Merge pull request #112 from phantomzone-org/bdd_multi_thread

Bdd multi thread
This commit is contained in:
Jean-Philippe Bossuat
2025-11-12 15:41:55 +01:00
committed by GitHub
29 changed files with 528 additions and 195 deletions

7
Cargo.lock generated
View File

@@ -323,6 +323,12 @@ version = "11.1.4"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b410bbe7e14ab526a0e86877eb47c6996a2bd7746f027ba551028c925390e4e9" checksum = "b410bbe7e14ab526a0e86877eb47c6996a2bd7746f027ba551028c925390e4e9"
[[package]]
name = "paste"
version = "1.0.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "57c0d7b74b563b49d38dae00a0c37d4d6de9b432382b2892f0574ddcae73fd0a"
[[package]] [[package]]
name = "plotters" name = "plotters"
version = "0.3.7" version = "0.3.7"
@@ -406,6 +412,7 @@ dependencies = [
"byteorder", "byteorder",
"criterion", "criterion",
"itertools 0.14.0", "itertools 0.14.0",
"paste",
"poulpy-backend", "poulpy-backend",
"poulpy-core", "poulpy-core",
"poulpy-hal", "poulpy-hal",

View File

@@ -1,5 +1,5 @@
use poulpy_hal::{ use poulpy_hal::{
api::{ModuleN, ScratchAvailable, ScratchTakeBasic, SvpPPolBytesOf, VecZnxDftBytesOf, VmpPMatBytesOf}, api::{ModuleN, ScratchAvailable, ScratchFromBytes, ScratchTakeBasic, SvpPPolBytesOf, VecZnxDftBytesOf, VmpPMatBytesOf},
layouts::{Backend, Scratch}, layouts::{Backend, Scratch},
}; };
@@ -7,7 +7,7 @@ use crate::{
dist::Distribution, dist::Distribution,
layouts::{ layouts::{
Degree, GGLWE, GGLWEInfos, GGLWELayout, GGSW, GGSWInfos, GLWE, GLWEAutomorphismKey, GLWEInfos, GLWEPlaintext, Degree, GGLWE, GGLWEInfos, GGLWELayout, GGSW, GGSWInfos, GLWE, GLWEAutomorphismKey, GLWEInfos, GLWEPlaintext,
GLWEPrepared, GLWEPublicKey, GLWESecret, GLWESecretTensor, GLWESwitchingKey, GLWETensorKey, Rank, GLWEPrepared, GLWEPublicKey, GLWESecret, GLWESecretTensor, GLWESwitchingKey, GLWETensorKey, LWE, LWEInfos, Rank,
prepared::{ prepared::{
GGLWEPrepared, GGSWPrepared, GLWEAutomorphismKeyPrepared, GLWEPublicKeyPrepared, GLWESecretPrepared, GGLWEPrepared, GGSWPrepared, GLWEAutomorphismKeyPrepared, GLWEPublicKeyPrepared, GLWESecretPrepared,
GLWESwitchingKeyPrepared, GLWETensorKeyPrepared, GLWESwitchingKeyPrepared, GLWETensorKeyPrepared,
@@ -17,8 +17,23 @@ use crate::{
pub trait ScratchTakeCore<B: Backend> pub trait ScratchTakeCore<B: Backend>
where where
Self: ScratchTakeBasic + ScratchAvailable, Self: ScratchTakeBasic + ScratchAvailable + ScratchFromBytes<B>,
{ {
fn take_lwe<A>(&mut self, infos: &A) -> (LWE<&mut [u8]>, &mut Self)
where
A: LWEInfos,
{
let (data, scratch) = self.take_zn(infos.n().into(), 1, infos.size());
(
LWE {
k: infos.k(),
base2k: infos.base2k(),
data,
},
scratch,
)
}
fn take_glwe<A>(&mut self, infos: &A) -> (GLWE<&mut [u8]>, &mut Self) fn take_glwe<A>(&mut self, infos: &A) -> (GLWE<&mut [u8]>, &mut Self)
where where
A: GLWEInfos, A: GLWEInfos,
@@ -367,4 +382,4 @@ where
} }
} }
impl<B: Backend> ScratchTakeCore<B> for Scratch<B> where Self: ScratchTakeBasic + ScratchAvailable {} impl<B: Backend> ScratchTakeCore<B> for Scratch<B> where Self: ScratchTakeBasic + ScratchAvailable + ScratchFromBytes<B> {}

View File

@@ -1,6 +1,6 @@
use crate::{ use crate::{
api::{ModuleN, SvpPPolBytesOf, VecZnxBigBytesOf, VecZnxDftBytesOf, VmpPMatBytesOf}, api::{ModuleN, SvpPPolBytesOf, VecZnxBigBytesOf, VecZnxDftBytesOf, VmpPMatBytesOf},
layouts::{Backend, MatZnx, ScalarZnx, Scratch, SvpPPol, VecZnx, VecZnxBig, VecZnxDft, VmpPMat}, layouts::{Backend, MatZnx, ScalarZnx, Scratch, SvpPPol, VecZnx, VecZnxBig, VecZnxDft, VmpPMat, Zn},
}; };
/// Allocates a new [crate::layouts::ScratchOwned] of `size` aligned bytes. /// Allocates a new [crate::layouts::ScratchOwned] of `size` aligned bytes.
@@ -28,7 +28,29 @@ pub trait TakeSlice {
fn take_slice<T>(&mut self, len: usize) -> (&mut [T], &mut Self); fn take_slice<T>(&mut self, len: usize) -> (&mut [T], &mut Self);
} }
impl<B: Backend> ScratchTakeBasic for Scratch<B> where Self: TakeSlice {} impl<BE: Backend> Scratch<BE>
where
Self: TakeSlice + ScratchAvailable + ScratchFromBytes<BE>,
{
pub fn split_at_mut(&mut self, len: usize) -> (&mut Scratch<BE>, &mut Self) {
let (take_slice, rem_slice) = self.take_slice(len);
(Self::from_bytes(take_slice), rem_slice)
}
pub fn split_mut(&mut self, n: usize, len: usize) -> (Vec<&mut Scratch<BE>>, &mut Self) {
assert!(self.available() >= n * len);
let mut scratches: Vec<&mut Scratch<BE>> = Vec::with_capacity(n);
let mut scratch: &mut Scratch<BE> = self;
for _ in 0..n {
let (tmp, scratch_new) = scratch.split_at_mut(len);
scratch = scratch_new;
scratches.push(tmp);
}
(scratches, scratch)
}
}
impl<B: Backend> ScratchTakeBasic for Scratch<B> where Self: TakeSlice + ScratchFromBytes<B> {}
pub trait ScratchTakeBasic pub trait ScratchTakeBasic
where where
@@ -47,6 +69,11 @@ where
(SvpPPol::from_data(take_slice, module.n(), cols), rem_slice) (SvpPPol::from_data(take_slice, module.n(), cols), rem_slice)
} }
fn take_zn(&mut self, n: usize, cols: usize, size: usize) -> (Zn<&mut [u8]>, &mut Self) {
let (take_slice, rem_slice) = self.take_slice(Zn::bytes_of(n, cols, size));
(Zn::from_data(take_slice, n, cols, size), rem_slice)
}
fn take_vec_znx(&mut self, n: usize, cols: usize, size: usize) -> (VecZnx<&mut [u8]>, &mut Self) { fn take_vec_znx(&mut self, n: usize, cols: usize, size: usize) -> (VecZnx<&mut [u8]>, &mut Self) {
let (take_slice, rem_slice) = self.take_slice(VecZnx::bytes_of(n, cols, size)); let (take_slice, rem_slice) = self.take_slice(VecZnx::bytes_of(n, cols, size));
(VecZnx::from_data(take_slice, n, cols, size), rem_slice) (VecZnx::from_data(take_slice, n, cols, size), rem_slice)

