From d8252214ee2bfa727c530d135cb8e25e64972fec Mon Sep 17 00:00:00 2001
From: nindanaoto <matsuoka.kotaro@gmail.com>
Date: Fri, 23 Aug 2024 13:49:57 +0000
Subject: [PATCH 1/2] Fix Xbyak and Randen

---
 thirdparties/randen/vector128.h               |  2 +-
 .../fft_processor_spqliox_aarch64.cpp         | 30 +++++++++++++++++--
 .../fft_processor_spqliox_aarch64.h           |  7 ++++-
 3 files changed, 35 insertions(+), 4 deletions(-)

diff --git a/thirdparties/randen/vector128.h b/thirdparties/randen/vector128.h
index 9a49f6d..b15dd97 100644
--- a/thirdparties/randen/vector128.h
+++ b/thirdparties/randen/vector128.h
@@ -29,7 +29,7 @@
 #define RANDEN_BIG_ENDIAN 1
 #include <altivec.h>
 
-#elif defined(__ARM_NEON) && defined(__ARM_FEATURE_CRYPTO)
+#elif defined(__ARM_NEON) && defined(__ARM_FEATURE_AES)
 
 #define RANDEN_ARM 1
 #include <arm_neon.h>
diff --git a/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.cpp b/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.cpp
index 6b4212e..d897db3 100644
--- a/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.cpp
+++ b/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.cpp
@@ -193,6 +193,17 @@ void FFT_Processor_Spqliox_AArch64::execute_reverse_int(double *res,
     for (size_t i = 0; i < N; i++) res[i] = real_inout[i];
 }
 
+void FFT_Processor_Spqliox_AArch64::execute_reverse_uint(double *res,
+                                                        const uint32_t *a)
+{
+    for (size_t i = 0; i < N; i++) real_inout[i] = (double)a[i];
+
+    ifft_(real_inout, NULL, NULL, &tables_reverse_, real_inout,
+          table_negation_reverse_.trig_tables);
+
+    for (size_t i = 0; i < N; i++) res[i] = real_inout[i];
+}
+
 void FFT_Processor_Spqliox_AArch64::execute_reverse_torus32(double *res,
                                                             const uint32_t *a)
 {
@@ -236,9 +247,24 @@ void FFT_Processor_Spqliox_AArch64::execute_direct_torus32_rescale(
     fft_(dst, sit, send, bla, &tables_direct_, real_inout,
          table_negation_forward_.trig_tables);
     for (int32_t i = 0; i < N; i++)
-        res[i] = uint32_t(int64_t(real_inout[i] / (Δ / 4)));
+        res[i] = uint32_t(int64_t(real_inout[i] / Δ));
 }
 
+void FFT_Processor_Spqliox_AArch64::execute_direct_torus64_rescale(
+    uint64_t *res, const double *a, const double Δ)
+{
+    const double *sit = a;
+    const double *send = a + N;
+    static const double _2sN = double(2) / double(N);
+    const double *bla = &_2sN;
+    double *dst = real_inout;
+    fft_(dst, sit, send, bla, &tables_direct_, real_inout,
+         table_negation_forward_.trig_tables);
+    for (int32_t i = 0; i < N; i++)
+        res[i] = uint64_t(int64_t(real_inout[i] / Δ));
+}
+
+
 void FFT_Processor_Spqliox_AArch64::execute_direct_torus64(uint64_t *res,
                                                            const double *a)
 {
@@ -267,4 +293,4 @@ void FFT_Processor_Spqliox_AArch64::execute_direct_torus64(uint64_t *res,
 FFT_Processor_Spqliox_AArch64::~FFT_Processor_Spqliox_AArch64() {}
 
 thread_local FFT_Processor_Spqliox_AArch64 fftplvl1(TFHEpp::lvl1param::n);
-thread_local FFT_Processor_Spqliox_AArch64 fftplvl2(TFHEpp::lvl2param::n);
\ No newline at end of file
+thread_local FFT_Processor_Spqliox_AArch64 fftplvl2(TFHEpp::lvl2param::n);
diff --git a/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.h b/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.h
index 73a53e8..3041af5 100644
--- a/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.h
+++ b/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.h
@@ -34,6 +34,8 @@ class FFT_Processor_Spqliox_AArch64 {
 
     void execute_reverse_int(double *res, const int32_t *a);
 
+    void execute_reverse_uint(double *res, const uint32_t *a);
+
     void execute_reverse_torus32(double *res, const uint32_t *a);
 
     void execute_direct_torus32(uint32_t *res, const double *a);
@@ -41,6 +43,9 @@ class FFT_Processor_Spqliox_AArch64 {
     void execute_direct_torus32_rescale(uint32_t *res, const double *a,
                                         const double Δ);
 
