fix and re-add rainbow

crypto_sign/rainbowIIIc-classic/META.yml

@@ -0,0 +1,18 @@
+name: Rainbow-IIIc-classic
+type: signature
+claimed-nist-level: 3
+length-public-key: 710640
+length-secret-key: 511448
+length-signature: 156
+nistkat-sha256: 199cf313d96a4fbb481c50e568ac0222ec955b3e20551d0fadbb6c5e97bd1ada
+testvectors-sha256: e738081bcc34228184645dd79237daabc89b7ed22172637b6c10f51dd1e417d9
+principal-submitters:
+  - Jintai Ding
+auxiliary-submitters:
+  - Ming-Shing Chen
+  - Albrecht Petzoldt
+  - Dieter Schmidt
+  - Bo-Yin Yang
+implementations:
+    - name: clean
+      version: https://github.com/fast-crypto-lab/rainbow-submission-round2/commit/af826fcb78f6af51a02d0352cff28a9690467bfd

crypto_sign/rainbowIIIc-classic/clean/LICENSE

@@ -0,0 +1,8 @@
+`Software implementation of Rainbow for NIST R2 submission' by Ming-Shing Chen
+
+To the extent possible under law, the person who associated CC0 with
+`Software implementation of Rainbow for NIST R2 submission' has waived all copyright and related or neighboring rights
+to `Software implementation of Rainbow for NIST R2 submission'.
+
+You should have received a copy of the CC0 legalcode along with this
+work.  If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.

crypto_sign/rainbowIIIc-classic/clean/Makefile

@@ -0,0 +1,20 @@
+# This Makefile can be used with GNU Make or BSD Make
+
+LIB=librainbowIIIc-classic_clean.a
+
+HEADERS = api.h blas_comm.h blas.h blas_u32.h gf.h parallel_matrix_op.h rainbow_blas.h rainbow_config.h rainbow.h rainbow_keypair_computation.h rainbow_keypair.h utils_hash.h utils_prng.h 
+OBJECTS =  blas_comm.o parallel_matrix_op.o rainbow.o rainbow_keypair.o rainbow_keypair_computation.o sign.o utils_hash.o utils_prng.o blas_u32.o gf.o
+
+CFLAGS=-O3 -Wall -Wconversion -Wextra -Wpedantic -Wvla -Werror -Wmissing-prototypes -Wredundant-decls -std=c99 -I../../../common $(EXTRAFLAGS)
+
+all: $(LIB)
+
+%.o: %.c $(HEADERS)
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(LIB): $(OBJECTS)
+	$(AR) -r $@ $(OBJECTS)
+
+clean:
+	$(RM) $(OBJECTS)
+	$(RM) $(LIB)

crypto_sign/rainbowIIIc-classic/clean/Makefile.Microsoft_nmake

@@ -0,0 +1,19 @@
+# This Makefile can be used with Microsoft Visual Studio's nmake using the command:
+#    nmake /f Makefile.Microsoft_nmake
+
+LIBRARY=librainbowIIIc-classic_clean.lib
+OBJECTS =  blas_comm.obj parallel_matrix_op.obj rainbow.obj rainbow_keypair.obj rainbow_keypair_computation.obj sign.obj utils_hash.obj utils_prng.obj blas_u32.obj gf.obj
+
+CFLAGS=/nologo /I ..\..\..\common /W4 /WX
+
+all: $(LIBRARY)
+
+# Make sure objects are recompiled if headers change.
+$(OBJECTS): *.h
+
+$(LIBRARY): $(OBJECTS)
+	LIB.EXE /NOLOGO /WX /OUT:$@ $**
+
+clean:
+    -DEL $(OBJECTS)
+    -DEL $(LIBRARY)

crypto_sign/rainbowIIIc-classic/clean/api.h

@@ -0,0 +1,32 @@
+#ifndef PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_API_H
+#define PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_API_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+#define PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_CRYPTO_SECRETKEYBYTES 511448
+#define PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_CRYPTO_PUBLICKEYBYTES 710640
+#define PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_CRYPTO_BYTES 156
+#define PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_CRYPTO_ALGNAME "RAINBOW(256,68,36,36) - classic"
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign_keypair(uint8_t *pk, uint8_t *sk);
+
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign_signature(
+    uint8_t *sig, size_t *siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *sk);
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign_verify(
+    const uint8_t *sig, size_t siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *pk);
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign(uint8_t *sm, size_t *smlen,
+        const uint8_t *m, size_t mlen,
+        const uint8_t *sk);
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign_open(uint8_t *m, size_t *mlen,
+        const uint8_t *sm, size_t smlen,
+        const uint8_t *pk);
+
+
+#endif

crypto_sign/rainbowIIIc-classic/clean/blas.h

@@ -0,0 +1,19 @@
+#ifndef _BLAS_H_
+#define _BLAS_H_
+/// @file blas.h
+/// @brief Defining the implementations for linear algebra functions depending on the machine architecture.
+///
+
+#include "blas_comm.h"
+#include "blas_u32.h"
+#include "rainbow_config.h"
+
+#define gf256v_predicated_add PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_predicated_add_u32
+#define gf256v_add PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_add_u32
+
+
+#define gf256v_mul_scalar PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_scalar_u32
+#define gf256v_madd PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_madd_u32
+
+
+#endif // _BLAS_H_

crypto_sign/rainbowIIIc-classic/clean/blas_comm.c

@@ -0,0 +1,144 @@
+/// @file blas_comm.c
+/// @brief The standard implementations for blas_comm.h
+///
+
+#include "blas_comm.h"
+#include "blas.h"
+#include "gf.h"
+#include "rainbow_config.h"
+
+#include <stdint.h>
+#include <string.h>
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(uint8_t *b, unsigned int _num_byte) {
+    for (size_t i = 0; i < _num_byte; i++) {
+        b[i] = 0;
+    }
+}
+/// @brief get an element from GF(256) vector .
+///
+/// @param[in]  a         - the input vector a.
+/// @param[in]  i         - the index in the vector a.
+/// @return  the value of the element.
+///
+uint8_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(const uint8_t *a, unsigned int i) {
+    return a[i];
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_is_zero(const uint8_t *a, unsigned int _num_byte) {
+    uint8_t r = 0;
+    while (_num_byte--) {
+        r |= a[0];
+        a++;
+    }
+    return (0 == r);
+}
+
+/// polynomial multplication
+/// School boook
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_polymul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int _num) {
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(c, _num * 2 - 1);
+    for (unsigned int i = 0; i < _num; i++) {
+        gf256v_madd(c + i, a, b[i], _num);
+    }
+}
+
+static void gf256mat_prod_ref(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b) {
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(c, n_A_vec_byte);
+    for (unsigned int i = 0; i < n_A_width; i++) {
+        gf256v_madd(c, matA, b[i], n_A_vec_byte);
+        matA += n_A_vec_byte;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_mul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int len_vec) {
+    unsigned int n_vec_byte = len_vec;
+    for (unsigned int k = 0; k < len_vec; k++) {
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(c, n_vec_byte);
+        const uint8_t *bk = b + n_vec_byte * k;
+        for (unsigned int i = 0; i < len_vec; i++) {
+            gf256v_madd(c, a + n_vec_byte * i, bk[i], n_vec_byte);
+        }
+        c += n_vec_byte;
+    }
+}
+
+static unsigned int gf256mat_gauss_elim_ref(uint8_t *mat, unsigned int h, unsigned int w) {
+    unsigned int r8 = 1;
+
+    for (unsigned int i = 0; i < h; i++) {
+        uint8_t *ai = mat + w * i;
+        unsigned int skip_len_align4 = i & ((unsigned int)~0x3);
+
+        for (unsigned int j = i + 1; j < h; j++) {
+            uint8_t *aj = mat + w * j;
+            gf256v_predicated_add(ai + skip_len_align4, !PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256_is_nonzero(ai[i]), aj + skip_len_align4, w - skip_len_align4);
+        }
+        r8 &= PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256_is_nonzero(ai[i]);
+        uint8_t pivot = ai[i];
+        pivot = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256_inv(pivot);
+        gf256v_mul_scalar(ai + skip_len_align4, pivot, w - skip_len_align4);
+        for (unsigned int j = 0; j < h; j++) {
+            if (i == j) {
+                continue;
+            }
+            uint8_t *aj = mat + w * j;
+            gf256v_madd(aj + skip_len_align4, ai + skip_len_align4, aj[i], w - skip_len_align4);
+        }
+    }
+
+    return r8;
+}
+
+static unsigned int gf256mat_solve_linear_eq_ref(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n) {
+    uint8_t mat[64 * 64];
+    for (unsigned int i = 0; i < n; i++) {
+        memcpy(mat + i * (n + 1), inp_mat + i * n, n);
+        mat[i * (n + 1) + n] = c_terms[i];
+    }
+    unsigned int r8 = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_gauss_elim(mat, n, n + 1);
+    for (unsigned int i = 0; i < n; i++) {
+        sol[i] = mat[i * (n + 1) + n];
+    }
+    return r8;
+}
+
+static inline void gf256mat_submat(uint8_t *mat2, unsigned int w2, unsigned int st, const uint8_t *mat, unsigned int w, unsigned int h) {
+    for (unsigned int i = 0; i < h; i++) {
+        for (unsigned int j = 0; j < w2; j++) {
+            mat2[i * w2 + j] = mat[i * w + st + j];
+        }
+    }
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_inv(uint8_t *inv_a, const uint8_t *a, unsigned int H, uint8_t *buffer) {
+    uint8_t *aa = buffer;
+    for (unsigned int i = 0; i < H; i++) {
+        uint8_t *ai = aa + i * 2 * H;
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(ai, 2 * H);
+        gf256v_add(ai, a + i * H, H);
+        ai[H + i] = 1;
+    }
+    unsigned int r8 = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_gauss_elim(aa, H, 2 * H);
+    gf256mat_submat(inv_a, H, H, aa, 2 * H, H);
+    return r8;
+}
+
+
+// choosing the implementations depends on the macros _BLAS_AVX2_ and _BLAS_SSE
+
+#define gf256mat_prod_impl gf256mat_prod_ref
+#define gf256mat_gauss_elim_impl gf256mat_gauss_elim_ref
+#define gf256mat_solve_linear_eq_impl gf256mat_solve_linear_eq_ref
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_prod(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b) {
+    gf256mat_prod_impl(c, matA, n_A_vec_byte, n_A_width, b);
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_gauss_elim(uint8_t *mat, unsigned int h, unsigned int w) {
+    return gf256mat_gauss_elim_impl(mat, h, w);
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_solve_linear_eq(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n) {
+    return gf256mat_solve_linear_eq_impl(sol, inp_mat, c_terms, n);
+}
+

crypto_sign/rainbowIIIc-classic/clean/blas_comm.h

@@ -0,0 +1,90 @@
+#ifndef _BLAS_COMM_H_
+#define _BLAS_COMM_H_
+/// @file blas_comm.h
+/// @brief Common functions for linear algebra.
+///
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+/// @brief set a vector to 0.
+///
+/// @param[in,out]   b      - the vector b.
+/// @param[in]  _num_byte   - number of bytes for the vector b.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(uint8_t *b, unsigned int _num_byte);
+
+/// @brief get an element from GF(256) vector .
+///
+/// @param[in]  a         - the input vector a.
+/// @param[in]  i         - the index in the vector a.
+/// @return  the value of the element.
+///
+uint8_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(const uint8_t *a, unsigned int i);
+
+/// @brief check if a vector is 0.
+///
+/// @param[in]   a          - the vector a.
+/// @param[in]  _num_byte   - number of bytes for the vector a.
+/// @return  1(true) if a is 0. 0(false) else.
+///
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_is_zero(const uint8_t *a, unsigned int _num_byte);
+
+/// @brief polynomial multiplication:  c = a*b
+///
+/// @param[out]   c         - the output polynomial c
+/// @param[in]   a          - the vector a.
+/// @param[in]   b          - the vector b.
+/// @param[in]  _num   - number of elements for the polynomials a and b.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_polymul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int _num);
+
+/// @brief matrix-vector multiplication:  c = matA * b , in GF(256)
+///
+/// @param[out]  c         - the output vector c
+/// @param[in]   matA      - a column-major matrix A.
+/// @param[in]   n_A_vec_byte  - the size of column vectors in bytes.
+/// @param[in]   n_A_width   - the width of matrix A.
+/// @param[in]   b          - the vector b.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_prod(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b);
+
+/// @brief matrix-matrix multiplication:  c = a * b , in GF(256)
+///
+/// @param[out]  c         - the output matrix c
+/// @param[in]   c         - a matrix a.
+/// @param[in]   b         - a matrix b.
+/// @param[in]   len_vec   - the length of column vectors.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_mul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int len_vec);
+
+/// @brief Gauss elimination for a matrix, in GF(256)
+///
+/// @param[in,out]  mat    - the matrix.
+/// @param[in]   h         - the height of the matrix.
+/// @param[in]   w         - the width of the matrix.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_gauss_elim(uint8_t *mat, unsigned int h, unsigned int w);
+
+/// @brief Solving linear equations, in GF(256)
+///
+/// @param[out]  sol       - the solutions.
+/// @param[in]   inp_mat   - the matrix parts of input equations.
+/// @param[in]   c_terms   - the constant terms of the input equations.
+/// @param[in]   n         - the number of equations.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_solve_linear_eq(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n);
+
+/// @brief Computing the inverse matrix, in GF(256)
+///
+/// @param[out]  inv_a     - the output of matrix a.
+/// @param[in]   a         - a matrix a.
+/// @param[in]   H         - height of matrix a, i.e., matrix a is an HxH matrix.
+/// @param[in]   buffer    - The buffer for computations. it has to be as large as 2 input matrixes.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_inv(uint8_t *inv_a, const uint8_t *a, unsigned int H, uint8_t *buffer);
+
+#endif // _BLAS_COMM_H_

crypto_sign/rainbowIIIc-classic/clean/blas_u32.c

@@ -0,0 +1,87 @@
+#include "blas_u32.h"
+#include "gf.h"
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_predicated_add_u32(uint8_t *accu_b, uint8_t predicate, const uint8_t *a, unsigned int _num_byte) {
+    uint32_t pr_u32 = ((uint32_t)0) - ((uint32_t)predicate);
+    uint8_t pr_u8 = pr_u32 & 0xff;
+
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *b_u32 = (uint32_t *)accu_b;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        b_u32[i] ^= (a_u32[i] & pr_u32);
+    }
+
+    a += (n_u32 << 2);
+    accu_b += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_b[i] ^= (a[i] & pr_u8);
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_add_u32(uint8_t *accu_b, const uint8_t *a, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *b_u32 = (uint32_t *)accu_b;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        b_u32[i] ^= a_u32[i];
+    }
+
+    a += (n_u32 << 2);
+    accu_b += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_b[i] ^= a[i];
+    }
+}
+
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_scalar_u32(uint8_t *a, uint8_t b, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *a_u32 = (uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        a_u32[i] = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_u32(a_u32[i], b);
+    }
+
+    union tmp_32 {
+        uint8_t u8[4];
+        uint32_t u32;
+    } t;
+    t.u32 = 0;
+    a += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        t.u8[i] = a[i];
+    }
+    t.u32 = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_u32(t.u32, b);
+    for (unsigned int i = 0; i < rem; i++) {
+        a[i] = t.u8[i];
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_madd_u32(uint8_t *accu_c, const uint8_t *a, uint8_t gf256_b, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *c_u32 = (uint32_t *)accu_c;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        c_u32[i] ^= PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_u32(a_u32[i], gf256_b);
+    }
+
+    union tmp_32 {
+        uint8_t u8[4];
+        uint32_t u32;
+    } t;
+    t.u32 = 0;
+    accu_c += (n_u32 << 2);
+    a += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        t.u8[i] = a[i];
+    }
+    t.u32 = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_u32(t.u32, gf256_b);
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_c[i] ^= t.u8[i];
+    }
+}
+

crypto_sign/rainbowIIIc-classic/clean/blas_u32.h

@@ -0,0 +1,18 @@
+#ifndef _BLAS_U32_H_
+#define _BLAS_U32_H_
+/// @file blas_u32.h
+/// @brief Inlined functions for implementing basic linear algebra functions for uint32 arch.
+///
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_predicated_add_u32(uint8_t *accu_b, uint8_t predicate, const uint8_t *a, unsigned int _num_byte);
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_add_u32(uint8_t *accu_b, const uint8_t *a, unsigned int _num_byte);
+
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_scalar_u32(uint8_t *a, uint8_t b, unsigned int _num_byte);
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_madd_u32(uint8_t *accu_c, const uint8_t *a, uint8_t gf256_b, unsigned int _num_byte);
+
+
+#endif // _BLAS_U32_H_

crypto_sign/rainbowIIIc-classic/clean/gf.c

@@ -0,0 +1,134 @@
+#include "gf.h"
+
+//// gf4 := gf2[x]/x^2+x+1
+static inline uint8_t gf4_mul_2(uint8_t a) {
+    uint8_t r = (uint8_t)(a << 1);
+    r ^= (uint8_t)((a >> 1) * 7);
+    return r;
+}
+
+static inline uint8_t gf4_mul(uint8_t a, uint8_t b) {
+    uint8_t r = (uint8_t)(a * (b & 1));
+    return r ^ (uint8_t)(gf4_mul_2(a) * (b >> 1));
+}
+
+static inline uint8_t gf4_squ(uint8_t a) {
+    return a ^ (a >> 1);
+}
+
+static inline uint32_t gf4v_mul_2_u32(uint32_t a) {
+    uint32_t bit0 = a & 0x55555555;
+    uint32_t bit1 = a & 0xaaaaaaaa;
+    return (bit0 << 1) ^ bit1 ^ (bit1 >> 1);
+}
+
+static inline uint32_t gf4v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t bit0_b = ((uint32_t)0) - ((uint32_t)(b & 1));
+    uint32_t bit1_b = ((uint32_t)0) - ((uint32_t)((b >> 1) & 1));
+    return (a & bit0_b) ^ (bit1_b & gf4v_mul_2_u32(a));
+}
+
+//// gf16 := gf4[y]/y^2+y+x
+static inline uint8_t gf16_mul(uint8_t a, uint8_t b) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = (a >> 2);
+    uint8_t b0 = b & 3;
+    uint8_t b1 = (b >> 2);
+    uint8_t a0b0 = gf4_mul(a0, b0);
+    uint8_t a1b1 = gf4_mul(a1, b1);
+    uint8_t a0b1_a1b0 = gf4_mul(a0 ^ a1, b0 ^ b1) ^ a0b0 ^ a1b1;
+    uint8_t a1b1_x2 = gf4_mul_2(a1b1);
+    return (uint8_t)((a0b1_a1b0 ^ a1b1) << 2 ^ a0b0 ^ a1b1_x2);
+}
+
+static inline uint8_t gf16_squ(uint8_t a) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = (a >> 2);
+    a1 = gf4_squ(a1);
+    uint8_t a1squ_x2 = gf4_mul_2(a1);
+    return (uint8_t)((a1 << 2) ^ a1squ_x2 ^ gf4_squ(a0));
+}
+
+// gf16 := gf4[y]/y^2+y+x
+uint32_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf16v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t axb0 = gf4v_mul_u32(a, b);
+    uint32_t axb1 = gf4v_mul_u32(a, b >> 2);
+    uint32_t a0b1 = (axb1 << 2) & 0xcccccccc;
+    uint32_t a1b1 = axb1 & 0xcccccccc;
+    uint32_t a1b1_2 = a1b1 >> 2;
+
+    return axb0 ^ a0b1 ^ a1b1 ^ gf4v_mul_2_u32(a1b1_2);
+}
+
+uint8_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256_is_nonzero(uint8_t a) {
+    unsigned int a8 = a;
+    unsigned int r = ((unsigned int)0) - a8;
+    r >>= 8;
+    return r & 1;
+}
+
+static inline uint8_t gf4_mul_3(uint8_t a) {
+    uint8_t msk = (uint8_t)((a - 2) >> 1);
+    return (uint8_t)((msk & ((int)a * 3)) | ((~msk) & ((int)a - 1)));
+}
+static inline uint8_t gf16_mul_8(uint8_t a) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = a >> 2;
+    return (uint8_t)((gf4_mul_2(a0 ^ a1) << 2) | gf4_mul_3(a1));
+}
+
+// gf256 := gf16[X]/X^2+X+xy
+static inline uint8_t gf256_mul(uint8_t a, uint8_t b) {
+    uint8_t a0 = a & 15;
+    uint8_t a1 = (a >> 4);
+    uint8_t b0 = b & 15;
+    uint8_t b1 = (b >> 4);
+    uint8_t a0b0 = gf16_mul(a0, b0);
+    uint8_t a1b1 = gf16_mul(a1, b1);
+    uint8_t a0b1_a1b0 = gf16_mul(a0 ^ a1, b0 ^ b1) ^ a0b0 ^ a1b1;
+    uint8_t a1b1_x8 = gf16_mul_8(a1b1);
+    return (uint8_t)((a0b1_a1b0 ^ a1b1) << 4 ^ a0b0 ^ a1b1_x8);
+}
+
+static inline uint8_t gf256_squ(uint8_t a) {
+    uint8_t a0 = a & 15;
+    uint8_t a1 = (a >> 4);
+    a1 = gf16_squ(a1);
+    uint8_t a1squ_x8 = gf16_mul_8(a1);
+    return (uint8_t)((a1 << 4) ^ a1squ_x8 ^ gf16_squ(a0));
+}
+
+uint8_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256_inv(uint8_t a) {
+    // 128+64+32+16+8+4+2 = 254
+    uint8_t a2 = gf256_squ(a);
+    uint8_t a4 = gf256_squ(a2);
+    uint8_t a8 = gf256_squ(a4);
+    uint8_t a4_2 = gf256_mul(a4, a2);
+    uint8_t a8_4_2 = gf256_mul(a4_2, a8);
+    uint8_t a64_ = gf256_squ(a8_4_2);
+    a64_ = gf256_squ(a64_);
+    a64_ = gf256_squ(a64_);
+    uint8_t a64_2 = gf256_mul(a64_, a8_4_2);
+    uint8_t a128_ = gf256_squ(a64_2);
+    return gf256_mul(a2, a128_);
+}
+
+static inline uint32_t gf4v_mul_3_u32(uint32_t a) {
+    uint32_t bit0 = a & 0x55555555;
+    uint32_t bit1 = a & 0xaaaaaaaa;
+    return (bit0 << 1) ^ bit0 ^ (bit1 >> 1);
+}
+static inline uint32_t gf16v_mul_8_u32(uint32_t a) {
+    uint32_t a1 = a & 0xcccccccc;
+    uint32_t a0 = (a << 2) & 0xcccccccc;
+    return gf4v_mul_2_u32(a0 ^ a1) | gf4v_mul_3_u32(a1 >> 2);
+}
+uint32_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t axb0 = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf16v_mul_u32(a, b);
+    uint32_t axb1 = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf16v_mul_u32(a, b >> 4);
+    uint32_t a0b1 = (axb1 << 4) & 0xf0f0f0f0;
+    uint32_t a1b1 = axb1 & 0xf0f0f0f0;
+    uint32_t a1b1_4 = a1b1 >> 4;
+
+    return axb0 ^ a0b1 ^ a1b1 ^ gf16v_mul_8_u32(a1b1_4);
+}

crypto_sign/rainbowIIIc-classic/clean/gf.h

@@ -0,0 +1,19 @@
+#ifndef _GF16_H_
+#define _GF16_H_
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+/// @file gf16.h
+/// @brief Library for arithmetics in GF(16) and GF(256)
+///
+
+uint32_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf16v_mul_u32(uint32_t a, uint8_t b);
+
+
+uint8_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256_is_nonzero(uint8_t a);
+uint8_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256_inv(uint8_t a);
+uint32_t PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_u32(uint32_t a, uint8_t b);
+
+
+#endif // _GF16_H_

crypto_sign/rainbowIIIc-classic/clean/parallel_matrix_op.c

@@ -0,0 +1,183 @@
+///  @file parallel_matrix_op.c
+///  @brief the standard implementations for functions in parallel_matrix_op.h
+///
+///  the standard implementations for functions in parallel_matrix_op.h
+///
+
+#include "parallel_matrix_op.h"
+#include "blas.h"
+#include "blas_comm.h"
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle(UT) matrix.
+///
+/// @param[in]  i_row     - the i-th row in an upper-triangle matrix.
+/// @param[in]  j_col     - the j-th column in an upper-triangle matrix.
+/// @param[in]  dim       - the dimension of the upper-triangle matrix, i.e., an dim x dim matrix.
+/// @return  the corresponding index in an array storage.
+///
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(unsigned int i_row, unsigned int j_col, unsigned int dim) {
+    return (dim + dim - i_row + 1) * i_row / 2 + j_col - i_row;
+}
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle or lower-triangle matrix.
+///
+/// @param[in]  i_row     - the i-th row in a triangle matrix.
+/// @param[in]  j_col     - the j-th column in a triangle matrix.
+/// @param[in]  dim       - the dimension of the triangle matrix, i.e., an dim x dim matrix.
+/// @return    the corresponding index in an array storage.
+///
+static inline unsigned int idx_of_2trimat(unsigned int i_row, unsigned int j_col, unsigned int n_var) {
+    if (i_row > j_col) {
+        return PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(j_col, i_row, n_var);
+    }
+    return PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(i_row, j_col, n_var);
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_UpperTrianglize(unsigned char *btriC, const unsigned char *bA, unsigned int Awidth, unsigned int size_batch) {
+    unsigned char *runningC = btriC;
+    unsigned int Aheight = Awidth;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < i; j++) {
+            unsigned int idx = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(j, i, Aheight);
+            gf256v_add(btriC + idx * size_batch, bA + size_batch * (i * Awidth + j), size_batch);
+        }
+        gf256v_add(runningC, bA + size_batch * (i * Awidth + i), size_batch * (Aheight - i));
+        runningC += size_batch * (Aheight - i);
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Awidth = Bheight;
+    unsigned int Aheight = Awidth;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (k < i) {
+                    continue;
+                }
+                gf256v_madd(bC, &btriA[(k - i) * size_batch], PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+        btriA += (Aheight - i) * size_batch;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_trimatTr_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Aheight = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (i < k) {
+                    continue;
+                }
+                gf256v_madd(bC, &btriA[size_batch * (PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(k, i, Aheight))], PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_2trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Aheight = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (i == k) {
+                    continue;
+                }
+                gf256v_madd(bC, &btriA[size_batch * (idx_of_2trimat(i, k, Aheight))], PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_matTr_madd_gf256(unsigned char *bC, const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Atr_height = Awidth;
+    unsigned int Atr_width = Aheight;
+    for (unsigned int i = 0; i < Atr_height; i++) {
+        for (unsigned int j = 0; j < Atr_width; j++) {
+            gf256v_madd(bC, &bB[j * Bwidth * size_batch], PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(&A_to_tr[size_Acolvec * i], j), size_batch * Bwidth);
+        }
+        bC += size_batch * Bwidth;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_bmatTr_madd_gf256(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    const unsigned char *bA = bA_to_tr;
+    unsigned int Aheight = Awidth_before_tr;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                gf256v_madd(bC, &bA[size_batch * (i + k * Aheight)], PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_mat_madd_gf256(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Awidth = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                gf256v_madd(bC, &bA[k * size_batch], PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+        bA += (Awidth) * size_batch;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_quad_trimat_eval_gf256(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch) {
+    unsigned char tmp[256];
+
+    unsigned char _x[256];
+    for (unsigned int i = 0; i < dim; i++) {
+        _x[i] = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(x, i);
+    }
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(y, size_batch);
+    for (unsigned int i = 0; i < dim; i++) {
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(tmp, size_batch);
+        for (unsigned int j = i; j < dim; j++) {
+            gf256v_madd(tmp, trimat, _x[j], size_batch);
+            trimat += size_batch;
+        }
+        gf256v_madd(y, tmp, _x[i], size_batch);
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_quad_recmat_eval_gf256(unsigned char *z, const unsigned char *y, unsigned int dim_y, const unsigned char *mat,
+        const unsigned char *x, unsigned dim_x, unsigned size_batch) {
+    unsigned char tmp[128];
+
+    unsigned char _x[128];
+    for (unsigned int i = 0; i < dim_x; i++) {
+        _x[i] = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(x, i);
+    }
+    unsigned char _y[128];
+    for (unsigned int i = 0; i < dim_y; i++) {
+        _y[i] = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele(y, i);
+    }
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(z, size_batch);
+    for (unsigned int i = 0; i < dim_y; i++) {
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(tmp, size_batch);
+        for (unsigned int j = 0; j < dim_x; j++) {
+            gf256v_madd(tmp, mat, _x[j], size_batch);
+            mat += size_batch;
+        }
+        gf256v_madd(z, tmp, _y[i], size_batch);
+    }
+}
+

