483 implement compiler specific optimizations for ntt and ringmult

benchmark/src/Lattice.cpp

@@ -53,8 +53,6 @@
 
 using namespace lbcrypto;
 
-namespace lbcrypto {
-
 template <typename E>
 static E makeElement(std::shared_ptr<lbcrypto::ILParamsImpl<typename E::Integer>> params) {
     typename E::Vector vec = makeVector<typename E::Vector>(params->GetRingDimension(), params->GetModulus());
@@ -114,8 +112,6 @@ static void GeneratePolys(std::map<usint, std::shared_ptr<P>>& parmArray, std::m
     }
 }
 
-}  // namespace lbcrypto
-
 std::map<usint, std::shared_ptr<ILNativeParams>> Nativeparms;
 std::map<usint, std::vector<NativePoly>> Nativepolys;
 

benchmark/src/poly-benchmark-16k.cpp

@@ -44,8 +44,6 @@
 
 using namespace lbcrypto;
 
-namespace lbcrypto {
-
 static std::vector<usint> tow_args({1, 2, 4, 8});
 
 static const usint DCRTBITS     = MAX_MODULUS_SIZE;
@@ -136,8 +134,6 @@ static void GenerateDCRTPolys(std::map<usint, std::shared_ptr<ILDCRTParams<BigIn
     }
 }
 
-}  // namespace lbcrypto
-
 std::shared_ptr<ILNativeParams> Nativeparms;
 std::map<usint, std::shared_ptr<ILDCRTParams<BigInteger>>> DCRTparms;
 
@@ -209,7 +205,6 @@ static void Native_mul(benchmark::State& state) {
         c = a->Times(*b);
     }
 }
-
 BENCHMARK(Native_mul)->Unit(benchmark::kMicrosecond);
 
 static void DCRT_mul(benchmark::State& state) {

benchmark/src/poly-benchmark-1k.cpp

@@ -44,8 +44,6 @@
 
 using namespace lbcrypto;
 
-namespace lbcrypto {
-
 static std::vector<usint> tow_args({1, 2, 4, 8});
 
 static const usint DCRTBITS     = MAX_MODULUS_SIZE;
@@ -136,8 +134,6 @@ static void GenerateDCRTPolys(std::map<usint, std::shared_ptr<ILDCRTParams<BigIn
     }
 }
 
-}  // namespace lbcrypto
-
 std::shared_ptr<ILNativeParams> Nativeparms;
 std::map<usint, std::shared_ptr<ILDCRTParams<BigInteger>>> DCRTparms;
 

benchmark/src/poly-benchmark-4k.cpp

@@ -44,8 +44,6 @@
 
 using namespace lbcrypto;
 
-namespace lbcrypto {
-
 static std::vector<usint> tow_args({1, 2, 4, 8});
 
 static const usint DCRTBITS     = MAX_MODULUS_SIZE;
@@ -136,8 +134,6 @@ static void GenerateDCRTPolys(std::map<usint, std::shared_ptr<ILDCRTParams<BigIn
     }
 }
 
-}  // namespace lbcrypto
-
 std::shared_ptr<ILNativeParams> Nativeparms;
 std::map<usint, std::shared_ptr<ILDCRTParams<BigInteger>>> DCRTparms;
 

benchmark/src/poly-benchmark-64k.cpp

@@ -45,8 +45,6 @@
 
 using namespace lbcrypto;
 
-namespace lbcrypto {
-
 static std::vector<usint> tow_args({1, 2, 4, 8});
 
 static const usint DCRTBITS     = MAX_MODULUS_SIZE;
@@ -137,8 +135,6 @@ static void GenerateDCRTPolys(std::map<usint, std::shared_ptr<ILDCRTParams<BigIn
     }
 }
 
-}  // namespace lbcrypto
-
 std::shared_ptr<ILNativeParams> Nativeparms;
 std::map<usint, std::shared_ptr<ILDCRTParams<BigInteger>>> DCRTparms;
 

benchmark/src/vechelper.h

@@ -44,17 +44,7 @@ using namespace lbcrypto;
 
 template <typename V>
 inline V makeVector(usint ringdim, const typename V::Integer& mod) {
-    DiscreteUniformGeneratorImpl<V> dug;
-    dug.SetModulus(mod);
-
-    return dug.GenerateVector(ringdim);
-}
-
-inline NativeVector makeNativeVector(usint ringdim, const NativeInteger& mod) {
-    DiscreteUniformGeneratorImpl<NativeVector> dug;
-    dug.SetModulus(mod);
-
-    return dug.GenerateVector(ringdim);
+    return DiscreteUniformGeneratorImpl<V>().GenerateVector(ringdim, mod);
 }
 
