVERROR_EXTERNAL;
+                        goto end;
+                    }
+                    if (num_devices == 0) {
+                        //find CPU device
+                        status = clGetDeviceIDs(gpu_env->platform_id, CL_DEVICE_TYPE_CPU,
+                                             0, NULL, &num_devices);
+                    }
+                    if (status != CL_SUCCESS) {
+                        av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL device ids: %s\n", opencl_errstr(status));
+                        ret = AVERROR(EINVAL);
+                        goto end;
+                    }
+                    if (num_devices)
+                       break;
+                    if (gpu_env->usr_spec_dev_info.platform_idx >= 0) {
+                        av_log(&openclutils, AV_LOG_ERROR, "Device number of user set platform is 0\n");
+                        ret = AVERROR_EXTERNAL;
+                        goto end;
+                    }
+                    i++;
+
+                }
+            }
+            if (!gpu_env->platform_id) {
+                av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL platforms\n");
+                ret = AVERROR_EXTERNAL;
+                goto end;
+            }
+            if (gpu_env->usr_spec_dev_info.dev_idx >= 0) {
+                if (num_devices < gpu_env->usr_spec_dev_info.dev_idx + 1) {
+                    av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL device idx in the user set platform\n");
+                    ret = AVERROR(EINVAL);
+                    goto end;
+                }
+            }
+
+           /*
+                    * Use available platform.
+                    */
+
+            av_log(&openclutils, AV_LOG_VERBOSE, "Platform Name: %s\n", platform_name);
+            cps[0] = CL_CONTEXT_PLATFORM;
+            cps[1] = (cl_context_properties)gpu_env->platform_id;
+            cps[2] = 0;
+            /* Check for GPU. */
+
+            for (i = 0; i < sizeof(device_type); i++) {
+                gpu_env->device_type = device_type[i];
+                gpu_env->context     = clCreateContextFromType(cps, gpu_env->device_type,
+                                                               NULL, NULL, &status);
+                if (status != CL_SUCCESS) {
+                    av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL context from device type: %s\n", opencl_errstr(status));
+                    ret = AVERROR_EXTERNAL;
+                    goto end;
+                }
+                if (gpu_env->context)
+                    break;
+            }
+            if (!gpu_env->context) {
+                av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL context from device type\n");
+                ret = AVERROR_EXTERNAL;
+                goto end;
+            }
+            /* Detect OpenCL devices. */
+            /* First, get the size of device list data */
+            status = clGetContextInfo(gpu_env->context, CL_CONTEXT_DEVICES,
+                                      0, NULL, &device_length);
+            if (status != CL_SUCCESS) {
+                av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL device length: %s\n", opencl_errstr(status));
+                ret = AVERROR_EXTERNAL;
+                goto end;
+            }
+            if (device_length == 0) {
+                av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL device length\n");
+                ret = AVERROR_EXTERNAL;
+                goto end;
+            }
+            /* Now allocate memory for device list based on the size we got earlier */
+            gpu_env->device_ids = av_mallocz(device_length);
+            if (!gpu_env->device_ids) {
+                ret = AVERROR(ENOMEM);
+                goto end;
+            }
+            /* Now, get the device list data */
+            status = clGetContextInfo(gpu_env->context, CL_CONTEXT_DEVICES, device_length,
