summaryrefslogtreecommitdiff
path: root/sys/nvcodec/gstcudacontext.c
blob: 35b03cf39921eff41443bd45d866e51d689e9dcf (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
/* GStreamer
 * Copyright (C) <2018-2019> Seungha Yang <seungha.yang@navercorp.com>
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Library General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This library 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
 * Library General Public License for more details.
 *
 * You should have received a copy of the GNU Library General Public
 * License along with this library; if not, write to the
 * Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
 * Boston, MA 02110-1301, USA.
 */

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#include "gstcudaloader.h"
#include "gstcudacontext.h"
#include "gstcudautils.h"

GST_DEBUG_CATEGORY_STATIC (gst_cuda_context_debug);
#define GST_CAT_DEFAULT gst_cuda_context_debug

enum
{
  PROP_0,
  PROP_DEVICE_ID
};

#define DEFAULT_DEVICE_ID -1

struct _GstCudaContextPrivate
{
  CUcontext context;
  CUdevice device;
  gint device_id;
};

#define gst_cuda_context_parent_class parent_class
G_DEFINE_TYPE_WITH_PRIVATE (GstCudaContext, gst_cuda_context, GST_TYPE_OBJECT);

static void gst_cuda_context_set_property (GObject * object, guint prop_id,
    const GValue * value, GParamSpec * pspec);
static void gst_cuda_context_get_property (GObject * object, guint prop_id,
    GValue * value, GParamSpec * pspec);
static void gst_cuda_context_constructed (GObject * object);
static void gst_cuda_context_finalize (GObject * object);

static void
gst_cuda_context_class_init (GstCudaContextClass * klass)
{
  GObjectClass *gobject_class = G_OBJECT_CLASS (klass);

  gobject_class->set_property = gst_cuda_context_set_property;
  gobject_class->get_property = gst_cuda_context_get_property;
  gobject_class->constructed = gst_cuda_context_constructed;
  gobject_class->finalize = gst_cuda_context_finalize;

  g_object_class_install_property (gobject_class, PROP_DEVICE_ID,
      g_param_spec_int ("cuda-device-id", "Cuda Device ID",
          "Set the GPU device to use for operations (-1 = auto)",
          -1, G_MAXINT, DEFAULT_DEVICE_ID,
          G_PARAM_CONSTRUCT_ONLY | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));

  GST_DEBUG_CATEGORY_INIT (gst_cuda_context_debug,
      "cudacontext", 0, "CUDA Context");
}

static void
gst_cuda_context_init (GstCudaContext * context)
{
  GstCudaContextPrivate *priv = gst_cuda_context_get_instance_private (context);

  priv->context = NULL;
  priv->device_id = DEFAULT_DEVICE_ID;

  context->priv = priv;
}

static void
gst_cuda_context_set_property (GObject * object, guint prop_id,
    const GValue * value, GParamSpec * pspec)
{
  GstCudaContext *context = GST_CUDA_CONTEXT (object);
  GstCudaContextPrivate *priv = context->priv;

  switch (prop_id) {
    case PROP_DEVICE_ID:
      priv->device_id = g_value_get_int (value);
      break;
    default:
      G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
      break;
  }
}

static void
gst_cuda_context_get_property (GObject * object, guint prop_id,
    GValue * value, GParamSpec * pspec)
{
  GstCudaContext *context = GST_CUDA_CONTEXT (object);
  GstCudaContextPrivate *priv = context->priv;

  switch (prop_id) {
    case PROP_DEVICE_ID:
      g_value_set_int (value, priv->device_id);
      break;
    default:
      G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
      break;
  }
}

static void
gst_cuda_context_constructed (GObject * object)
{
  static gsize once = 0;
  GstCudaContext *context = GST_CUDA_CONTEXT (object);
  GstCudaContextPrivate *priv = context->priv;
  CUcontext cuda_ctx, old_ctx;
  gboolean ret = TRUE;
  CUdevice cdev = 0, cuda_dev = -1;
  gint dev_count = 0;
  gchar name[256];
  gint min = 0, maj = 0;
  gint i;

  if (g_once_init_enter (&once)) {
    if (CuInit (0) != CUDA_SUCCESS) {
      GST_ERROR_OBJECT (context, "Failed to cuInit");
      ret = FALSE;
    }
    g_once_init_leave (&once, ret);

