Check cuda API and kernel calls for successful execution
commit7e78c06c70d66c014577c9d42943028f4ae9ed56
authorTobias Grosser <tobias@grosser.es>
Sun, 7 Aug 2011 22:55:27 +0000 (7 23:55 +0100)
committerSven Verdoolaege <skimo@kotnet.org>
Mon, 8 Aug 2011 07:33:08 +0000 (8 09:33 +0200)
treef9b1814160a3d475305e6f81bcc39d2911d77d34
parentfc2f14974ffec88c80cd12f2ea31a1b0bbf103a5
Check cuda API and kernel calls for successful execution

Cuda API calls give a return value that specifies if they were executed
sucessfully and a cuda kernel sets a global error buffer, if a problem
occurred. We add runtime checks to check these error conditions.  Before
this patch, we could not even notice, if a kernel was not executed at
all.

Signed-off-by: Tobias Grosser <tobias@grosser.es>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
cuda.c