+                                      gpu_env->device_ids, NULL);
+            if (status != CL_SUCCESS) {
+                av_log(&openclutils, AV_LOG_ERROR, "Could not get OpenCL context info: %s\n", opencl_errstr(status));
+                ret = AVERROR_EXTERNAL;
+                goto end;
+            }
+            /* Create OpenCL command queue. */
+            i = 0;
+            if (gpu_env->usr_spec_dev_info.dev_idx >= 0) {
+                i = gpu_env->usr_spec_dev_info.dev_idx;
+            }
+            gpu_env->command_queue = clCreateCommandQueue(gpu_env->context, gpu_env->device_ids[i],
+                                                          0, &status);
+            if (status != CL_SUCCESS) {
+                av_log(&openclutils, AV_LOG_ERROR, "Could not create OpenCL command queue: %s\n", opencl_errstr(status));
+                ret = AVERROR_EXTERNAL;
+                goto end;
+            }
+        }
+    }
+end:
+    av_free(platform_ids);
+    return ret;
+}
+
+static int compile_kernel_file(GPUEnv *gpu_env, const char *build_options)
+{
+    cl_int status;
+    char *temp, *source_str = NULL;
+    size_t source_str_len = 0;
+    int i, ret = 0;
+
+    for (i = 0; i < gpu_env->kernel_code_count; i++) {
+        if (!gpu_env->kernel_code[i].is_compiled)
+            source_str_len += strlen(gpu_env->kernel_code[i].kernel_string);
+    }
+    if (!source_str_len) {
+        return 0;
+    }
+    source_str = av_mallocz(source_str_len + 1);
+    if (!source_str) {
+        return AVERROR(ENOMEM);
+    }
+    temp = source_str;
+    for (i = 0; i < gpu_env->kernel_code_count; i++) {
+        if (!gpu_env->kernel_code[i].is_compiled) {
+            memcpy(temp, gpu_env->kernel_code[i].kernel_string,
+                        strlen(gpu_env->kernel_code[i].kernel_string));
+            gpu_env->kernel_code[i].is_compiled = 1;
+            temp += strlen(gpu_env->kernel_code[i].kernel_string);
+        }
+    }
+    /* create a CL program using the kernel source */
+    gpu_env->programs[gpu_env->program_count] = clCreateProgramWithSource(gpu_env->context,
+                                                           1, (const char **)(&source_str),
+                                                                   &source_str_len, &status);
+    if(status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not create OpenCL program with source code: %s\n",
+               opencl_errstr(status));
+        ret = AVERROR_EXTERNAL;
+        goto end;
+    }
+    if (!gpu_env->programs[gpu_env->program_count]) {
+        av_log(&openclutils, AV_LOG_ERROR, "Created program is NULL\n");
+        ret = AVERROR_EXTERNAL;
+        goto end;
+    }
+    i = 0;
+    if (gpu_env->usr_spec_dev_info.dev_idx >= 0)
+        i = gpu_env->usr_spec_dev_info.dev_idx;
+    /* create a cl program executable for all the devices specified */
+    if (!gpu_env->is_user_created)
+        status = clBuildProgram(gpu_env->programs[gpu_env->program_count], 1, &gpu_env->device_ids[i],
+                                build_options, NULL, NULL);
+    else
+        status = clBuildProgram(gpu_env->programs[gpu_env->program_count], 1, &(gpu_env->device_id),
+                                 build_options, NULL, NULL);
+
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not compile OpenCL kernel: %s\n", opencl_errstr(status));
+        ret = AVERROR_EXTERNAL;
+        goto end;
+    }
+    gpu_env->program_count++;
+end:
+    av_free(source_str);
+    return ret;
+}
+
+int av_opencl_init(AVDictionary *options, AVOpenCLExternalEnv *ext_opencl_env)
+{
+    int ret = 0;
+    AVDictionaryEntry *opt_build_entry;
+    AVDictionaryEntry *opt_platform_entry;
+    AVDictionaryEntry *opt_device_entry;
+    LOCK_OPENCL
+    if (!gpu_env.init_count) {
+        opt_platform_entry = av_dict_get(options, "platform_idx", NULL, 0);
+        opt_device_entry   = av_dict_get(options, "device_idx", NULL, 0);
+        /*initialize devices, context, command_queue*/
