-
Notifications
You must be signed in to change notification settings - Fork 0
jacksongod/OramGPU
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
THIS is ORAM GPU project The project maps Oblivious RAM algorithm to GPGPU programming. oramstart is the initial verision: 1. Oram Block size: 64 Bytes. Number of Blocks: 2^12 = 4K. Total oram size: 4K*64Byte = 256KB 2. ORAM tree: 12 level complete tree. Leaf level has 2^11 = 2K nodes. Each nodes has 2 blocks. Total capacity of tree: (4K-1)*2*64 Bytes = ~512KB. 3. On-chip stash with capacity of 256 oram blocks. 256*64B = 16KB 4. On-chip meta stash/meta tree (data structure which contains oram block ID). size of meta stash = 256* 2B = 512 Byte. size of meta tree = (4K-1) *2 * 2B = 16 KB (could be moved off chip) 5. without encryption/decryption so far. 6. Only 1 SM is untilized. Todo: Add encryption/decryption. Profile/optimizing design. Goal: get 10x speed up. Profile with full SMs running. Optimize if necessary. Moving stash off-chip which Allows larger oram size. USE make to build binary ,needs nvcc to compile. if in tacc stampede machine, do "module load cuda", "make " the program, run "sbatch jobscrp" to send executable to the compute node use "watch showq -u" to track the status of the job. [email protected]
About
No description, website, or topics provided.
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published