View File

@@ -28,8 +28,8 @@ pub use zn::*;
pub use znx_base::*; pub use znx_base::*;
pub trait Data = PartialEq + Eq + Sized + Default; pub trait Data = PartialEq + Eq + Sized + Default;
pub trait DataRef = Data + AsRef<[u8]>; pub trait DataRef = Data + AsRef<[u8]> + Sync;
pub trait DataMut = DataRef + AsMut<[u8]>; pub trait DataMut = DataRef + AsMut<[u8]> + Send;
pub trait ToOwnedDeep { pub trait ToOwnedDeep {
type Owned; type Owned;

View File

@@ -13,7 +13,7 @@ use crate::{
}; };
#[allow(clippy::missing_safety_doc)] #[allow(clippy::missing_safety_doc)]
pub trait Backend: Sized { pub trait Backend: Sized + Sync + Send {
type ScalarBig: Copy + Zero + Display + Debug + Pod; type ScalarBig: Copy + Zero + Display + Debug + Pod;
type ScalarPrep: Copy + Zero + Display + Debug + Pod; type ScalarPrep: Copy + Zero + Display + Debug + Pod;
type Handle: 'static; type Handle: 'static;

View File

@@ -17,7 +17,7 @@ criterion = {workspace = true}
itertools = "0.14.0" itertools = "0.14.0"
byteorder = "1.5.0" byteorder = "1.5.0"
rand = "0.9.2" rand = "0.9.2"
paste = "1.0.15"
[[bench]] [[bench]]
name = "circuit_bootstrapping" name = "circuit_bootstrapping"

View File

@@ -13,17 +13,14 @@ use crate::tfhe::bdd_arithmetic::{
BitSize, ExecuteBDDCircuit, FheUint, FheUintPrepared, GetBitCircuitInfo, GetGGSWBit, UnsignedInteger, circuits, BitSize, ExecuteBDDCircuit, FheUint, FheUintPrepared, GetBitCircuitInfo, GetGGSWBit, UnsignedInteger, circuits,
}; };
impl<T: UnsignedInteger, BE: Backend> ExecuteBDDCircuit2WTo1W<T, BE> for Module<BE> where impl<BE: Backend> ExecuteBDDCircuit2WTo1W<BE> for Module<BE> where Self: Sized + ExecuteBDDCircuit<BE> + GLWEPacking<BE> + GLWECopy
Self: Sized + ExecuteBDDCircuit<T, BE> + GLWEPacking<BE> + GLWECopy {}
{
}
pub trait ExecuteBDDCircuit2WTo1W<T: UnsignedInteger, BE: Backend> pub trait ExecuteBDDCircuit2WTo1W<BE: Backend>
where where
Self: Sized + ModuleLogN + ExecuteBDDCircuit<T, BE> + GLWEPacking<BE> + GLWECopy, Self: Sized + ModuleLogN + ExecuteBDDCircuit<BE> + GLWEPacking<BE> + GLWECopy,
{ {
/// Operations Z x Z -> Z fn execute_bdd_circuit_2w_to_1w<R, C, A, B, K, H, T>(
fn execute_bdd_circuit_2w_to_1w<R, C, A, B, K, H>(
&self, &self,
out: &mut FheUint<R, T>, out: &mut FheUint<R, T>,
circuit: &C, circuit: &C,
@@ -32,7 +29,32 @@ where
key: &H, key: &H,
scratch: &mut Scratch<BE>, scratch: &mut Scratch<BE>,
) where ) where
C: GetBitCircuitInfo<T>, T: UnsignedInteger,
C: GetBitCircuitInfo,
R: DataMut,
A: DataRef,
B: DataRef,
K: GGLWEPreparedToRef<BE> + GetGaloisElement + GGLWEInfos,
H: GLWEAutomorphismKeyHelper<K, BE>,
Scratch<BE>: ScratchTakeCore<BE>,
{
self.execute_bdd_circuit_2w_to_1w_multi_thread(1, out, circuit, a, b, key, scratch);
}
#[allow(clippy::too_many_arguments)]
/// Operations Z x Z -> Z
fn execute_bdd_circuit_2w_to_1w_multi_thread<R, C, A, B, K, H, T>(
&self,
threads: usize,
out: &mut FheUint<R, T>,
circuit: &C,
a: &FheUintPrepared<A, T, BE>,
b: &FheUintPrepared<B, T, BE>,
key: &H,
scratch: &mut Scratch<BE>,
) where
T: UnsignedInteger,
C: GetBitCircuitInfo,
R: DataMut, R: DataMut,
A: DataRef, A: DataRef,
B: DataRef, B: DataRef,
@@ -50,7 +72,7 @@ where
let (mut out_bits, scratch_1) = scratch.take_glwe_slice(T::BITS as usize, out); let (mut out_bits, scratch_1) = scratch.take_glwe_slice(T::BITS as usize, out);
// Evaluates out[i] = circuit[i](a, b) // Evaluates out[i] = circuit[i](a, b)
self.execute_bdd_circuit(&mut out_bits, &helper, circuit, scratch_1); self.execute_bdd_circuit_multi_thread(threads, &mut out_bits, &helper, circuit, scratch_1);
// Repacks the bits // Repacks the bits
out.pack(self, out_bits, key, scratch_1); out.pack(self, out_bits, key, scratch_1);
@@ -100,22 +122,43 @@ where
#[macro_export] #[macro_export]
macro_rules! define_bdd_2w_to_1w_trait { macro_rules! define_bdd_2w_to_1w_trait {
($(#[$meta:meta])* $vis:vis $trait_name:ident, $method_name:ident) => { ($(#[$meta:meta])* $vis:vis $trait_name:ident, $method_name:ident) => {
$(#[$meta])* paste::paste! {
$vis trait $trait_name<T: UnsignedInteger, BE: Backend> { $(#[$meta])*
fn $method_name<A, M, K, H, B>( $vis trait $trait_name<T: UnsignedInteger, BE: Backend> {
&mut self,
module: &M, /// Single-threaded version
a: &FheUintPrepared<A, T, BE>, fn $method_name<A, M, K, H, B>(
b: &FheUintPrepared<B, T, BE>, &mut self,
key: &H, module: &M,
scratch: &mut Scratch<BE>, a: &FheUintPrepared<A, T, BE>,
) where b: &FheUintPrepared<B, T, BE>,
M: ExecuteBDDCircuit2WTo1W<T, BE>, key: &H,
A: DataRef, scratch: &mut Scratch<BE>,
B: DataRef, ) where
K: GGLWEPreparedToRef<BE> + GetGaloisElement + GGLWEInfos, M: ExecuteBDDCircuit2WTo1W<BE>,
H: GLWEAutomorphismKeyHelper<K, BE>, A: DataRef,
Scratch<BE>: ScratchTakeCore<BE>; B: DataRef,
K: GGLWEPreparedToRef<BE> + GetGaloisElement + GGLWEInfos,
H: GLWEAutomorphismKeyHelper<K, BE>,
Scratch<BE>: ScratchTakeCore<BE>;
/// Multithreaded version same vis, method_name + "_multi_thread"
fn [<$method_name _multi_thread>]<A, M, K, H, B>(
&mut self,
threads: usize,
module: &M,
a: &FheUintPrepared<A, T, BE>,
b: &FheUintPrepared<B, T, BE>,
key: &H,
scratch: &mut Scratch<BE>,
) where
M: ExecuteBDDCircuit2WTo1W<BE>,
A: DataRef,
B: DataRef,
K: GGLWEPreparedToRef<BE> + GetGaloisElement + GGLWEInfos,
H: GLWEAutomorphismKeyHelper<K, BE>,
Scratch<BE>: ScratchTakeCore<BE>;
}
} }
}; };
} }
@@ -123,23 +166,45 @@ macro_rules! define_bdd_2w_to_1w_trait {
#[macro_export] #[macro_export]
macro_rules! impl_bdd_2w_to_1w_trait { macro_rules! impl_bdd_2w_to_1w_trait {
($trait_name:ident, $method_name:ident, $ty:ty, $circuit_ty:ty, $output_circuits:path) => { ($trait_name:ident, $method_name:ident, $ty:ty, $circuit_ty:ty, $output_circuits:path) => {
impl<D: DataMut, BE: Backend> $trait_name<$ty, BE> for FheUint<D, $ty> { paste::paste! {
fn $method_name<A, M, K, H, B>( impl<D: DataMut, BE: Backend> $trait_name<$ty, BE> for FheUint<D, $ty> {
&mut self,
module: &M, fn $method_name<A, M, K, H, B>(
a: &FheUintPrepared<A, $ty, BE>, &mut self,
b: &FheUintPrepared<B, $ty, BE>, module: &M,
key: &H, a: &FheUintPrepared<A, $ty, BE>,
scratch: &mut Scratch<BE>, b: &FheUintPrepared<B, $ty, BE>,
) where key: &H,
M: ExecuteBDDCircuit2WTo1W<$ty, BE>, scratch: &mut Scratch<BE>,
A: DataRef, ) where
B: DataRef, M: ExecuteBDDCircuit2WTo1W<BE>,
K: GGLWEPreparedToRef<BE> + GetGaloisElement + GGLWEInfos, A: DataRef,
H: GLWEAutomorphismKeyHelper<K, BE>, B: DataRef,
Scratch<BE>: ScratchTakeCore<BE>, K: GGLWEPreparedToRef<BE> + GetGaloisElement + GGLWEInfos,
{ H: GLWEAutomorphismKeyHelper<K, BE>,
module.execute_bdd_circuit_2w_to_1w(self, &$output_circuits, a, b, key, scratch) Scratch<BE>: ScratchTakeCore<BE>,
{
module.execute_bdd_circuit_2w_to_1w(self, &$output_circuits, a, b, key, scratch)
}
fn [<$method_name _multi_thread>]<A, M, K, H, B>(
&mut self,
threads: usize,
module: &M,
a: &FheUintPrepared<A, $ty, BE>,
b: &FheUintPrepared<B, $ty, BE>,
key: &H,
scratch: &mut Scratch<BE>,
) where
M: ExecuteBDDCircuit2WTo1W<BE>,
A: DataRef,
B: DataRef,
K: GGLWEPreparedToRef<BE> + GetGaloisElement + GGLWEInfos,
H: GLWEAutomorphismKeyHelper<K, BE>,
Scratch<BE>: ScratchTakeCore<BE>,
{
module.execute_bdd_circuit_2w_to_1w_multi_thread(threads, self, &$output_circuits, a, b, key, scratch)
}
} }
} }
}; };