+        gpu_env.usr_spec_dev_info.platform_idx = -1;
+        gpu_env.usr_spec_dev_info.dev_idx = -1;
+        if (opt_platform_entry) {
+            gpu_env.usr_spec_dev_info.platform_idx = strtol(opt_platform_entry->value, NULL, 10);
+        }
+        if (opt_device_entry) {
+            gpu_env.usr_spec_dev_info.dev_idx = strtol(opt_device_entry->value, NULL, 10);
+        }
+        ret = init_opencl_env(&gpu_env, ext_opencl_env);
+        if (ret < 0)
+            goto end;
+    }
+    /*initialize program, kernel_name, kernel_count*/
+    opt_build_entry = av_dict_get(options, "build_options", NULL, 0);
+    if (opt_build_entry)
+        ret = compile_kernel_file(&gpu_env, opt_build_entry->value);
+    else
+        ret = compile_kernel_file(&gpu_env, NULL);
+    if (ret < 0)
+        goto end;
+    av_assert1(gpu_env.kernel_code_count > 0);
+    gpu_env.init_count++;
+
+end:
+    UNLOCK_OPENCL
+    return ret;
+}
+
+void av_opencl_uninit(void)
+{
+    cl_int status;
+    int i;
+    LOCK_OPENCL
+    gpu_env.init_count--;
+    if (gpu_env.is_user_created)
+        goto end;
+    if ((gpu_env.init_count > 0) || (gpu_env.kernel_count > 0))
+        goto end;
+    for (i = 0; i < gpu_env.program_count; i++) {
+        if (gpu_env.programs[i]) {
+            status = clReleaseProgram(gpu_env.programs[i]);
+            if (status != CL_SUCCESS) {
+                av_log(&openclutils, AV_LOG_ERROR, "Could not release OpenCL program: %s\n", opencl_errstr(status));
+            }
+            gpu_env.programs[i] = NULL;
+        }
+    }
+    if (gpu_env.command_queue) {
+        status = clReleaseCommandQueue(gpu_env.command_queue);
+        if (status != CL_SUCCESS) {
+            av_log(&openclutils, AV_LOG_ERROR, "Could not release OpenCL command queue: %s\n", opencl_errstr(status));
+        }
+        gpu_env.command_queue = NULL;
+    }
+    if (gpu_env.context) {
+        status = clReleaseContext(gpu_env.context);
+        if (status != CL_SUCCESS) {
+            av_log(&openclutils, AV_LOG_ERROR, "Could not release OpenCL context: %s\n", opencl_errstr(status));
+        }
+        gpu_env.context = NULL;
+    }
+    av_freep(&(gpu_env.device_ids));
+end:
+    UNLOCK_OPENCL
+}
+
+int av_opencl_buffer_create(cl_mem *cl_buf, size_t cl_buf_size, int flags, void *host_ptr)
+{
+    cl_int status;
+    *cl_buf = clCreateBuffer(gpu_env.context, flags, cl_buf_size, host_ptr, &status);
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not create OpenCL buffer: %s\n", opencl_errstr(status));
+        return AVERROR_EXTERNAL;
+    }
+    return 0;
+}
+
+void av_opencl_buffer_release(cl_mem *cl_buf)
+{
+    cl_int status = 0;
+    if (!cl_buf)
+        return;
+    status = clReleaseMemObject(*cl_buf);
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not release OpenCL buffer: %s\n", opencl_errstr(status));
+    }
+    memset(cl_buf, 0, sizeof(*cl_buf));
+}
+
+int av_opencl_buffer_write(cl_mem dst_cl_buf, uint8_t *src_buf, size_t buf_size)
+{
+    cl_int status;
+    void *mapped = clEnqueueMapBuffer(gpu_env.command_queue, dst_cl_buf,
+                                      CL_TRUE,CL_MAP_WRITE, 0, sizeof(uint8_t) * buf_size,
+                                      0, NULL, NULL, &status);
+
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not map OpenCL buffer: %s\n", opencl_errstr(status));
+        return AVERROR_EXTERNAL;
+    }
+    memcpy(mapped, src_buf, buf_size);
+
+    status = clEnqueueUnmapMemObject(gpu_env.command_queue, dst_cl_buf, mapped, 0, NULL, NULL);
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not unmap OpenCL buffer: %s\n", opencl_errstr(status));
+        return AVERROR_EXTERNAL;
+    }
+    return 0;
+}
+
+int av_opencl_buffer_read(uint8_t *dst_buf, cl_mem src_cl_buf, size_t buf_size)
+{
+    cl_int status;
+    void *mapped = clEnqueueMapBuffer(gpu_env.command_queue, src_cl_buf,
+                                      CL_TRUE,CL_MAP_READ, 0, buf_size,