crypto_sign/rainbowIIIc-classic/clean/parallel_matrix_op.h

@@ -0,0 +1,260 @@
+#ifndef _P_MATRIX_OP_H_
+#define _P_MATRIX_OP_H_
+///  @file  parallel_matrix_op.h
+///  @brief Librarys for operations of batched matrixes.
+///
+///
+
+////////////////  Section:  triangle matrix <-> rectangle matrix   ///////////////////////////////////
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle(UT) matrix.
+///
+/// @param[in]  i_row     - the i-th row in an upper-triangle matrix.
+/// @param[in]  j_col     - the j-th column in an upper-triangle matrix.
+/// @param[in]  dim       - the dimension of the upper-triangle matrix, i.e., an dim x dim matrix.
+/// @return  the corresponding index in an array storage.
+///
+unsigned int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(unsigned int i_row, unsigned int j_col, unsigned int dim);
+
+///
+/// @brief  Upper trianglize a rectangle matrix to the corresponding upper-trangle matrix.
+///
+/// @param[out]  btriC    - the batched upper-trianglized matrix C.
+/// @param[in]   bA       - a batched retangle matrix A.
+/// @param[in]   bwidth   - the width of the batched matrix A, i.e., A is a Awidth x Awidth matrix.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_UpperTrianglize(unsigned char *btriC, const unsigned char *bA, unsigned int Awidth, unsigned int size_batch);
+
+////////////////////  Section:  matrix multiplications  ///////////////////////////////
+
+///
+/// @brief  bC += btriA * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA^Tr * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. A will be transposed while multiplying.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_trimatTr_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA^Tr * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A, which will be transposed while multiplying.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_trimatTr_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC +=  (btriA + btriA^Tr) *B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. The operand for multiplication is (btriA + btriA^Tr).
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_2trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC +=  (btriA + btriA^Tr) *B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. The operand for multiplication is (btriA + btriA^Tr).
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_2trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += A^Tr * bB  , in GF(16)
+///
+/// @param[out]  bC           - the batched matrix C.
+/// @param[in]   A_to_tr      - a column-major matrix A. The operand for multiplication is A^Tr.
+/// @param[in]   Aheight      - the height of A.
+/// @param[in]   size_Acolvec    - the size of a column vector in A.
+/// @param[in]   Awidth           - the width of A.
+/// @param[in]   bB          - a batched matrix B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_matTr_madd_gf16(unsigned char *bC,
+        const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += A^Tr * bB  , in GF(256)
+///
+/// @param[out]  bC           - the batched matrix C.
+/// @param[in]   A_to_tr      - a column-major matrix A. The operand for multiplication is A^Tr.
+/// @param[in]   Aheight      - the height of A.
+/// @param[in]   size_Acolvec    - the size of a column vector in A.
+/// @param[in]   Awidth           - the width of A.
+/// @param[in]   bB          - a batched matrix B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_matTr_madd_gf256(unsigned char *bC,
+        const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA^Tr * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA_to_tr   - a batched matrix A. The operand for multiplication is (bA^Tr).
+/// @param[in]   Awidth_befor_tr     - the width of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_bmatTr_madd_gf16(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA^Tr * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA_to_tr   - a batched matrix A. The operand for multiplication is (bA^Tr).
+/// @param[in]   Awidth_befor_tr     - the width of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_bmatTr_madd_gf256(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA         - a batched matrix A.
+/// @param[in]   Aheigh     - the height of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_mat_madd_gf16(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA         - a batched matrix A.
+/// @param[in]   Aheigh     - the height of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_mat_madd_gf256(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+////////////////////  Section: "quadratric" matrix evaluation  ///////////////////////////////
+
+///
+/// @brief  y =  x^Tr * trimat * x  , in GF(16)
+///
+/// @param[out]  y          - the returned batched element y.
+/// @param[in]   trimat     - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim        - the dimension of matrix trimat (and x).
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_quad_trimat_eval_gf16(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch);
+
+///
+/// @brief  y =  x^Tr * trimat * x  , in GF(256)
+///
+/// @param[out]  y          - the returned batched element y.
+/// @param[in]   trimat     - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim        - the dimension of matrix trimat (and x).
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_quad_trimat_eval_gf256(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch);
+
+///
+/// @brief  z =  y^Tr * mat * x  , in GF(16)
+///
+/// @param[out]  z          - the returned batched element z.
+/// @param[in]   y          - an input vector y.
+/// @param[in]   dim_y      - the length of y.
+/// @param[in]   mat        - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim_x        - the length of x.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_quad_recmat_eval_gf16(unsigned char *z, const unsigned char *y, unsigned int dim_y,
+        const unsigned char *mat, const unsigned char *x, unsigned int dim_x, unsigned int size_batch);
+
+///
+/// @brief  z =  y^Tr * mat * x  , in GF(256)
+///
+/// @param[out]  z          - the returned batched element z.
+/// @param[in]   y          - an input vector y.
+/// @param[in]   dim_y      - the length of y.
+/// @param[in]   mat        - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim_x        - the length of x.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_quad_recmat_eval_gf256(unsigned char *z, const unsigned char *y, unsigned int dim_y,
+        const unsigned char *mat, const unsigned char *x, unsigned int dim_x, unsigned int size_batch);
+
+#endif // _P_MATRIX_OP_H_

crypto_sign/rainbowIIIc-classic/clean/rainbow.c

@@ -0,0 +1,169 @@
+/// @file rainbow.c
+/// @brief The standard implementations for functions in rainbow.h
+///
+
+#include "rainbow.h"
+#include "blas.h"
+#include "rainbow_blas.h"
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+#include "utils_hash.h"
+#include "utils_prng.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#define MAX_ATTEMPT_FRMAT 128
+#define _MAX_O ((_O1 > _O2) ? _O1 : _O2)
+#define _MAX_O_BYTE ((_O1_BYTE > _O2_BYTE) ? _O1_BYTE : _O2_BYTE)
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_rainbow_sign(uint8_t *signature, const sk_t *sk, const uint8_t *_digest) {
+    uint8_t mat_l1[_O1 * _O1_BYTE];
+    uint8_t mat_l2[_O2 * _O2_BYTE];
+    uint8_t mat_buffer[2 * _MAX_O * _MAX_O_BYTE];
+
+    // setup PRNG
+    prng_t prng_sign;
+    uint8_t prng_preseed[LEN_SKSEED + _HASH_LEN];
+    memcpy(prng_preseed, sk->sk_seed, LEN_SKSEED);
+    memcpy(prng_preseed + LEN_SKSEED, _digest, _HASH_LEN); // prng_preseed = sk_seed || digest
+    uint8_t prng_seed[_HASH_LEN];
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(prng_seed, _HASH_LEN, prng_preseed, _HASH_LEN + LEN_SKSEED);
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_set(&prng_sign, prng_seed, _HASH_LEN); // seed = H( sk_seed || digest )
+    for (unsigned int i = 0; i < LEN_SKSEED + _HASH_LEN; i++) {
+        prng_preseed[i] ^= prng_preseed[i]; // clean
+    }
+    for (unsigned int i = 0; i < _HASH_LEN; i++) {
+        prng_seed[i] ^= prng_seed[i]; // clean
+    }
+
+    // roll vinegars.
+    uint8_t vinegar[_V1_BYTE];
+    unsigned int n_attempt = 0;
+    unsigned int l1_succ = 0;
+    while (!l1_succ) {
+        if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+            break;
+        }
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(&prng_sign, vinegar, _V1_BYTE);   // generating vinegars
+        gfmat_prod(mat_l1, sk->l1_F2, _O1 * _O1_BYTE, _V1, vinegar); // generating the linear equations for layer 1
+        l1_succ = gfmat_inv(mat_l1, mat_l1, _O1, mat_buffer);        // check if the linear equation solvable
+        n_attempt++;
+    }
+
+    // Given the vinegars, pre-compute variables needed for layer 2
+    uint8_t r_l1_F1[_O1_BYTE] = {0};
+    uint8_t r_l2_F1[_O2_BYTE] = {0};
+    batch_quad_trimat_eval(r_l1_F1, sk->l1_F1, vinegar, _V1, _O1_BYTE);
+    batch_quad_trimat_eval(r_l2_F1, sk->l2_F1, vinegar, _V1, _O2_BYTE);
+    uint8_t mat_l2_F3[_O2 * _O2_BYTE];
+    uint8_t mat_l2_F2[_O1 * _O2_BYTE];
+    gfmat_prod(mat_l2_F3, sk->l2_F3, _O2 * _O2_BYTE, _V1, vinegar);
+    gfmat_prod(mat_l2_F2, sk->l2_F2, _O1 * _O2_BYTE, _V1, vinegar);
+
+    // Some local variables.
+    uint8_t _z[_PUB_M_BYTE];
+    uint8_t y[_PUB_M_BYTE];
+    uint8_t *x_v1 = vinegar;
+    uint8_t x_o1[_O1_BYTE];
+    uint8_t x_o2[_O1_BYTE];
+
+    uint8_t digest_salt[_HASH_LEN + _SALT_BYTE];
+    memcpy(digest_salt, _digest, _HASH_LEN);
+    uint8_t *salt = digest_salt + _HASH_LEN;
+
+    uint8_t temp_o[_MAX_O_BYTE + 32] = {0};
+    unsigned int succ = 0;
+    while (!succ) {
+        if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+            break;
+        }
+        // The computation:  H(digest||salt)  -->   z   --S-->   y  --C-map-->   x   --T-->   w
+
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(&prng_sign, salt, _SALT_BYTE);                         // roll the salt
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(_z, _PUB_M_BYTE, digest_salt, _HASH_LEN + _SALT_BYTE); // H(digest||salt)
+
+        //  y = S^-1 * z
+        memcpy(y, _z, _PUB_M_BYTE); // identity part of S
+        gfmat_prod(temp_o, sk->s1, _O1_BYTE, _O2, _z + _O1_BYTE);
+        gf256v_add(y, temp_o, _O1_BYTE);
+
+        // Central Map:
+        // layer 1: calculate x_o1
+        memcpy(temp_o, r_l1_F1, _O1_BYTE);
+        gf256v_add(temp_o, y, _O1_BYTE);
+        gfmat_prod(x_o1, mat_l1, _O1_BYTE, _O1, temp_o);
+
+        // layer 2: calculate x_o2
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_set_zero(temp_o, _O2_BYTE);
+        gfmat_prod(temp_o, mat_l2_F2, _O2_BYTE, _O1, x_o1);             // F2
+        batch_quad_trimat_eval(mat_l2, sk->l2_F5, x_o1, _O1, _O2_BYTE); // F5
+        gf256v_add(temp_o, mat_l2, _O2_BYTE);
+        gf256v_add(temp_o, r_l2_F1, _O2_BYTE); // F1
+        gf256v_add(temp_o, y + _O1_BYTE, _O2_BYTE);
+
+        // generate the linear equations of the 2nd layer
+        gfmat_prod(mat_l2, sk->l2_F6, _O2 * _O2_BYTE, _O1, x_o1); // F6
+        gf256v_add(mat_l2, mat_l2_F3, _O2 * _O2_BYTE);            // F3
+        succ = gfmat_inv(mat_l2, mat_l2, _O2, mat_buffer);
+        gfmat_prod(x_o2, mat_l2, _O2_BYTE, _O2, temp_o); // solve l2 eqs
+
+        n_attempt++;
+    };
+    //  w = T^-1 * y
+    uint8_t w[_PUB_N_BYTE];
+    // identity part of T.
+    memcpy(w, x_v1, _V1_BYTE);
+    memcpy(w + _V1_BYTE, x_o1, _O1_BYTE);
+    memcpy(w + _V2_BYTE, x_o2, _O2_BYTE);
+    // Computing the t1 part.
+    gfmat_prod(y, sk->t1, _V1_BYTE, _O1, x_o1);
+    gf256v_add(w, y, _V1_BYTE);
+    // Computing the t4 part.
+    gfmat_prod(y, sk->t4, _V1_BYTE, _O2, x_o2);
+    gf256v_add(w, y, _V1_BYTE);
+    // Computing the t3 part.
+    gfmat_prod(y, sk->t3, _O1_BYTE, _O2, x_o2);
+    gf256v_add(w + _V1_BYTE, y, _O1_BYTE);
+
+    memset(signature, 0, _SIGNATURE_BYTE); // set the output 0
+    // clean
+    memset(&prng_sign, 0, sizeof(prng_t));
+    memset(vinegar, 0, _V1_BYTE);
+    memset(r_l1_F1, 0, _O1_BYTE);
+    memset(r_l2_F1, 0, _O2_BYTE);
+    memset(_z, 0, _PUB_M_BYTE);
+    memset(y, 0, _PUB_M_BYTE);
+    memset(x_o1, 0, _O1_BYTE);
+    memset(x_o2, 0, _O2_BYTE);
+    memset(temp_o, 0, sizeof(temp_o));
+
+    // return: copy w and salt to the signature.
+    if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+        return -1;
+    }
+    gf256v_add(signature, w, _PUB_N_BYTE);
+    gf256v_add(signature + _PUB_N_BYTE, salt, _SALT_BYTE);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_rainbow_verify(const uint8_t *digest, const uint8_t *signature, const pk_t *pk) {
+    unsigned char digest_ck[_PUB_M_BYTE];
+    // public_map( digest_ck , pk , signature ); Evaluating the quadratic public polynomials.
+    batch_quad_trimat_eval(digest_ck, pk->pk, signature, _PUB_N, _PUB_M_BYTE);
+
+    unsigned char correct[_PUB_M_BYTE];
+    unsigned char digest_salt[_HASH_LEN + _SALT_BYTE];
+    memcpy(digest_salt, digest, _HASH_LEN);
+    memcpy(digest_salt + _HASH_LEN, signature + _PUB_N_BYTE, _SALT_BYTE);
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(correct, _PUB_M_BYTE, digest_salt, _HASH_LEN + _SALT_BYTE); // H( digest || salt )
+
+    // check consistancy.
+    unsigned char cc = 0;
+    for (unsigned int i = 0; i < _PUB_M_BYTE; i++) {
+        cc |= (digest_ck[i] ^ correct[i]);
+    }
+    return (0 == cc) ? 0 : -1;
+}
+
+

crypto_sign/rainbowIIIc-classic/clean/rainbow.h

@@ -0,0 +1,33 @@
+#ifndef _RAINBOW_H_
+#define _RAINBOW_H_
+/// @file rainbow.h
+/// @brief APIs for rainbow.
+///
+
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+
+#include <stdint.h>
+
+///
+/// @brief Signing function for classical secret key.
+///
+/// @param[out] signature - the signature.
+/// @param[in]  sk        - the secret key.
+/// @param[in]  digest    - the digest.
+///
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_rainbow_sign(uint8_t *signature, const sk_t *sk, const uint8_t *digest);
+
+///
+/// @brief Verifying function.
+///
+/// @param[in]  digest    - the digest.
+/// @param[in]  signature - the signature.
+/// @param[in]  pk        - the public key.
+/// @return 0 for successful verified. -1 for failed verification.
+///
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_rainbow_verify(const uint8_t *digest, const uint8_t *signature, const pk_t *pk);
+
+
+
+#endif // _RAINBOW_H_

crypto_sign/rainbowIIIc-classic/clean/rainbow_blas.h

@@ -0,0 +1,31 @@
+#ifndef _RAINBOW_BLAS_H_
+#define _RAINBOW_BLAS_H_
+/// @file rainbow_blas.h
+/// @brief Defining the functions used in rainbow.c acconding to the definitions in rainbow_config.h
+///
+///  Defining the functions used in rainbow.c acconding to the definitions in rainbow_config.h
+
+#include "blas.h"
+#include "parallel_matrix_op.h"
+#include "rainbow_config.h"
+
+
+#define gfv_get_ele PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_get_ele
+#define gfv_mul_scalar PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_mul_scalar
+#define gfv_madd PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256v_madd
+
+#define gfmat_prod PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_prod
+#define gfmat_inv PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_gf256mat_inv
+
+#define batch_trimat_madd PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_trimat_madd_gf256
+#define batch_trimatTr_madd PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_trimatTr_madd_gf256
+#define batch_2trimat_madd PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_2trimat_madd_gf256
+#define batch_matTr_madd PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_matTr_madd_gf256
+#define batch_bmatTr_madd PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_bmatTr_madd_gf256
+#define batch_mat_madd PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_mat_madd_gf256
+
+#define batch_quad_trimat_eval PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_quad_trimat_eval_gf256
+#define batch_quad_recmat_eval PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_batch_quad_recmat_eval_gf256
+
+
+#endif // _RAINBOW_BLAS_H_

crypto_sign/rainbowIIIc-classic/clean/rainbow_config.h

@@ -0,0 +1,46 @@
+#ifndef _H_RAINBOW_CONFIG_H_
+#define _H_RAINBOW_CONFIG_H_
+
+/// @file rainbow_config.h
+/// @brief Defining the parameters of the Rainbow and the corresponding constants.
+///
+
+#define _GFSIZE 256
+#define _V1 68
+#define _O1 36
+#define _O2 36
+#define _HASH_LEN 48
+
+
+#define _V2 ((_V1) + (_O1))
+
+/// size of N, in # of gf elements.
+#define _PUB_N (_V1 + _O1 + _O2)
+
+/// size of M, in # gf elements.
+#define _PUB_M (_O1 + _O2)
+
+/// size of variables, in # bytes.
+
+// GF256
+#define _V1_BYTE (_V1)
+#define _V2_BYTE (_V2)
+#define _O1_BYTE (_O1)
+#define _O2_BYTE (_O2)
+#define _PUB_N_BYTE (_PUB_N)
+#define _PUB_M_BYTE (_PUB_M)
+
+
+/// length of seed for public key, in # bytes
+#define LEN_PKSEED 32
+
+/// length of seed for secret key, in # bytes
+#define LEN_SKSEED 32
+
+/// length of salt for a signature, in # bytes
+#define _SALT_BYTE 16
+
+/// length of a signature
+#define _SIGNATURE_BYTE (_PUB_N_BYTE + _SALT_BYTE)
+
+#endif //  _H_RAINBOW_CONFIG_H_

crypto_sign/rainbowIIIc-classic/clean/rainbow_keypair.c

@@ -0,0 +1,126 @@
+/// @file rainbow_keypair.c
+/// @brief implementations of functions in rainbow_keypair.h
+///
+
+#include "rainbow_keypair.h"
+#include "blas.h"
+#include "blas_comm.h"
+#include "rainbow_blas.h"
+#include "rainbow_keypair_computation.h"
+#include "utils_prng.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+static void generate_S_T(unsigned char *s_and_t, prng_t *prng0) {
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, s_and_t, _O1_BYTE * _O2); // S1
+    s_and_t += _O1_BYTE * _O2;
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, s_and_t, _V1_BYTE * _O1); // T1
+    s_and_t += _V1_BYTE * _O1;
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, s_and_t, _V1_BYTE * _O2); // T2
+    s_and_t += _V1_BYTE * _O2;
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, s_and_t, _O1_BYTE * _O2); // T3
+}
+
+static unsigned int generate_l1_F12(unsigned char *sk, prng_t *prng0) {
+    unsigned int n_byte_generated = 0;
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, sk, _O1_BYTE * N_TRIANGLE_TERMS(_V1)); // l1_F1
+    sk += _O1_BYTE * N_TRIANGLE_TERMS(_V1);
+    n_byte_generated += _O1_BYTE * N_TRIANGLE_TERMS(_V1);
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, sk, _O1_BYTE * _V1 * _O1); // l1_F2
+    n_byte_generated += _O1_BYTE * _V1 * _O1;
+    return n_byte_generated;
+}
+
+static unsigned int generate_l2_F12356(unsigned char *sk, prng_t *prng0) {
+    unsigned int n_byte_generated = 0;
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * N_TRIANGLE_TERMS(_V1)); // l2_F1
+    sk += _O2_BYTE * N_TRIANGLE_TERMS(_V1);
+    n_byte_generated += _O2_BYTE * N_TRIANGLE_TERMS(_V1);
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _V1 * _O1); // l2_F2
+    sk += _O2_BYTE * _V1 * _O1;
+    n_byte_generated += _O2_BYTE * _V1 * _O1;
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _V1 * _O2); // l2_F3
+    sk += _O2_BYTE * _V1 * _O1;
+    n_byte_generated += _O2_BYTE * _V1 * _O1;
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * N_TRIANGLE_TERMS(_O1)); // l2_F5
+    sk += _O2_BYTE * N_TRIANGLE_TERMS(_O1);
+    n_byte_generated += _O2_BYTE * N_TRIANGLE_TERMS(_O1);
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _O1 * _O2); // l2_F6
+    n_byte_generated += _O2_BYTE * _O1 * _O2;
+
+    return n_byte_generated;
+}
+
+static void generate_B1_B2(unsigned char *sk, prng_t *prng0) {
+    sk += generate_l1_F12(sk, prng0);
+    generate_l2_F12356(sk, prng0);
+}
+
+static void calculate_t4(unsigned char *t2_to_t4, const unsigned char *t1, const unsigned char *t3) {
+    //  t4 = T_sk.t1 * T_sk.t3 - T_sk.t2
+    unsigned char temp[_V1_BYTE + 32];
+    unsigned char *t4 = t2_to_t4;
+    for (unsigned int i = 0; i < _O2; i++) { /// t3 width
+        gfmat_prod(temp, t1, _V1_BYTE, _O1, t3);
+        gf256v_add(t4, temp, _V1_BYTE);
+        t4 += _V1_BYTE;
+        t3 += _O1_BYTE;
+    }
+}
+
+static void obsfucate_l1_polys(unsigned char *l1_polys, const unsigned char *l2_polys, unsigned int n_terms, const unsigned char *s1) {
+    unsigned char temp[_O1_BYTE + 32];
+    while (n_terms--) {
+        gfmat_prod(temp, s1, _O1_BYTE, _O2, l2_polys);
+        gf256v_add(l1_polys, temp, _O1_BYTE);
+        l1_polys += _O1_BYTE;
+        l2_polys += _O2_BYTE;
+    }
+}
+
+///////////////////  Classic //////////////////////////////////
+
+static void _generate_secretkey(sk_t *sk, const unsigned char *sk_seed) {
+    memcpy(sk->sk_seed, sk_seed, LEN_SKSEED);
+
+    // set up prng
+    prng_t prng0;
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_set(&prng0, sk_seed, LEN_SKSEED);
+
+    // generating secret key with prng.
+    generate_S_T(sk->s1, &prng0);
+    generate_B1_B2(sk->l1_F1, &prng0);
+
+    // clean prng
+    memset(&prng0, 0, sizeof(prng_t));
+}
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_generate_keypair(pk_t *rpk, sk_t *sk, const unsigned char *sk_seed) {
+    _generate_secretkey(sk, sk_seed);
+
+    // set up a temporary structure ext_cpk_t for calculating public key.
+    ext_cpk_t pk;
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_calculate_Q_from_F(&pk, sk, sk); // compute the public key in ext_cpk_t format.
+    calculate_t4(sk->t4, sk->t1, sk->t3);
+
+    obsfucate_l1_polys(pk.l1_Q1, pk.l2_Q1, N_TRIANGLE_TERMS(_V1), sk->s1);
+    obsfucate_l1_polys(pk.l1_Q2, pk.l2_Q2, _V1 * _O1, sk->s1);
+    obsfucate_l1_polys(pk.l1_Q3, pk.l2_Q3, _V1 * _O2, sk->s1);
+    obsfucate_l1_polys(pk.l1_Q5, pk.l2_Q5, N_TRIANGLE_TERMS(_O1), sk->s1);
+    obsfucate_l1_polys(pk.l1_Q6, pk.l2_Q6, _O1 * _O2, sk->s1);
+    obsfucate_l1_polys(pk.l1_Q9, pk.l2_Q9, N_TRIANGLE_TERMS(_O2), sk->s1);
+    // so far, the pk contains the full pk but in ext_cpk_t format.
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_extcpk_to_pk(rpk, &pk); // convert the public key from ext_cpk_t to pk_t.
+}
+
+
+

