A block device exposing NVIDIA GPUs' RAM implemented with CUDA
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
cudaramd
kmod
.gitignore
Makefile.am
README
bootstrap
configure.in

README

cudaram - block device in CUDA RAM

###
### Build
###

$ ./bootstrap
$ ./configure [--with-kernel=path_to_kernel] [--with-cuda=path_to_cuda_toolkit] [--enable-debug]
$ make

###
### Use
###
- Load the module
# insmod kmod/cudaram.ko
- /dev/cudaram* /dev/cudaramctl* should be created
- Start the daemon, the params are cudaram_id and capacity_in_MB
# ./cudaramd/cudaramd 0 400
- Use the block device, e.g. create an ext2 fs on it
# mkfs.ext2 /dev/cudaram0
- And mount it
# mount /dev/cudaram0 /mnt/foo

###
### Stop using
###
- Unmount
# umount /dev/cudaram0
- Kill the daemon
# ^C or kill
- Unload the module
# rmmod cudaram