+                                      0, NULL, NULL, &status);
+
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not map OpenCL buffer: %s\n", opencl_errstr(status));
+        return AVERROR_EXTERNAL;
+    }
+    memcpy(dst_buf, mapped, buf_size);
+
+    status = clEnqueueUnmapMemObject(gpu_env.command_queue, src_cl_buf, mapped, 0, NULL, NULL);
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not unmap OpenCL buffer: %s\n", opencl_errstr(status));
+        return AVERROR_EXTERNAL;
+    }
+    return 0;
+}
+
+int av_opencl_buffer_write_image(cl_mem dst_cl_buf, size_t cl_buffer_size, int dst_cl_offset,
+                                                    uint8_t **src_data, int *plane_size, int plane_num)
+{
+    int i, buffer_size = 0;
+    uint8_t *temp;
+    cl_int status;
+    void *mapped;
+    if ((unsigned int)plane_num > 8) {
+        return AVERROR(EINVAL);
+    }
+    for (i = 0;i < plane_num;i++) {
+        buffer_size += plane_size[i];
+    }
+    if (buffer_size > cl_buffer_size) {
+        av_log(&openclutils, AV_LOG_ERROR, "Cannot write image to OpenCL buffer: buffer too small\n");
+        return AVERROR(EINVAL);
+    }
+    mapped = clEnqueueMapBuffer(gpu_env.command_queue, dst_cl_buf,
+                                      CL_TRUE,CL_MAP_WRITE, 0, buffer_size + dst_cl_offset,
+                                      0, NULL, NULL, &status);
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not map OpenCL buffer: %s\n", opencl_errstr(status));
+        return AVERROR_EXTERNAL;
+    }
+    temp = mapped;
+    temp += dst_cl_offset;
+    for (i = 0; i < plane_num; i++) {
+        memcpy(temp, src_data[i], plane_size[i]);
+        temp += plane_size[i];
+    }
+    status = clEnqueueUnmapMemObject(gpu_env.command_queue, dst_cl_buf, mapped, 0, NULL, NULL);
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not unmap OpenCL buffer: %s\n", opencl_errstr(status));
+        return AVERROR_EXTERNAL;
+    }
+    return 0;
+}
+
+int av_opencl_buffer_read_image(uint8_t **dst_data, int *plane_size, int plane_num,
+                                       cl_mem src_cl_buf, size_t cl_buffer_size)
+{
+    int i,buffer_size = 0,ret = 0;
+    uint8_t *temp;
+    void *mapped;
+    cl_int status;
+    if ((unsigned int)plane_num > 8) {
+        return AVERROR(EINVAL);
+    }
+    for (i = 0;i < plane_num;i++) {
+        buffer_size += plane_size[i];
+    }
+    if (buffer_size > cl_buffer_size) {
+        av_log(&openclutils, AV_LOG_ERROR, "Cannot write image to CPU buffer: OpenCL buffer too small\n");
+        return AVERROR(EINVAL);
+    }
+    mapped = clEnqueueMapBuffer(gpu_env.command_queue, src_cl_buf,
+                                      CL_TRUE,CL_MAP_READ, 0, buffer_size,
+                                      0, NULL, NULL, &status);
+
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not map OpenCL buffer: %s\n", opencl_errstr(status));
+        return AVERROR_EXTERNAL;
+    }
+    temp = mapped;
+    if (ret >= 0) {
+        for (i = 0;i < plane_num;i++) {
+            memcpy(dst_data[i], temp, plane_size[i]);
+            temp += plane_size[i];
+        }
+    }
+    status = clEnqueueUnmapMemObject(gpu_env.command_queue, src_cl_buf, mapped, 0, NULL, NULL);
+    if (status != CL_SUCCESS) {
+        av_log(&openclutils, AV_LOG_ERROR, "Could not unmap OpenCL buffer: %s\n", opencl_errstr(status));
+        return AVERROR_EXTERNAL;
+    }
+    return 0;
+}
diff --git a/libavutil/opencl.h b/libavutil/opencl.h
new file mode 100644 (file)
index 0000000..81ba1e4
--- /dev/null
@@ -0,0 +1,200 @@
+/*
+ * Copyright (C) 2012 Peng Gao <peng@multicorewareinc.com>
+ * Copyright (C) 2012 Li   Cao <li@multicorewareinc.com>
+ * Copyright (C) 2012 Wei  Gao <weigao@multicorewareinc.com>
+ *
+ * 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
+ */
+
+/**
+ * @file
+ * OpenCL wrapper
+ *
+ * This interface is considered still experimental and its API and ABI may
+ * change without prior notice.