View File

@@ -1,15 +1,17 @@
use std::marker::PhantomData; use std::marker::PhantomData;
use std::thread;
use poulpy_core::layouts::{ use poulpy_core::layouts::{
Base2K, Dnum, Dsize, GGSWInfos, GGSWPreparedFactory, GLWEInfos, LWEInfos, Rank, TorusPrecision, prepared::GGSWPrepared, Base2K, Dnum, Dsize, GGSWInfos, GGSWPreparedFactory, GLWEInfos, LWEInfos, Rank, TorusPrecision, prepared::GGSWPrepared,
}; };
use poulpy_core::layouts::{ use poulpy_core::layouts::{
GGLWEInfos, GGLWEPreparedToRef, GGSWPreparedToMut, GGSWPreparedToRef, GLWEAutomorphismKeyHelper, GetGaloisElement, LWE, GGLWEInfos, GGLWEPreparedToRef, GGSW, GGSWLayout, GGSWPreparedToMut, GGSWPreparedToRef, GLWEAutomorphismKeyHelper,
GetGaloisElement, LWE,
}; };
use poulpy_core::{GLWECopy, GLWEDecrypt, GLWEPacking, LWEFromGLWE}; use poulpy_core::{GLWECopy, GLWEDecrypt, GLWEPacking, LWEFromGLWE};
use poulpy_core::{GGSWEncryptSk, ScratchTakeCore, layouts::GLWESecretPreparedToRef}; use poulpy_core::{GGSWEncryptSk, ScratchTakeCore, layouts::GLWESecretPreparedToRef};
use poulpy_hal::api::ModuleLogN; use poulpy_hal::api::{ModuleLogN, ScratchAvailable, ScratchFromBytes};
use poulpy_hal::layouts::{Backend, Data, DataRef, Module}; use poulpy_hal::layouts::{Backend, Data, DataRef, Module};
use poulpy_hal::{ use poulpy_hal::{
@@ -21,7 +23,7 @@ use poulpy_hal::{
use crate::tfhe::bdd_arithmetic::{BDDKey, BDDKeyHelper, BDDKeyInfos, BDDKeyPrepared, BDDKeyPreparedFactory, FheUint, ToBits}; use crate::tfhe::bdd_arithmetic::{BDDKey, BDDKeyHelper, BDDKeyInfos, BDDKeyPrepared, BDDKeyPreparedFactory, FheUint, ToBits};
use crate::tfhe::bdd_arithmetic::{Cmux, FromBits, ScratchTakeBDD, UnsignedInteger}; use crate::tfhe::bdd_arithmetic::{Cmux, FromBits, ScratchTakeBDD, UnsignedInteger};
use crate::tfhe::blind_rotation::BlindRotationAlgo; use crate::tfhe::blind_rotation::BlindRotationAlgo;
use crate::tfhe::circuit_bootstrapping::CirtuitBootstrappingExecute; use crate::tfhe::circuit_bootstrapping::{CircuitBootstrappingKeyInfos, CirtuitBootstrappingExecute};
/// A prepared FHE ciphertext encrypting the bits of an [UnsignedInteger]. /// A prepared FHE ciphertext encrypting the bits of an [UnsignedInteger].
pub struct FheUintPrepared<D: Data, T: UnsignedInteger, B: Backend> { pub struct FheUintPrepared<D: Data, T: UnsignedInteger, B: Backend> {
@@ -31,7 +33,7 @@ pub struct FheUintPrepared<D: Data, T: UnsignedInteger, B: Backend> {
impl<T: UnsignedInteger, BE: Backend> FheUintPreparedFactory<T, BE> for Module<BE> where Self: Sized + GGSWPreparedFactory<BE> {} impl<T: UnsignedInteger, BE: Backend> FheUintPreparedFactory<T, BE> for Module<BE> where Self: Sized + GGSWPreparedFactory<BE> {}
pub trait GetGGSWBit<BE: Backend> { pub trait GetGGSWBit<BE: Backend>: Sync {
fn get_bit(&self, bit: usize) -> GGSWPrepared<&[u8], BE>; fn get_bit(&self, bit: usize) -> GGSWPrepared<&[u8], BE>;
} }
@@ -219,12 +221,20 @@ impl<D: DataMut, BRA: BlindRotationAlgo, BE: Backend> BDDKeyPrepared<D, BRA, BE>
} }
} }
pub trait FheUintPrepare<BRA: BlindRotationAlgo, T: UnsignedInteger, BE: Backend> { pub trait FheUintPrepare<BRA: BlindRotationAlgo, BE: Backend> {
fn fhe_uint_prepare_tmp_bytes<R, A>(&self, block_size: usize, extension_factor: usize, res_infos: &R, infos: &A) -> usize fn fhe_uint_prepare_tmp_bytes<R, A, B>(
&self,
block_size: usize,
extension_factor: usize,
res_infos: &R,
bits_infos: &A,
bdd_infos: &B,
) -> usize
where where
R: GGSWInfos, R: GGSWInfos,
A: BDDKeyInfos; A: GLWEInfos,
fn fhe_uint_prepare<DM, DB, DK, K>( B: BDDKeyInfos;
fn fhe_uint_prepare<DM, DB, DK, K, T: UnsignedInteger>(
&self, &self,
res: &mut FheUintPrepared<DM, T, BE>, res: &mut FheUintPrepared<DM, T, BE>,
bits: &FheUint<DB, T>, bits: &FheUint<DB, T>,
@@ -234,79 +244,120 @@ pub trait FheUintPrepare<BRA: BlindRotationAlgo, T: UnsignedInteger, BE: Backend
DM: DataMut, DM: DataMut,
DB: DataRef, DB: DataRef,
DK: DataRef, DK: DataRef,
K: BDDKeyHelper<DK, BRA, BE>; K: BDDKeyHelper<DK, BRA, BE> + BDDKeyInfos,
fn fhe_uint_prepare_custom<DM, DB, DK, K>( Scratch<BE>: ScratchFromBytes<BE>,
{
self.fhe_uint_prepare_custom(res, bits, 0, T::BITS as usize, key, scratch);
}
fn fhe_uint_prepare_custom<DM, DB, DK, K, T: UnsignedInteger>(
&self, &self,
res: &mut FheUintPrepared<DM, T, BE>, res: &mut FheUintPrepared<DM, T, BE>,
bits: &FheUint<DB, T>, bits: &FheUint<DB, T>,
bit_start: usize, bit_start: usize,
bit_end: usize, bit_count: usize,
key: &K, key: &K,
scratch: &mut Scratch<BE>, scratch: &mut Scratch<BE>,
) where ) where
DM: DataMut, DM: DataMut,
DB: DataRef, DB: DataRef,
DK: DataRef, DK: DataRef,
K: BDDKeyHelper<DK, BRA, BE>; K: BDDKeyHelper<DK, BRA, BE> + BDDKeyInfos,
{
self.fhe_uint_prepare_custom_multi_thread(1, res, bits, bit_start, bit_count, key, scratch)
}
#[allow(clippy::too_many_arguments)]
fn fhe_uint_prepare_custom_multi_thread<DM, DB, DK, K, T: UnsignedInteger>(
&self,
threads: usize,
res: &mut FheUintPrepared<DM, T, BE>,
bits: &FheUint<DB, T>,
bit_start: usize,
bit_count: usize,
key: &K,
scratch: &mut Scratch<BE>,
) where
DM: DataMut,
DB: DataRef,
DK: DataRef,
K: BDDKeyHelper<DK, BRA, BE> + BDDKeyInfos;
} }