    if (!ret)
      return;
  }

  if (!gst_cuda_result (CuDeviceGetCount (&dev_count)) || dev_count == 0) {
    GST_WARNING ("No CUDA devices detected");
    return;
  }

  for (i = 0; i < dev_count; ++i) {
    if (gst_cuda_result (CuDeviceGet (&cdev, i)) &&
        gst_cuda_result (CuDeviceGetName (name, sizeof (name), cdev)) &&
        gst_cuda_result (CuDeviceGetAttribute (&maj,
                CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR, cdev)) &&
        gst_cuda_result (CuDeviceGetAttribute (&min,
                CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR, cdev))) {
      GST_INFO ("GPU #%d supports NVENC: %s (%s) (Compute SM %d.%d)", i,
          (((maj << 4) + min) >= 0x30) ? "yes" : "no", name, maj, min);
      if (priv->device_id == -1 || priv->device_id == cdev) {
        priv->device_id = cuda_dev = cdev;
      }
    }
  }

  if (cuda_dev == -1) {
    GST_WARNING ("Device with id %d does not exist", priv->device_id);
    return;
  }

  GST_DEBUG ("Creating cuda context for device index %d", cuda_dev);

  if (!gst_cuda_result (CuCtxCreate (&cuda_ctx, 0, cuda_dev))) {
    GST_WARNING ("Failed to create CUDA context for cuda device %d", cuda_dev);
    return;
  }

  if (!gst_cuda_result (CuCtxPopCurrent (&old_ctx))) {
    return;
  }

  GST_INFO ("Created CUDA context %p with device-id %d", cuda_ctx, cuda_dev);

  priv->context = cuda_ctx;
  priv->device = cuda_dev;
}

static void
gst_cuda_context_finalize (GObject * object)
{
  GstCudaContext *context = GST_CUDA_CONTEXT_CAST (object);
  GstCudaContextPrivate *priv = context->priv;

  if (priv->context) {
    GST_DEBUG_OBJECT (context, "Destroying CUDA context %p", priv->context);
    gst_cuda_result (CuCtxDestroy (priv->context));
  }

  G_OBJECT_CLASS (parent_class)->finalize (object);
}

/**
 * gst_cuda_context_new:
 * @device_id: device-id for creating #GstCudaContext or -1 for auto selection
 *
 * Create #GstCudaContext with given device_id. If the @device_id was not -1
 * but was out of range (e.g., exceed the number of device),
 * #GstCudaContext will not be created.
 *
 * Returns: a new #GstCudaContext or %NULL on failure
 */
GstCudaContext *
gst_cuda_context_new (gint device_id)
{
  GstCudaContext *self =
      g_object_new (GST_TYPE_CUDA_CONTEXT, "cuda-device-id", device_id, NULL);

  gst_object_ref_sink (self);

  if (!self->priv->context) {
    GST_ERROR ("Failed to create CUDA context");
    gst_clear_object (&self);
  }

  return self;
}

/**
 * gst_cuda_context_push:
 * @ctx: a #GstCudaContext to push current thread
 *
 * Pushes the given @ctx onto the CPU thread's stack of current contexts.
 * The specified context becomes the CPU thread's current context,
 * so all CUDA functions that operate on the current context are affected.
 *
 * Returns: %TRUE if @ctx was pushed without error.
 */
gboolean
gst_cuda_context_push (GstCudaContext * ctx)
{
  g_return_val_if_fail (ctx, FALSE);
  g_return_val_if_fail (GST_IS_CUDA_CONTEXT (ctx), FALSE);

  return gst_cuda_result (CuCtxPushCurrent (ctx->priv->context));
}

/**
 * gst_cuda_context_pop:
 *
 * Pops the current CUDA context from CPU thread
 *
 * Returns: %TRUE if @ctx was pushed without error.
 */
gboolean
gst_cuda_context_pop (CUcontext * cuda_ctx)
{
  return gst_cuda_result (CuCtxPopCurrent (cuda_ctx));
}

/**
 * gst_cuda_context_get_handle:
 * @ctx: a #GstCudaContext
 *
 * Get CUDA device context. Caller must not modify and/or destroy
 * returned device context.
 *
 * Returns: the #CUcontext of @ctx
 */
gpointer
gst_cuda_context_get_handle (GstCudaContext * ctx)
{
  g_return_val_if_fail (ctx, NULL);
  g_return_val_if_fail (GST_IS_CUDA_CONTEXT (ctx), NULL);

  return ctx->priv->context;
}