+ */
+
+#ifndef LIBAVUTIL_OPENCLWRAPPER_H
+#define LIBAVUTIL_OPENCLWRAPPER_H
+
+#include <CL/cl.h>
+#include "config.h"
+#include "dict.h"
+
+#define AV_OPENCL_KERNEL( ... )# __VA_ARGS__
+
+#define AV_OPENCL_MAX_KERNEL_NAME_SIZE 150
+
+typedef struct {
+    cl_command_queue command_queue;
+    cl_kernel kernel;
+    char kernel_name[AV_OPENCL_MAX_KERNEL_NAME_SIZE];
+} AVOpenCLKernelEnv;
+
+typedef struct {
+    cl_platform_id platform_id;
+    cl_device_type device_type;
+    cl_context context;
+    cl_device_id *device_ids;
+    cl_device_id  device_id;
+    cl_command_queue command_queue;
+    char *platform_name;
+} AVOpenCLExternalEnv;
+
+/**
+ * Allocate OpenCL external environment.
+ *
+ * It must be freed with av_opencl_free_external_env().
+ *
+ * @return pointer to allocated OpenCL external environment
+ */
+AVOpenCLExternalEnv *av_opencl_alloc_external_env(void);
+
+/**
+ * Free OpenCL external environment.
+ *
+ * @param ext_opencl_env pointer to OpenCL external environment created by av_opencl_alloc_external_env()
+ */
+void av_opencl_free_external_env(AVOpenCLExternalEnv **ext_opencl_env);
+
+/**
+ * Register kernel code.
+ *
+ *  The registered kernel code is stored in a global context, and compiled
+ *  in the runtime environment when av_opencl_init() is called.
+ *
+ * @param kernel_code    kernel code to be compiled in the OpenCL runtime environment
+ * @return  >=0 on success, a negative error code in case of failure
+ */
+int av_opencl_register_kernel_code(const char *kernel_code);
+
+/**
+ * Initialize the run time OpenCL environment and compile the kernel code registered with
+ * av_opencl_register_kernel_code().
+ *
+ * Currently, the only accepted option is "build_options", used to set
+ * options to compile registered kernels code. See reference "OpenCL
+ * Specification Version: 1.2 chapter 5.6.4".
+ *
+ * @param options                  dictionary of key/value options
+ * @param ext_opencl_env      external OpenCL environment, created by an
+ *                                          application program, ignored if set to NULL
+ * @return  >=0 on success, a negative error code in case of failure
+ */
+ int av_opencl_init(AVDictionary *options, AVOpenCLExternalEnv *ext_opencl_env);
+
+/**
+ * Create kernel object in the specified kernel environment.
+ *
+ * @param env                 pointer to kernel environment which is filled with the environment,
+ *                                   used to run the kernel
+ * @param kernel_name      kernel function name
+ * @return >=0 on success, a negative error code in case of failure
+ */
+int av_opencl_create_kernel(AVOpenCLKernelEnv *env, const char *kernel_name);
+
+/**
+ * Create OpenCL buffer.
