mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-11-24 10:41:04 +00:00
4cc73ff9d6
Add util functions for runtime CUDA kernel source compilation
using NVRTC library. Like other nvcodec dependent libraries,
NVRTC library will be loaded via g_module_open.
Note that the NVRTC library naming is not g_module_open friendly
on Windows.
(i.e., nvrtc64_{CUDA major version}{CUDA minor version}.dll).
So users can specify the dll name using GST_NVCODEC_NVRTC_LIBNAME
environment.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1633>
105 lines
2.6 KiB
C
105 lines
2.6 KiB
C
/* GStreamer
|
|
* Copyright (C) 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 "gstcudanvrtc.h"
|
|
|
|
GST_DEBUG_CATEGORY_STATIC (gst_cuda_nvrtc_debug);
|
|
#define GST_CAT_DEFAULT gst_cuda_nvrtc_debug
|
|
|
|
static void
|
|
_init_debug (void)
|
|
{
|
|
static volatile gsize once_init = 0;
|
|
|
|
if (g_once_init_enter (&once_init)) {
|
|
|
|
GST_DEBUG_CATEGORY_INIT (gst_cuda_nvrtc_debug, "cudanvrtc", 0,
|
|
"CUDA runtime compiler");
|
|
g_once_init_leave (&once_init, 1);
|
|
}
|
|
}
|
|
|
|
gchar *
|
|
gst_cuda_nvrtc_compile (const gchar * source)
|
|
{
|
|
nvrtcProgram prog;
|
|
nvrtcResult ret;
|
|
const gchar *opts[] = { "--gpu-architecture=compute_30" };
|
|
gsize ptx_size;
|
|
gchar *ptx = NULL;
|
|
|
|
g_return_val_if_fail (source != NULL, FALSE);
|
|
|
|
_init_debug ();
|
|
|
|
GST_TRACE ("CUDA kernel source \n%s", source);
|
|
|
|
ret = NvrtcCreateProgram (&prog, source, NULL, 0, NULL, NULL);
|
|
if (ret != NVRTC_SUCCESS) {
|
|
GST_ERROR ("couldn't create nvrtc program, ret %d", ret);
|
|
return NULL;
|
|
}
|
|
|
|
ret = NvrtcCompileProgram (prog, 1, opts);
|
|
if (ret != NVRTC_SUCCESS) {
|
|
gsize log_size;
|
|
|
|
GST_ERROR ("couldn't compile nvrtc program, ret %d", ret);
|
|
if (NvrtcGetProgramLogSize (prog, &log_size) == NVRTC_SUCCESS &&
|
|
log_size > 0) {
|
|
gchar *compile_log = g_alloca (log_size);
|
|
if (NvrtcGetProgramLog (prog, compile_log) == NVRTC_SUCCESS) {
|
|
GST_ERROR ("nvrtc compile log %s", compile_log);
|
|
}
|
|
}
|
|
|
|
goto error;
|
|
}
|
|
|
|
ret = NvrtcGetPTXSize (prog, &ptx_size);
|
|
if (ret != NVRTC_SUCCESS) {
|
|
GST_ERROR ("unknown ptx size, ret %d", ret);
|
|
|
|
goto error;
|
|
}
|
|
|
|
ptx = g_malloc0 (ptx_size);
|
|
ret = NvrtcGetPTX (prog, ptx);
|
|
if (ret != NVRTC_SUCCESS) {
|
|
GST_ERROR ("couldn't get ptx, ret %d", ret);
|
|
g_free (ptx);
|
|
|
|
goto error;
|
|
}
|
|
|
|
NvrtcDestroyProgram (&prog);
|
|
|
|
GST_TRACE ("compiled CUDA PTX %s\n", ptx);
|
|
|
|
return ptx;
|
|
|
|
error:
|
|
NvrtcDestroyProgram (&prog);
|
|
|
|
return NULL;
|
|
}
|