impl<BRA: BlindRotationAlgo, BE: Backend, T: UnsignedInteger> FheUintPrepare<BRA, T, BE> for Module<BE> impl<BRA: BlindRotationAlgo, BE: Backend> FheUintPrepare<BRA, BE> for Module<BE>
where where
Self: LWEFromGLWE<BE> + CirtuitBootstrappingExecute<BRA, BE> + GGSWPreparedFactory<BE>, Self: LWEFromGLWE<BE> + CirtuitBootstrappingExecute<BRA, BE> + GGSWPreparedFactory<BE>,
Scratch<BE>: ScratchTakeCore<BE>, Scratch<BE>: ScratchTakeCore<BE>,
{ {
fn fhe_uint_prepare_tmp_bytes<R, A>(&self, block_size: usize, extension_factor: usize, res_infos: &R, bdd_infos: &A) -> usize fn fhe_uint_prepare_tmp_bytes<R, A, B>(
&self,
block_size: usize,
extension_factor: usize,
res_infos: &R,
bits_infos: &A,
bdd_infos: &B,
) -> usize
where where
R: GGSWInfos, R: GGSWInfos,
A: BDDKeyInfos, A: GLWEInfos,
B: BDDKeyInfos,
{ {
self.circuit_bootstrapping_execute_tmp_bytes( self.circuit_bootstrapping_execute_tmp_bytes(
block_size, block_size,
extension_factor, extension_factor,
res_infos, res_infos,
&bdd_infos.cbt_infos(), &bdd_infos.cbt_infos(),
) ) + GGSW::bytes_of_from_infos(res_infos)
+ LWE::bytes_of_from_infos(bits_infos)
} }
fn fhe_uint_prepare<DM, DB, DK, K>( fn fhe_uint_prepare_custom_multi_thread<DM, DB, DK, K, T: UnsignedInteger>(
&self,
res: &mut FheUintPrepared<DM, T, BE>,
bits: &FheUint<DB, T>,
key: &K,
scratch: &mut Scratch<BE>,
) where
DM: DataMut,
DB: DataRef,
DK: DataRef,
K: BDDKeyHelper<DK, BRA, BE>,
{
self.fhe_uint_prepare_custom(res, bits, 0, T::BITS as usize, key, scratch);
}
fn fhe_uint_prepare_custom<DM, DB, DK, K>(
&self, &self,
threads: usize,
res: &mut FheUintPrepared<DM, T, BE>, res: &mut FheUintPrepared<DM, T, BE>,
bits: &FheUint<DB, T>, bits: &FheUint<DB, T>,
bit_start: usize, bit_start: usize,
bit_end: usize, bit_count: usize,
key: &K, key: &K,
scratch: &mut Scratch<BE>, scratch: &mut Scratch<BE>,
) where ) where
DM: DataMut, DM: DataMut,
DB: DataRef, DB: DataRef,
DK: DataRef, DK: DataRef,
K: BDDKeyHelper<DK, BRA, BE>, K: BDDKeyHelper<DK, BRA, BE> + BDDKeyInfos,
{ {
let bit_end = bit_start + bit_count;
let (cbt, ks) = key.get_cbt_key(); let (cbt, ks) = key.get_cbt_key();
let mut lwe: LWE<Vec<u8>> = LWE::alloc_from_infos(bits); //TODO: add TakeLWE assert!(bit_end <= T::BITS as usize);
let (mut tmp_ggsw, scratch_1) = scratch.take_ggsw(res);
for (bit, dst) in res.bits[bit_start..bit_end].iter_mut().enumerate() { let scratch_thread_size = self.fhe_uint_prepare_tmp_bytes(cbt.block_size(), 1, res, bits, key);
// TODO: set the rest of the bits to a prepared zero GGSW
bits.get_bit_lwe(self, bit, &mut lwe, ks, scratch_1); assert!(scratch.available() >= threads * scratch_thread_size);
cbt.execute_to_constant(self, &mut tmp_ggsw, &lwe, 1, 1, scratch_1);
dst.prepare(self, &tmp_ggsw, scratch_1); let chunk_size: usize = bit_count.div_ceil(threads);
}
let (mut scratches, _) = scratch.split_mut(threads, scratch_thread_size);
let ggsw_infos: &GGSWLayout = &res.ggsw_layout();
thread::scope(|scope| {
for (thread_index, (scratch_thread, res_bits_chunk)) in scratches
.iter_mut()
.zip(res.bits[bit_start..bit_end].chunks_mut(chunk_size))
.enumerate()
{
let start: usize = bit_start + thread_index * chunk_size;
scope.spawn(move || {
let (mut tmp_ggsw, scratch_1) = scratch_thread.take_ggsw(ggsw_infos);
let (mut tmp_lwe, scratch_2) = scratch_1.take_lwe(bits);
for (local_bit, dst) in res_bits_chunk.iter_mut().enumerate() {
bits.get_bit_lwe(self, start + local_bit, &mut tmp_lwe, ks, scratch_2);
cbt.execute_to_constant(self, &mut tmp_ggsw, &tmp_lwe, 1, 1, scratch_2);
dst.prepare(self, &tmp_ggsw, scratch_2);
}
});
}
});
for i in 0..bit_start { for i in 0..bit_start {
res.bits[i].zero(self); res.bits[i].zero(self);
@@ -324,8 +375,8 @@ impl<D: DataMut, T: UnsignedInteger, BE: Backend> FheUintPrepared<D, T, BE> {
BRA: BlindRotationAlgo, BRA: BlindRotationAlgo,
O: DataRef, O: DataRef,
DK: DataRef, DK: DataRef,
K: BDDKeyHelper<DK, BRA, BE>, K: BDDKeyHelper<DK, BRA, BE> + BDDKeyInfos,
M: FheUintPrepare<BRA, T, BE>, M: FheUintPrepare<BRA, BE>,
Scratch<BE>: ScratchTakeCore<BE>, Scratch<BE>: ScratchTakeCore<BE>,
{ {
module.fhe_uint_prepare(self, other, key, scratch); module.fhe_uint_prepare(self, other, key, scratch);
@@ -342,10 +393,31 @@ impl<D: DataMut, T: UnsignedInteger, BE: Backend> FheUintPrepared<D, T, BE> {
BRA: BlindRotationAlgo, BRA: BlindRotationAlgo,
O: DataRef, O: DataRef,
DK: DataRef, DK: DataRef,
K: BDDKeyHelper<DK, BRA, BE>, K: BDDKeyHelper<DK, BRA, BE> + BDDKeyInfos,
M: FheUintPrepare<BRA, T, BE>, M: FheUintPrepare<BRA, BE>,
Scratch<BE>: ScratchTakeCore<BE>, Scratch<BE>: ScratchTakeCore<BE>,
{ {
module.fhe_uint_prepare_custom(self, other, bit_start, bit_end, key, scratch); module.fhe_uint_prepare_custom(self, other, bit_start, bit_end, key, scratch);
} }
#[allow(clippy::too_many_arguments)]
pub fn prepare_custom_multi_thread<BRA, M, O, K, DK>(
&mut self,
threads: usize,
module: &M,
other: &FheUint<O, T>,
bit_start: usize,
bit_end: usize,
key: &K,
scratch: &mut Scratch<BE>,
) where
BRA: BlindRotationAlgo,
O: DataRef,
DK: DataRef,
K: BDDKeyHelper<DK, BRA, BE> + BDDKeyInfos,
M: FheUintPrepare<BRA, BE>,
Scratch<BE>: ScratchTakeCore<BE>,
{
module.fhe_uint_prepare_custom_multi_thread(threads, self, other, bit_start, bit_end, key, scratch);
}
} }