+ *
+ * The buffer is used to save the data used or created by an OpenCL
+ * kernel.
+ * The created buffer must be released with av_opencl_buffer_release().
+ *
+ * See clCreateBuffer() function reference for more information about
+ * the parameters.
+ *
+ * @param cl_buf       pointer to OpenCL buffer
+ * @param cl_buf_size  size in bytes of the OpenCL buffer to create
+ * @param flags        flags used to control buffer attributes
+ * @param host_ptr     host pointer of the OpenCL buffer
+ * @return >=0 on success, a negative error code in case of failure
+ */
+int av_opencl_buffer_create(cl_mem *cl_buf, size_t cl_buf_size, int flags, void *host_ptr);
+
+/**
+ * Write OpenCL buffer with data from src_buf.
+ *
+ * @param dst_cl_buf        pointer to OpenCL destination buffer
+ * @param src_buf           pointer to source buffer
+ * @param buf_size          size in bytes of the source and destination buffers
+ * @return >=0 on success, a negative error code in case of failure
+ */
+int av_opencl_buffer_write(cl_mem dst_cl_buf, uint8_t *src_buf, size_t buf_size);
+
+/**
+ * Read data from OpenCL buffer to memory buffer.
+ *
+ * @param dst_buf           pointer to destination buffer (CPU memory)
+ * @param src_cl_buf        pointer to source OpenCL buffer
+ * @param buf_size          size in bytes of the source and destination buffers
+ * @return >=0 on success, a negative error code in case of failure
+ */
+int av_opencl_buffer_read(uint8_t *dst_buf, cl_mem src_cl_buf, size_t buf_size);
+
+/**
+ * Write image data from memory to OpenCL buffer.
+ *
+ * The source must be an array of pointers to image plane buffers.
+ *
+ * @param dst_cl_buf         pointer to destination OpenCL buffer
+ * @param dst_cl_buf_size    size in bytes of OpenCL buffer
+ * @param dst_cl_buf_offset  the offset of the OpenCL buffer start position
+ * @param src_data           array of pointers to source plane buffers
+ * @param src_plane_sizes    array of sizes in bytes of the source plane buffers
+ * @param src_plane_num      number of source image planes
+ * @return >=0 on success, a negative error code in case of failure
+ */
+int av_opencl_buffer_write_image(cl_mem dst_cl_buf, size_t cl_buffer_size, int dst_cl_offset,
+                                        uint8_t **src_data, int *plane_size, int plane_num);
+/**
+ * Read image data from OpenCL buffer.
+ *
+ * src buffer is OpenCL buffer, dst buffer is frame buffer(data[0],data[1]....).
+ *
+ * @param dst_data           array of pointers to destination plane buffers
+ * @param dst_plane_sizes    array of pointers to destination plane buffers
+ * @param dst_plane_num      number of destination image planes
+ * @param src_cl_buf         pointer to source OpenCL buffer
+ * @param src_cl_buf_size    size in bytes of OpenCL buffer
+ * @return >=0 on success, a negative error code in case of failure
+ */
+
+int av_opencl_buffer_read_image(uint8_t **dst_data, int *plane_size, int plane_num,
+                                       cl_mem src_cl_buf, size_t cl_buffer_size);
+/**
+ * Release OpenCL buffer.
+ *
+ * @param cl_buf  pointer to OpenCL buffer to release, which was previously filled with av_opencl_buffer_create()
+ */
+void av_opencl_buffer_release(cl_mem *cl_buf);
+
+/**
+ *  Release kernel object.
+ *
+ * @param env  kernel environment where the kernel object was created with av_opencl_create_kernel
+ */
+void av_opencl_release_kernel(AVOpenCLKernelEnv *env);
+
+/**
+ *  Release OpenCL environment.
+ *
+ * The OpenCL environment is effectively released only if all the created
+ * kernels had been released with av_opencl_release_kernel().
+ */
+void av_opencl_uninit(void);
+
+#endif/*LIBAVUTIL_OPENCL_H*/