crypto_sign/rainbowIIIc-classic/clean/rainbow_keypair.h

@@ -0,0 +1,61 @@
+#ifndef _RAINBOW_KEYPAIR_H_
+#define _RAINBOW_KEYPAIR_H_
+/// @file rainbow_keypair.h
+/// @brief Formats of key pairs and functions for generating key pairs.
+/// Formats of key pairs and functions for generating key pairs.
+///
+
+#include "rainbow_config.h"
+
+#define N_TRIANGLE_TERMS(n_var) ((n_var) * ((n_var) + 1) / 2)
+
+/// @brief public key for classic rainbow
+///
+///  public key for classic rainbow
+///
+typedef struct rainbow_publickey {
+    unsigned char pk[(_PUB_M_BYTE)*N_TRIANGLE_TERMS(_PUB_N)];
+} pk_t;
+
+/// @brief secret key for classic rainbow
+///
+/// secret key for classic rainbow
+///
+typedef struct rainbow_secretkey {
+    ///
+    /// seed for generating secret key.
+    /// Generating S, T, and F for classic rainbow.
+    /// Generating S and T only for cyclic rainbow.
+    unsigned char sk_seed[LEN_SKSEED];
+
+    unsigned char s1[_O1_BYTE * _O2]; ///< part of S map
+    unsigned char t1[_V1_BYTE * _O1]; ///< part of T map
+    unsigned char t4[_V1_BYTE * _O2]; ///< part of T map
+    unsigned char t3[_O1_BYTE * _O2]; ///< part of T map
+
+    unsigned char l1_F1[_O1_BYTE * N_TRIANGLE_TERMS(_V1)]; ///< part of C-map, F1, Layer1
+    unsigned char l1_F2[_O1_BYTE * _V1 * _O1];             ///< part of C-map, F2, Layer1
+
+    unsigned char l2_F1[_O2_BYTE * N_TRIANGLE_TERMS(_V1)]; ///< part of C-map, F1, Layer2
+    unsigned char l2_F2[_O2_BYTE * _V1 * _O1];             ///< part of C-map, F2, Layer2
+
+    unsigned char l2_F3[_O2_BYTE * _V1 * _O2];             ///< part of C-map, F3, Layer2
+    unsigned char l2_F5[_O2_BYTE * N_TRIANGLE_TERMS(_O1)]; ///< part of C-map, F5, Layer2
+    unsigned char l2_F6[_O2_BYTE * _O1 * _O2];             ///< part of C-map, F6, Layer2
+} sk_t;
+
+
+///
+/// @brief Generate key pairs for classic rainbow.
+///
+/// @param[out] pk        - the public key.
+/// @param[out] sk        - the secret key.
+/// @param[in]  sk_seed   - seed for generating the secret key.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_generate_keypair(pk_t *pk, sk_t *sk, const unsigned char *sk_seed);
+
+
+
+
+
+#endif //  _RAINBOW_KEYPAIR_H_

crypto_sign/rainbowIIIc-classic/clean/rainbow_keypair_computation.c

@@ -0,0 +1,189 @@
+/// @file rainbow_keypair_computation.c
+/// @brief Implementations for functions in rainbow_keypair_computation.h
+///
+
+#include "rainbow_keypair_computation.h"
+#include "blas.h"
+#include "blas_comm.h"
+#include "rainbow_blas.h"
+#include "rainbow_keypair.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_extcpk_to_pk(pk_t *pk, const ext_cpk_t *cpk) {
+    const unsigned char *idx_l1 = cpk->l1_Q1;
+    const unsigned char *idx_l2 = cpk->l2_Q1;
+    for (unsigned int i = 0; i < _V1; i++) {
+        for (unsigned int j = i; j < _V1; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q2;
+    idx_l2 = cpk->l2_Q2;
+    for (unsigned int i = 0; i < _V1; i++) {
+        for (unsigned int j = _V1; j < _V1 + _O1; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q3;
+    idx_l2 = cpk->l2_Q3;
+    for (unsigned int i = 0; i < _V1; i++) {
+        for (unsigned int j = _V1 + _O1; j < _PUB_N; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q5;
+    idx_l2 = cpk->l2_Q5;
+    for (unsigned int i = _V1; i < _V1 + _O1; i++) {
+        for (unsigned int j = i; j < _V1 + _O1; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q6;
+    idx_l2 = cpk->l2_Q6;
+    for (unsigned int i = _V1; i < _V1 + _O1; i++) {
+        for (unsigned int j = _V1 + _O1; j < _PUB_N; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q9;
+    idx_l2 = cpk->l2_Q9;
+    for (unsigned int i = _V1 + _O1; i < _PUB_N; i++) {
+        for (unsigned int j = i; j < _PUB_N; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+}
+
+static void calculate_Q_from_F_ref(ext_cpk_t *Qs, const sk_t *Fs, const sk_t *Ts) {
+    /*
+          Layer 1
+          Computing :
+          Q_pk.l1_F1s[i] = F_sk.l1_F1s[i]
+
+          Q_pk.l1_F2s[i] = (F1* T1 + F2) + F1tr * t1
+          Q_pk.l1_F5s[i] = UT( T1tr* (F1 * T1 + F2) )
+      */
+    const unsigned char *t2 = Ts->t4;
+
+    memcpy(Qs->l1_Q1, Fs->l1_F1, _O1_BYTE * N_TRIANGLE_TERMS(_V1));
+
+    memcpy(Qs->l1_Q2, Fs->l1_F2, _O1_BYTE * _V1 * _O1);
+    batch_trimat_madd(Qs->l1_Q2, Fs->l1_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O1_BYTE); // F1*T1 + F2
+
+    memset(Qs->l1_Q3, 0, _O1_BYTE * _V1 * _O2);
+    memset(Qs->l1_Q5, 0, _O1_BYTE * N_TRIANGLE_TERMS(_O1));
+    memset(Qs->l1_Q6, 0, _O1_BYTE * _O1 * _O2);
+    memset(Qs->l1_Q9, 0, _O1_BYTE * N_TRIANGLE_TERMS(_O2));
+
+    // l1_Q5 : _O1_BYTE * _O1 * _O1
+    // l1_Q9 : _O1_BYTE * _O2 * _O2
+    // l2_Q5 : _O2_BYTE * _V1 * _O1
+    // l2_Q9 : _O2_BYTE * _V1 * _O2
+
+    unsigned char tempQ[_O1_BYTE * _O1 * _O1 + 32];
+
+    memset(tempQ, 0, _O1_BYTE * _O1 * _O1);                                        // l1_Q5
+    batch_matTr_madd(tempQ, Ts->t1, _V1, _V1_BYTE, _O1, Qs->l1_Q2, _O1, _O1_BYTE); // t1_tr*(F1*T1 + F2)
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_UpperTrianglize(Qs->l1_Q5, tempQ, _O1, _O1_BYTE);            // UT( ... )   // Q5
+
+    batch_trimatTr_madd(Qs->l1_Q2, Fs->l1_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O1_BYTE); // Q2
+    /*
+          Computing:
+          F1_T2     = F1 * t2
+          F2_T3     = F2 * t3
+          F1_F1T_T2 + F2_T3 = F1_T2 + F2_T3 + F1tr * t2
+          Q_pk.l1_F3s[i] =         F1_F1T_T2 + F2_T3
+          Q_pk.l1_F6s[i] = T1tr* ( F1_F1T_T2 + F2_T3 ) + F2tr * t2
+          Q_pk.l1_F9s[i] = UT( T2tr* ( F1_T2 + F2_T3 ) )
+      */
+    batch_trimat_madd(Qs->l1_Q3, Fs->l1_F1, t2, _V1, _V1_BYTE, _O2, _O1_BYTE);       // F1*T2
+    batch_mat_madd(Qs->l1_Q3, Fs->l1_F2, _V1, Ts->t3, _O1, _O1_BYTE, _O2, _O1_BYTE); // F1_T2 + F2_T3
+
+    memset(tempQ, 0, _O1_BYTE * _O2 * _O2);                                    // l1_Q9
+    batch_matTr_madd(tempQ, t2, _V1, _V1_BYTE, _O2, Qs->l1_Q3, _O2, _O1_BYTE); // T2tr * ( F1_T2 + F2_T3 )
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_UpperTrianglize(Qs->l1_Q9, tempQ, _O2, _O1_BYTE);        // Q9
+
+    batch_trimatTr_madd(Qs->l1_Q3, Fs->l1_F1, t2, _V1, _V1_BYTE, _O2, _O1_BYTE); // F1_F1T_T2 + F2_T3  // Q3
+
+    batch_bmatTr_madd(Qs->l1_Q6, Fs->l1_F2, _O1, t2, _V1, _V1_BYTE, _O2, _O1_BYTE);    // F2tr*T2
+    batch_matTr_madd(Qs->l1_Q6, Ts->t1, _V1, _V1_BYTE, _O1, Qs->l1_Q3, _O2, _O1_BYTE); // Q6
+
+    /*
+          layer 2
+          Computing:
+          Q1 = F1
+          Q2 = F1_F1T*T1 + F2
+          Q5 = UT( T1tr( F1*T1 + F2 )  + F5 )
+      */
+    memcpy(Qs->l2_Q1, Fs->l2_F1, _O2_BYTE * N_TRIANGLE_TERMS(_V1));
+
+    memcpy(Qs->l2_Q2, Fs->l2_F2, _O2_BYTE * _V1 * _O1);
+    batch_trimat_madd(Qs->l2_Q2, Fs->l2_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O2_BYTE); // F1*T1 + F2
+
+    memcpy(Qs->l2_Q5, Fs->l2_F5, _O2_BYTE * N_TRIANGLE_TERMS(_O1));
+    memset(tempQ, 0, _O2_BYTE * _O1 * _O1);                                        // l2_Q5
+    batch_matTr_madd(tempQ, Ts->t1, _V1, _V1_BYTE, _O1, Qs->l2_Q2, _O1, _O2_BYTE); // t1_tr*(F1*T1 + F2)
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_UpperTrianglize(Qs->l2_Q5, tempQ, _O1, _O2_BYTE);            // UT( ... )   // Q5
+
+    batch_trimatTr_madd(Qs->l2_Q2, Fs->l2_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O2_BYTE); // Q2
+
+    /*
+          Computing:
+          F1_T2     = F1 * t2
+          F2_T3     = F2 * t3
+          F1_F1T_T2 + F2_T3 = F1_T2 + F2_T3 + F1tr * t2
+
+          Q3 =        F1_F1T*T2 + F2*T3 + F3
+          Q9 = UT( T2tr*( F1*T2 + F2*T3 + F3 )  +      T3tr*( F5*T3 + F6 ) )
+          Q6 = T1tr*( F1_F1T*T2 + F2*T3 + F3 )  + F2Tr*T2 + F5_F5T*T3 + F6
+      */
+    memcpy(Qs->l2_Q3, Fs->l2_F3, _O2_BYTE * _V1 * _O2);
+    batch_trimat_madd(Qs->l2_Q3, Fs->l2_F1, t2, _V1, _V1_BYTE, _O2, _O2_BYTE);       // F1*T2 + F3
+    batch_mat_madd(Qs->l2_Q3, Fs->l2_F2, _V1, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE); // F1_T2 + F2_T3 + F3
+
+    memset(tempQ, 0, _O2_BYTE * _O2 * _O2);                                    // l2_Q9
+    batch_matTr_madd(tempQ, t2, _V1, _V1_BYTE, _O2, Qs->l2_Q3, _O2, _O2_BYTE); // T2tr * ( ..... )
+
+    memcpy(Qs->l2_Q6, Fs->l2_F6, _O2_BYTE * _O1 * _O2);
+
+    batch_trimat_madd(Qs->l2_Q6, Fs->l2_F5, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE); // F5*T3 + F6
+    batch_matTr_madd(tempQ, Ts->t3, _O1, _O1_BYTE, _O2, Qs->l2_Q6, _O2, _O2_BYTE); // T2tr*( ..... ) + T3tr*( ..... )
+    memset(Qs->l2_Q9, 0, _O2_BYTE * N_TRIANGLE_TERMS(_O2));
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_UpperTrianglize(Qs->l2_Q9, tempQ, _O2, _O2_BYTE); // Q9
+
+    batch_trimatTr_madd(Qs->l2_Q3, Fs->l2_F1, t2, _V1, _V1_BYTE, _O2, _O2_BYTE); // F1_F1T_T2 + F2_T3 + F3 // Q3
+
+    batch_bmatTr_madd(Qs->l2_Q6, Fs->l2_F2, _O1, t2, _V1, _V1_BYTE, _O2, _O2_BYTE);    //  F5*T3 + F6 +  F2tr*T2
+    batch_trimatTr_madd(Qs->l2_Q6, Fs->l2_F5, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE);   //   F2tr*T2 + F5_F5T*T3 + F6
+    batch_matTr_madd(Qs->l2_Q6, Ts->t1, _V1, _V1_BYTE, _O1, Qs->l2_Q3, _O2, _O2_BYTE); // Q6
+}
+#define calculate_Q_from_F_impl calculate_Q_from_F_ref
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_calculate_Q_from_F(ext_cpk_t *Qs, const sk_t *Fs, const sk_t *Ts) {
+    calculate_Q_from_F_impl(Qs, Fs, Ts);
+}

crypto_sign/rainbowIIIc-classic/clean/rainbow_keypair_computation.h

@@ -0,0 +1,53 @@
+#ifndef _RAINBOW_KEYPAIR_COMP_H_
+#define _RAINBOW_KEYPAIR_COMP_H_
+/// @file rainbow_keypair_computation.h
+/// @brief Functions for calculating pk/sk while generating keys.
+///
+/// Defining an internal structure of public key.
+/// Functions for calculating pk/sk for key generation.
+///
+
+#include "rainbow_keypair.h"
+
+/// @brief The (internal use) public key for rainbow
+///
+/// The (internal use) public key for rainbow. The public
+/// polynomials are divided into l1_Q1, l1_Q2, ... l1_Q9,
+/// l2_Q1, .... , l2_Q9.
+///
+typedef struct rainbow_extend_publickey {
+    unsigned char l1_Q1[_O1_BYTE * N_TRIANGLE_TERMS(_V1)];
+    unsigned char l1_Q2[_O1_BYTE * _V1 * _O1];
+    unsigned char l1_Q3[_O1_BYTE * _V1 * _O2];
+    unsigned char l1_Q5[_O1_BYTE * N_TRIANGLE_TERMS(_O1)];
+    unsigned char l1_Q6[_O1_BYTE * _O1 * _O2];
+    unsigned char l1_Q9[_O1_BYTE * N_TRIANGLE_TERMS(_O2)];
+
+    unsigned char l2_Q1[_O2_BYTE * N_TRIANGLE_TERMS(_V1)];
+    unsigned char l2_Q2[_O2_BYTE * _V1 * _O1];
+    unsigned char l2_Q3[_O2_BYTE * _V1 * _O2];
+    unsigned char l2_Q5[_O2_BYTE * N_TRIANGLE_TERMS(_O1)];
+    unsigned char l2_Q6[_O2_BYTE * _O1 * _O2];
+    unsigned char l2_Q9[_O2_BYTE * N_TRIANGLE_TERMS(_O2)];
+} ext_cpk_t;
+
+///
+/// @brief converting formats of public keys : from ext_cpk_t version to pk_t
+///
+/// @param[out] pk       - the classic public key.
+/// @param[in]  cpk      - the internel public key.
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_extcpk_to_pk(pk_t *pk, const ext_cpk_t *cpk);
+/////////////////////////////////////////////////
+
+///
+/// @brief Computing public key from secret key
+///
+/// @param[out] Qs       - the public key
+/// @param[in]  Fs       - parts of the secret key: l1_F1, l1_F2, l2_F1, l2_F2, l2_F3, l2_F5, l2_F6
+/// @param[in]  Ts       - parts of the secret key: T1, T4, T3
+///
+void PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_calculate_Q_from_F(ext_cpk_t *Qs, const sk_t *Fs, const sk_t *Ts);
+
+
+#endif // _RAINBOW_KEYPAIR_COMP_H_

crypto_sign/rainbowIIIc-classic/clean/sign.c

@@ -0,0 +1,74 @@
+///  @file  sign.c
+///  @brief the implementations for functions in api.h
+///
+///
+
+#include "api.h"
+#include "rainbow.h"
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+#include "randombytes.h"
+#include "utils_hash.h"
+#include <stdlib.h>
+#include <string.h>
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign_keypair(unsigned char *pk, unsigned char *sk) {
+    unsigned char sk_seed[LEN_SKSEED] = {0};
+    randombytes(sk_seed, LEN_SKSEED);
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_generate_keypair((pk_t *)pk, (sk_t *)sk, sk_seed);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign(unsigned char *sm, size_t *smlen, const unsigned char *m, size_t mlen, const unsigned char *sk) {
+    unsigned char digest[_HASH_LEN];
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(digest, _HASH_LEN, m, mlen);
+
+    memcpy(sm, m, mlen);
+    smlen[0] = mlen + _SIGNATURE_BYTE;
+
+    return PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_rainbow_sign(sm + mlen, (const sk_t *)sk, digest);
+}
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign_open(unsigned char *m, size_t *mlen, const unsigned char *sm, size_t smlen, const unsigned char *pk) {
+    int rc;
+    if (_SIGNATURE_BYTE > smlen) {
+        rc = -1;
+    } else {
+        *mlen = smlen - _SIGNATURE_BYTE;
+
+        unsigned char digest[_HASH_LEN];
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(digest, _HASH_LEN, sm, *mlen);
+
+        rc = PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_rainbow_verify(digest, sm + mlen[0], (const pk_t *)pk);
+    }
+    if (!rc) {
+        memcpy(m, sm, smlen - _SIGNATURE_BYTE);
+    } else { // bad signature
+        *mlen = (size_t) -1;
+        memset(m, 0, smlen);
+    }
+    return rc;
+}
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign_signature(
+    uint8_t *sig, size_t *siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *sk) {
+    unsigned char digest[_HASH_LEN];
+
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(digest, _HASH_LEN, m, mlen);
+    *siglen = _SIGNATURE_BYTE;
+    return PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_rainbow_sign(sig, (const sk_t *)sk, digest);
+}
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_crypto_sign_verify(
+    const uint8_t *sig, size_t siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *pk) {
+    if (siglen != _SIGNATURE_BYTE) {
+        return -1;
+    }
+    unsigned char digest[_HASH_LEN];
+    PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(digest, _HASH_LEN, m, mlen);
+    return PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_rainbow_verify(digest, sig, (const pk_t *)pk);
+}

crypto_sign/rainbowIIIc-classic/clean/utils_hash.c

@@ -0,0 +1,50 @@
+/// @file utils_hash.c
+/// @brief the adapter for SHA2 families.
+///
+///
+
+#include "utils_hash.h"
+#include "rainbow_config.h"
+#include "sha2.h"
+
+static inline int _hash(unsigned char *digest, const unsigned char *m, size_t mlen) {
+    sha384(digest, m, mlen);
+    return 0;
+}
+
+static inline int expand_hash(unsigned char *digest, size_t n_digest, const unsigned char *hash) {
+    if (_HASH_LEN >= n_digest) {
+        for (size_t i = 0; i < n_digest; i++) {
+            digest[i] = hash[i];
+        }
+        return 0;
+    }
+    for (size_t i = 0; i < _HASH_LEN; i++) {
+        digest[i] = hash[i];
+    }
+    n_digest -= _HASH_LEN;
+
+    while (_HASH_LEN <= n_digest) {
+        _hash(digest + _HASH_LEN, digest, _HASH_LEN);
+
+        n_digest -= _HASH_LEN;
+        digest += _HASH_LEN;
+    }
+    unsigned char temp[_HASH_LEN];
+    if (n_digest) {
+        _hash(temp, digest, _HASH_LEN);
+        for (size_t i = 0; i < n_digest; i++) {
+            digest[_HASH_LEN + i] = temp[i];
+        }
+    }
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(unsigned char *digest,
+        size_t len_digest,
+        const unsigned char *m,
+        size_t mlen) {
+    unsigned char buf[_HASH_LEN];
+    _hash(buf, m, mlen);
+    return expand_hash(digest, len_digest, buf);
+}

crypto_sign/rainbowIIIc-classic/clean/utils_hash.h

@@ -0,0 +1,11 @@
+#ifndef _UTILS_HASH_H_
+#define _UTILS_HASH_H_
+/// @file utils_hash.h
+/// @brief the interface for adapting hash functions.
+///
+
+#include <stddef.h>
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(unsigned char *digest, size_t len_digest, const unsigned char *m, size_t mlen);
+
+#endif // _UTILS_HASH_H_

crypto_sign/rainbowIIIc-classic/clean/utils_prng.c

@@ -0,0 +1,95 @@
+/// @file utils_prng.c
+/// @brief The implementation of PRNG related functions.
+///
+
+#include "utils_prng.h"
+#include "aes.h"
+#include "randombytes.h"
+#include "utils_hash.h"
+#include <stdlib.h>
+#include <string.h>
+
+static void prng_update(const unsigned char *provided_data,
+                        unsigned char *Key,
+                        unsigned char *V) {
+    unsigned char temp[48];
+    aes256ctx ctx;
+    aes256_keyexp(&ctx, Key);
+    for (int i = 0; i < 3; i++) {
+        //increment V
+        for (int j = 15; j >= 0; j--) {
+            if (V[j] == 0xff) {
+                V[j] = 0x00;
+            } else {
+                V[j]++;
+                break;
+            }
+        }
+        aes256_ecb(temp + 16 * i, V, 1, &ctx);
+    }
+    if (provided_data != NULL) {
+        for (int i = 0; i < 48; i++) {
+            temp[i] ^= provided_data[i];
+        }
+    }
+    memcpy(Key, temp, 32);
+    memcpy(V, temp + 32, 16);
+}
+static void randombytes_init_with_state(prng_t *state,
+                                        unsigned char *entropy_input_48bytes) {
+    memset(state->Key, 0x00, 32);
+    memset(state->V, 0x00, 16);
+    prng_update(entropy_input_48bytes, state->Key, state->V);
+}
+
+static int randombytes_with_state(prng_t *state,
+                                  unsigned char *x,
+                                  size_t xlen) {
+
+    unsigned char block[16];
+    int i = 0;
+
+    aes256ctx ctx;
+    aes256_keyexp(&ctx, state->Key);
+
+    while (xlen > 0) {
+        //increment V
+        for (int j = 15; j >= 0; j--) {
+            if (state->V[j] == 0xff) {
+                state->V[j] = 0x00;
+            } else {
+                state->V[j]++;
+                break;
+            }
+        }
+        aes256_ecb(block, state->V, 1, &ctx);
+        if (xlen > 15) {
+            memcpy(x + i, block, 16);
+            i += 16;
+            xlen -= 16;
+        } else {
+            memcpy(x + i, block, xlen);
+            xlen = 0;
+        }
+    }
+    prng_update(NULL, state->Key, state->V);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_set(prng_t *ctx, const void *prng_seed, unsigned long prng_seedlen) {
+    unsigned char seed[48];
+    if (prng_seedlen >= 48) {
+        memcpy(seed, prng_seed, 48);
+    } else {
+        memcpy(seed, prng_seed, prng_seedlen);
+        PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_hash_msg(seed + prng_seedlen, 48 - (unsigned)prng_seedlen, (const unsigned char *)prng_seed, prng_seedlen);
+    }
+
+    randombytes_init_with_state(ctx, seed);
+
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng_t *ctx, unsigned char *out, unsigned long outlen) {
+    return randombytes_with_state(ctx, out, outlen);
+}

crypto_sign/rainbowIIIc-classic/clean/utils_prng.h

@@ -0,0 +1,18 @@
+#ifndef _UTILS_PRNG_H_
+#define _UTILS_PRNG_H_
+/// @file utils_prng.h
+/// @brief the interface for adapting PRNG functions.
+///
+///
+
+#include "randombytes.h"
+
+typedef struct {
+    unsigned char Key[32];
+    unsigned char V[16];
+} prng_t;
+
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_set(prng_t *ctx, const void *prng_seed, unsigned long prng_seedlen);
+int PQCLEAN_RAINBOWIIICCLASSIC_CLEAN_prng_gen(prng_t *ctx, unsigned char *out, unsigned long outlen);
+
+#endif // _UTILS_PRNG_H_

crypto_sign/rainbowIIIc-cyclic-compressed/META.yml

@@ -0,0 +1,18 @@
+name: Rainbow-IIIc-cyclic-compressed
+type: signature
+claimed-nist-level: 3
+length-public-key: 206744
+length-secret-key: 64
+length-signature: 156
+nistkat-sha256: 1ad6d22a9e98c3e05a6aceb5b892dd75908924733aadfe074b6556e1dbd881c0
+testvectors-sha256: 40df2d3b2eb52aada14469c95e6890c486eaf22dcfca9604bbf528a0b7b75070
+principal-submitters:
+  - Jintai Ding
+auxiliary-submitters:
+  - Ming-Shing Chen
+  - Albrecht Petzoldt
+  - Dieter Schmidt
+  - Bo-Yin Yang
+implementations:
+    - name: clean
+      version: https://github.com/fast-crypto-lab/rainbow-submission-round2/commit/af826fcb78f6af51a02d0352cff28a9690467bfd

crypto_sign/rainbowIIIc-cyclic-compressed/clean/LICENSE

@@ -0,0 +1,8 @@
+`Software implementation of Rainbow for NIST R2 submission' by Ming-Shing Chen
+
+To the extent possible under law, the person who associated CC0 with
+`Software implementation of Rainbow for NIST R2 submission' has waived all copyright and related or neighboring rights
+to `Software implementation of Rainbow for NIST R2 submission'.
+
+You should have received a copy of the CC0 legalcode along with this
+work.  If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.

crypto_sign/rainbowIIIc-cyclic-compressed/clean/Makefile

@@ -0,0 +1,20 @@
+# This Makefile can be used with GNU Make or BSD Make
+
+LIB=librainbowIIIc-cyclic-compressed_clean.a
+
+HEADERS = api.h blas_comm.h blas.h blas_u32.h gf.h parallel_matrix_op.h rainbow_blas.h rainbow_config.h rainbow.h rainbow_keypair_computation.h rainbow_keypair.h utils_hash.h utils_prng.h 
+OBJECTS =  blas_comm.o parallel_matrix_op.o rainbow.o rainbow_keypair.o rainbow_keypair_computation.o sign.o utils_hash.o utils_prng.o blas_u32.o gf.o
+
+CFLAGS=-O3 -Wall -Wconversion -Wextra -Wpedantic -Wvla -Werror -Wmissing-prototypes -Wredundant-decls -std=c99 -I../../../common $(EXTRAFLAGS)
+
+all: $(LIB)
+
+%.o: %.c $(HEADERS)
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(LIB): $(OBJECTS)
+	$(AR) -r $@ $(OBJECTS)
+
+clean:
+	$(RM) $(OBJECTS)
+	$(RM) $(LIB)

crypto_sign/rainbowIIIc-cyclic-compressed/clean/Makefile.Microsoft_nmake

@@ -0,0 +1,19 @@
+# This Makefile can be used with Microsoft Visual Studio's nmake using the command:
+#    nmake /f Makefile.Microsoft_nmake
+
+LIBRARY=librainbowIIIc-cyclic-compressed_clean.lib
+OBJECTS =  blas_comm.obj parallel_matrix_op.obj rainbow.obj rainbow_keypair.obj rainbow_keypair_computation.obj sign.obj utils_hash.obj utils_prng.obj blas_u32.obj gf.obj
+
+CFLAGS=/nologo /I ..\..\..\common /W4 /WX
+
+all: $(LIBRARY)
+
+# Make sure objects are recompiled if headers change.
+$(OBJECTS): *.h
+
+$(LIBRARY): $(OBJECTS)
+	LIB.EXE /NOLOGO /WX /OUT:$@ $**
+
+clean:
+    -DEL $(OBJECTS)
+    -DEL $(LIBRARY)

crypto_sign/rainbowIIIc-cyclic-compressed/clean/api.h

@@ -0,0 +1,32 @@
+#ifndef PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_API_H
+#define PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_API_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+#define PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_CRYPTO_SECRETKEYBYTES 64
+#define PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_CRYPTO_PUBLICKEYBYTES 206744
+#define PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_CRYPTO_BYTES 156
+#define PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_CRYPTO_ALGNAME "RAINBOW(256,68,36,36) - cyclic compressed"
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign_keypair(uint8_t *pk, uint8_t *sk);
+
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign_signature(
+    uint8_t *sig, size_t *siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *sk);
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign_verify(
+    const uint8_t *sig, size_t siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *pk);
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign(uint8_t *sm, size_t *smlen,
+        const uint8_t *m, size_t mlen,
+        const uint8_t *sk);
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign_open(uint8_t *m, size_t *mlen,
+        const uint8_t *sm, size_t smlen,
+        const uint8_t *pk);
+
+
+#endif

crypto_sign/rainbowIIIc-cyclic-compressed/clean/blas.h

@@ -0,0 +1,19 @@
+#ifndef _BLAS_H_
+#define _BLAS_H_
+/// @file blas.h
+/// @brief Defining the implementations for linear algebra functions depending on the machine architecture.
+///
+
+#include "blas_comm.h"
+#include "blas_u32.h"
+#include "rainbow_config.h"
+
+#define gf256v_predicated_add PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_predicated_add_u32
+#define gf256v_add PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_add_u32
+
+
+#define gf256v_mul_scalar PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_scalar_u32
+#define gf256v_madd PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_madd_u32
+
+
+#endif // _BLAS_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/blas_comm.c