View File

@@ -10,7 +10,7 @@ use poulpy_core::layouts::{Base2K, Dnum, Dsize, Rank, TorusPrecision};
use poulpy_core::layouts::{GGSW, GLWESecretPreparedToRef}; use poulpy_core::layouts::{GGSW, GLWESecretPreparedToRef};
use poulpy_core::{ use poulpy_core::{
LWEFromGLWE, ScratchTakeCore, LWEFromGLWE, ScratchTakeCore,
layouts::{GGSWInfos, GGSWPreparedFactory, GLWEInfos, LWE, LWEInfos}, layouts::{GGSWInfos, GGSWPreparedFactory, GLWEInfos, LWEInfos},
}; };
use poulpy_hal::api::ModuleN; use poulpy_hal::api::ModuleN;
@@ -125,10 +125,12 @@ where
DR0: DataRef, DR0: DataRef,
DR1: DataRef, DR1: DataRef,
{ {
let mut lwe: LWE<Vec<u8>> = LWE::alloc_from_infos(bits); //TODO: add TakeLWE let (_, scratch_1) = scratch.take_ggsw(res);
let (mut tmp_lwe, scratch_2) = scratch_1.take_lwe(bits);
for (bit, dst) in res.bits.iter_mut().enumerate() { for (bit, dst) in res.bits.iter_mut().enumerate() {
bits.get_bit_lwe(self, bit, &mut lwe, &key.ks, scratch); bits.get_bit_lwe(self, bit, &mut tmp_lwe, &key.ks, scratch_2);
key.cbt.execute_to_constant(self, dst, &lwe, 1, 1, scratch); key.cbt
.execute_to_constant(self, dst, &tmp_lwe, 1, 1, scratch_2);
} }
} }
} }