+    void execute_direct_torus64_rescale(uint64_t *res, const double *a,
+                                        const double Δ);
+
     void execute_reverse_torus64(double *res, const uint64_t *a);
 
     void execute_direct_torus64(uint64_t *res, const double *a);
@@ -49,4 +54,4 @@ class FFT_Processor_Spqliox_AArch64 {
 };
 
 extern thread_local FFT_Processor_Spqliox_AArch64 fftplvl1;
-extern thread_local FFT_Processor_Spqliox_AArch64 fftplvl2;
\ No newline at end of file
+extern thread_local FFT_Processor_Spqliox_AArch64 fftplvl2;

From 5207698deab2b53d9a0e59331526716e679eef86 Mon Sep 17 00:00:00 2001
From: nindanaoto <matsuoka.kotaro@gmail.com>
Date: Fri, 23 Aug 2024 13:50:45 +0000
Subject: [PATCH 2/2] clang formatted

---
 include/mulfft.hpp                            | 21 +++++++++++--------
 .../fft_processor_spqliox_aarch64.cpp         |  3 +--
 2 files changed, 13 insertions(+), 11 deletions(-)

diff --git a/include/mulfft.hpp b/include/mulfft.hpp
index 6b8e43a..50c48ac 100644
--- a/include/mulfft.hpp
+++ b/include/mulfft.hpp
@@ -152,10 +152,10 @@ inline void TwistIFFTUInt(PolynomialInFD<P> &res, const Polynomial<P> &a)
         if constexpr (std::is_same_v<typename P::T, uint32_t>)
             fftplvl1.execute_reverse_uint(res.data(), a.data());
         // if constexpr (std::is_same_v<typename P::T, uint64_t>)
-            // fftplvl1.execute_reverse_torus64(res.data(), a.data());
+        // fftplvl1.execute_reverse_torus64(res.data(), a.data());
     }
     // else if constexpr (std::is_same_v<typename P::T, uint64_t>)
-        // fftplvl2.execute_reverse_torus64(res.data(), a.data());
+    // fftplvl2.execute_reverse_torus64(res.data(), a.data());
     else
         static_assert(false_v<typename P::T>, "Undefined TwistIFFT!");
 }
@@ -316,8 +316,7 @@ inline void PolyMul(Polynomial<P> &res, const Polynomial<P> &a,
 }
 
 template <class P>
-inline void PolyMulRescaleUnsigned(Polynomial<P> &res,
-                                   const Polynomial<P> &a,
+inline void PolyMulRescaleUnsigned(Polynomial<P> &res, const Polynomial<P> &a,
                                    const Polynomial<P> &b)
 {
     // if constexpr (std::is_same_v<typename P::T, uint32_t>) {
@@ -350,16 +349,20 @@ inline void PolyMulNaive(Polynomial<P> &res, const Polynomial<P> &a,
 }
 
 template <class P>
-inline void PolyMulNaieveRescaleUnsigned(Polynomial<P> &res, const Polynomial<P> &a,
-                                 const Polynomial<P> &b)
+inline void PolyMulNaieveRescaleUnsigned(Polynomial<P> &res,
+                                         const Polynomial<P> &a,
+                                         const Polynomial<P> &b)
 {
     for (int i = 0; i < P::n; i++) {
         __int128_t ri = 0;
         for (int j = 0; j <= i; j++)
-            ri += static_cast<__int128_t>(a[j]) * static_cast<__int128_t>(b[i - j]);
+            ri += static_cast<__int128_t>(a[j]) *
+                  static_cast<__int128_t>(b[i - j]);
         for (int j = i + 1; j < P::n; j++)
-            ri -= static_cast<__int128_t>(a[j]) * static_cast<__int128_t>(b[P::n + i - j]);
-        // res[i] = static_cast<typename P::T>((ri) >> (std::numeric_limits<typename P::T>::digits - 3));
+            ri -= static_cast<__int128_t>(a[j]) *
+                  static_cast<__int128_t>(b[P::n + i - j]);
+        // res[i] = static_cast<typename P::T>((ri) >>
+        // (std::numeric_limits<typename P::T>::digits - 3));
         res[i] = static_cast<typename P::T>((ri) >> 29);
     }
 }
diff --git a/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.cpp b/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.cpp
index d897db3..a9466e0 100644
--- a/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.cpp
+++ b/thirdparties/spqliox_aarch64/fft_processor_spqliox_aarch64.cpp
@@ -194,7 +194,7 @@ void FFT_Processor_Spqliox_AArch64::execute_reverse_int(double *res,
 }
 
 void FFT_Processor_Spqliox_AArch64::execute_reverse_uint(double *res,
-                                                        const uint32_t *a)
+                                                         const uint32_t *a)
 {
     for (size_t i = 0; i < N; i++) real_inout[i] = (double)a[i];
 
@@ -264,7 +264,6 @@ void FFT_Processor_Spqliox_AArch64::execute_direct_torus64_rescale(
         res[i] = uint64_t(int64_t(real_inout[i] / Δ));
 }
 
-
 void FFT_Processor_Spqliox_AArch64::execute_direct_torus64(uint64_t *res,
                                                            const double *a)
 {