mirror of
https://github.com/arnaucube/poulpy.git
synced 2026-02-10 21:26:41 +01:00
updated repo for publishing (#74)
This commit is contained in:
committed by
GitHub
parent
0be569eca0
commit
62eb87cc07
131
poulpy-core/src/layouts/prepared/gglwe_tsk.rs
Normal file
131
poulpy-core/src/layouts/prepared/gglwe_tsk.rs
Normal file
@@ -0,0 +1,131 @@
|
||||
use poulpy_backend::hal::{
|
||||
api::{VmpPMatAlloc, VmpPMatAllocBytes, VmpPrepare},
|
||||
layouts::{Backend, Data, DataMut, DataRef, Module, Scratch, VmpPMat},
|
||||
};
|
||||
|
||||
use crate::layouts::{
|
||||
GGLWETensorKey, Infos,
|
||||
prepared::{GGLWESwitchingKeyPrepared, Prepare, PrepareAlloc},
|
||||
};
|
||||
|
||||
#[derive(PartialEq, Eq)]
|
||||
pub struct GGLWETensorKeyPrepared<D: Data, B: Backend> {
|
||||
pub(crate) keys: Vec<GGLWESwitchingKeyPrepared<D, B>>,
|
||||
}
|
||||
|
||||
impl<B: Backend> GGLWETensorKeyPrepared<Vec<u8>, B> {
|
||||
pub fn alloc(module: &Module<B>, n: usize, basek: usize, k: usize, rows: usize, digits: usize, rank: usize) -> Self
|
||||
where
|
||||
Module<B>: VmpPMatAlloc<B>,
|
||||
{
|
||||
let mut keys: Vec<GGLWESwitchingKeyPrepared<Vec<u8>, B>> = Vec::new();
|
||||
let pairs: usize = (((rank + 1) * rank) >> 1).max(1);
|
||||
(0..pairs).for_each(|_| {
|
||||
keys.push(GGLWESwitchingKeyPrepared::alloc(
|
||||
module, n, basek, k, rows, digits, 1, rank,
|
||||
));
|
||||
});
|
||||
Self { keys }
|
||||
}
|
||||
|
||||
pub fn bytes_of(module: &Module<B>, n: usize, basek: usize, k: usize, rows: usize, digits: usize, rank: usize) -> usize
|
||||
where
|
||||
Module<B>: VmpPMatAllocBytes,
|
||||
{
|
||||
let pairs: usize = (((rank + 1) * rank) >> 1).max(1);
|
||||
pairs * GGLWESwitchingKeyPrepared::bytes_of(module, n, basek, k, rows, digits, 1, rank)
|
||||
}
|
||||
}
|
||||
|
||||
impl<D: Data, B: Backend> Infos for GGLWETensorKeyPrepared<D, B> {
|
||||
type Inner = VmpPMat<D, B>;
|
||||
|
||||
fn inner(&self) -> &Self::Inner {
|
||||
self.keys[0].inner()
|
||||
}
|
||||
|
||||
fn basek(&self) -> usize {
|
||||
self.keys[0].basek()
|
||||
}
|
||||
|
||||
fn k(&self) -> usize {
|
||||
self.keys[0].k()
|
||||
}
|
||||
}
|
||||
|
||||
impl<D: Data, B: Backend> GGLWETensorKeyPrepared<D, B> {
|
||||
pub fn rank(&self) -> usize {
|
||||
self.keys[0].rank()
|
||||
}
|
||||
|
||||
pub fn rank_in(&self) -> usize {
|
||||
self.keys[0].rank_in()
|
||||
}
|
||||
|
||||
pub fn rank_out(&self) -> usize {
|
||||
self.keys[0].rank_out()
|
||||
}
|
||||
|
||||
pub fn digits(&self) -> usize {
|
||||
self.keys[0].digits()
|
||||
}
|
||||
}
|
||||
|
||||
impl<D: DataMut, B: Backend> GGLWETensorKeyPrepared<D, B> {
|
||||
// Returns a mutable reference to GLWESwitchingKey_{s}(s[i] * s[j])
|
||||
pub fn at_mut(&mut self, mut i: usize, mut j: usize) -> &mut GGLWESwitchingKeyPrepared<D, B> {
|
||||
if i > j {
|
||||
std::mem::swap(&mut i, &mut j);
|
||||
};
|
||||
let rank: usize = self.rank();
|
||||
&mut self.keys[i * rank + j - (i * (i + 1) / 2)]
|
||||
}
|
||||
}
|
||||
|
||||
impl<D: DataRef, B: Backend> GGLWETensorKeyPrepared<D, B> {
|
||||
// Returns a reference to GLWESwitchingKey_{s}(s[i] * s[j])
|
||||
pub fn at(&self, mut i: usize, mut j: usize) -> &GGLWESwitchingKeyPrepared<D, B> {
|
||||
if i > j {
|
||||
std::mem::swap(&mut i, &mut j);
|
||||
};
|
||||
let rank: usize = self.rank();
|
||||
&self.keys[i * rank + j - (i * (i + 1) / 2)]
|
||||
}
|
||||
}
|
||||
|
||||
impl<D: DataMut, DR: DataRef, B: Backend> Prepare<B, GGLWETensorKey<DR>> for GGLWETensorKeyPrepared<D, B>
|
||||
where
|
||||
Module<B>: VmpPrepare<B>,
|
||||
{
|
||||
fn prepare(&mut self, module: &Module<B>, other: &GGLWETensorKey<DR>, scratch: &mut Scratch<B>) {
|
||||
#[cfg(debug_assertions)]
|
||||
{
|
||||
assert_eq!(self.keys.len(), other.keys.len());
|
||||
}
|
||||
self.keys
|
||||
.iter_mut()
|
||||
.zip(other.keys.iter())
|
||||
.for_each(|(a, b)| {
|
||||
a.prepare(module, b, scratch);
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
impl<D: DataRef, B: Backend> PrepareAlloc<B, GGLWETensorKeyPrepared<Vec<u8>, B>> for GGLWETensorKey<D>
|
||||
where
|
||||
Module<B>: VmpPMatAlloc<B> + VmpPrepare<B>,
|
||||
{
|
||||
fn prepare_alloc(&self, module: &Module<B>, scratch: &mut Scratch<B>) -> GGLWETensorKeyPrepared<Vec<u8>, B> {
|
||||
let mut tsk_prepared: GGLWETensorKeyPrepared<Vec<u8>, B> = GGLWETensorKeyPrepared::alloc(
|
||||
module,
|
||||
self.n(),
|
||||
self.basek(),
|
||||
self.k(),
|
||||
self.rows(),
|
||||
self.digits(),
|
||||
self.rank(),
|
||||
);
|
||||
tsk_prepared.prepare(module, self, scratch);
|
||||
tsk_prepared
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user