View File

@@ -1,4 +1,5 @@
use core::panic; use core::panic;
use std::thread;
use itertools::Itertools; use itertools::Itertools;
use poulpy_core::{ use poulpy_core::{
@@ -6,17 +7,20 @@ use poulpy_core::{
layouts::{GGSWInfos, GGSWPrepared, GLWE, GLWEInfos, GLWEToMut, GLWEToRef, LWEInfos, prepared::GGSWPreparedToRef}, layouts::{GGSWInfos, GGSWPrepared, GLWE, GLWEInfos, GLWEToMut, GLWEToRef, LWEInfos, prepared::GGSWPreparedToRef},
}; };
use poulpy_hal::{ use poulpy_hal::{
api::{ScratchTakeBasic, VecZnxBigAddSmallInplace, VecZnxBigNormalize, VecZnxBigNormalizeTmpBytes, VecZnxDftBytesOf}, api::{
ScratchAvailable, ScratchTakeBasic, VecZnxBigAddSmallInplace, VecZnxBigNormalize, VecZnxBigNormalizeTmpBytes,
VecZnxDftBytesOf,
},
layouts::{Backend, DataMut, Module, Scratch, VecZnxBig, ZnxZero}, layouts::{Backend, DataMut, Module, Scratch, VecZnxBig, ZnxZero},
}; };
use crate::tfhe::bdd_arithmetic::{GetGGSWBit, UnsignedInteger}; use crate::tfhe::bdd_arithmetic::GetGGSWBit;
pub trait BitCircuitInfo { pub trait BitCircuitInfo: Sync {
fn info(&self) -> (&[Node], usize); fn info(&self) -> (&[Node], usize);
} }
pub trait GetBitCircuitInfo<T: UnsignedInteger> { pub trait GetBitCircuitInfo: Sync {
fn input_size(&self) -> usize; fn input_size(&self) -> usize;
fn output_size(&self) -> usize; fn output_size(&self) -> usize;
fn get_circuit(&self, bit: usize) -> (&[Node], usize); fn get_circuit(&self, bit: usize) -> (&[Node], usize);
@@ -34,7 +38,7 @@ pub trait BitCircuitFamily {
pub struct Circuit<C: BitCircuitInfo, const N: usize>(pub [C; N]); pub struct Circuit<C: BitCircuitInfo, const N: usize>(pub [C; N]);
impl<C, T: UnsignedInteger, const N: usize> GetBitCircuitInfo<T> for Circuit<C, N> impl<C, const N: usize> GetBitCircuitInfo for Circuit<C, N>
where where
C: BitCircuitInfo + BitCircuitFamily, C: BitCircuitInfo + BitCircuitFamily,
{ {
@@ -49,11 +53,31 @@ where
} }
} }
pub trait ExecuteBDDCircuit<T: UnsignedInteger, BE: Backend> { pub trait ExecuteBDDCircuit<BE: Backend> {
fn execute_bdd_circuit_tmp_bytes<R, G>(&self, res_infos: &R, state_size: usize, ggsw_infos: &G) -> usize
where
R: GLWEInfos,
G: GGSWInfos;
fn execute_bdd_circuit<C, G, O>(&self, out: &mut [GLWE<O>], inputs: &G, circuit: &C, scratch: &mut Scratch<BE>) fn execute_bdd_circuit<C, G, O>(&self, out: &mut [GLWE<O>], inputs: &G, circuit: &C, scratch: &mut Scratch<BE>)
where where
G: GetGGSWBit<BE> + BitSize, G: GetGGSWBit<BE> + BitSize,
C: GetBitCircuitInfo<T>, C: GetBitCircuitInfo,
O: DataMut,
{
self.execute_bdd_circuit_multi_thread(1, out, inputs, circuit, scratch);
}
fn execute_bdd_circuit_multi_thread<C, G, O>(
&self,
threads: usize,
out: &mut [GLWE<O>],
inputs: &G,
circuit: &C,
scratch: &mut Scratch<BE>,
) where
G: GetGGSWBit<BE> + BitSize,
C: GetBitCircuitInfo,
O: DataMut; O: DataMut;
} }
@@ -61,15 +85,29 @@ pub trait BitSize {
fn bit_size(&self) -> usize; fn bit_size(&self) -> usize;
} }
impl<T: UnsignedInteger, BE: Backend> ExecuteBDDCircuit<T, BE> for Module<BE> impl<BE: Backend> ExecuteBDDCircuit<BE> for Module<BE>
where where
Self: Cmux<BE> + GLWECopy, Self: Cmux<BE> + GLWECopy,
Scratch<BE>: ScratchTakeCore<BE>, Scratch<BE>: ScratchTakeCore<BE>,
{ {
fn execute_bdd_circuit<C, G, O>(&self, out: &mut [GLWE<O>], inputs: &G, circuit: &C, scratch: &mut Scratch<BE>) fn execute_bdd_circuit_tmp_bytes<R, G>(&self, res_infos: &R, state_size: usize, ggsw_infos: &G) -> usize
where where
R: GLWEInfos,
G: GGSWInfos,
{
2 * state_size * GLWE::bytes_of_from_infos(res_infos) + self.cmux_tmp_bytes(res_infos, res_infos, ggsw_infos)
}
fn execute_bdd_circuit_multi_thread<C, G, O>(
&self,
threads: usize,
out: &mut [GLWE<O>],
inputs: &G,
circuit: &C,
scratch: &mut Scratch<BE>,
) where
G: GetGGSWBit<BE> + BitSize, G: GetGGSWBit<BE> + BitSize,
C: GetBitCircuitInfo<T>, C: GetBitCircuitInfo,
O: DataMut, O: DataMut,
{ {
#[cfg(debug_assertions)] #[cfg(debug_assertions)]
@@ -88,66 +126,43 @@ where
); );
} }
for (i, out_i) in out.iter_mut().enumerate().take(circuit.output_size()) { let mut max_state_size = 0;
let (nodes, max_inter_state) = circuit.get_circuit(i); for i in 0..circuit.output_size() {
let (_, state_size) = circuit.get_circuit(i);
max_state_size = max_state_size.max(state_size)
}
if max_inter_state == 0 { let scratch_thread_size: usize = self.execute_bdd_circuit_tmp_bytes(&out[0], max_state_size, &inputs.get_bit(0));
out_i.data_mut().zero();
} else {
assert!(nodes.len().is_multiple_of(max_inter_state));
let (mut level, scratch_1) = scratch.take_glwe_slice(max_inter_state * 2, out_i); assert!(
scratch.available() >= threads * scratch_thread_size,
"scratch.available(): {} < threads:{threads} * scratch_thread_size: {scratch_thread_size}",
scratch.available()
);
level.iter_mut().for_each(|ct| ct.data_mut().zero()); let (mut scratches, _) = scratch.split_mut(threads, scratch_thread_size);
// TODO: implement API on GLWE let chunk_size: usize = circuit.output_size().div_ceil(threads);
level[1]
.data_mut()
.encode_coeff_i64(out_i.base2k().into(), 0, 2, 0, 1);
let mut level_ref = level.iter_mut().collect_vec(); thread::scope(|scope| {
let (mut prev_level, mut next_level) = level_ref.split_at_mut(max_inter_state); for (scratch_thread, out_chunk) in scratches
.iter_mut()
.zip(out[..circuit.output_size()].chunks_mut(chunk_size))
{
// Capture chunk + thread scratch by move
scope.spawn(move || {
for (idx, out_i) in out_chunk.iter_mut().enumerate() {
let (nodes, state_size) = circuit.get_circuit(idx);
let (all_but_last, last) = nodes.split_at(nodes.len() - max_inter_state); if state_size == 0 {
out_i.data_mut().zero();
for nodes_lvl in all_but_last.chunks_exact(max_inter_state) { } else {
for (j, node) in nodes_lvl.iter().enumerate() { eval_level(self, out_i, inputs, nodes, state_size, *scratch_thread);
match node {
Node::Cmux(in_idx, hi_idx, lo_idx) => {
self.cmux(
next_level[j],
prev_level[*hi_idx],
prev_level[*lo_idx],
&inputs.get_bit(*in_idx),
scratch_1,
);
}
Node::Copy => self.glwe_copy(next_level[j], prev_level[j]), /* Update BDD circuits to order Cmux -> Copy -> None so that mem swap can be used */
Node::None => {}
} }
} }
});
(prev_level, next_level) = (next_level, prev_level);
}
// Last chunck of max_inter_state Nodes is always structured as
// [CMUX, NONE, NONE, ..., NONE]
match &last[0] {
Node::Cmux(in_idx, hi_idx, lo_idx) => {
self.cmux(
out_i,
prev_level[*hi_idx],
prev_level[*lo_idx],
&inputs.get_bit(*in_idx),
scratch_1,
);
}
_ => {
panic!("invalid last node, should be CMUX")
}
}
} }
} });
for out_i in out.iter_mut().skip(circuit.output_size()) { for out_i in out.iter_mut().skip(circuit.output_size()) {
out_i.data_mut().zero(); out_i.data_mut().zero();
@@ -155,6 +170,74 @@ where
} }
} }
fn eval_level<M, R, G, BE: Backend>(
module: &M,
res: &mut R,
inputs: &G,
nodes: &[Node],
state_size: usize,
scratch: &mut Scratch<BE>,
) where
M: Cmux<BE> + GLWECopy,
R: GLWEToMut,
G: GetGGSWBit<BE> + BitSize,
Scratch<BE>: ScratchTakeCore<BE>,
{
assert!(nodes.len().is_multiple_of(state_size));
let res: &mut GLWE<&mut [u8]> = &mut res.to_mut();
let (mut level, scratch_1) = scratch.take_glwe_slice(state_size * 2, res);
level.iter_mut().for_each(|ct| ct.data_mut().zero());
// TODO: implement API on GLWE
level[1]
.data_mut()
.encode_coeff_i64(res.base2k().into(), 0, 2, 0, 1);
let mut level_ref: Vec<&mut GLWE<&mut [u8]>> = level.iter_mut().collect_vec();
let (mut prev_level, mut next_level) = level_ref.split_at_mut(state_size);
let (all_but_last, last) = nodes.split_at(nodes.len() - state_size);
for nodes_lvl in all_but_last.chunks_exact(state_size) {
for (j, node) in nodes_lvl.iter().enumerate() {
match node {
Node::Cmux(in_idx, hi_idx, lo_idx) => {
module.cmux(
next_level[j],
prev_level[*hi_idx],
prev_level[*lo_idx],
&inputs.get_bit(*in_idx),
scratch_1,
);
}
Node::Copy => module.glwe_copy(next_level[j], prev_level[j]), /* Update BDD circuits to order Cmux -> Copy -> None so that mem swap can be used */
Node::None => {}
}
}
(prev_level, next_level) = (next_level, prev_level);
}
// Last chunck of max_inter_state Nodes is always structured as
// [CMUX, NONE, NONE, ..., NONE]
match &last[0] {
Node::Cmux(in_idx, hi_idx, lo_idx) => {
module.cmux(
res,
prev_level[*hi_idx],
prev_level[*lo_idx],
&inputs.get_bit(*in_idx),
scratch_1,
);
}
_ => {
panic!("invalid last node, should be CMUX")
}
}
}
impl<const N: usize> BitCircuit<N> { impl<const N: usize> BitCircuit<N> {
pub const fn new(nodes: [Node; N], max_inter_state: usize) -> Self { pub const fn new(nodes: [Node; N], max_inter_state: usize) -> Self {
Self { Self {