@@ -0,0 +1,144 @@
+/// @file blas_comm.c
+/// @brief The standard implementations for blas_comm.h
+///
+
+#include "blas_comm.h"
+#include "blas.h"
+#include "gf.h"
+#include "rainbow_config.h"
+
+#include <stdint.h>
+#include <string.h>
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(uint8_t *b, unsigned int _num_byte) {
+    for (size_t i = 0; i < _num_byte; i++) {
+        b[i] = 0;
+    }
+}
+/// @brief get an element from GF(256) vector .
+///
+/// @param[in]  a         - the input vector a.
+/// @param[in]  i         - the index in the vector a.
+/// @return  the value of the element.
+///
+uint8_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(const uint8_t *a, unsigned int i) {
+    return a[i];
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_is_zero(const uint8_t *a, unsigned int _num_byte) {
+    uint8_t r = 0;
+    while (_num_byte--) {
+        r |= a[0];
+        a++;
+    }
+    return (0 == r);
+}
+
+/// polynomial multplication
+/// School boook
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_polymul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int _num) {
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(c, _num * 2 - 1);
+    for (unsigned int i = 0; i < _num; i++) {
+        gf256v_madd(c + i, a, b[i], _num);
+    }
+}
+
+static void gf256mat_prod_ref(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b) {
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(c, n_A_vec_byte);
+    for (unsigned int i = 0; i < n_A_width; i++) {
+        gf256v_madd(c, matA, b[i], n_A_vec_byte);
+        matA += n_A_vec_byte;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_mul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int len_vec) {
+    unsigned int n_vec_byte = len_vec;
+    for (unsigned int k = 0; k < len_vec; k++) {
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(c, n_vec_byte);
+        const uint8_t *bk = b + n_vec_byte * k;
+        for (unsigned int i = 0; i < len_vec; i++) {
+            gf256v_madd(c, a + n_vec_byte * i, bk[i], n_vec_byte);
+        }
+        c += n_vec_byte;
+    }
+}
+
+static unsigned int gf256mat_gauss_elim_ref(uint8_t *mat, unsigned int h, unsigned int w) {
+    unsigned int r8 = 1;
+
+    for (unsigned int i = 0; i < h; i++) {
+        uint8_t *ai = mat + w * i;
+        unsigned int skip_len_align4 = i & ((unsigned int)~0x3);
+
+        for (unsigned int j = i + 1; j < h; j++) {
+            uint8_t *aj = mat + w * j;
+            gf256v_predicated_add(ai + skip_len_align4, !PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256_is_nonzero(ai[i]), aj + skip_len_align4, w - skip_len_align4);
+        }
+        r8 &= PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256_is_nonzero(ai[i]);
+        uint8_t pivot = ai[i];
+        pivot = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256_inv(pivot);
+        gf256v_mul_scalar(ai + skip_len_align4, pivot, w - skip_len_align4);
+        for (unsigned int j = 0; j < h; j++) {
+            if (i == j) {
+                continue;
+            }
+            uint8_t *aj = mat + w * j;
+            gf256v_madd(aj + skip_len_align4, ai + skip_len_align4, aj[i], w - skip_len_align4);
+        }
+    }
+
+    return r8;
+}
+
+static unsigned int gf256mat_solve_linear_eq_ref(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n) {
+    uint8_t mat[64 * 64];
+    for (unsigned int i = 0; i < n; i++) {
+        memcpy(mat + i * (n + 1), inp_mat + i * n, n);
+        mat[i * (n + 1) + n] = c_terms[i];
+    }
+    unsigned int r8 = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_gauss_elim(mat, n, n + 1);
+    for (unsigned int i = 0; i < n; i++) {
+        sol[i] = mat[i * (n + 1) + n];
+    }
+    return r8;
+}
+
+static inline void gf256mat_submat(uint8_t *mat2, unsigned int w2, unsigned int st, const uint8_t *mat, unsigned int w, unsigned int h) {
+    for (unsigned int i = 0; i < h; i++) {
+        for (unsigned int j = 0; j < w2; j++) {
+            mat2[i * w2 + j] = mat[i * w + st + j];
+        }
+    }
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_inv(uint8_t *inv_a, const uint8_t *a, unsigned int H, uint8_t *buffer) {
+    uint8_t *aa = buffer;
+    for (unsigned int i = 0; i < H; i++) {
+        uint8_t *ai = aa + i * 2 * H;
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(ai, 2 * H);
+        gf256v_add(ai, a + i * H, H);
+        ai[H + i] = 1;
+    }
+    unsigned int r8 = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_gauss_elim(aa, H, 2 * H);
+    gf256mat_submat(inv_a, H, H, aa, 2 * H, H);
+    return r8;
+}
+
+
+// choosing the implementations depends on the macros _BLAS_AVX2_ and _BLAS_SSE
+
+#define gf256mat_prod_impl gf256mat_prod_ref
+#define gf256mat_gauss_elim_impl gf256mat_gauss_elim_ref
+#define gf256mat_solve_linear_eq_impl gf256mat_solve_linear_eq_ref
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_prod(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b) {
+    gf256mat_prod_impl(c, matA, n_A_vec_byte, n_A_width, b);
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_gauss_elim(uint8_t *mat, unsigned int h, unsigned int w) {
+    return gf256mat_gauss_elim_impl(mat, h, w);
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_solve_linear_eq(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n) {
+    return gf256mat_solve_linear_eq_impl(sol, inp_mat, c_terms, n);
+}
+

crypto_sign/rainbowIIIc-cyclic-compressed/clean/blas_comm.h

@@ -0,0 +1,90 @@
+#ifndef _BLAS_COMM_H_
+#define _BLAS_COMM_H_
+/// @file blas_comm.h
+/// @brief Common functions for linear algebra.
+///
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+/// @brief set a vector to 0.
+///
+/// @param[in,out]   b      - the vector b.
+/// @param[in]  _num_byte   - number of bytes for the vector b.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(uint8_t *b, unsigned int _num_byte);
+
+/// @brief get an element from GF(256) vector .
+///
+/// @param[in]  a         - the input vector a.
+/// @param[in]  i         - the index in the vector a.
+/// @return  the value of the element.
+///
+uint8_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(const uint8_t *a, unsigned int i);
+
+/// @brief check if a vector is 0.
+///
+/// @param[in]   a          - the vector a.
+/// @param[in]  _num_byte   - number of bytes for the vector a.
+/// @return  1(true) if a is 0. 0(false) else.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_is_zero(const uint8_t *a, unsigned int _num_byte);
+
+/// @brief polynomial multiplication:  c = a*b
+///
+/// @param[out]   c         - the output polynomial c
+/// @param[in]   a          - the vector a.
+/// @param[in]   b          - the vector b.
+/// @param[in]  _num   - number of elements for the polynomials a and b.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_polymul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int _num);
+
+/// @brief matrix-vector multiplication:  c = matA * b , in GF(256)
+///
+/// @param[out]  c         - the output vector c
+/// @param[in]   matA      - a column-major matrix A.
+/// @param[in]   n_A_vec_byte  - the size of column vectors in bytes.
+/// @param[in]   n_A_width   - the width of matrix A.
+/// @param[in]   b          - the vector b.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_prod(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b);
+
+/// @brief matrix-matrix multiplication:  c = a * b , in GF(256)
+///
+/// @param[out]  c         - the output matrix c
+/// @param[in]   c         - a matrix a.
+/// @param[in]   b         - a matrix b.
+/// @param[in]   len_vec   - the length of column vectors.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_mul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int len_vec);
+
+/// @brief Gauss elimination for a matrix, in GF(256)
+///
+/// @param[in,out]  mat    - the matrix.
+/// @param[in]   h         - the height of the matrix.
+/// @param[in]   w         - the width of the matrix.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_gauss_elim(uint8_t *mat, unsigned int h, unsigned int w);
+
+/// @brief Solving linear equations, in GF(256)
+///
+/// @param[out]  sol       - the solutions.
+/// @param[in]   inp_mat   - the matrix parts of input equations.
+/// @param[in]   c_terms   - the constant terms of the input equations.
+/// @param[in]   n         - the number of equations.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_solve_linear_eq(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n);
+
+/// @brief Computing the inverse matrix, in GF(256)
+///
+/// @param[out]  inv_a     - the output of matrix a.
+/// @param[in]   a         - a matrix a.
+/// @param[in]   H         - height of matrix a, i.e., matrix a is an HxH matrix.
+/// @param[in]   buffer    - The buffer for computations. it has to be as large as 2 input matrixes.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_inv(uint8_t *inv_a, const uint8_t *a, unsigned int H, uint8_t *buffer);
+
+#endif // _BLAS_COMM_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/blas_u32.c

@@ -0,0 +1,87 @@
+#include "blas_u32.h"
+#include "gf.h"
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_predicated_add_u32(uint8_t *accu_b, uint8_t predicate, const uint8_t *a, unsigned int _num_byte) {
+    uint32_t pr_u32 = ((uint32_t)0) - ((uint32_t)predicate);
+    uint8_t pr_u8 = pr_u32 & 0xff;
+
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *b_u32 = (uint32_t *)accu_b;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        b_u32[i] ^= (a_u32[i] & pr_u32);
+    }
+
+    a += (n_u32 << 2);
+    accu_b += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_b[i] ^= (a[i] & pr_u8);
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_add_u32(uint8_t *accu_b, const uint8_t *a, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *b_u32 = (uint32_t *)accu_b;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        b_u32[i] ^= a_u32[i];
+    }
+
+    a += (n_u32 << 2);
+    accu_b += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_b[i] ^= a[i];
+    }
+}
+
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_scalar_u32(uint8_t *a, uint8_t b, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *a_u32 = (uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        a_u32[i] = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_u32(a_u32[i], b);
+    }
+
+    union tmp_32 {
+        uint8_t u8[4];
+        uint32_t u32;
+    } t;
+    t.u32 = 0;
+    a += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        t.u8[i] = a[i];
+    }
+    t.u32 = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_u32(t.u32, b);
+    for (unsigned int i = 0; i < rem; i++) {
+        a[i] = t.u8[i];
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_madd_u32(uint8_t *accu_c, const uint8_t *a, uint8_t gf256_b, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *c_u32 = (uint32_t *)accu_c;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        c_u32[i] ^= PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_u32(a_u32[i], gf256_b);
+    }
+
+    union tmp_32 {
+        uint8_t u8[4];
+        uint32_t u32;
+    } t;
+    t.u32 = 0;
+    accu_c += (n_u32 << 2);
+    a += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        t.u8[i] = a[i];
+    }
+    t.u32 = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_u32(t.u32, gf256_b);
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_c[i] ^= t.u8[i];
+    }
+}
+

crypto_sign/rainbowIIIc-cyclic-compressed/clean/blas_u32.h

@@ -0,0 +1,18 @@
+#ifndef _BLAS_U32_H_
+#define _BLAS_U32_H_
+/// @file blas_u32.h
+/// @brief Inlined functions for implementing basic linear algebra functions for uint32 arch.
+///
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_predicated_add_u32(uint8_t *accu_b, uint8_t predicate, const uint8_t *a, unsigned int _num_byte);
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_add_u32(uint8_t *accu_b, const uint8_t *a, unsigned int _num_byte);
+
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_scalar_u32(uint8_t *a, uint8_t b, unsigned int _num_byte);
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_madd_u32(uint8_t *accu_c, const uint8_t *a, uint8_t gf256_b, unsigned int _num_byte);
+
+
+#endif // _BLAS_U32_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/gf.c

@@ -0,0 +1,134 @@
+#include "gf.h"
+
+//// gf4 := gf2[x]/x^2+x+1
+static inline uint8_t gf4_mul_2(uint8_t a) {
+    uint8_t r = (uint8_t)(a << 1);
+    r ^= (uint8_t)((a >> 1) * 7);
+    return r;
+}
+
+static inline uint8_t gf4_mul(uint8_t a, uint8_t b) {
+    uint8_t r = (uint8_t)(a * (b & 1));
+    return r ^ (uint8_t)(gf4_mul_2(a) * (b >> 1));
+}
+
+static inline uint8_t gf4_squ(uint8_t a) {
+    return a ^ (a >> 1);
+}
+
+static inline uint32_t gf4v_mul_2_u32(uint32_t a) {
+    uint32_t bit0 = a & 0x55555555;
+    uint32_t bit1 = a & 0xaaaaaaaa;
+    return (bit0 << 1) ^ bit1 ^ (bit1 >> 1);
+}
+
+static inline uint32_t gf4v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t bit0_b = ((uint32_t)0) - ((uint32_t)(b & 1));
+    uint32_t bit1_b = ((uint32_t)0) - ((uint32_t)((b >> 1) & 1));
+    return (a & bit0_b) ^ (bit1_b & gf4v_mul_2_u32(a));
+}
+
+//// gf16 := gf4[y]/y^2+y+x
+static inline uint8_t gf16_mul(uint8_t a, uint8_t b) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = (a >> 2);
+    uint8_t b0 = b & 3;
+    uint8_t b1 = (b >> 2);
+    uint8_t a0b0 = gf4_mul(a0, b0);
+    uint8_t a1b1 = gf4_mul(a1, b1);
+    uint8_t a0b1_a1b0 = gf4_mul(a0 ^ a1, b0 ^ b1) ^ a0b0 ^ a1b1;
+    uint8_t a1b1_x2 = gf4_mul_2(a1b1);
+    return (uint8_t)((a0b1_a1b0 ^ a1b1) << 2 ^ a0b0 ^ a1b1_x2);
+}
+
+static inline uint8_t gf16_squ(uint8_t a) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = (a >> 2);
+    a1 = gf4_squ(a1);
+    uint8_t a1squ_x2 = gf4_mul_2(a1);
+    return (uint8_t)((a1 << 2) ^ a1squ_x2 ^ gf4_squ(a0));
+}
+
+// gf16 := gf4[y]/y^2+y+x
+uint32_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf16v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t axb0 = gf4v_mul_u32(a, b);
+    uint32_t axb1 = gf4v_mul_u32(a, b >> 2);
+    uint32_t a0b1 = (axb1 << 2) & 0xcccccccc;
+    uint32_t a1b1 = axb1 & 0xcccccccc;
+    uint32_t a1b1_2 = a1b1 >> 2;
+
+    return axb0 ^ a0b1 ^ a1b1 ^ gf4v_mul_2_u32(a1b1_2);
+}
+
+uint8_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256_is_nonzero(uint8_t a) {
+    unsigned int a8 = a;
+    unsigned int r = ((unsigned int)0) - a8;
+    r >>= 8;
+    return r & 1;
+}
+
+static inline uint8_t gf4_mul_3(uint8_t a) {
+    uint8_t msk = (uint8_t)((a - 2) >> 1);
+    return (uint8_t)((msk & ((int)a * 3)) | ((~msk) & ((int)a - 1)));
+}
+static inline uint8_t gf16_mul_8(uint8_t a) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = a >> 2;
+    return (uint8_t)((gf4_mul_2(a0 ^ a1) << 2) | gf4_mul_3(a1));
+}
+
+// gf256 := gf16[X]/X^2+X+xy
+static inline uint8_t gf256_mul(uint8_t a, uint8_t b) {
+    uint8_t a0 = a & 15;
+    uint8_t a1 = (a >> 4);
+    uint8_t b0 = b & 15;
+    uint8_t b1 = (b >> 4);
+    uint8_t a0b0 = gf16_mul(a0, b0);
+    uint8_t a1b1 = gf16_mul(a1, b1);
+    uint8_t a0b1_a1b0 = gf16_mul(a0 ^ a1, b0 ^ b1) ^ a0b0 ^ a1b1;
+    uint8_t a1b1_x8 = gf16_mul_8(a1b1);
+    return (uint8_t)((a0b1_a1b0 ^ a1b1) << 4 ^ a0b0 ^ a1b1_x8);
+}
+
+static inline uint8_t gf256_squ(uint8_t a) {
+    uint8_t a0 = a & 15;
+    uint8_t a1 = (a >> 4);
+    a1 = gf16_squ(a1);
+    uint8_t a1squ_x8 = gf16_mul_8(a1);
+    return (uint8_t)((a1 << 4) ^ a1squ_x8 ^ gf16_squ(a0));
+}
+
+uint8_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256_inv(uint8_t a) {
+    // 128+64+32+16+8+4+2 = 254
+    uint8_t a2 = gf256_squ(a);
+    uint8_t a4 = gf256_squ(a2);
+    uint8_t a8 = gf256_squ(a4);
+    uint8_t a4_2 = gf256_mul(a4, a2);
+    uint8_t a8_4_2 = gf256_mul(a4_2, a8);
+    uint8_t a64_ = gf256_squ(a8_4_2);
+    a64_ = gf256_squ(a64_);
+    a64_ = gf256_squ(a64_);
+    uint8_t a64_2 = gf256_mul(a64_, a8_4_2);
+    uint8_t a128_ = gf256_squ(a64_2);
+    return gf256_mul(a2, a128_);
+}
+
+static inline uint32_t gf4v_mul_3_u32(uint32_t a) {
+    uint32_t bit0 = a & 0x55555555;
+    uint32_t bit1 = a & 0xaaaaaaaa;
+    return (bit0 << 1) ^ bit0 ^ (bit1 >> 1);
+}
+static inline uint32_t gf16v_mul_8_u32(uint32_t a) {
+    uint32_t a1 = a & 0xcccccccc;
+    uint32_t a0 = (a << 2) & 0xcccccccc;
+    return gf4v_mul_2_u32(a0 ^ a1) | gf4v_mul_3_u32(a1 >> 2);
+}
+uint32_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t axb0 = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf16v_mul_u32(a, b);
+    uint32_t axb1 = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf16v_mul_u32(a, b >> 4);
+    uint32_t a0b1 = (axb1 << 4) & 0xf0f0f0f0;
+    uint32_t a1b1 = axb1 & 0xf0f0f0f0;
+    uint32_t a1b1_4 = a1b1 >> 4;
+
+    return axb0 ^ a0b1 ^ a1b1 ^ gf16v_mul_8_u32(a1b1_4);
+}

crypto_sign/rainbowIIIc-cyclic-compressed/clean/gf.h

@@ -0,0 +1,19 @@
+#ifndef _GF16_H_
+#define _GF16_H_
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+/// @file gf16.h
+/// @brief Library for arithmetics in GF(16) and GF(256)
+///
+
+uint32_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf16v_mul_u32(uint32_t a, uint8_t b);
+
+
+uint8_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256_is_nonzero(uint8_t a);
+uint8_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256_inv(uint8_t a);
+uint32_t PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_u32(uint32_t a, uint8_t b);
+
+
+#endif // _GF16_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/parallel_matrix_op.c

@@ -0,0 +1,183 @@
+///  @file parallel_matrix_op.c
+///  @brief the standard implementations for functions in parallel_matrix_op.h
+///
+///  the standard implementations for functions in parallel_matrix_op.h
+///
+
+#include "parallel_matrix_op.h"
+#include "blas.h"
+#include "blas_comm.h"
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle(UT) matrix.
+///
+/// @param[in]  i_row     - the i-th row in an upper-triangle matrix.
+/// @param[in]  j_col     - the j-th column in an upper-triangle matrix.
+/// @param[in]  dim       - the dimension of the upper-triangle matrix, i.e., an dim x dim matrix.
+/// @return  the corresponding index in an array storage.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(unsigned int i_row, unsigned int j_col, unsigned int dim) {
+    return (dim + dim - i_row + 1) * i_row / 2 + j_col - i_row;
+}
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle or lower-triangle matrix.
+///
+/// @param[in]  i_row     - the i-th row in a triangle matrix.
+/// @param[in]  j_col     - the j-th column in a triangle matrix.
+/// @param[in]  dim       - the dimension of the triangle matrix, i.e., an dim x dim matrix.
+/// @return    the corresponding index in an array storage.
+///
+static inline unsigned int idx_of_2trimat(unsigned int i_row, unsigned int j_col, unsigned int n_var) {
+    if (i_row > j_col) {
+        return PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(j_col, i_row, n_var);
+    }
+    return PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(i_row, j_col, n_var);
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_UpperTrianglize(unsigned char *btriC, const unsigned char *bA, unsigned int Awidth, unsigned int size_batch) {
+    unsigned char *runningC = btriC;
+    unsigned int Aheight = Awidth;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < i; j++) {
+            unsigned int idx = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(j, i, Aheight);
+            gf256v_add(btriC + idx * size_batch, bA + size_batch * (i * Awidth + j), size_batch);
+        }
+        gf256v_add(runningC, bA + size_batch * (i * Awidth + i), size_batch * (Aheight - i));
+        runningC += size_batch * (Aheight - i);
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Awidth = Bheight;
+    unsigned int Aheight = Awidth;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (k < i) {
+                    continue;
+                }
+                gf256v_madd(bC, &btriA[(k - i) * size_batch], PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+        btriA += (Aheight - i) * size_batch;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_trimatTr_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Aheight = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (i < k) {
+                    continue;
+                }
+                gf256v_madd(bC, &btriA[size_batch * (PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(k, i, Aheight))], PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_2trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Aheight = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (i == k) {
+                    continue;
+                }
+                gf256v_madd(bC, &btriA[size_batch * (idx_of_2trimat(i, k, Aheight))], PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_matTr_madd_gf256(unsigned char *bC, const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Atr_height = Awidth;
+    unsigned int Atr_width = Aheight;
+    for (unsigned int i = 0; i < Atr_height; i++) {
+        for (unsigned int j = 0; j < Atr_width; j++) {
+            gf256v_madd(bC, &bB[j * Bwidth * size_batch], PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(&A_to_tr[size_Acolvec * i], j), size_batch * Bwidth);
+        }
+        bC += size_batch * Bwidth;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_bmatTr_madd_gf256(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    const unsigned char *bA = bA_to_tr;
+    unsigned int Aheight = Awidth_before_tr;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                gf256v_madd(bC, &bA[size_batch * (i + k * Aheight)], PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_mat_madd_gf256(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Awidth = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                gf256v_madd(bC, &bA[k * size_batch], PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+        bA += (Awidth) * size_batch;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_quad_trimat_eval_gf256(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch) {
+    unsigned char tmp[256];
+
+    unsigned char _x[256];
+    for (unsigned int i = 0; i < dim; i++) {
+        _x[i] = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(x, i);
+    }
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(y, size_batch);
+    for (unsigned int i = 0; i < dim; i++) {
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(tmp, size_batch);
+        for (unsigned int j = i; j < dim; j++) {
+            gf256v_madd(tmp, trimat, _x[j], size_batch);
+            trimat += size_batch;
+        }
+        gf256v_madd(y, tmp, _x[i], size_batch);
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_quad_recmat_eval_gf256(unsigned char *z, const unsigned char *y, unsigned int dim_y, const unsigned char *mat,
+        const unsigned char *x, unsigned dim_x, unsigned size_batch) {
+    unsigned char tmp[128];
+
+    unsigned char _x[128];
+    for (unsigned int i = 0; i < dim_x; i++) {
+        _x[i] = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(x, i);
+    }
+    unsigned char _y[128];
+    for (unsigned int i = 0; i < dim_y; i++) {
+        _y[i] = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele(y, i);
+    }
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(z, size_batch);
+    for (unsigned int i = 0; i < dim_y; i++) {
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(tmp, size_batch);
+        for (unsigned int j = 0; j < dim_x; j++) {
+            gf256v_madd(tmp, mat, _x[j], size_batch);
+            mat += size_batch;
+        }
+        gf256v_madd(z, tmp, _y[i], size_batch);
+    }
+}
+

crypto_sign/rainbowIIIc-cyclic-compressed/clean/parallel_matrix_op.h

@@ -0,0 +1,260 @@
+#ifndef _P_MATRIX_OP_H_
+#define _P_MATRIX_OP_H_
+///  @file  parallel_matrix_op.h
+///  @brief Librarys for operations of batched matrixes.
+///
+///
+
+////////////////  Section:  triangle matrix <-> rectangle matrix   ///////////////////////////////////
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle(UT) matrix.
+///
+/// @param[in]  i_row     - the i-th row in an upper-triangle matrix.
+/// @param[in]  j_col     - the j-th column in an upper-triangle matrix.
+/// @param[in]  dim       - the dimension of the upper-triangle matrix, i.e., an dim x dim matrix.
+/// @return  the corresponding index in an array storage.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(unsigned int i_row, unsigned int j_col, unsigned int dim);
+
+///
+/// @brief  Upper trianglize a rectangle matrix to the corresponding upper-trangle matrix.
+///
+/// @param[out]  btriC    - the batched upper-trianglized matrix C.
+/// @param[in]   bA       - a batched retangle matrix A.
+/// @param[in]   bwidth   - the width of the batched matrix A, i.e., A is a Awidth x Awidth matrix.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_UpperTrianglize(unsigned char *btriC, const unsigned char *bA, unsigned int Awidth, unsigned int size_batch);
+
+////////////////////  Section:  matrix multiplications  ///////////////////////////////
+
+///
+/// @brief  bC += btriA * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA^Tr * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. A will be transposed while multiplying.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_trimatTr_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA^Tr * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A, which will be transposed while multiplying.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_trimatTr_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC +=  (btriA + btriA^Tr) *B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. The operand for multiplication is (btriA + btriA^Tr).
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_2trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC +=  (btriA + btriA^Tr) *B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. The operand for multiplication is (btriA + btriA^Tr).
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_2trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += A^Tr * bB  , in GF(16)
+///
+/// @param[out]  bC           - the batched matrix C.
+/// @param[in]   A_to_tr      - a column-major matrix A. The operand for multiplication is A^Tr.
+/// @param[in]   Aheight      - the height of A.
+/// @param[in]   size_Acolvec    - the size of a column vector in A.
+/// @param[in]   Awidth           - the width of A.
+/// @param[in]   bB          - a batched matrix B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_matTr_madd_gf16(unsigned char *bC,
+        const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += A^Tr * bB  , in GF(256)
+///
+/// @param[out]  bC           - the batched matrix C.
+/// @param[in]   A_to_tr      - a column-major matrix A. The operand for multiplication is A^Tr.
+/// @param[in]   Aheight      - the height of A.
+/// @param[in]   size_Acolvec    - the size of a column vector in A.
+/// @param[in]   Awidth           - the width of A.
+/// @param[in]   bB          - a batched matrix B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_matTr_madd_gf256(unsigned char *bC,
+        const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA^Tr * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA_to_tr   - a batched matrix A. The operand for multiplication is (bA^Tr).
+/// @param[in]   Awidth_befor_tr     - the width of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_bmatTr_madd_gf16(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA^Tr * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA_to_tr   - a batched matrix A. The operand for multiplication is (bA^Tr).
+/// @param[in]   Awidth_befor_tr     - the width of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_bmatTr_madd_gf256(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA         - a batched matrix A.
+/// @param[in]   Aheigh     - the height of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_mat_madd_gf16(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA         - a batched matrix A.
+/// @param[in]   Aheigh     - the height of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_mat_madd_gf256(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+////////////////////  Section: "quadratric" matrix evaluation  ///////////////////////////////
+
+///
+/// @brief  y =  x^Tr * trimat * x  , in GF(16)
+///
+/// @param[out]  y          - the returned batched element y.
+/// @param[in]   trimat     - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim        - the dimension of matrix trimat (and x).
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_quad_trimat_eval_gf16(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch);
+
+///
+/// @brief  y =  x^Tr * trimat * x  , in GF(256)
+///
+/// @param[out]  y          - the returned batched element y.
+/// @param[in]   trimat     - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim        - the dimension of matrix trimat (and x).
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_quad_trimat_eval_gf256(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch);
+
+///
+/// @brief  z =  y^Tr * mat * x  , in GF(16)
+///
+/// @param[out]  z          - the returned batched element z.
+/// @param[in]   y          - an input vector y.
+/// @param[in]   dim_y      - the length of y.
+/// @param[in]   mat        - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim_x        - the length of x.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_quad_recmat_eval_gf16(unsigned char *z, const unsigned char *y, unsigned int dim_y,
+        const unsigned char *mat, const unsigned char *x, unsigned int dim_x, unsigned int size_batch);
+
+///
+/// @brief  z =  y^Tr * mat * x  , in GF(256)
+///
+/// @param[out]  z          - the returned batched element z.
+/// @param[in]   y          - an input vector y.
+/// @param[in]   dim_y      - the length of y.
+/// @param[in]   mat        - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim_x        - the length of x.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_quad_recmat_eval_gf256(unsigned char *z, const unsigned char *y, unsigned int dim_y,
+        const unsigned char *mat, const unsigned char *x, unsigned int dim_x, unsigned int size_batch);
+
+#endif // _P_MATRIX_OP_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/rainbow.c