 #endif

src/binfhe/lib/lwe-pke.cpp

@@ -332,6 +332,7 @@ LWESwitchingKey LWEEncryptionScheme::KeySwitchGen(const std::shared_ptr<LWECrypt
 // TODO (cpascoe/dsuponit): this pragma needs to be revised as it may have to be removed completely
 // #if !defined(__MINGW32__) && !defined(__MINGW64__)
 // #pragma omp parallel for num_threads(N)
+// #pragma omp parallel for num_threads(OpenFHEParallelControls.GetThreadLimit(N))
 // #endif
     for (size_t i = 0; i < N; ++i) {
         std::vector<std::vector<NativeVector>> vector1A;

src/binfhe/lib/rgsw-acc-cggi.cpp

@@ -45,9 +45,9 @@ RingGSWACCKey RingGSWAccumulatorCGGI::KeyGenAcc(const std::shared_ptr<RingGSWCry
     auto& ek00 = (*ek)[0][0];
     auto& ek01 = (*ek)[0][1];
 
-    // handles ternary secrets using signed mod 3 arithmetic; 0 -> {0,0}, 1 ->
-    // {1,0}, -1 -> {0,1}
-#pragma omp parallel for
+    // handles ternary secrets using signed mod 3 arithmetic
+    // 0 -> {0,0}, 1 -> {1,0}, -1 -> {0,1}
+#pragma omp parallel for num_threads(OpenFHEParallelControls.GetThreadLimit(n))
     for (uint32_t i = 0; i < n; ++i) {
         auto s  = sv[i].ConvertToInt();
         ek00[i] = KeyGenCGGI(params, skNTT, s == 1 ? 1 : 0);
@@ -112,6 +112,7 @@ void RingGSWAccumulatorCGGI::AddToAccCGGI(const std::shared_ptr<RingGSWCryptoPar
 
     SignedDigitDecompose(params, ct, dct);
 
+#pragma omp parallel for num_threads(OpenFHEParallelControls.GetThreadLimit(digitsG2))
     for (uint32_t i = 0; i < digitsG2; ++i)
         dct[i].SetFormat(Format::EVALUATION);
 

src/binfhe/lib/rgsw-acc-dm.cpp

@@ -38,25 +38,21 @@ namespace lbcrypto {
 // Key generation as described in Section 4 of https://eprint.iacr.org/2014/816
 RingGSWACCKey RingGSWAccumulatorDM::KeyGenAcc(const std::shared_ptr<RingGSWCryptoParams>& params,
                                               const NativePoly& skNTT, ConstLWEPrivateKey& LWEsk) const {
-    auto sv     = LWEsk->GetElement();
-    int32_t mod = sv.GetModulus().ConvertToInt();
-
-    int32_t modHalf = mod >> 1;
-
-    uint32_t baseR                            = params->GetBaseR();
-    const std::vector<NativeInteger>& digitsR = params->GetDigitsR();
-    uint32_t n                                = sv.GetLength();
-    RingGSWACCKey ek                          = std::make_shared<RingGSWACCKeyImpl>(n, baseR, digitsR.size());
-
-#pragma omp parallel for
-    for (size_t i = 0; i < n; ++i) {
-        for (size_t j = 1; j < baseR; ++j) {
+    auto sv{LWEsk->GetElement()};
+    auto mod{sv.GetModulus().ConvertToInt<int32_t>()};
+    auto modHalf{mod >> 1};
+    uint32_t n(sv.GetLength());
+    int32_t baseR(params->GetBaseR());
+    const auto& digitsR = params->GetDigitsR();
+    RingGSWACCKey ek    = std::make_shared<RingGSWACCKeyImpl>(n, baseR, digitsR.size());
+
+#pragma omp parallel for num_threads(OpenFHEParallelControls.GetThreadLimit(n))
+    for (uint32_t i = 0; i < n; ++i) {
+        for (int32_t j = 1; j < baseR; ++j) {
             for (size_t k = 0; k < digitsR.size(); ++k) {
-                int32_t s = (int32_t)sv[i].ConvertToInt();
-                if (s > modHalf) {
-                    s -= mod;
-                }
-                (*ek)[i][j][k] = KeyGenDM(params, skNTT, s * j * (int32_t)digitsR[k].ConvertToInt());
+                auto s{sv[i].ConvertToInt<int32_t>()};
+                (*ek)[i][j][k] =
+                    KeyGenDM(params, skNTT, (s > modHalf ? s - mod : s) * j * digitsR[k].ConvertToInt<int32_t>());
             }
         }
     }
@@ -132,6 +128,7 @@ void RingGSWAccumulatorDM::AddToAccDM(const std::shared_ptr<RingGSWCryptoParams>
 
     SignedDigitDecompose(params, ct, dct);
 