View File

@@ -1,4 +1,5 @@
use crate::tfhe::bdd_arithmetic::FheUintPreparedDebug; use crate::tfhe::bdd_arithmetic::FheUintPreparedDebug;
use crate::tfhe::circuit_bootstrapping::CircuitBootstrappingKeyInfos;
use crate::tfhe::{ use crate::tfhe::{
bdd_arithmetic::{FheUint, UnsignedInteger}, bdd_arithmetic::{FheUint, UnsignedInteger},
blind_rotation::{BlindRotationAlgo, BlindRotationKey, BlindRotationKeyFactory}, blind_rotation::{BlindRotationAlgo, BlindRotationKey, BlindRotationKeyFactory},
@@ -8,7 +9,7 @@ use crate::tfhe::{
}, },
}; };
use poulpy_core::layouts::{GLWEAutomorphismKeyHelper, GLWEAutomorphismKeyPrepared}; use poulpy_core::layouts::{GGLWEInfos, GLWEAutomorphismKeyHelper, GLWEAutomorphismKeyPrepared};
use poulpy_core::{ use poulpy_core::{
GLWEToLWESwitchingKeyEncryptSk, GetDistribution, ScratchTakeCore, GLWEToLWESwitchingKeyEncryptSk, GetDistribution, ScratchTakeCore,
layouts::{ layouts::{
@@ -135,6 +136,25 @@ where
pub(crate) ks: GLWEToLWEKeyPrepared<D, BE>, pub(crate) ks: GLWEToLWEKeyPrepared<D, BE>,
} }
impl<D: DataRef, BRA: BlindRotationAlgo, BE: Backend> BDDKeyInfos for BDDKeyPrepared<D, BRA, BE> {
fn cbt_infos(&self) -> CircuitBootstrappingKeyLayout {
CircuitBootstrappingKeyLayout {
layout_brk: self.cbt.brk_infos(),
layout_atk: self.cbt.atk_infos(),
layout_tsk: self.cbt.tsk_infos(),
}
}
fn ks_infos(&self) -> GLWEToLWEKeyLayout {
GLWEToLWEKeyLayout {
n: self.ks.n(),
base2k: self.ks.base2k(),
k: self.ks.k(),
rank_in: self.ks.rank_in(),
dnum: self.ks.dnum(),
}
}
}
impl<D: DataRef, BRA: BlindRotationAlgo, BE: Backend> GLWEAutomorphismKeyHelper<GLWEAutomorphismKeyPrepared<D, BE>, BE> impl<D: DataRef, BRA: BlindRotationAlgo, BE: Backend> GLWEAutomorphismKeyHelper<GLWEAutomorphismKeyPrepared<D, BE>, BE>
for BDDKeyPrepared<D, BRA, BE> for BDDKeyPrepared<D, BRA, BE>
{ {

View File

@@ -16,7 +16,7 @@ pub use key::*;
pub mod tests; pub mod tests;
pub trait UnsignedInteger: Copy + 'static { pub trait UnsignedInteger: Copy + Sync + Send + 'static {
const BITS: u32; const BITS: u32;
const LOG_BITS: u32; const LOG_BITS: u32;
const LOG_BYTES: u32; const LOG_BYTES: u32;

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -75,6 +75,14 @@ where
} }
impl<BRA: BlindRotationAlgo, BE: Backend> TestContext<BRA, BE> { impl<BRA: BlindRotationAlgo, BE: Backend> TestContext<BRA, BE> {
pub fn glwe_infos(&self) -> GLWELayout {
TEST_GLWE_INFOS
}
pub fn ggsw_infos(&self) -> GGSWLayout {
TEST_GGSW_INFOS
}
pub fn new() -> Self pub fn new() -> Self
where where
Module<BE>: ModuleNew<BE> Module<BE>: ModuleNew<BE>

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,
@@ -67,8 +67,10 @@ where
let mut c_enc_prep_debug: FheUintPreparedDebug<Vec<u8>, u32> = let mut c_enc_prep_debug: FheUintPreparedDebug<Vec<u8>, u32> =
FheUintPreparedDebug::<Vec<u8>, u32>::alloc_from_infos(module, &ggsw_infos); FheUintPreparedDebug::<Vec<u8>, u32>::alloc_from_infos(module, &ggsw_infos);
let mut scratch_2 = ScratchOwned::alloc(module.fhe_uint_prepare_tmp_bytes(7, 1, &c_enc_prep_debug, &c_enc, bdd_key_prepared));
// GGSW(value) // GGSW(value)
c_enc_prep_debug.prepare(module, &c_enc, bdd_key_prepared, scratch.borrow()); c_enc_prep_debug.prepare(module, &c_enc, bdd_key_prepared, scratch_2.borrow());
let max_noise = |col_i: usize| { let max_noise = |col_i: usize| {
let mut noise: f64 = -(ggsw_infos.size() as f64 * TEST_BASE2K as f64) + SIGMA.log2() + 2.0; let mut noise: f64 = -(ggsw_infos.size() as f64 * TEST_BASE2K as f64) + SIGMA.log2() + 2.0;

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -30,8 +30,8 @@ where
+ BDDKeyEncryptSk<BRA, BE> + BDDKeyEncryptSk<BRA, BE>
+ BDDKeyPreparedFactory<BRA, BE> + BDDKeyPreparedFactory<BRA, BE>
+ GGSWNoise<BE> + GGSWNoise<BE>
+ FheUintPrepare<BRA, u32, BE> + FheUintPrepare<BRA, BE>
+ ExecuteBDDCircuit2WTo1W<u32, BE> + ExecuteBDDCircuit2WTo1W<BE>
+ GLWEEncryptSk<BE>, + GLWEEncryptSk<BE>,
BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>, BlindRotationKey<Vec<u8>, BRA>: BlindRotationKeyFactory<BRA>,
ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>, ScratchOwned<BE>: ScratchOwnedAlloc<BE> + ScratchOwnedBorrow<BE>,

View File

@@ -11,7 +11,7 @@ use poulpy_hal::layouts::{Backend, DataMut, DataRef, Scratch, ZnxView};
use crate::tfhe::blind_rotation::{BlindRotationKeyInfos, BlindRotationKeyPrepared, LookUpTableRotationDirection, LookupTable}; use crate::tfhe::blind_rotation::{BlindRotationKeyInfos, BlindRotationKeyPrepared, LookUpTableRotationDirection, LookupTable};
pub trait BlindRotationAlgo {} pub trait BlindRotationAlgo: Sync {}
pub trait BlindRotationExecute<BRA: BlindRotationAlgo, BE: Backend> { pub trait BlindRotationExecute<BRA: BlindRotationAlgo, BE: Backend> {
fn blind_rotation_execute_tmp_bytes<G, B>( fn blind_rotation_execute_tmp_bytes<G, B>(