@@ -0,0 +1,180 @@
+/// @file rainbow.c
+/// @brief The standard implementations for functions in rainbow.h
+///
+
+#include "rainbow.h"
+#include "blas.h"
+#include "rainbow_blas.h"
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+#include "utils_hash.h"
+#include "utils_prng.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#define MAX_ATTEMPT_FRMAT 128
+#define _MAX_O ((_O1 > _O2) ? _O1 : _O2)
+#define _MAX_O_BYTE ((_O1_BYTE > _O2_BYTE) ? _O1_BYTE : _O2_BYTE)
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_sign(uint8_t *signature, const sk_t *sk, const uint8_t *_digest) {
+    uint8_t mat_l1[_O1 * _O1_BYTE];
+    uint8_t mat_l2[_O2 * _O2_BYTE];
+    uint8_t mat_buffer[2 * _MAX_O * _MAX_O_BYTE];
+
+    // setup PRNG
+    prng_t prng_sign;
+    uint8_t prng_preseed[LEN_SKSEED + _HASH_LEN];
+    memcpy(prng_preseed, sk->sk_seed, LEN_SKSEED);
+    memcpy(prng_preseed + LEN_SKSEED, _digest, _HASH_LEN); // prng_preseed = sk_seed || digest
+    uint8_t prng_seed[_HASH_LEN];
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(prng_seed, _HASH_LEN, prng_preseed, _HASH_LEN + LEN_SKSEED);
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_set(&prng_sign, prng_seed, _HASH_LEN); // seed = H( sk_seed || digest )
+    for (unsigned int i = 0; i < LEN_SKSEED + _HASH_LEN; i++) {
+        prng_preseed[i] ^= prng_preseed[i]; // clean
+    }
+    for (unsigned int i = 0; i < _HASH_LEN; i++) {
+        prng_seed[i] ^= prng_seed[i]; // clean
+    }
+
+    // roll vinegars.
+    uint8_t vinegar[_V1_BYTE];
+    unsigned int n_attempt = 0;
+    unsigned int l1_succ = 0;
+    while (!l1_succ) {
+        if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+            break;
+        }
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(&prng_sign, vinegar, _V1_BYTE);   // generating vinegars
+        gfmat_prod(mat_l1, sk->l1_F2, _O1 * _O1_BYTE, _V1, vinegar); // generating the linear equations for layer 1
+        l1_succ = gfmat_inv(mat_l1, mat_l1, _O1, mat_buffer);        // check if the linear equation solvable
+        n_attempt++;
+    }
+
+    // Given the vinegars, pre-compute variables needed for layer 2
+    uint8_t r_l1_F1[_O1_BYTE] = {0};
+    uint8_t r_l2_F1[_O2_BYTE] = {0};
+    batch_quad_trimat_eval(r_l1_F1, sk->l1_F1, vinegar, _V1, _O1_BYTE);
+    batch_quad_trimat_eval(r_l2_F1, sk->l2_F1, vinegar, _V1, _O2_BYTE);
+    uint8_t mat_l2_F3[_O2 * _O2_BYTE];
+    uint8_t mat_l2_F2[_O1 * _O2_BYTE];
+    gfmat_prod(mat_l2_F3, sk->l2_F3, _O2 * _O2_BYTE, _V1, vinegar);
+    gfmat_prod(mat_l2_F2, sk->l2_F2, _O1 * _O2_BYTE, _V1, vinegar);
+
+    // Some local variables.
+    uint8_t _z[_PUB_M_BYTE];
+    uint8_t y[_PUB_M_BYTE];
+    uint8_t *x_v1 = vinegar;
+    uint8_t x_o1[_O1_BYTE];
+    uint8_t x_o2[_O1_BYTE];
+
+    uint8_t digest_salt[_HASH_LEN + _SALT_BYTE];
+    memcpy(digest_salt, _digest, _HASH_LEN);
+    uint8_t *salt = digest_salt + _HASH_LEN;
+
+    uint8_t temp_o[_MAX_O_BYTE + 32] = {0};
+    unsigned int succ = 0;
+    while (!succ) {
+        if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+            break;
+        }
+        // The computation:  H(digest||salt)  -->   z   --S-->   y  --C-map-->   x   --T-->   w
+
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(&prng_sign, salt, _SALT_BYTE);                         // roll the salt
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(_z, _PUB_M_BYTE, digest_salt, _HASH_LEN + _SALT_BYTE); // H(digest||salt)
+
+        //  y = S^-1 * z
+        memcpy(y, _z, _PUB_M_BYTE); // identity part of S
+        gfmat_prod(temp_o, sk->s1, _O1_BYTE, _O2, _z + _O1_BYTE);
+        gf256v_add(y, temp_o, _O1_BYTE);
+
+        // Central Map:
+        // layer 1: calculate x_o1
+        memcpy(temp_o, r_l1_F1, _O1_BYTE);
+        gf256v_add(temp_o, y, _O1_BYTE);
+        gfmat_prod(x_o1, mat_l1, _O1_BYTE, _O1, temp_o);
+
+        // layer 2: calculate x_o2
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_set_zero(temp_o, _O2_BYTE);
+        gfmat_prod(temp_o, mat_l2_F2, _O2_BYTE, _O1, x_o1);             // F2
+        batch_quad_trimat_eval(mat_l2, sk->l2_F5, x_o1, _O1, _O2_BYTE); // F5
+        gf256v_add(temp_o, mat_l2, _O2_BYTE);
+        gf256v_add(temp_o, r_l2_F1, _O2_BYTE); // F1
+        gf256v_add(temp_o, y + _O1_BYTE, _O2_BYTE);
+
+        // generate the linear equations of the 2nd layer
+        gfmat_prod(mat_l2, sk->l2_F6, _O2 * _O2_BYTE, _O1, x_o1); // F6
+        gf256v_add(mat_l2, mat_l2_F3, _O2 * _O2_BYTE);            // F3
+        succ = gfmat_inv(mat_l2, mat_l2, _O2, mat_buffer);
+        gfmat_prod(x_o2, mat_l2, _O2_BYTE, _O2, temp_o); // solve l2 eqs
+
+        n_attempt++;
+    };
+    //  w = T^-1 * y
+    uint8_t w[_PUB_N_BYTE];
+    // identity part of T.
+    memcpy(w, x_v1, _V1_BYTE);
+    memcpy(w + _V1_BYTE, x_o1, _O1_BYTE);
+    memcpy(w + _V2_BYTE, x_o2, _O2_BYTE);
+    // Computing the t1 part.
+    gfmat_prod(y, sk->t1, _V1_BYTE, _O1, x_o1);
+    gf256v_add(w, y, _V1_BYTE);
+    // Computing the t4 part.
+    gfmat_prod(y, sk->t4, _V1_BYTE, _O2, x_o2);
+    gf256v_add(w, y, _V1_BYTE);
+    // Computing the t3 part.
+    gfmat_prod(y, sk->t3, _O1_BYTE, _O2, x_o2);
+    gf256v_add(w + _V1_BYTE, y, _O1_BYTE);
+
+    memset(signature, 0, _SIGNATURE_BYTE); // set the output 0
+    // clean
+    memset(&prng_sign, 0, sizeof(prng_t));
+    memset(vinegar, 0, _V1_BYTE);
+    memset(r_l1_F1, 0, _O1_BYTE);
+    memset(r_l2_F1, 0, _O2_BYTE);
+    memset(_z, 0, _PUB_M_BYTE);
+    memset(y, 0, _PUB_M_BYTE);
+    memset(x_o1, 0, _O1_BYTE);
+    memset(x_o2, 0, _O2_BYTE);
+    memset(temp_o, 0, sizeof(temp_o));
+
+    // return: copy w and salt to the signature.
+    if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+        return -1;
+    }
+    gf256v_add(signature, w, _PUB_N_BYTE);
+    gf256v_add(signature + _PUB_N_BYTE, salt, _SALT_BYTE);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_verify(const uint8_t *digest, const uint8_t *signature, const pk_t *pk) {
+    unsigned char digest_ck[_PUB_M_BYTE];
+    // public_map( digest_ck , pk , signature ); Evaluating the quadratic public polynomials.
+    batch_quad_trimat_eval(digest_ck, pk->pk, signature, _PUB_N, _PUB_M_BYTE);
+
+    unsigned char correct[_PUB_M_BYTE];
+    unsigned char digest_salt[_HASH_LEN + _SALT_BYTE];
+    memcpy(digest_salt, digest, _HASH_LEN);
+    memcpy(digest_salt + _HASH_LEN, signature + _PUB_N_BYTE, _SALT_BYTE);
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(correct, _PUB_M_BYTE, digest_salt, _HASH_LEN + _SALT_BYTE); // H( digest || salt )
+
+    // check consistancy.
+    unsigned char cc = 0;
+    for (unsigned int i = 0; i < _PUB_M_BYTE; i++) {
+        cc |= (digest_ck[i] ^ correct[i]);
+    }
+    return (0 == cc) ? 0 : -1;
+}
+
+///////////////  cyclic version  ///////////////////////////
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_sign_cyclic(uint8_t *signature, const csk_t *csk, const uint8_t *digest) {
+    unsigned char sk[sizeof(sk_t) + 32];
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_generate_secretkey_cyclic((sk_t *)sk, csk->pk_seed, csk->sk_seed); // generating classic secret key.
+    return PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_sign(signature, (sk_t *)sk, digest);
+}
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_verify_cyclic(const uint8_t *digest, const uint8_t *signature, const cpk_t *_pk) {
+    unsigned char pk[sizeof(pk_t) + 32];
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_cpk_to_pk((pk_t *)pk, _pk); // generating classic public key.
+    return PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_verify(digest, signature, (pk_t *)pk);
+}

crypto_sign/rainbowIIIc-cyclic-compressed/clean/rainbow.h

@@ -0,0 +1,50 @@
+#ifndef _RAINBOW_H_
+#define _RAINBOW_H_
+/// @file rainbow.h
+/// @brief APIs for rainbow.
+///
+
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+
+#include <stdint.h>
+
+///
+/// @brief Signing function for classical secret key.
+///
+/// @param[out] signature - the signature.
+/// @param[in]  sk        - the secret key.
+/// @param[in]  digest    - the digest.
+///
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_sign(uint8_t *signature, const sk_t *sk, const uint8_t *digest);
+
+///
+/// @brief Verifying function.
+///
+/// @param[in]  digest    - the digest.
+/// @param[in]  signature - the signature.
+/// @param[in]  pk        - the public key.
+/// @return 0 for successful verified. -1 for failed verification.
+///
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_verify(const uint8_t *digest, const uint8_t *signature, const pk_t *pk);
+
+///
+/// @brief Signing function for compressed secret key of the cyclic rainbow.
+///
+/// @param[out] signature - the signature.
+/// @param[in]  sk        - the compressed secret key.
+/// @param[in]  digest    - the digest.
+///
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_sign_cyclic(uint8_t *signature, const csk_t *sk, const uint8_t *digest);
+
+///
+/// @brief Verifying function for cyclic public keys.
+///
+/// @param[in]  digest    - the digest.
+/// @param[in]  signature - the signature.
+/// @param[in]  pk        - the public key of cyclic rainbow.
+/// @return 0 for successful verified. -1 for failed verification.
+///
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_verify_cyclic(const uint8_t *digest, const uint8_t *signature, const cpk_t *pk);
+
+#endif // _RAINBOW_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/rainbow_blas.h

@@ -0,0 +1,31 @@
+#ifndef _RAINBOW_BLAS_H_
+#define _RAINBOW_BLAS_H_
+/// @file rainbow_blas.h
+/// @brief Defining the functions used in rainbow.c acconding to the definitions in rainbow_config.h
+///
+///  Defining the functions used in rainbow.c acconding to the definitions in rainbow_config.h
+
+#include "blas.h"
+#include "parallel_matrix_op.h"
+#include "rainbow_config.h"
+
+
+#define gfv_get_ele PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_get_ele
+#define gfv_mul_scalar PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_mul_scalar
+#define gfv_madd PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256v_madd
+
+#define gfmat_prod PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_prod
+#define gfmat_inv PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_gf256mat_inv
+
+#define batch_trimat_madd PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_trimat_madd_gf256
+#define batch_trimatTr_madd PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_trimatTr_madd_gf256
+#define batch_2trimat_madd PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_2trimat_madd_gf256
+#define batch_matTr_madd PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_matTr_madd_gf256
+#define batch_bmatTr_madd PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_bmatTr_madd_gf256
+#define batch_mat_madd PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_mat_madd_gf256
+
+#define batch_quad_trimat_eval PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_quad_trimat_eval_gf256
+#define batch_quad_recmat_eval PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_batch_quad_recmat_eval_gf256
+
+
+#endif // _RAINBOW_BLAS_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/rainbow_config.h

@@ -0,0 +1,46 @@
+#ifndef _H_RAINBOW_CONFIG_H_
+#define _H_RAINBOW_CONFIG_H_
+
+/// @file rainbow_config.h
+/// @brief Defining the parameters of the Rainbow and the corresponding constants.
+///
+
+#define _GFSIZE 256
+#define _V1 68
+#define _O1 36
+#define _O2 36
+#define _HASH_LEN 48
+
+
+#define _V2 ((_V1) + (_O1))
+
+/// size of N, in # of gf elements.
+#define _PUB_N (_V1 + _O1 + _O2)
+
+/// size of M, in # gf elements.
+#define _PUB_M (_O1 + _O2)
+
+/// size of variables, in # bytes.
+
+// GF256
+#define _V1_BYTE (_V1)
+#define _V2_BYTE (_V2)
+#define _O1_BYTE (_O1)
+#define _O2_BYTE (_O2)
+#define _PUB_N_BYTE (_PUB_N)
+#define _PUB_M_BYTE (_PUB_M)
+
+
+/// length of seed for public key, in # bytes
+#define LEN_PKSEED 32
+
+/// length of seed for secret key, in # bytes
+#define LEN_SKSEED 32
+
+/// length of salt for a signature, in # bytes
+#define _SALT_BYTE 16
+
+/// length of a signature
+#define _SIGNATURE_BYTE (_PUB_N_BYTE + _SALT_BYTE)
+
+#endif //  _H_RAINBOW_CONFIG_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/rainbow_keypair.c

@@ -0,0 +1,188 @@
+/// @file rainbow_keypair.c
+/// @brief implementations of functions in rainbow_keypair.h
+///
+
+#include "rainbow_keypair.h"
+#include "blas.h"
+#include "blas_comm.h"
+#include "rainbow_blas.h"
+#include "rainbow_keypair_computation.h"
+#include "utils_prng.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+static void generate_S_T(unsigned char *s_and_t, prng_t *prng0) {
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, s_and_t, _O1_BYTE * _O2); // S1
+    s_and_t += _O1_BYTE * _O2;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, s_and_t, _V1_BYTE * _O1); // T1
+    s_and_t += _V1_BYTE * _O1;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, s_and_t, _V1_BYTE * _O2); // T2
+    s_and_t += _V1_BYTE * _O2;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, s_and_t, _O1_BYTE * _O2); // T3
+}
+
+static unsigned int generate_l1_F12(unsigned char *sk, prng_t *prng0) {
+    unsigned int n_byte_generated = 0;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, sk, _O1_BYTE * N_TRIANGLE_TERMS(_V1)); // l1_F1
+    sk += _O1_BYTE * N_TRIANGLE_TERMS(_V1);
+    n_byte_generated += _O1_BYTE * N_TRIANGLE_TERMS(_V1);
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, sk, _O1_BYTE * _V1 * _O1); // l1_F2
+    n_byte_generated += _O1_BYTE * _V1 * _O1;
+    return n_byte_generated;
+}
+
+static unsigned int generate_l2_F12356(unsigned char *sk, prng_t *prng0) {
+    unsigned int n_byte_generated = 0;
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, sk, _O2_BYTE * N_TRIANGLE_TERMS(_V1)); // l2_F1
+    sk += _O2_BYTE * N_TRIANGLE_TERMS(_V1);
+    n_byte_generated += _O2_BYTE * N_TRIANGLE_TERMS(_V1);
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _V1 * _O1); // l2_F2
+    sk += _O2_BYTE * _V1 * _O1;
+    n_byte_generated += _O2_BYTE * _V1 * _O1;
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _V1 * _O2); // l2_F3
+    sk += _O2_BYTE * _V1 * _O1;
+    n_byte_generated += _O2_BYTE * _V1 * _O1;
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, sk, _O2_BYTE * N_TRIANGLE_TERMS(_O1)); // l2_F5
+    sk += _O2_BYTE * N_TRIANGLE_TERMS(_O1);
+    n_byte_generated += _O2_BYTE * N_TRIANGLE_TERMS(_O1);
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _O1 * _O2); // l2_F6
+    n_byte_generated += _O2_BYTE * _O1 * _O2;
+
+    return n_byte_generated;
+}
+
+static void generate_B1_B2(unsigned char *sk, prng_t *prng0) {
+    sk += generate_l1_F12(sk, prng0);
+    generate_l2_F12356(sk, prng0);
+}
+
+static void calculate_t4(unsigned char *t2_to_t4, const unsigned char *t1, const unsigned char *t3) {
+    //  t4 = T_sk.t1 * T_sk.t3 - T_sk.t2
+    unsigned char temp[_V1_BYTE + 32];
+    unsigned char *t4 = t2_to_t4;
+    for (unsigned int i = 0; i < _O2; i++) { /// t3 width
+        gfmat_prod(temp, t1, _V1_BYTE, _O1, t3);
+        gf256v_add(t4, temp, _V1_BYTE);
+        t4 += _V1_BYTE;
+        t3 += _O1_BYTE;
+    }
+}
+
+static void obsfucate_l1_polys(unsigned char *l1_polys, const unsigned char *l2_polys, unsigned int n_terms, const unsigned char *s1) {
+    unsigned char temp[_O1_BYTE + 32];
+    while (n_terms--) {
+        gfmat_prod(temp, s1, _O1_BYTE, _O2, l2_polys);
+        gf256v_add(l1_polys, temp, _O1_BYTE);
+        l1_polys += _O1_BYTE;
+        l2_polys += _O2_BYTE;
+    }
+}
+
+///////////////////  Classic //////////////////////////////////
+
+
+/////////////////////   Cyclic   //////////////////////////////////
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_generate_keypair_cyclic(cpk_t *pk, sk_t *sk, const unsigned char *pk_seed, const unsigned char *sk_seed) {
+    memcpy(pk->pk_seed, pk_seed, LEN_PKSEED);
+    memcpy(sk->sk_seed, sk_seed, LEN_SKSEED);
+
+    // prng for sk
+    prng_t prng;
+    prng_t *prng0 = &prng;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_set(prng0, sk_seed, LEN_SKSEED);
+    generate_S_T(sk->s1, prng0); // S,T:  only a part of sk
+
+    unsigned char t2[sizeof(sk->t4)];
+    memcpy(t2, sk->t4, _V1_BYTE * _O2);   // temporarily store t2
+    calculate_t4(sk->t4, sk->t1, sk->t3); // t2 <- t4
+
+    // prng for pk
+    sk_t inst_Qs;
+    sk_t *Qs = &inst_Qs;
+    prng_t *prng1 = &prng;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_set(prng1, pk_seed, LEN_PKSEED);
+    generate_B1_B2(Qs->l1_F1, prng1); // generating l1_Q1, l1_Q2, l2_Q1, l2_Q2, l2_Q3, l2_Q5, l2_Q6
+    obsfucate_l1_polys(Qs->l1_F1, Qs->l2_F1, N_TRIANGLE_TERMS(_V1), sk->s1);
+    obsfucate_l1_polys(Qs->l1_F2, Qs->l2_F2, _V1 * _O1, sk->s1);
+    // so far, the Qs contains l1_F1, l1_F2, l2_F1, l2_F2, l2_F3, l2_F5, l2_F6.
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_calculate_F_from_Q(sk, Qs, sk); // calcuate the rest parts of secret key from Qs and S,T
+
+    unsigned char t4[sizeof(sk->t4)];
+    memcpy(t4, sk->t4, _V1_BYTE * _O2);                      // temporarily store t4
+    memcpy(sk->t4, t2, _V1_BYTE * _O2);                      // restore t2
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_calculate_Q_from_F_cyclic(pk, sk, sk); // calculate the rest parts of public key: l1_Q3, l1_Q5, l1_Q6, l1_Q9, l2_Q9
+    memcpy(sk->t4, t4, _V1_BYTE * _O2);                      // restore t4
+
+    obsfucate_l1_polys(pk->l1_Q3, Qs->l2_F3, _V1 * _O2, sk->s1);
+    obsfucate_l1_polys(pk->l1_Q5, Qs->l2_F5, N_TRIANGLE_TERMS(_O1), sk->s1);
+    obsfucate_l1_polys(pk->l1_Q6, Qs->l2_F6, _O1 * _O2, sk->s1);
+    obsfucate_l1_polys(pk->l1_Q9, pk->l2_Q9, N_TRIANGLE_TERMS(_O2), sk->s1);
+
+    // clean
+    memset(&prng, 0, sizeof(prng_t));
+}
+
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_generate_compact_keypair_cyclic(cpk_t *pk, csk_t *rsk, const unsigned char *pk_seed, const unsigned char *sk_seed) {
+    memcpy(rsk->pk_seed, pk_seed, LEN_PKSEED);
+    memcpy(rsk->sk_seed, sk_seed, LEN_SKSEED);
+    sk_t sk;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_generate_keypair_cyclic(pk, &sk, pk_seed, sk_seed);
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_generate_secretkey_cyclic(sk_t *sk, const unsigned char *pk_seed, const unsigned char *sk_seed) {
+    memcpy(sk->sk_seed, sk_seed, LEN_SKSEED);
+
+    // prng for sk
+    prng_t prng0;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_set(&prng0, sk_seed, LEN_SKSEED);
+    generate_S_T(sk->s1, &prng0);
+    calculate_t4(sk->t4, sk->t1, sk->t3);
+
+    // prng for pk
+    sk_t inst_Qs;
+    sk_t *Qs = &inst_Qs;
+    prng_t prng1;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_set(&prng1, pk_seed, LEN_PKSEED);
+    generate_B1_B2(Qs->l1_F1, &prng1);
+
+    obsfucate_l1_polys(Qs->l1_F1, Qs->l2_F1, N_TRIANGLE_TERMS(_V1), sk->s1);
+    obsfucate_l1_polys(Qs->l1_F2, Qs->l2_F2, _V1 * _O1, sk->s1);
+
+    // calcuate the parts of sk according to pk.
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_calculate_F_from_Q(sk, Qs, sk);
+
+    // clean prng for sk
+    memset(&prng0, 0, sizeof(prng_t));
+}
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_cpk_to_pk(pk_t *rpk, const cpk_t *cpk) {
+    // procedure:  cpk_t --> extcpk_t  --> pk_t
+
+    // convert from cpk_t to extcpk_t
+    ext_cpk_t pk;
+
+    // setup prng
+    prng_t prng0;
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_set(&prng0, cpk->pk_seed, LEN_SKSEED);
+
+    // generating parts of key with prng
+    generate_l1_F12(pk.l1_Q1, &prng0);
+    // copying parts of key from input. l1_Q3, l1_Q5, l1_Q6, l1_Q9
+    memcpy(pk.l1_Q3, cpk->l1_Q3, _O1_BYTE * (_V1 * _O2 + N_TRIANGLE_TERMS(_O1) + _O1 * _O2 + N_TRIANGLE_TERMS(_O2)));
+
+    // generating parts of key with prng
+    generate_l2_F12356(pk.l2_Q1, &prng0);
+    // copying parts of key from input: l2_Q9
+    memcpy(pk.l2_Q9, cpk->l2_Q9, _O2_BYTE * N_TRIANGLE_TERMS(_O2));
+
+    // convert from extcpk_t to pk_t
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_extcpk_to_pk(rpk, &pk);
+}