+#pragma omp parallel for num_threads(OpenFHEParallelControls.GetThreadLimit(digitsG2))
     for (uint32_t j = 0; j < digitsG2; ++j)
         dct[j].SetFormat(Format::EVALUATION);
 

src/binfhe/lib/rgsw-acc-lmkcdey.cpp

@@ -35,41 +35,35 @@
 
 namespace lbcrypto {
 
-// TODO: optimize this
-
 // Key generation as described in https://eprint.iacr.org/2022/198
 RingGSWACCKey RingGSWAccumulatorLMKCDEY::KeyGenAcc(const std::shared_ptr<RingGSWCryptoParams>& params,
                                                    const NativePoly& skNTT, ConstLWEPrivateKey& LWEsk) const {
-    auto sv              = LWEsk->GetElement();
-    int32_t mod          = sv.GetModulus().ConvertToInt();
-    int32_t modHalf      = mod >> 1;
-    uint32_t N           = params->GetN();
-    size_t n             = sv.GetLength();
-    uint32_t numAutoKeys = params->GetNumAutoKeys();
+    auto sv{LWEsk->GetElement()};
+    auto mod{sv.GetModulus().ConvertToInt<int32_t>()};
+    auto modHalf{mod >> 1};
+    uint32_t N{params->GetN()};
+    size_t n{sv.GetLength()};
+    uint32_t numAutoKeys{params->GetNumAutoKeys()};
 
     // dim2, 0: for RGSW(X^si), 1: for automorphism keys
     // only w automorphism keys required
     // allocates (n - w) more memory for pointer (not critical for performance)
     RingGSWACCKey ek = std::make_shared<RingGSWACCKeyImpl>(1, 2, n);
 
-#pragma omp parallel for
+#pragma omp parallel for num_threads(OpenFHEParallelControls.GetThreadLimit(n))
     for (size_t i = 0; i < n; ++i) {
-        int32_t s = (int32_t)sv[i].ConvertToInt();
-        if (s > modHalf) {
-            s -= mod;
-        }
-
-        (*ek)[0][0][i] = KeyGenLMKCDEY(params, skNTT, s);
+        auto s{sv[i].ConvertToInt<int32_t>()};
+        (*ek)[0][0][i] = KeyGenLMKCDEY(params, skNTT, s > modHalf ? s - mod : s);
     }
 
     NativeInteger gen = NativeInteger(5);
 
     (*ek)[0][1][0] = KeyGenAuto(params, skNTT, 2 * N - gen.ConvertToInt());
 
     // m_window: window size, consider parameterization in the future
-#pragma omp parallel for
+#pragma omp parallel for num_threads(OpenFHEParallelControls.GetThreadLimit(numAutoKeys))
     for (uint32_t i = 1; i <= numAutoKeys; ++i)
-        (*ek)[0][1][i] = KeyGenAuto(params, skNTT, gen.ModExp(i, 2 * N).ConvertToInt());
+        (*ek)[0][1][i] = KeyGenAuto(params, skNTT, gen.ModExp(i, 2 * N).ConvertToInt<LWEPlaintext>());
     return ek;
 }
 
@@ -247,6 +241,7 @@ void RingGSWAccumulatorLMKCDEY::AddToAccLMKCDEY(const std::shared_ptr<RingGSWCry
     SignedDigitDecompose(params, ct, dct);
 
     // calls digitsG2 NTTs
+#pragma omp parallel for num_threads(OpenFHEParallelControls.GetThreadLimit(digitsG2))
     for (uint32_t d = 0; d < digitsG2; ++d)
         dct[d].SetFormat(Format::EVALUATION);
 
@@ -281,6 +276,7 @@ void RingGSWAccumulatorLMKCDEY::Automorphism(const std::shared_ptr<RingGSWCrypto
 
     SignedDigitDecompose(params, cta, dcta);
 
+#pragma omp parallel for num_threads(OpenFHEParallelControls.GetThreadLimit(digitsG))
     for (uint32_t d = 0; d < digitsG; ++d)
         dcta[d].SetFormat(Format::EVALUATION);
 

src/core/include/lattice/hal/dcrtpoly-interface.h

@@ -1390,13 +1390,18 @@ class DCRTPolyInterface : public ILElement<DerivedType, BigVecType> {
         const std::vector<NativeInteger>& BModq, const std::vector<NativeInteger>& BModqPrecon) = 0;
 
     /**
-   * @brief Convert from Coefficient to CRT or vice versa; calls FFT and inverse
-   * FFT.
+   * @brief Convert from Coefficient to CRT or vice versa; calls FFT and inverse FFT.
    *
    * @warning use @see SetFormat(format) instead
    */
     void SwitchFormat() override = 0;
 
+    /**
+   * @brief Sets format to value without calling FFT. Only use if you know what you're doing.
+   *
+   */
+    virtual void OverrideFormat(const Format f) = 0;
+
     /**
    * @brief Switch modulus and adjust the values
    *