View File

@@ -188,8 +188,7 @@ impl<D: DataRef, BRT: BlindRotationAlgo> BlindRotationKeyInfos for BlindRotation
} }
impl<D: DataRef, BRT: BlindRotationAlgo> BlindRotationKey<D, BRT> { impl<D: DataRef, BRT: BlindRotationAlgo> BlindRotationKey<D, BRT> {
#[allow(dead_code)] pub fn block_size(&self) -> usize {
fn block_size(&self) -> usize {
match self.dist { match self.dist {
Distribution::BinaryBlock(value) => value, Distribution::BinaryBlock(value) => value,
_ => 1, _ => 1,

View File

@@ -1,15 +1,15 @@
use std::collections::HashMap; use std::collections::HashMap;
use poulpy_hal::{ use poulpy_hal::{
api::{ModuleLogN, ModuleN, ScratchOwnedAlloc, ScratchOwnedBorrow}, api::{ModuleLogN, ModuleN, ScratchAvailable, ScratchOwnedAlloc, ScratchOwnedBorrow},
layouts::{Backend, DataRef, Module, Scratch, ScratchOwned, ToOwnedDeep}, layouts::{Backend, DataRef, Module, Scratch, ScratchOwned, ToOwnedDeep},
}; };
use poulpy_core::{ use poulpy_core::{
GGSWFromGGLWE, GLWEDecrypt, GLWEPacking, GLWERotate, GLWETrace, ScratchTakeCore, GGSWFromGGLWE, GLWEDecrypt, GLWEPacking, GLWERotate, GLWETrace, ScratchTakeCore,
layouts::{ layouts::{
Dsize, GGLWEInfos, GGLWELayout, GGLWEPreparedToRef, GGSWInfos, GGSWToMut, GLWEAutomorphismKeyHelper, GLWEInfos, Dsize, GGLWE, GGLWEInfos, GGLWELayout, GGLWEPreparedToRef, GGSWInfos, GGSWToMut, GLWEAutomorphismKeyHelper, GLWEInfos,
GLWESecretPreparedFactory, GLWEToMut, GLWEToRef, GetGaloisElement, LWEInfos, LWEToRef, GLWESecretPreparedFactory, GLWEToMut, GLWEToRef, GetGaloisElement, LWEInfos, LWEToRef, Rank,
}, },
}; };
@@ -132,6 +132,16 @@ where
R: GGSWInfos, R: GGSWInfos,
A: CircuitBootstrappingKeyInfos, A: CircuitBootstrappingKeyInfos,
{ {
let gglwe_infos: GGLWELayout = GGLWELayout {
n: res_infos.n(),
base2k: res_infos.base2k(),
k: res_infos.k(),
dnum: res_infos.dnum(),
dsize: Dsize(1),
rank_in: res_infos.rank().max(Rank(1)),
rank_out: res_infos.rank(),
};
self.blind_rotation_execute_tmp_bytes( self.blind_rotation_execute_tmp_bytes(
block_size, block_size,
extension_factor, extension_factor,
@@ -140,6 +150,8 @@ where
) )
.max(self.glwe_trace_tmp_bytes(res_infos, res_infos, &cbt_infos.atk_infos())) .max(self.glwe_trace_tmp_bytes(res_infos, res_infos, &cbt_infos.atk_infos()))
.max(self.ggsw_from_gglwe_tmp_bytes(res_infos, &cbt_infos.tsk_infos())) .max(self.ggsw_from_gglwe_tmp_bytes(res_infos, &cbt_infos.tsk_infos()))
+ GLWE::bytes_of_from_infos(res_infos)
+ GGLWE::bytes_of_from_infos(&gglwe_infos)
} }
fn circuit_bootstrapping_execute_to_constant<R, L, D>( fn circuit_bootstrapping_execute_to_constant<R, L, D>(
@@ -155,6 +167,10 @@ where
L: LWEToRef + LWEInfos, L: LWEToRef + LWEInfos,
D: DataRef, D: DataRef,
{ {
assert!(
scratch.available() >= self.circuit_bootstrapping_execute_tmp_bytes(key.block_size(), extension_factor, res, key)
);
circuit_bootstrap_core( circuit_bootstrap_core(
false, false,
self, self,
@@ -182,6 +198,10 @@ where
L: LWEToRef + LWEInfos, L: LWEToRef + LWEInfos,
D: DataRef, D: DataRef,
{ {
assert!(
scratch.available() >= self.circuit_bootstrapping_execute_tmp_bytes(key.block_size(), extension_factor, res, key)
);
circuit_bootstrap_core( circuit_bootstrap_core(
true, true,
self, self,

View File

@@ -19,6 +19,7 @@ use crate::tfhe::blind_rotation::{
}; };
pub trait CircuitBootstrappingKeyInfos { pub trait CircuitBootstrappingKeyInfos {
fn block_size(&self) -> usize;
fn brk_infos(&self) -> BlindRotationKeyLayout; fn brk_infos(&self) -> BlindRotationKeyLayout;
fn atk_infos(&self) -> GLWEAutomorphismKeyLayout; fn atk_infos(&self) -> GLWEAutomorphismKeyLayout;
fn tsk_infos(&self) -> GGLWEToGGSWKeyLayout; fn tsk_infos(&self) -> GGLWEToGGSWKeyLayout;
@@ -32,6 +33,10 @@ pub struct CircuitBootstrappingKeyLayout {
} }
impl CircuitBootstrappingKeyInfos for CircuitBootstrappingKeyLayout { impl CircuitBootstrappingKeyInfos for CircuitBootstrappingKeyLayout {
fn block_size(&self) -> usize {
unimplemented!("unimplemented for CircuitBootstrappingKeyLayout")
}
fn atk_infos(&self) -> GLWEAutomorphismKeyLayout { fn atk_infos(&self) -> GLWEAutomorphismKeyLayout {
self.layout_atk self.layout_atk
} }
@@ -164,6 +169,10 @@ where
} }
impl<D: DataRef, BRA: BlindRotationAlgo> CircuitBootstrappingKeyInfos for CircuitBootstrappingKey<D, BRA> { impl<D: DataRef, BRA: BlindRotationAlgo> CircuitBootstrappingKeyInfos for CircuitBootstrappingKey<D, BRA> {
fn block_size(&self) -> usize {
self.brk.block_size()
}
fn atk_infos(&self) -> GLWEAutomorphismKeyLayout { fn atk_infos(&self) -> GLWEAutomorphismKeyLayout {
let (_, atk) = self.atk.iter().next().expect("atk is empty"); let (_, atk) = self.atk.iter().next().expect("atk is empty");
GLWEAutomorphismKeyLayout { GLWEAutomorphismKeyLayout {

View File

@@ -122,6 +122,10 @@ impl<D: DataRef, BRA: BlindRotationAlgo, BE: Backend> GLWEAutomorphismKeyHelper<
} }
impl<D: DataRef, BRA: BlindRotationAlgo, B: Backend> CircuitBootstrappingKeyInfos for CircuitBootstrappingKeyPrepared<D, BRA, B> { impl<D: DataRef, BRA: BlindRotationAlgo, B: Backend> CircuitBootstrappingKeyInfos for CircuitBootstrappingKeyPrepared<D, BRA, B> {
fn block_size(&self) -> usize {
self.brk.block_size()
}
fn atk_infos(&self) -> GLWEAutomorphismKeyLayout { fn atk_infos(&self) -> GLWEAutomorphismKeyLayout {
let (_, atk) = self.atk.iter().next().expect("atk is empty"); let (_, atk) = self.atk.iter().next().expect("atk is empty");
GLWEAutomorphismKeyLayout { GLWEAutomorphismKeyLayout {