crypto_sign/rainbowIIIc-cyclic-compressed/clean/rainbow_keypair.h

@@ -0,0 +1,111 @@
+#ifndef _RAINBOW_KEYPAIR_H_
+#define _RAINBOW_KEYPAIR_H_
+/// @file rainbow_keypair.h
+/// @brief Formats of key pairs and functions for generating key pairs.
+/// Formats of key pairs and functions for generating key pairs.
+///
+
+#include "rainbow_config.h"
+
+#define N_TRIANGLE_TERMS(n_var) ((n_var) * ((n_var) + 1) / 2)
+
+/// @brief public key for classic rainbow
+///
+///  public key for classic rainbow
+///
+typedef struct rainbow_publickey {
+    unsigned char pk[(_PUB_M_BYTE)*N_TRIANGLE_TERMS(_PUB_N)];
+} pk_t;
+
+/// @brief secret key for classic rainbow
+///
+/// secret key for classic rainbow
+///
+typedef struct rainbow_secretkey {
+    ///
+    /// seed for generating secret key.
+    /// Generating S, T, and F for classic rainbow.
+    /// Generating S and T only for cyclic rainbow.
+    unsigned char sk_seed[LEN_SKSEED];
+
+    unsigned char s1[_O1_BYTE * _O2]; ///< part of S map
+    unsigned char t1[_V1_BYTE * _O1]; ///< part of T map
+    unsigned char t4[_V1_BYTE * _O2]; ///< part of T map
+    unsigned char t3[_O1_BYTE * _O2]; ///< part of T map
+
+    unsigned char l1_F1[_O1_BYTE * N_TRIANGLE_TERMS(_V1)]; ///< part of C-map, F1, Layer1
+    unsigned char l1_F2[_O1_BYTE * _V1 * _O1];             ///< part of C-map, F2, Layer1
+
+    unsigned char l2_F1[_O2_BYTE * N_TRIANGLE_TERMS(_V1)]; ///< part of C-map, F1, Layer2
+    unsigned char l2_F2[_O2_BYTE * _V1 * _O1];             ///< part of C-map, F2, Layer2
+
+    unsigned char l2_F3[_O2_BYTE * _V1 * _O2];             ///< part of C-map, F3, Layer2
+    unsigned char l2_F5[_O2_BYTE * N_TRIANGLE_TERMS(_O1)]; ///< part of C-map, F5, Layer2
+    unsigned char l2_F6[_O2_BYTE * _O1 * _O2];             ///< part of C-map, F6, Layer2
+} sk_t;
+
+/// @brief public key for cyclic rainbow
+///
+///  public key for cyclic rainbow
+///
+typedef struct rainbow_publickey_cyclic {
+    unsigned char pk_seed[LEN_PKSEED]; ///< seed for generating l1_Q1,l1_Q2,l2_Q1,l2_Q2,l2_Q3,l2_Q5,l2_Q6
+
+    unsigned char l1_Q3[_O1_BYTE * _V1 * _O2];             ///< Q3, layer1
+    unsigned char l1_Q5[_O1_BYTE * N_TRIANGLE_TERMS(_O1)]; ///< Q5, layer1
+    unsigned char l1_Q6[_O1_BYTE * _O1 * _O2];             ///< Q6, layer1
+    unsigned char l1_Q9[_O1_BYTE * N_TRIANGLE_TERMS(_O2)]; ///< Q9, layer1
+
+    unsigned char l2_Q9[_O2_BYTE * N_TRIANGLE_TERMS(_O2)]; ///< Q9, layer2
+} cpk_t;
+
+/// @brief compressed secret key for cyclic rainbow
+///
+/// compressed secret key for cyclic rainbow
+///
+typedef struct rainbow_secretkey_cyclic {
+    unsigned char pk_seed[LEN_PKSEED]; ///< seed for generating a part of public key.
+    unsigned char sk_seed[LEN_SKSEED]; ///< seed for generating a part of secret key.
+} csk_t;
+
+
+///
+/// @brief Generate key pairs for cyclic rainbow.
+///
+/// @param[out] pk        - the public key.
+/// @param[out] sk        - the secret key.
+/// @param[in]  pk_seed   - seed for generating parts of public key.
+/// @param[in]  sk_seed   - seed for generating secret key.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_generate_keypair_cyclic(cpk_t *pk, sk_t *sk, const unsigned char *pk_seed, const unsigned char *sk_seed);
+
+///
+/// @brief Generate compressed key pairs for cyclic rainbow.
+///
+/// @param[out] pk        - the public key.
+/// @param[out] sk        - the compressed secret key.
+/// @param[in]  pk_seed   - seed for generating parts of the public key.
+/// @param[in]  sk_seed   - seed for generating the secret key.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_generate_compact_keypair_cyclic(cpk_t *pk, csk_t *sk, const unsigned char *pk_seed, const unsigned char *sk_seed);
+
+///
+/// @brief Generate secret key for cyclic rainbow.
+///
+/// @param[out] sk        - the secret key.
+/// @param[in]  pk_seed   - seed for generating parts of the pbulic key.
+/// @param[in]  sk_seed   - seed for generating the secret key.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_generate_secretkey_cyclic(sk_t *sk, const unsigned char *pk_seed, const unsigned char *sk_seed);
+
+////////////////////////////////////
+
+///
+/// @brief converting formats of public keys : from cyclic version to classic key
+///
+/// @param[out] pk       - the classic public key.
+/// @param[in]  cpk      - the cyclic  public key.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_cpk_to_pk(pk_t *pk, const cpk_t *cpk);
+
+#endif //  _RAINBOW_KEYPAIR_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/rainbow_keypair_computation.c

@@ -0,0 +1,213 @@
+/// @file rainbow_keypair_computation.c
+/// @brief Implementations for functions in rainbow_keypair_computation.h
+///
+
+#include "rainbow_keypair_computation.h"
+#include "blas.h"
+#include "blas_comm.h"
+#include "rainbow_blas.h"
+#include "rainbow_keypair.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_extcpk_to_pk(pk_t *pk, const ext_cpk_t *cpk) {
+    const unsigned char *idx_l1 = cpk->l1_Q1;
+    const unsigned char *idx_l2 = cpk->l2_Q1;
+    for (unsigned int i = 0; i < _V1; i++) {
+        for (unsigned int j = i; j < _V1; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q2;
+    idx_l2 = cpk->l2_Q2;
+    for (unsigned int i = 0; i < _V1; i++) {
+        for (unsigned int j = _V1; j < _V1 + _O1; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q3;
+    idx_l2 = cpk->l2_Q3;
+    for (unsigned int i = 0; i < _V1; i++) {
+        for (unsigned int j = _V1 + _O1; j < _PUB_N; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q5;
+    idx_l2 = cpk->l2_Q5;
+    for (unsigned int i = _V1; i < _V1 + _O1; i++) {
+        for (unsigned int j = i; j < _V1 + _O1; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q6;
+    idx_l2 = cpk->l2_Q6;
+    for (unsigned int i = _V1; i < _V1 + _O1; i++) {
+        for (unsigned int j = _V1 + _O1; j < _PUB_N; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q9;
+    idx_l2 = cpk->l2_Q9;
+    for (unsigned int i = _V1 + _O1; i < _PUB_N; i++) {
+        for (unsigned int j = i; j < _PUB_N; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+}
+
+static void calculate_F_from_Q_ref(sk_t *Fs, const sk_t *Qs, sk_t *Ts) {
+    // Layer 1
+    // F_sk.l1_F1s[i] = Q_pk.l1_F1s[i]
+    memcpy(Fs->l1_F1, Qs->l1_F1, _O1_BYTE * N_TRIANGLE_TERMS(_V1));
+
+    // F_sk.l1_F2s[i] = ( Q_pk.l1_F1s[i] + Q_pk.l1_F1s[i].transpose() ) * T_sk.t1 + Q_pk.l1_F2s[i]
+    memcpy(Fs->l1_F2, Qs->l1_F2, _O1_BYTE * _V1 * _O1);
+    batch_2trimat_madd(Fs->l1_F2, Qs->l1_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O1_BYTE);
+
+    /*
+          Layer 2
+          computations:
+
+          F_sk.l2_F1s[i] = Q_pk.l2_F1s[i]
+
+          Q1_T1 = Q_pk.l2_F1s[i]*T_sk.t1
+          F_sk.l2_F2s[i] =              Q1_T1 + Q_pk.l2_F2s[i]     + Q_pk.l2_F1s[i].transpose() * T_sk.t1
+          F_sk.l2_F5s[i] = UT( t1_tr* ( Q1_T1 + Q_pk.l2_F2s[i] ) ) + Q_pk.l2_F5s[i]
+
+          Q1_Q1T_T4 =  (Q_pk.l2_F1s[i] + Q_pk.l2_F1s[i].transpose()) * t4
+          #Q1_Q1T_T4 =  Q1_Q1T * t4
+          Q2_T3 = Q_pk.l2_F2s[i]*T_sk.t3
+          F_sk.l2_F3s[i] =           Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i]
+          F_sk.l2_F6s[i] = t1_tr * ( Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i] )
+                          +  Q_pk.l2_F2s[i].transpose() * t4
+                          + (Q_pk.l2_F5s[i] + Q_pk.l2_F5s[i].transpose())*T_sk.t3   + Q_pk.l2_F6s[i]
+
+      */
+    memcpy(Fs->l2_F1, Qs->l2_F1, _O2_BYTE * N_TRIANGLE_TERMS(_V1)); // F_sk.l2_F1s[i] = Q_pk.l2_F1s[i]
+
+    // F_sk.l2_F2s[i] =              Q1_T1 + Q_pk.l2_F2s[i]     + Q_pk.l2_F1s[i].transpose() * T_sk.t1
+    // F_sk.l2_F5s[i] = UT( t1_tr* ( Q1_T1 + Q_pk.l2_F2s[i] ) ) + Q_pk.l2_F5s[i]
+    memcpy(Fs->l2_F2, Qs->l2_F2, _O2_BYTE * _V1 * _O1);
+    batch_trimat_madd(Fs->l2_F2, Qs->l2_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O2_BYTE); // Q1_T1+ Q2
+
+    unsigned char tempQ[_O1 * _O1 * _O2_BYTE + 32];
+    memset(tempQ, 0, _O1 * _O1 * _O2_BYTE);
+    batch_matTr_madd(tempQ, Ts->t1, _V1, _V1_BYTE, _O1, Fs->l2_F2, _O1, _O2_BYTE); // t1_tr*(Q1_T1+Q2)
+    memcpy(Fs->l2_F5, Qs->l2_F5, _O2_BYTE * N_TRIANGLE_TERMS(_O1));                // F5
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_UpperTrianglize(Fs->l2_F5, tempQ, _O1, _O2_BYTE);            // UT( ... )
+
+    batch_trimatTr_madd(Fs->l2_F2, Qs->l2_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O2_BYTE); // F2 = Q1_T1 + Q2 + Q1^tr*t1
+
+    // Q1_Q1T_T4 =  (Q_pk.l2_F1s[i] + Q_pk.l2_F1s[i].transpose()) * t4
+    // Q2_T3 = Q_pk.l2_F2s[i]*T_sk.t3
+    // F_sk.l2_F3s[i] =           Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i]
+    memcpy(Fs->l2_F3, Qs->l2_F3, _V1 * _O2 * _O2_BYTE);
+    batch_2trimat_madd(Fs->l2_F3, Qs->l2_F1, Ts->t4, _V1, _V1_BYTE, _O2, _O2_BYTE);  // Q1_Q1T_T4
+    batch_mat_madd(Fs->l2_F3, Qs->l2_F2, _V1, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE); // Q2_T3
+
+    // F_sk.l2_F6s[i] = t1_tr * ( Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i] )
+    //                +  Q_pk.l2_F2s[i].transpose() * t4
+    //                + (Q_pk.l2_F5s[i] + Q_pk.l2_F5s[i].transpose())*T_sk.t3   + Q_pk.l2_F6s[i]
+    memcpy(Fs->l2_F6, Qs->l2_F6, _O1 * _O2 * _O2_BYTE);
+    batch_matTr_madd(Fs->l2_F6, Ts->t1, _V1, _V1_BYTE, _O1, Fs->l2_F3, _O2, _O2_BYTE); // t1_tr * ( Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i] )
+    batch_2trimat_madd(Fs->l2_F6, Qs->l2_F5, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE);    // (Q_pk.l2_F5s[i] + Q_pk.l2_F5s[i].transpose())*T_sk.t3
+    batch_bmatTr_madd(Fs->l2_F6, Qs->l2_F2, _O1, Ts->t4, _V1, _V1_BYTE, _O2, _O2_BYTE);
+}
+
+static void calculate_Q_from_F_cyclic_ref(cpk_t *Qs, const sk_t *Fs, const sk_t *Ts) {
+    // Layer 1: Computing Q5, Q3, Q6, Q9
+
+    //  Q_pk.l1_F5s[i] = UT( T1tr* (F1 * T1 + F2) )
+    const unsigned char *t2 = Ts->t4;
+    sk_t tempQ;
+    memcpy(tempQ.l1_F2, Fs->l1_F2, _O1_BYTE * _V1 * _O1);
+    batch_trimat_madd(tempQ.l1_F2, Fs->l1_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O1_BYTE); // F1*T1 + F2
+    memset(tempQ.l2_F1, 0, sizeof(tempQ.l2_F1));
+    memset(tempQ.l2_F2, 0, sizeof(tempQ.l2_F2));
+    batch_matTr_madd(tempQ.l2_F1, Ts->t1, _V1, _V1_BYTE, _O1, tempQ.l1_F2, _O1, _O1_BYTE); // T1tr*(F1*T1 + F2)
+    memset(Qs->l1_Q5, 0, _O1_BYTE * N_TRIANGLE_TERMS(_O1));
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_UpperTrianglize(Qs->l1_Q5, tempQ.l2_F1, _O1, _O1_BYTE); // UT( ... )   // Q5
+
+    /*
+          F1_T2     = F1 * t2
+          F2_T3     = F2 * t3
+          F1_F1T_T2 + F2_T3 = F1_T2 + F2_T3 + F1tr * t2
+          Q_pk.l1_F3s[i] =         F1_F1T_T2 + F2_T3
+          Q_pk.l1_F6s[i] = T1tr* ( F1_F1T_T2 + F2_T3 ) + F2tr * t2
+          Q_pk.l1_F9s[i] = UT( T2tr* ( F1_T2 + F2_T3 ) )
+      */
+    memset(Qs->l1_Q3, 0, _O1_BYTE * _V1 * _O2);
+    memset(Qs->l1_Q6, 0, _O1_BYTE * _O1 * _O2);
+    memset(Qs->l1_Q9, 0, _O1_BYTE * N_TRIANGLE_TERMS(_O2));
+
+    batch_trimat_madd(Qs->l1_Q3, Fs->l1_F1, t2, _V1, _V1_BYTE, _O2, _O1_BYTE);       // F1*T2
+    batch_mat_madd(Qs->l1_Q3, Fs->l1_F2, _V1, Ts->t3, _O1, _O1_BYTE, _O2, _O1_BYTE); // F1_T2 + F2_T3
+
+    memset(tempQ.l1_F2, 0, _O1_BYTE * _V1 * _O2);                                    // should be F3. assuming: _O1 >= _O2
+    batch_matTr_madd(tempQ.l1_F2, t2, _V1, _V1_BYTE, _O2, Qs->l1_Q3, _O2, _O1_BYTE); // T2tr * ( F1_T2 + F2_T3 )
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_UpperTrianglize(Qs->l1_Q9, tempQ.l1_F2, _O2, _O1_BYTE);        // Q9
+
+    batch_trimatTr_madd(Qs->l1_Q3, Fs->l1_F1, t2, _V1, _V1_BYTE, _O2, _O1_BYTE); // F1_F1T_T2 + F2_T3  // Q3
+
+    batch_bmatTr_madd(Qs->l1_Q6, Fs->l1_F2, _O1, t2, _V1, _V1_BYTE, _O2, _O1_BYTE);    // F2tr*T2
+    batch_matTr_madd(Qs->l1_Q6, Ts->t1, _V1, _V1_BYTE, _O1, Qs->l1_Q3, _O2, _O1_BYTE); // Q6
+    /*
+          Layer 2
+          Computing Q9:
+
+          F1_T2     = F1 * t2
+          F2_T3     = F2 * t3
+          Q9 = UT( T2tr*( F1*T2 + F2*T3 + F3 )  +  T3tr*( F5*T3 + F6 ) )
+      */
+    sk_t tempQ2;
+    memcpy(tempQ2.l2_F3, Fs->l2_F3, _O2_BYTE * _V1 * _O2);                              /// F3 actually.
+    batch_trimat_madd(tempQ2.l2_F3, Fs->l2_F1, t2, _V1, _V1_BYTE, _O2, _O2_BYTE);       // F1*T2 + F3
+    batch_mat_madd(tempQ2.l2_F3, Fs->l2_F2, _V1, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE); // F1_T2 + F2_T3 + F3
+
+    memset(tempQ.l2_F3, 0, _O2_BYTE * _V1 * _O2);
+    batch_matTr_madd(tempQ.l2_F3, t2, _V1, _V1_BYTE, _O2, tempQ2.l2_F3, _O2, _O2_BYTE); // T2tr * ( ..... )
+
+    memcpy(tempQ.l2_F6, Fs->l2_F6, _O2_BYTE * _O1 * _O2);
+    batch_trimat_madd(tempQ.l2_F6, Fs->l2_F5, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE); // F5*T3 + F6
+
+    batch_matTr_madd(tempQ.l2_F3, Ts->t3, _O1, _O1_BYTE, _O2, tempQ.l2_F6, _O2, _O2_BYTE); // T2tr*( ..... ) + T3tr*( ..... )
+    memset(Qs->l2_Q9, 0, _O2_BYTE * N_TRIANGLE_TERMS(_O2));
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_UpperTrianglize(Qs->l2_Q9, tempQ.l2_F3, _O2, _O2_BYTE); // Q9
+}
+
+// Choosing implementations depends on the macros: _BLAS_SSE_ and _BLAS_AVX2_
+#define calculate_F_from_Q_impl calculate_F_from_Q_ref
+#define calculate_Q_from_F_cyclic_impl calculate_Q_from_F_cyclic_ref
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_calculate_F_from_Q(sk_t *Fs, const sk_t *Qs, sk_t *Ts) {
+    calculate_F_from_Q_impl(Fs, Qs, Ts);
+}
+
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_calculate_Q_from_F_cyclic(cpk_t *Qs, const sk_t *Fs, const sk_t *Ts) {
+    calculate_Q_from_F_cyclic_impl(Qs, Fs, Ts);
+}

crypto_sign/rainbowIIIc-cyclic-compressed/clean/rainbow_keypair_computation.h

@@ -0,0 +1,71 @@
+#ifndef _RAINBOW_KEYPAIR_COMP_H_
+#define _RAINBOW_KEYPAIR_COMP_H_
+/// @file rainbow_keypair_computation.h
+/// @brief Functions for calculating pk/sk while generating keys.
+///
+/// Defining an internal structure of public key.
+/// Functions for calculating pk/sk for key generation.
+///
+
+#include "rainbow_keypair.h"
+
+/// @brief The (internal use) public key for rainbow
+///
+/// The (internal use) public key for rainbow. The public
+/// polynomials are divided into l1_Q1, l1_Q2, ... l1_Q9,
+/// l2_Q1, .... , l2_Q9.
+///
+typedef struct rainbow_extend_publickey {
+    unsigned char l1_Q1[_O1_BYTE * N_TRIANGLE_TERMS(_V1)];
+    unsigned char l1_Q2[_O1_BYTE * _V1 * _O1];
+    unsigned char l1_Q3[_O1_BYTE * _V1 * _O2];
+    unsigned char l1_Q5[_O1_BYTE * N_TRIANGLE_TERMS(_O1)];
+    unsigned char l1_Q6[_O1_BYTE * _O1 * _O2];
+    unsigned char l1_Q9[_O1_BYTE * N_TRIANGLE_TERMS(_O2)];
+
+    unsigned char l2_Q1[_O2_BYTE * N_TRIANGLE_TERMS(_V1)];
+    unsigned char l2_Q2[_O2_BYTE * _V1 * _O1];
+    unsigned char l2_Q3[_O2_BYTE * _V1 * _O2];
+    unsigned char l2_Q5[_O2_BYTE * N_TRIANGLE_TERMS(_O1)];
+    unsigned char l2_Q6[_O2_BYTE * _O1 * _O2];
+    unsigned char l2_Q9[_O2_BYTE * N_TRIANGLE_TERMS(_O2)];
+} ext_cpk_t;
+
+///
+/// @brief converting formats of public keys : from ext_cpk_t version to pk_t
+///
+/// @param[out] pk       - the classic public key.
+/// @param[in]  cpk      - the internel public key.
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_extcpk_to_pk(pk_t *pk, const ext_cpk_t *cpk);
+/////////////////////////////////////////////////
+
+///
+/// @brief Computing public key from secret key
+///
+/// @param[out] Qs       - the public key
+/// @param[in]  Fs       - parts of the secret key: l1_F1, l1_F2, l2_F1, l2_F2, l2_F3, l2_F5, l2_F6
+/// @param[in]  Ts       - parts of the secret key: T1, T4, T3
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_calculate_Q_from_F(ext_cpk_t *Qs, const sk_t *Fs, const sk_t *Ts);
+
+
+///
+/// @brief Computing parts of the sk from parts of pk and sk
+///
+/// @param[out] Fs       - parts of the sk: l1_F1, l1_F2, l2_F1, l2_F2, l2_F3, l2_F5, l2_F6
+/// @param[in]  Qs       - parts of the pk: l1_Q1, l1_Q2, l2_Q1, l2_Q2, l2_Q3, l2_Q5, l2_Q6
+/// @param[in]  Ts       - parts of the sk: T1, T4, T3
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_calculate_F_from_Q(sk_t *Fs, const sk_t *Qs, sk_t *Ts);
+
+///
+/// @brief Computing parts of the pk from the secret key
+///
+/// @param[out] Qs       - parts of the pk: l1_Q3, l1_Q5, l2_Q6, l1_Q9, l2_Q9
+/// @param[in]  Fs       - parts of the sk: l1_F1, l1_F2, l2_F1, l2_F2, l2_F3, l2_F5, l2_F6
+/// @param[in]  Ts       - parts of the sk: T1, T4, T3
+///
+void PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_calculate_Q_from_F_cyclic(cpk_t *Qs, const sk_t *Fs, const sk_t *Ts);
+
+#endif // _RAINBOW_KEYPAIR_COMP_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/sign.c

@@ -0,0 +1,76 @@
+///  @file  sign.c
+///  @brief the implementations for functions in api.h
+///
+///
+
+#include "api.h"
+#include "rainbow.h"
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+#include "randombytes.h"
+#include "utils_hash.h"
+#include <stdlib.h>
+#include <string.h>
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign_keypair(unsigned char *pk, unsigned char *sk) {
+    unsigned char sk_seed[LEN_SKSEED] = {0};
+    randombytes(sk_seed, LEN_SKSEED);
+
+    unsigned char pk_seed[LEN_PKSEED] = {0};
+    randombytes(pk_seed, LEN_PKSEED);
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_generate_compact_keypair_cyclic((cpk_t *)pk, (csk_t *)sk, pk_seed, sk_seed);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign(unsigned char *sm, size_t *smlen, const unsigned char *m, size_t mlen, const unsigned char *sk) {
+    unsigned char digest[_HASH_LEN];
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(digest, _HASH_LEN, m, mlen);
+
+    memcpy(sm, m, mlen);
+    smlen[0] = mlen + _SIGNATURE_BYTE;
+
+    return PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_sign_cyclic(sm + mlen, (const csk_t *)sk, digest);
+}
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign_open(unsigned char *m, size_t *mlen, const unsigned char *sm, size_t smlen, const unsigned char *pk) {
+    int rc;
+    if (_SIGNATURE_BYTE > smlen) {
+        rc = -1;
+    } else {
+        *mlen = smlen - _SIGNATURE_BYTE;
+
+        unsigned char digest[_HASH_LEN];
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(digest, _HASH_LEN, sm, *mlen);
+
+        rc = PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_verify_cyclic(digest, sm + mlen[0], (const cpk_t *)pk);
+    }
+    if (!rc) {
+        memcpy(m, sm, smlen - _SIGNATURE_BYTE);
+    } else { // bad signature
+        *mlen = (size_t) -1;
+        memset(m, 0, smlen);
+    }
+    return rc;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign_signature(
+    uint8_t *sig, size_t *siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *sk) {
+    unsigned char digest[_HASH_LEN];
+
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(digest, _HASH_LEN, m, mlen);
+    *siglen = _SIGNATURE_BYTE;
+    return PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_sign_cyclic(sig, (const csk_t *)sk, digest);
+}
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_crypto_sign_verify(
+    const uint8_t *sig, size_t siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *pk) {
+    if (siglen != _SIGNATURE_BYTE) {
+        return -1;
+    }
+    unsigned char digest[_HASH_LEN];
+    PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(digest, _HASH_LEN, m, mlen);
+    return PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_rainbow_verify_cyclic(digest, sig, (const cpk_t *)pk);
+}

crypto_sign/rainbowIIIc-cyclic-compressed/clean/utils_hash.c

@@ -0,0 +1,50 @@
+/// @file utils_hash.c
+/// @brief the adapter for SHA2 families.
+///
+///
+
+#include "utils_hash.h"
+#include "rainbow_config.h"
+#include "sha2.h"
+
+static inline int _hash(unsigned char *digest, const unsigned char *m, size_t mlen) {
+    sha384(digest, m, mlen);
+    return 0;
+}
+
+static inline int expand_hash(unsigned char *digest, size_t n_digest, const unsigned char *hash) {
+    if (_HASH_LEN >= n_digest) {
+        for (size_t i = 0; i < n_digest; i++) {
+            digest[i] = hash[i];
+        }
+        return 0;
+    }
+    for (size_t i = 0; i < _HASH_LEN; i++) {
+        digest[i] = hash[i];
+    }
+    n_digest -= _HASH_LEN;
+
+    while (_HASH_LEN <= n_digest) {
+        _hash(digest + _HASH_LEN, digest, _HASH_LEN);
+
+        n_digest -= _HASH_LEN;
+        digest += _HASH_LEN;
+    }
+    unsigned char temp[_HASH_LEN];
+    if (n_digest) {
+        _hash(temp, digest, _HASH_LEN);
+        for (size_t i = 0; i < n_digest; i++) {
+            digest[_HASH_LEN + i] = temp[i];
+        }
+    }
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(unsigned char *digest,
+        size_t len_digest,
+        const unsigned char *m,
+        size_t mlen) {
+    unsigned char buf[_HASH_LEN];
+    _hash(buf, m, mlen);
+    return expand_hash(digest, len_digest, buf);
+}

crypto_sign/rainbowIIIc-cyclic-compressed/clean/utils_hash.h

@@ -0,0 +1,11 @@
+#ifndef _UTILS_HASH_H_
+#define _UTILS_HASH_H_
+/// @file utils_hash.h
+/// @brief the interface for adapting hash functions.
+///
+
+#include <stddef.h>
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(unsigned char *digest, size_t len_digest, const unsigned char *m, size_t mlen);
+
+#endif // _UTILS_HASH_H_

crypto_sign/rainbowIIIc-cyclic-compressed/clean/utils_prng.c

@@ -0,0 +1,95 @@
+/// @file utils_prng.c
+/// @brief The implementation of PRNG related functions.
+///
+
+#include "utils_prng.h"
+#include "aes.h"
+#include "randombytes.h"
+#include "utils_hash.h"
+#include <stdlib.h>
+#include <string.h>
+
+static void prng_update(const unsigned char *provided_data,
+                        unsigned char *Key,
+                        unsigned char *V) {
+    unsigned char temp[48];
+    aes256ctx ctx;
+    aes256_keyexp(&ctx, Key);
+    for (int i = 0; i < 3; i++) {
+        //increment V
+        for (int j = 15; j >= 0; j--) {
+            if (V[j] == 0xff) {
+                V[j] = 0x00;
+            } else {
+                V[j]++;
+                break;
+            }
+        }
+        aes256_ecb(temp + 16 * i, V, 1, &ctx);
+    }
+    if (provided_data != NULL) {
+        for (int i = 0; i < 48; i++) {
+            temp[i] ^= provided_data[i];
+        }
+    }
+    memcpy(Key, temp, 32);
+    memcpy(V, temp + 32, 16);
+}
+static void randombytes_init_with_state(prng_t *state,
+                                        unsigned char *entropy_input_48bytes) {
+    memset(state->Key, 0x00, 32);
+    memset(state->V, 0x00, 16);
+    prng_update(entropy_input_48bytes, state->Key, state->V);
+}
+
+static int randombytes_with_state(prng_t *state,
+                                  unsigned char *x,
+                                  size_t xlen) {
+
+    unsigned char block[16];
+    int i = 0;
+
+    aes256ctx ctx;
+    aes256_keyexp(&ctx, state->Key);
+
+    while (xlen > 0) {
+        //increment V
+        for (int j = 15; j >= 0; j--) {
+            if (state->V[j] == 0xff) {
+                state->V[j] = 0x00;
+            } else {
+                state->V[j]++;
+                break;
+            }
+        }
+        aes256_ecb(block, state->V, 1, &ctx);
+        if (xlen > 15) {
+            memcpy(x + i, block, 16);
+            i += 16;
+            xlen -= 16;
+        } else {
+            memcpy(x + i, block, xlen);
+            xlen = 0;
+        }
+    }
+    prng_update(NULL, state->Key, state->V);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_set(prng_t *ctx, const void *prng_seed, unsigned long prng_seedlen) {
+    unsigned char seed[48];
+    if (prng_seedlen >= 48) {
+        memcpy(seed, prng_seed, 48);
+    } else {
+        memcpy(seed, prng_seed, prng_seedlen);
+        PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_hash_msg(seed + prng_seedlen, 48 - (unsigned)prng_seedlen, (const unsigned char *)prng_seed, prng_seedlen);
+    }
+
+    randombytes_init_with_state(ctx, seed);
+
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng_t *ctx, unsigned char *out, unsigned long outlen) {
+    return randombytes_with_state(ctx, out, outlen);
+}

crypto_sign/rainbowIIIc-cyclic-compressed/clean/utils_prng.h

@@ -0,0 +1,18 @@
+#ifndef _UTILS_PRNG_H_
+#define _UTILS_PRNG_H_
+/// @file utils_prng.h
+/// @brief the interface for adapting PRNG functions.
+///
+///
+
+#include "randombytes.h"
+
+typedef struct {
+    unsigned char Key[32];
+    unsigned char V[16];
+} prng_t;
+
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_set(prng_t *ctx, const void *prng_seed, unsigned long prng_seedlen);
+int PQCLEAN_RAINBOWIIICCYCLICCOMPRESSED_CLEAN_prng_gen(prng_t *ctx, unsigned char *out, unsigned long outlen);
+
+#endif // _UTILS_PRNG_H_

crypto_sign/rainbowIIIc-cyclic/META.yml

@@ -0,0 +1,18 @@
+name: Rainbow-IIIc-cyclic
+type: signature
+claimed-nist-level: 3
+length-public-key: 206744
+length-secret-key: 511448
+length-signature: 156
+nistkat-sha256: 607fa94312778210c443431974087ce99357494ab16c8ad5a8418784b811223d
+testvectors-sha256:  51559c38aefe636abd36094741be3b5f65de9e09cf9a5d37866b2e8d0eb58d22
+principal-submitters:
+  - Jintai Ding
+auxiliary-submitters:
+  - Ming-Shing Chen
+  - Albrecht Petzoldt
+  - Dieter Schmidt
+  - Bo-Yin Yang
+implementations:
+    - name: clean
+      version: https://github.com/fast-crypto-lab/rainbow-submission-round2/commit/af826fcb78f6af51a02d0352cff28a9690467bfd

crypto_sign/rainbowIIIc-cyclic/clean/LICENSE

@@ -0,0 +1,8 @@
+`Software implementation of Rainbow for NIST R2 submission' by Ming-Shing Chen
+
+To the extent possible under law, the person who associated CC0 with
+`Software implementation of Rainbow for NIST R2 submission' has waived all copyright and related or neighboring rights
+to `Software implementation of Rainbow for NIST R2 submission'.
+
+You should have received a copy of the CC0 legalcode along with this
+work.  If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.

crypto_sign/rainbowIIIc-cyclic/clean/Makefile

@@ -0,0 +1,20 @@
+# This Makefile can be used with GNU Make or BSD Make
+
+LIB=librainbowIIIc-cyclic_clean.a
+
+HEADERS = api.h blas_comm.h blas.h blas_u32.h gf.h parallel_matrix_op.h rainbow_blas.h rainbow_config.h rainbow.h rainbow_keypair_computation.h rainbow_keypair.h utils_hash.h utils_prng.h 
+OBJECTS =  blas_comm.o parallel_matrix_op.o rainbow.o rainbow_keypair.o rainbow_keypair_computation.o sign.o utils_hash.o utils_prng.o blas_u32.o gf.o
+
+CFLAGS=-O3 -Wall -Wconversion -Wextra -Wpedantic -Wvla -Werror -Wmissing-prototypes -Wredundant-decls -std=c99 -I../../../common $(EXTRAFLAGS)
+
+all: $(LIB)
+
+%.o: %.c $(HEADERS)
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(LIB): $(OBJECTS)
+	$(AR) -r $@ $(OBJECTS)
+
+clean:
+	$(RM) $(OBJECTS)
+	$(RM) $(LIB)

crypto_sign/rainbowIIIc-cyclic/clean/Makefile.Microsoft_nmake

@@ -0,0 +1,19 @@
+# This Makefile can be used with Microsoft Visual Studio's nmake using the command:
+#    nmake /f Makefile.Microsoft_nmake
+
+LIBRARY=librainbowIIIc-cyclic_clean.lib
+OBJECTS =  blas_comm.obj parallel_matrix_op.obj rainbow.obj rainbow_keypair.obj rainbow_keypair_computation.obj sign.obj utils_hash.obj utils_prng.obj blas_u32.obj gf.obj
+
+CFLAGS=/nologo /I ..\..\..\common /W4 /WX
+
+all: $(LIBRARY)
+
+# Make sure objects are recompiled if headers change.
+$(OBJECTS): *.h
+
+$(LIBRARY): $(OBJECTS)
+	LIB.EXE /NOLOGO /WX /OUT:$@ $**
+
+clean:
+    -DEL $(OBJECTS)
+    -DEL $(LIBRARY)

crypto_sign/rainbowIIIc-cyclic/clean/api.h

@@ -0,0 +1,32 @@
+#ifndef PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_API_H
+#define PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_API_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+#define PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_CRYPTO_SECRETKEYBYTES 511448
+#define PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_CRYPTO_PUBLICKEYBYTES 206744
+#define PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_CRYPTO_BYTES 156
+#define PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_CRYPTO_ALGNAME "RAINBOW(256,68,36,36) - cyclic"
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign_keypair(uint8_t *pk, uint8_t *sk);
+
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign_signature(
+    uint8_t *sig, size_t *siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *sk);
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign_verify(
+    const uint8_t *sig, size_t siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *pk);
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign(uint8_t *sm, size_t *smlen,
+        const uint8_t *m, size_t mlen,
+        const uint8_t *sk);
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign_open(uint8_t *m, size_t *mlen,
+        const uint8_t *sm, size_t smlen,
+        const uint8_t *pk);
+
+
+#endif

crypto_sign/rainbowIIIc-cyclic/clean/blas.h

@@ -0,0 +1,19 @@
+#ifndef _BLAS_H_
+#define _BLAS_H_
+/// @file blas.h
+/// @brief Defining the implementations for linear algebra functions depending on the machine architecture.
+///
+
+#include "blas_comm.h"
+#include "blas_u32.h"
+#include "rainbow_config.h"
+
+#define gf256v_predicated_add PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_predicated_add_u32
+#define gf256v_add PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_add_u32
+
+
+#define gf256v_mul_scalar PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_scalar_u32
+#define gf256v_madd PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_madd_u32
+
+
+#endif // _BLAS_H_

crypto_sign/rainbowIIIc-cyclic/clean/blas_comm.c

@@ -0,0 +1,144 @@
+/// @file blas_comm.c
+/// @brief The standard implementations for blas_comm.h
+///
+
+#include "blas_comm.h"
+#include "blas.h"
+#include "gf.h"
+#include "rainbow_config.h"
+
+#include <stdint.h>
+#include <string.h>
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(uint8_t *b, unsigned int _num_byte) {
+    for (size_t i = 0; i < _num_byte; i++) {
+        b[i] = 0;
+    }
+}
+/// @brief get an element from GF(256) vector .
+///
+/// @param[in]  a         - the input vector a.
+/// @param[in]  i         - the index in the vector a.
+/// @return  the value of the element.
+///
+uint8_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(const uint8_t *a, unsigned int i) {
+    return a[i];
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_is_zero(const uint8_t *a, unsigned int _num_byte) {
+    uint8_t r = 0;
+    while (_num_byte--) {
+        r |= a[0];
+        a++;
+    }
+    return (0 == r);
+}
+
+/// polynomial multplication
+/// School boook
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_polymul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int _num) {
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(c, _num * 2 - 1);
+    for (unsigned int i = 0; i < _num; i++) {
+        gf256v_madd(c + i, a, b[i], _num);
+    }
+}
+
+static void gf256mat_prod_ref(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b) {
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(c, n_A_vec_byte);
+    for (unsigned int i = 0; i < n_A_width; i++) {
+        gf256v_madd(c, matA, b[i], n_A_vec_byte);
+        matA += n_A_vec_byte;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_mul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int len_vec) {
+    unsigned int n_vec_byte = len_vec;
+    for (unsigned int k = 0; k < len_vec; k++) {
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(c, n_vec_byte);
+        const uint8_t *bk = b + n_vec_byte * k;
+        for (unsigned int i = 0; i < len_vec; i++) {
+            gf256v_madd(c, a + n_vec_byte * i, bk[i], n_vec_byte);
+        }
+        c += n_vec_byte;
+    }
+}
+
+static unsigned int gf256mat_gauss_elim_ref(uint8_t *mat, unsigned int h, unsigned int w) {
+    unsigned int r8 = 1;
+
+    for (unsigned int i = 0; i < h; i++) {
+        uint8_t *ai = mat + w * i;
+        unsigned int skip_len_align4 = i & ((unsigned int)~0x3);
+
+        for (unsigned int j = i + 1; j < h; j++) {
+            uint8_t *aj = mat + w * j;
+            gf256v_predicated_add(ai + skip_len_align4, !PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256_is_nonzero(ai[i]), aj + skip_len_align4, w - skip_len_align4);
+        }
+        r8 &= PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256_is_nonzero(ai[i]);
+        uint8_t pivot = ai[i];
+        pivot = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256_inv(pivot);
+        gf256v_mul_scalar(ai + skip_len_align4, pivot, w - skip_len_align4);
+        for (unsigned int j = 0; j < h; j++) {
+            if (i == j) {
+                continue;
+            }
+            uint8_t *aj = mat + w * j;
+            gf256v_madd(aj + skip_len_align4, ai + skip_len_align4, aj[i], w - skip_len_align4);
+        }
+    }
+
+    return r8;
+}
+
+static unsigned int gf256mat_solve_linear_eq_ref(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n) {
+    uint8_t mat[64 * 64];
+    for (unsigned int i = 0; i < n; i++) {
+        memcpy(mat + i * (n + 1), inp_mat + i * n, n);
+        mat[i * (n + 1) + n] = c_terms[i];
+    }
+    unsigned int r8 = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_gauss_elim(mat, n, n + 1);
+    for (unsigned int i = 0; i < n; i++) {
+        sol[i] = mat[i * (n + 1) + n];
+    }
+    return r8;
+}
+
+static inline void gf256mat_submat(uint8_t *mat2, unsigned int w2, unsigned int st, const uint8_t *mat, unsigned int w, unsigned int h) {
+    for (unsigned int i = 0; i < h; i++) {
+        for (unsigned int j = 0; j < w2; j++) {
+            mat2[i * w2 + j] = mat[i * w + st + j];
+        }
+    }
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_inv(uint8_t *inv_a, const uint8_t *a, unsigned int H, uint8_t *buffer) {
+    uint8_t *aa = buffer;
+    for (unsigned int i = 0; i < H; i++) {
+        uint8_t *ai = aa + i * 2 * H;
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(ai, 2 * H);
+        gf256v_add(ai, a + i * H, H);
+        ai[H + i] = 1;
+    }
+    unsigned int r8 = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_gauss_elim(aa, H, 2 * H);
+    gf256mat_submat(inv_a, H, H, aa, 2 * H, H);
+    return r8;
+}
+
+
+// choosing the implementations depends on the macros _BLAS_AVX2_ and _BLAS_SSE
+
+#define gf256mat_prod_impl gf256mat_prod_ref
+#define gf256mat_gauss_elim_impl gf256mat_gauss_elim_ref
+#define gf256mat_solve_linear_eq_impl gf256mat_solve_linear_eq_ref
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_prod(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b) {
+    gf256mat_prod_impl(c, matA, n_A_vec_byte, n_A_width, b);
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_gauss_elim(uint8_t *mat, unsigned int h, unsigned int w) {
+    return gf256mat_gauss_elim_impl(mat, h, w);
+}
+
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_solve_linear_eq(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n) {
+    return gf256mat_solve_linear_eq_impl(sol, inp_mat, c_terms, n);
+}
+

crypto_sign/rainbowIIIc-cyclic/clean/blas_comm.h

@@ -0,0 +1,90 @@
+#ifndef _BLAS_COMM_H_
+#define _BLAS_COMM_H_
+/// @file blas_comm.h
+/// @brief Common functions for linear algebra.
+///
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+/// @brief set a vector to 0.
+///
+/// @param[in,out]   b      - the vector b.
+/// @param[in]  _num_byte   - number of bytes for the vector b.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(uint8_t *b, unsigned int _num_byte);
+
+/// @brief get an element from GF(256) vector .
+///
+/// @param[in]  a         - the input vector a.
+/// @param[in]  i         - the index in the vector a.
+/// @return  the value of the element.
+///
+uint8_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(const uint8_t *a, unsigned int i);
+
+/// @brief check if a vector is 0.
+///
+/// @param[in]   a          - the vector a.
+/// @param[in]  _num_byte   - number of bytes for the vector a.
+/// @return  1(true) if a is 0. 0(false) else.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_is_zero(const uint8_t *a, unsigned int _num_byte);
+
+/// @brief polynomial multiplication:  c = a*b
+///
+/// @param[out]   c         - the output polynomial c
+/// @param[in]   a          - the vector a.
+/// @param[in]   b          - the vector b.
+/// @param[in]  _num   - number of elements for the polynomials a and b.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_polymul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int _num);
+
+/// @brief matrix-vector multiplication:  c = matA * b , in GF(256)
+///
+/// @param[out]  c         - the output vector c
+/// @param[in]   matA      - a column-major matrix A.
+/// @param[in]   n_A_vec_byte  - the size of column vectors in bytes.
+/// @param[in]   n_A_width   - the width of matrix A.
+/// @param[in]   b          - the vector b.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_prod(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b);
+
+/// @brief matrix-matrix multiplication:  c = a * b , in GF(256)
+///
+/// @param[out]  c         - the output matrix c
+/// @param[in]   c         - a matrix a.
+/// @param[in]   b         - a matrix b.
+/// @param[in]   len_vec   - the length of column vectors.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_mul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int len_vec);
+
+/// @brief Gauss elimination for a matrix, in GF(256)
+///
+/// @param[in,out]  mat    - the matrix.
+/// @param[in]   h         - the height of the matrix.
+/// @param[in]   w         - the width of the matrix.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_gauss_elim(uint8_t *mat, unsigned int h, unsigned int w);
+
+/// @brief Solving linear equations, in GF(256)
+///
+/// @param[out]  sol       - the solutions.
+/// @param[in]   inp_mat   - the matrix parts of input equations.
+/// @param[in]   c_terms   - the constant terms of the input equations.
+/// @param[in]   n         - the number of equations.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_solve_linear_eq(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n);
+
+/// @brief Computing the inverse matrix, in GF(256)
+///
+/// @param[out]  inv_a     - the output of matrix a.
+/// @param[in]   a         - a matrix a.
+/// @param[in]   H         - height of matrix a, i.e., matrix a is an HxH matrix.
+/// @param[in]   buffer    - The buffer for computations. it has to be as large as 2 input matrixes.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_inv(uint8_t *inv_a, const uint8_t *a, unsigned int H, uint8_t *buffer);
+
+#endif // _BLAS_COMM_H_

crypto_sign/rainbowIIIc-cyclic/clean/blas_u32.c

@@ -0,0 +1,87 @@
+#include "blas_u32.h"
+#include "gf.h"
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_predicated_add_u32(uint8_t *accu_b, uint8_t predicate, const uint8_t *a, unsigned int _num_byte) {
+    uint32_t pr_u32 = ((uint32_t)0) - ((uint32_t)predicate);
+    uint8_t pr_u8 = pr_u32 & 0xff;
+
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *b_u32 = (uint32_t *)accu_b;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        b_u32[i] ^= (a_u32[i] & pr_u32);
+    }
+
+    a += (n_u32 << 2);
+    accu_b += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_b[i] ^= (a[i] & pr_u8);
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_add_u32(uint8_t *accu_b, const uint8_t *a, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *b_u32 = (uint32_t *)accu_b;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        b_u32[i] ^= a_u32[i];
+    }
+
+    a += (n_u32 << 2);
+    accu_b += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_b[i] ^= a[i];
+    }
+}
+
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_scalar_u32(uint8_t *a, uint8_t b, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *a_u32 = (uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        a_u32[i] = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_u32(a_u32[i], b);
+    }
+
+    union tmp_32 {
+        uint8_t u8[4];
+        uint32_t u32;
+    } t;
+    t.u32 = 0;
+    a += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        t.u8[i] = a[i];
+    }
+    t.u32 = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_u32(t.u32, b);
+    for (unsigned int i = 0; i < rem; i++) {
+        a[i] = t.u8[i];
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_madd_u32(uint8_t *accu_c, const uint8_t *a, uint8_t gf256_b, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *c_u32 = (uint32_t *)accu_c;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        c_u32[i] ^= PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_u32(a_u32[i], gf256_b);
+    }
+
+    union tmp_32 {
+        uint8_t u8[4];
+        uint32_t u32;
+    } t;
+    t.u32 = 0;
+    accu_c += (n_u32 << 2);
+    a += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        t.u8[i] = a[i];
+    }
+    t.u32 = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_u32(t.u32, gf256_b);
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_c[i] ^= t.u8[i];
+    }
+}
+

crypto_sign/rainbowIIIc-cyclic/clean/blas_u32.h

@@ -0,0 +1,18 @@
+#ifndef _BLAS_U32_H_
+#define _BLAS_U32_H_
+/// @file blas_u32.h
+/// @brief Inlined functions for implementing basic linear algebra functions for uint32 arch.
+///
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_predicated_add_u32(uint8_t *accu_b, uint8_t predicate, const uint8_t *a, unsigned int _num_byte);
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_add_u32(uint8_t *accu_b, const uint8_t *a, unsigned int _num_byte);
+
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_scalar_u32(uint8_t *a, uint8_t b, unsigned int _num_byte);
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_madd_u32(uint8_t *accu_c, const uint8_t *a, uint8_t gf256_b, unsigned int _num_byte);
+
+
+#endif // _BLAS_U32_H_

crypto_sign/rainbowIIIc-cyclic/clean/gf.c

@@ -0,0 +1,134 @@
+#include "gf.h"
+
+//// gf4 := gf2[x]/x^2+x+1
+static inline uint8_t gf4_mul_2(uint8_t a) {
+    uint8_t r = (uint8_t)(a << 1);
+    r ^= (uint8_t)((a >> 1) * 7);
+    return r;
+}
+
+static inline uint8_t gf4_mul(uint8_t a, uint8_t b) {
+    uint8_t r = (uint8_t)(a * (b & 1));
+    return r ^ (uint8_t)(gf4_mul_2(a) * (b >> 1));
+}
+
+static inline uint8_t gf4_squ(uint8_t a) {
+    return a ^ (a >> 1);
+}
+
+static inline uint32_t gf4v_mul_2_u32(uint32_t a) {
+    uint32_t bit0 = a & 0x55555555;
+    uint32_t bit1 = a & 0xaaaaaaaa;
+    return (bit0 << 1) ^ bit1 ^ (bit1 >> 1);
+}
+
+static inline uint32_t gf4v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t bit0_b = ((uint32_t)0) - ((uint32_t)(b & 1));
+    uint32_t bit1_b = ((uint32_t)0) - ((uint32_t)((b >> 1) & 1));
+    return (a & bit0_b) ^ (bit1_b & gf4v_mul_2_u32(a));
+}
+
+//// gf16 := gf4[y]/y^2+y+x
+static inline uint8_t gf16_mul(uint8_t a, uint8_t b) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = (a >> 2);
+    uint8_t b0 = b & 3;
+    uint8_t b1 = (b >> 2);
+    uint8_t a0b0 = gf4_mul(a0, b0);
+    uint8_t a1b1 = gf4_mul(a1, b1);
+    uint8_t a0b1_a1b0 = gf4_mul(a0 ^ a1, b0 ^ b1) ^ a0b0 ^ a1b1;
+    uint8_t a1b1_x2 = gf4_mul_2(a1b1);
+    return (uint8_t)((a0b1_a1b0 ^ a1b1) << 2 ^ a0b0 ^ a1b1_x2);
+}
+
+static inline uint8_t gf16_squ(uint8_t a) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = (a >> 2);
+    a1 = gf4_squ(a1);
+    uint8_t a1squ_x2 = gf4_mul_2(a1);
+    return (uint8_t)((a1 << 2) ^ a1squ_x2 ^ gf4_squ(a0));
+}
+
+// gf16 := gf4[y]/y^2+y+x
+uint32_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf16v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t axb0 = gf4v_mul_u32(a, b);
+    uint32_t axb1 = gf4v_mul_u32(a, b >> 2);
+    uint32_t a0b1 = (axb1 << 2) & 0xcccccccc;
+    uint32_t a1b1 = axb1 & 0xcccccccc;
+    uint32_t a1b1_2 = a1b1 >> 2;
+
+    return axb0 ^ a0b1 ^ a1b1 ^ gf4v_mul_2_u32(a1b1_2);
+}
+
+uint8_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256_is_nonzero(uint8_t a) {
+    unsigned int a8 = a;
+    unsigned int r = ((unsigned int)0) - a8;
+    r >>= 8;
+    return r & 1;
+}
+
+static inline uint8_t gf4_mul_3(uint8_t a) {
+    uint8_t msk = (uint8_t)((a - 2) >> 1);
+    return (uint8_t)((msk & ((int)a * 3)) | ((~msk) & ((int)a - 1)));
+}
+static inline uint8_t gf16_mul_8(uint8_t a) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = a >> 2;
+    return (uint8_t)((gf4_mul_2(a0 ^ a1) << 2) | gf4_mul_3(a1));
+}
+
+// gf256 := gf16[X]/X^2+X+xy
+static inline uint8_t gf256_mul(uint8_t a, uint8_t b) {
+    uint8_t a0 = a & 15;
+    uint8_t a1 = (a >> 4);
+    uint8_t b0 = b & 15;
+    uint8_t b1 = (b >> 4);
+    uint8_t a0b0 = gf16_mul(a0, b0);
+    uint8_t a1b1 = gf16_mul(a1, b1);
+    uint8_t a0b1_a1b0 = gf16_mul(a0 ^ a1, b0 ^ b1) ^ a0b0 ^ a1b1;
+    uint8_t a1b1_x8 = gf16_mul_8(a1b1);
+    return (uint8_t)((a0b1_a1b0 ^ a1b1) << 4 ^ a0b0 ^ a1b1_x8);
+}
+
+static inline uint8_t gf256_squ(uint8_t a) {
+    uint8_t a0 = a & 15;
+    uint8_t a1 = (a >> 4);
+    a1 = gf16_squ(a1);
+    uint8_t a1squ_x8 = gf16_mul_8(a1);
+    return (uint8_t)((a1 << 4) ^ a1squ_x8 ^ gf16_squ(a0));
+}
+
+uint8_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256_inv(uint8_t a) {
+    // 128+64+32+16+8+4+2 = 254
+    uint8_t a2 = gf256_squ(a);
+    uint8_t a4 = gf256_squ(a2);
+    uint8_t a8 = gf256_squ(a4);
+    uint8_t a4_2 = gf256_mul(a4, a2);
+    uint8_t a8_4_2 = gf256_mul(a4_2, a8);
+    uint8_t a64_ = gf256_squ(a8_4_2);
+    a64_ = gf256_squ(a64_);
+    a64_ = gf256_squ(a64_);
+    uint8_t a64_2 = gf256_mul(a64_, a8_4_2);
+    uint8_t a128_ = gf256_squ(a64_2);
+    return gf256_mul(a2, a128_);
+}
+
+static inline uint32_t gf4v_mul_3_u32(uint32_t a) {
+    uint32_t bit0 = a & 0x55555555;
+    uint32_t bit1 = a & 0xaaaaaaaa;
+    return (bit0 << 1) ^ bit0 ^ (bit1 >> 1);
+}
+static inline uint32_t gf16v_mul_8_u32(uint32_t a) {
+    uint32_t a1 = a & 0xcccccccc;
+    uint32_t a0 = (a << 2) & 0xcccccccc;
+    return gf4v_mul_2_u32(a0 ^ a1) | gf4v_mul_3_u32(a1 >> 2);
+}
+uint32_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t axb0 = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf16v_mul_u32(a, b);
+    uint32_t axb1 = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf16v_mul_u32(a, b >> 4);
+    uint32_t a0b1 = (axb1 << 4) & 0xf0f0f0f0;
+    uint32_t a1b1 = axb1 & 0xf0f0f0f0;
+    uint32_t a1b1_4 = a1b1 >> 4;
+
+    return axb0 ^ a0b1 ^ a1b1 ^ gf16v_mul_8_u32(a1b1_4);
+}

crypto_sign/rainbowIIIc-cyclic/clean/gf.h

@@ -0,0 +1,19 @@
+#ifndef _GF16_H_
+#define _GF16_H_
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+/// @file gf16.h
+/// @brief Library for arithmetics in GF(16) and GF(256)
+///
+
+uint32_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf16v_mul_u32(uint32_t a, uint8_t b);
+
+
+uint8_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256_is_nonzero(uint8_t a);
+uint8_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256_inv(uint8_t a);
+uint32_t PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_u32(uint32_t a, uint8_t b);
+
+
+#endif // _GF16_H_

crypto_sign/rainbowIIIc-cyclic/clean/parallel_matrix_op.c

@@ -0,0 +1,183 @@
+///  @file parallel_matrix_op.c
+///  @brief the standard implementations for functions in parallel_matrix_op.h
+///
+///  the standard implementations for functions in parallel_matrix_op.h
+///
+
+#include "parallel_matrix_op.h"
+#include "blas.h"
+#include "blas_comm.h"
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle(UT) matrix.
+///
+/// @param[in]  i_row     - the i-th row in an upper-triangle matrix.
+/// @param[in]  j_col     - the j-th column in an upper-triangle matrix.
+/// @param[in]  dim       - the dimension of the upper-triangle matrix, i.e., an dim x dim matrix.
+/// @return  the corresponding index in an array storage.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(unsigned int i_row, unsigned int j_col, unsigned int dim) {
+    return (dim + dim - i_row + 1) * i_row / 2 + j_col - i_row;
+}
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle or lower-triangle matrix.
+///
+/// @param[in]  i_row     - the i-th row in a triangle matrix.
+/// @param[in]  j_col     - the j-th column in a triangle matrix.
+/// @param[in]  dim       - the dimension of the triangle matrix, i.e., an dim x dim matrix.
+/// @return    the corresponding index in an array storage.
+///
+static inline unsigned int idx_of_2trimat(unsigned int i_row, unsigned int j_col, unsigned int n_var) {
+    if (i_row > j_col) {
+        return PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(j_col, i_row, n_var);
+    }
+    return PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(i_row, j_col, n_var);
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_UpperTrianglize(unsigned char *btriC, const unsigned char *bA, unsigned int Awidth, unsigned int size_batch) {
+    unsigned char *runningC = btriC;
+    unsigned int Aheight = Awidth;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < i; j++) {
+            unsigned int idx = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(j, i, Aheight);
+            gf256v_add(btriC + idx * size_batch, bA + size_batch * (i * Awidth + j), size_batch);
+        }
+        gf256v_add(runningC, bA + size_batch * (i * Awidth + i), size_batch * (Aheight - i));
+        runningC += size_batch * (Aheight - i);
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Awidth = Bheight;
+    unsigned int Aheight = Awidth;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (k < i) {
+                    continue;
+                }
+                gf256v_madd(bC, &btriA[(k - i) * size_batch], PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+        btriA += (Aheight - i) * size_batch;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_trimatTr_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Aheight = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (i < k) {
+                    continue;
+                }
+                gf256v_madd(bC, &btriA[size_batch * (PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(k, i, Aheight))], PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_2trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Aheight = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (i == k) {
+                    continue;
+                }
+                gf256v_madd(bC, &btriA[size_batch * (idx_of_2trimat(i, k, Aheight))], PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_matTr_madd_gf256(unsigned char *bC, const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Atr_height = Awidth;
+    unsigned int Atr_width = Aheight;
+    for (unsigned int i = 0; i < Atr_height; i++) {
+        for (unsigned int j = 0; j < Atr_width; j++) {
+            gf256v_madd(bC, &bB[j * Bwidth * size_batch], PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(&A_to_tr[size_Acolvec * i], j), size_batch * Bwidth);
+        }
+        bC += size_batch * Bwidth;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_bmatTr_madd_gf256(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    const unsigned char *bA = bA_to_tr;
+    unsigned int Aheight = Awidth_before_tr;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                gf256v_madd(bC, &bA[size_batch * (i + k * Aheight)], PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_mat_madd_gf256(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Awidth = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                gf256v_madd(bC, &bA[k * size_batch], PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+        bA += (Awidth) * size_batch;
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_quad_trimat_eval_gf256(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch) {
+    unsigned char tmp[256];
+
+    unsigned char _x[256];
+    for (unsigned int i = 0; i < dim; i++) {
+        _x[i] = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(x, i);
+    }
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(y, size_batch);
+    for (unsigned int i = 0; i < dim; i++) {
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(tmp, size_batch);
+        for (unsigned int j = i; j < dim; j++) {
+            gf256v_madd(tmp, trimat, _x[j], size_batch);
+            trimat += size_batch;
+        }
+        gf256v_madd(y, tmp, _x[i], size_batch);
+    }
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_quad_recmat_eval_gf256(unsigned char *z, const unsigned char *y, unsigned int dim_y, const unsigned char *mat,
+        const unsigned char *x, unsigned dim_x, unsigned size_batch) {
+    unsigned char tmp[128];
+
+    unsigned char _x[128];
+    for (unsigned int i = 0; i < dim_x; i++) {
+        _x[i] = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(x, i);
+    }
+    unsigned char _y[128];
+    for (unsigned int i = 0; i < dim_y; i++) {
+        _y[i] = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele(y, i);
+    }
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(z, size_batch);
+    for (unsigned int i = 0; i < dim_y; i++) {
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(tmp, size_batch);
+        for (unsigned int j = 0; j < dim_x; j++) {
+            gf256v_madd(tmp, mat, _x[j], size_batch);
+            mat += size_batch;
+        }
+        gf256v_madd(z, tmp, _y[i], size_batch);
+    }
+}
+

crypto_sign/rainbowIIIc-cyclic/clean/parallel_matrix_op.h

@@ -0,0 +1,260 @@
+#ifndef _P_MATRIX_OP_H_
+#define _P_MATRIX_OP_H_
+///  @file  parallel_matrix_op.h
+///  @brief Librarys for operations of batched matrixes.
+///
+///
+
+////////////////  Section:  triangle matrix <-> rectangle matrix   ///////////////////////////////////
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle(UT) matrix.
+///
+/// @param[in]  i_row     - the i-th row in an upper-triangle matrix.
+/// @param[in]  j_col     - the j-th column in an upper-triangle matrix.
+/// @param[in]  dim       - the dimension of the upper-triangle matrix, i.e., an dim x dim matrix.
+/// @return  the corresponding index in an array storage.
+///
+unsigned int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(unsigned int i_row, unsigned int j_col, unsigned int dim);
+
+///
+/// @brief  Upper trianglize a rectangle matrix to the corresponding upper-trangle matrix.
+///
+/// @param[out]  btriC    - the batched upper-trianglized matrix C.
+/// @param[in]   bA       - a batched retangle matrix A.
+/// @param[in]   bwidth   - the width of the batched matrix A, i.e., A is a Awidth x Awidth matrix.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_UpperTrianglize(unsigned char *btriC, const unsigned char *bA, unsigned int Awidth, unsigned int size_batch);
+
+////////////////////  Section:  matrix multiplications  ///////////////////////////////
+
+///
+/// @brief  bC += btriA * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA^Tr * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. A will be transposed while multiplying.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_trimatTr_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA^Tr * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A, which will be transposed while multiplying.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_trimatTr_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC +=  (btriA + btriA^Tr) *B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. The operand for multiplication is (btriA + btriA^Tr).
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_2trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC +=  (btriA + btriA^Tr) *B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. The operand for multiplication is (btriA + btriA^Tr).
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_2trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += A^Tr * bB  , in GF(16)
+///
+/// @param[out]  bC           - the batched matrix C.
+/// @param[in]   A_to_tr      - a column-major matrix A. The operand for multiplication is A^Tr.
+/// @param[in]   Aheight      - the height of A.
+/// @param[in]   size_Acolvec    - the size of a column vector in A.
+/// @param[in]   Awidth           - the width of A.
+/// @param[in]   bB          - a batched matrix B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_matTr_madd_gf16(unsigned char *bC,
+        const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += A^Tr * bB  , in GF(256)
+///
+/// @param[out]  bC           - the batched matrix C.
+/// @param[in]   A_to_tr      - a column-major matrix A. The operand for multiplication is A^Tr.
+/// @param[in]   Aheight      - the height of A.
+/// @param[in]   size_Acolvec    - the size of a column vector in A.
+/// @param[in]   Awidth           - the width of A.
+/// @param[in]   bB          - a batched matrix B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_matTr_madd_gf256(unsigned char *bC,
+        const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA^Tr * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA_to_tr   - a batched matrix A. The operand for multiplication is (bA^Tr).
+/// @param[in]   Awidth_befor_tr     - the width of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_bmatTr_madd_gf16(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA^Tr * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA_to_tr   - a batched matrix A. The operand for multiplication is (bA^Tr).
+/// @param[in]   Awidth_befor_tr     - the width of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_bmatTr_madd_gf256(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA         - a batched matrix A.
+/// @param[in]   Aheigh     - the height of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_mat_madd_gf16(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA         - a batched matrix A.
+/// @param[in]   Aheigh     - the height of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_mat_madd_gf256(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+////////////////////  Section: "quadratric" matrix evaluation  ///////////////////////////////
+
+///
+/// @brief  y =  x^Tr * trimat * x  , in GF(16)
+///
+/// @param[out]  y          - the returned batched element y.
+/// @param[in]   trimat     - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim        - the dimension of matrix trimat (and x).
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_quad_trimat_eval_gf16(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch);
+
+///
+/// @brief  y =  x^Tr * trimat * x  , in GF(256)
+///
+/// @param[out]  y          - the returned batched element y.
+/// @param[in]   trimat     - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim        - the dimension of matrix trimat (and x).
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_quad_trimat_eval_gf256(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch);
+
+///
+/// @brief  z =  y^Tr * mat * x  , in GF(16)
+///
+/// @param[out]  z          - the returned batched element z.
+/// @param[in]   y          - an input vector y.
+/// @param[in]   dim_y      - the length of y.
+/// @param[in]   mat        - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim_x        - the length of x.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_quad_recmat_eval_gf16(unsigned char *z, const unsigned char *y, unsigned int dim_y,
+        const unsigned char *mat, const unsigned char *x, unsigned int dim_x, unsigned int size_batch);
+
+///
+/// @brief  z =  y^Tr * mat * x  , in GF(256)
+///
+/// @param[out]  z          - the returned batched element z.
+/// @param[in]   y          - an input vector y.
+/// @param[in]   dim_y      - the length of y.
+/// @param[in]   mat        - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim_x        - the length of x.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_quad_recmat_eval_gf256(unsigned char *z, const unsigned char *y, unsigned int dim_y,
+        const unsigned char *mat, const unsigned char *x, unsigned int dim_x, unsigned int size_batch);
+
+#endif // _P_MATRIX_OP_H_

crypto_sign/rainbowIIIc-cyclic/clean/rainbow.c

@@ -0,0 +1,174 @@
+/// @file rainbow.c
+/// @brief The standard implementations for functions in rainbow.h
+///
+
+#include "rainbow.h"
+#include "blas.h"
+#include "rainbow_blas.h"
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+#include "utils_hash.h"
+#include "utils_prng.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#define MAX_ATTEMPT_FRMAT 128
+#define _MAX_O ((_O1 > _O2) ? _O1 : _O2)
+#define _MAX_O_BYTE ((_O1_BYTE > _O2_BYTE) ? _O1_BYTE : _O2_BYTE)
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_sign(uint8_t *signature, const sk_t *sk, const uint8_t *_digest) {
+    uint8_t mat_l1[_O1 * _O1_BYTE];
+    uint8_t mat_l2[_O2 * _O2_BYTE];
+    uint8_t mat_buffer[2 * _MAX_O * _MAX_O_BYTE];
+
+    // setup PRNG
+    prng_t prng_sign;
+    uint8_t prng_preseed[LEN_SKSEED + _HASH_LEN];
+    memcpy(prng_preseed, sk->sk_seed, LEN_SKSEED);
+    memcpy(prng_preseed + LEN_SKSEED, _digest, _HASH_LEN); // prng_preseed = sk_seed || digest
+    uint8_t prng_seed[_HASH_LEN];
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(prng_seed, _HASH_LEN, prng_preseed, _HASH_LEN + LEN_SKSEED);
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_set(&prng_sign, prng_seed, _HASH_LEN); // seed = H( sk_seed || digest )
+    for (unsigned int i = 0; i < LEN_SKSEED + _HASH_LEN; i++) {
+        prng_preseed[i] ^= prng_preseed[i]; // clean
+    }
+    for (unsigned int i = 0; i < _HASH_LEN; i++) {
+        prng_seed[i] ^= prng_seed[i]; // clean
+    }
+
+    // roll vinegars.
+    uint8_t vinegar[_V1_BYTE];
+    unsigned int n_attempt = 0;
+    unsigned int l1_succ = 0;
+    while (!l1_succ) {
+        if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+            break;
+        }
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(&prng_sign, vinegar, _V1_BYTE);   // generating vinegars
+        gfmat_prod(mat_l1, sk->l1_F2, _O1 * _O1_BYTE, _V1, vinegar); // generating the linear equations for layer 1
+        l1_succ = gfmat_inv(mat_l1, mat_l1, _O1, mat_buffer);        // check if the linear equation solvable
+        n_attempt++;
+    }
+
+    // Given the vinegars, pre-compute variables needed for layer 2
+    uint8_t r_l1_F1[_O1_BYTE] = {0};
+    uint8_t r_l2_F1[_O2_BYTE] = {0};
+    batch_quad_trimat_eval(r_l1_F1, sk->l1_F1, vinegar, _V1, _O1_BYTE);
+    batch_quad_trimat_eval(r_l2_F1, sk->l2_F1, vinegar, _V1, _O2_BYTE);
+    uint8_t mat_l2_F3[_O2 * _O2_BYTE];
+    uint8_t mat_l2_F2[_O1 * _O2_BYTE];
+    gfmat_prod(mat_l2_F3, sk->l2_F3, _O2 * _O2_BYTE, _V1, vinegar);
+    gfmat_prod(mat_l2_F2, sk->l2_F2, _O1 * _O2_BYTE, _V1, vinegar);
+
+    // Some local variables.
+    uint8_t _z[_PUB_M_BYTE];
+    uint8_t y[_PUB_M_BYTE];
+    uint8_t *x_v1 = vinegar;
+    uint8_t x_o1[_O1_BYTE];
+    uint8_t x_o2[_O1_BYTE];
+
+    uint8_t digest_salt[_HASH_LEN + _SALT_BYTE];
+    memcpy(digest_salt, _digest, _HASH_LEN);
+    uint8_t *salt = digest_salt + _HASH_LEN;
+
+    uint8_t temp_o[_MAX_O_BYTE + 32] = {0};
+    unsigned int succ = 0;
+    while (!succ) {
+        if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+            break;
+        }
+        // The computation:  H(digest||salt)  -->   z   --S-->   y  --C-map-->   x   --T-->   w
+
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(&prng_sign, salt, _SALT_BYTE);                         // roll the salt
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(_z, _PUB_M_BYTE, digest_salt, _HASH_LEN + _SALT_BYTE); // H(digest||salt)
+
+        //  y = S^-1 * z
+        memcpy(y, _z, _PUB_M_BYTE); // identity part of S
+        gfmat_prod(temp_o, sk->s1, _O1_BYTE, _O2, _z + _O1_BYTE);
+        gf256v_add(y, temp_o, _O1_BYTE);
+
+        // Central Map:
+        // layer 1: calculate x_o1
+        memcpy(temp_o, r_l1_F1, _O1_BYTE);
+        gf256v_add(temp_o, y, _O1_BYTE);
+        gfmat_prod(x_o1, mat_l1, _O1_BYTE, _O1, temp_o);
+
+        // layer 2: calculate x_o2
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_set_zero(temp_o, _O2_BYTE);
+        gfmat_prod(temp_o, mat_l2_F2, _O2_BYTE, _O1, x_o1);             // F2
+        batch_quad_trimat_eval(mat_l2, sk->l2_F5, x_o1, _O1, _O2_BYTE); // F5
+        gf256v_add(temp_o, mat_l2, _O2_BYTE);
+        gf256v_add(temp_o, r_l2_F1, _O2_BYTE); // F1
+        gf256v_add(temp_o, y + _O1_BYTE, _O2_BYTE);
+
+        // generate the linear equations of the 2nd layer
+        gfmat_prod(mat_l2, sk->l2_F6, _O2 * _O2_BYTE, _O1, x_o1); // F6
+        gf256v_add(mat_l2, mat_l2_F3, _O2 * _O2_BYTE);            // F3
+        succ = gfmat_inv(mat_l2, mat_l2, _O2, mat_buffer);
+        gfmat_prod(x_o2, mat_l2, _O2_BYTE, _O2, temp_o); // solve l2 eqs
+
+        n_attempt++;
+    };
+    //  w = T^-1 * y
+    uint8_t w[_PUB_N_BYTE];
+    // identity part of T.
+    memcpy(w, x_v1, _V1_BYTE);
+    memcpy(w + _V1_BYTE, x_o1, _O1_BYTE);
+    memcpy(w + _V2_BYTE, x_o2, _O2_BYTE);
+    // Computing the t1 part.
+    gfmat_prod(y, sk->t1, _V1_BYTE, _O1, x_o1);
+    gf256v_add(w, y, _V1_BYTE);
+    // Computing the t4 part.
+    gfmat_prod(y, sk->t4, _V1_BYTE, _O2, x_o2);
+    gf256v_add(w, y, _V1_BYTE);
+    // Computing the t3 part.
+    gfmat_prod(y, sk->t3, _O1_BYTE, _O2, x_o2);
+    gf256v_add(w + _V1_BYTE, y, _O1_BYTE);
+
+    memset(signature, 0, _SIGNATURE_BYTE); // set the output 0
+    // clean
+    memset(&prng_sign, 0, sizeof(prng_t));
+    memset(vinegar, 0, _V1_BYTE);
+    memset(r_l1_F1, 0, _O1_BYTE);
+    memset(r_l2_F1, 0, _O2_BYTE);
+    memset(_z, 0, _PUB_M_BYTE);
+    memset(y, 0, _PUB_M_BYTE);
+    memset(x_o1, 0, _O1_BYTE);
+    memset(x_o2, 0, _O2_BYTE);
+    memset(temp_o, 0, sizeof(temp_o));
+
+    // return: copy w and salt to the signature.
+    if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+        return -1;
+    }
+    gf256v_add(signature, w, _PUB_N_BYTE);
+    gf256v_add(signature + _PUB_N_BYTE, salt, _SALT_BYTE);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_verify(const uint8_t *digest, const uint8_t *signature, const pk_t *pk) {
+    unsigned char digest_ck[_PUB_M_BYTE];
+    // public_map( digest_ck , pk , signature ); Evaluating the quadratic public polynomials.
+    batch_quad_trimat_eval(digest_ck, pk->pk, signature, _PUB_N, _PUB_M_BYTE);
+
+    unsigned char correct[_PUB_M_BYTE];
+    unsigned char digest_salt[_HASH_LEN + _SALT_BYTE];
+    memcpy(digest_salt, digest, _HASH_LEN);
+    memcpy(digest_salt + _HASH_LEN, signature + _PUB_N_BYTE, _SALT_BYTE);
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(correct, _PUB_M_BYTE, digest_salt, _HASH_LEN + _SALT_BYTE); // H( digest || salt )
+
+    // check consistancy.
+    unsigned char cc = 0;
+    for (unsigned int i = 0; i < _PUB_M_BYTE; i++) {
+        cc |= (digest_ck[i] ^ correct[i]);
+    }
+    return (0 == cc) ? 0 : -1;
+}
+
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_verify_cyclic(const uint8_t *digest, const uint8_t *signature, const cpk_t *_pk) {
+    unsigned char pk[sizeof(pk_t) + 32];
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_cpk_to_pk((pk_t *)pk, _pk); // generating classic public key.
+    return PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_verify(digest, signature, (pk_t *)pk);
+}

crypto_sign/rainbowIIIc-cyclic/clean/rainbow.h

@@ -0,0 +1,42 @@
+#ifndef _RAINBOW_H_
+#define _RAINBOW_H_
+/// @file rainbow.h
+/// @brief APIs for rainbow.
+///
+
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+
+#include <stdint.h>
+
+///
+/// @brief Signing function for classical secret key.
+///
+/// @param[out] signature - the signature.
+/// @param[in]  sk        - the secret key.
+/// @param[in]  digest    - the digest.
+///
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_sign(uint8_t *signature, const sk_t *sk, const uint8_t *digest);
+
+///
+/// @brief Verifying function.
+///
+/// @param[in]  digest    - the digest.
+/// @param[in]  signature - the signature.
+/// @param[in]  pk        - the public key.
+/// @return 0 for successful verified. -1 for failed verification.
+///
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_verify(const uint8_t *digest, const uint8_t *signature, const pk_t *pk);
+
+
+///
+/// @brief Verifying function for cyclic public keys.
+///
+/// @param[in]  digest    - the digest.
+/// @param[in]  signature - the signature.
+/// @param[in]  pk        - the public key of cyclic rainbow.
+/// @return 0 for successful verified. -1 for failed verification.
+///
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_verify_cyclic(const uint8_t *digest, const uint8_t *signature, const cpk_t *pk);
+
+#endif // _RAINBOW_H_

crypto_sign/rainbowIIIc-cyclic/clean/rainbow_blas.h

@@ -0,0 +1,31 @@
+#ifndef _RAINBOW_BLAS_H_
+#define _RAINBOW_BLAS_H_
+/// @file rainbow_blas.h
+/// @brief Defining the functions used in rainbow.c acconding to the definitions in rainbow_config.h
+///
+///  Defining the functions used in rainbow.c acconding to the definitions in rainbow_config.h
+
+#include "blas.h"
+#include "parallel_matrix_op.h"
+#include "rainbow_config.h"
+
+
+#define gfv_get_ele PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_get_ele
+#define gfv_mul_scalar PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_mul_scalar
+#define gfv_madd PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256v_madd
+
+#define gfmat_prod PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_prod
+#define gfmat_inv PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_gf256mat_inv
+
+#define batch_trimat_madd PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_trimat_madd_gf256
+#define batch_trimatTr_madd PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_trimatTr_madd_gf256
+#define batch_2trimat_madd PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_2trimat_madd_gf256
+#define batch_matTr_madd PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_matTr_madd_gf256
+#define batch_bmatTr_madd PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_bmatTr_madd_gf256
+#define batch_mat_madd PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_mat_madd_gf256
+
+#define batch_quad_trimat_eval PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_quad_trimat_eval_gf256
+#define batch_quad_recmat_eval PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_batch_quad_recmat_eval_gf256
+
+
+#endif // _RAINBOW_BLAS_H_

crypto_sign/rainbowIIIc-cyclic/clean/rainbow_config.h

@@ -0,0 +1,46 @@
+#ifndef _H_RAINBOW_CONFIG_H_
+#define _H_RAINBOW_CONFIG_H_
+
+/// @file rainbow_config.h
+/// @brief Defining the parameters of the Rainbow and the corresponding constants.
+///
+
+#define _GFSIZE 256
+#define _V1 68
+#define _O1 36
+#define _O2 36
+#define _HASH_LEN 48
+
+
+#define _V2 ((_V1) + (_O1))
+
+/// size of N, in # of gf elements.
+#define _PUB_N (_V1 + _O1 + _O2)
+
+/// size of M, in # gf elements.
+#define _PUB_M (_O1 + _O2)
+
+/// size of variables, in # bytes.
+
+// GF256
+#define _V1_BYTE (_V1)
+#define _V2_BYTE (_V2)
+#define _O1_BYTE (_O1)
+#define _O2_BYTE (_O2)
+#define _PUB_N_BYTE (_PUB_N)
+#define _PUB_M_BYTE (_PUB_M)
+
+
+/// length of seed for public key, in # bytes
+#define LEN_PKSEED 32
+
+/// length of seed for secret key, in # bytes
+#define LEN_SKSEED 32
+
+/// length of salt for a signature, in # bytes
+#define _SALT_BYTE 16
+
+/// length of a signature
+#define _SIGNATURE_BYTE (_PUB_N_BYTE + _SALT_BYTE)
+
+#endif //  _H_RAINBOW_CONFIG_H_

crypto_sign/rainbowIIIc-cyclic/clean/rainbow_keypair.c

@@ -0,0 +1,157 @@
+/// @file rainbow_keypair.c
+/// @brief implementations of functions in rainbow_keypair.h
+///
+
+#include "rainbow_keypair.h"
+#include "blas.h"
+#include "blas_comm.h"
+#include "rainbow_blas.h"
+#include "rainbow_keypair_computation.h"
+#include "utils_prng.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+static void generate_S_T(unsigned char *s_and_t, prng_t *prng0) {
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, s_and_t, _O1_BYTE * _O2); // S1
+    s_and_t += _O1_BYTE * _O2;
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, s_and_t, _V1_BYTE * _O1); // T1
+    s_and_t += _V1_BYTE * _O1;
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, s_and_t, _V1_BYTE * _O2); // T2
+    s_and_t += _V1_BYTE * _O2;
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, s_and_t, _O1_BYTE * _O2); // T3
+}
+
+static unsigned int generate_l1_F12(unsigned char *sk, prng_t *prng0) {
+    unsigned int n_byte_generated = 0;
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, sk, _O1_BYTE * N_TRIANGLE_TERMS(_V1)); // l1_F1
+    sk += _O1_BYTE * N_TRIANGLE_TERMS(_V1);
+    n_byte_generated += _O1_BYTE * N_TRIANGLE_TERMS(_V1);
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, sk, _O1_BYTE * _V1 * _O1); // l1_F2
+    n_byte_generated += _O1_BYTE * _V1 * _O1;
+    return n_byte_generated;
+}
+
+static unsigned int generate_l2_F12356(unsigned char *sk, prng_t *prng0) {
+    unsigned int n_byte_generated = 0;
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * N_TRIANGLE_TERMS(_V1)); // l2_F1
+    sk += _O2_BYTE * N_TRIANGLE_TERMS(_V1);
+    n_byte_generated += _O2_BYTE * N_TRIANGLE_TERMS(_V1);
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _V1 * _O1); // l2_F2
+    sk += _O2_BYTE * _V1 * _O1;
+    n_byte_generated += _O2_BYTE * _V1 * _O1;
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _V1 * _O2); // l2_F3
+    sk += _O2_BYTE * _V1 * _O1;
+    n_byte_generated += _O2_BYTE * _V1 * _O1;
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * N_TRIANGLE_TERMS(_O1)); // l2_F5
+    sk += _O2_BYTE * N_TRIANGLE_TERMS(_O1);
+    n_byte_generated += _O2_BYTE * N_TRIANGLE_TERMS(_O1);
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _O1 * _O2); // l2_F6
+    n_byte_generated += _O2_BYTE * _O1 * _O2;
+
+    return n_byte_generated;
+}
+
+static void generate_B1_B2(unsigned char *sk, prng_t *prng0) {
+    sk += generate_l1_F12(sk, prng0);
+    generate_l2_F12356(sk, prng0);
+}
+
+static void calculate_t4(unsigned char *t2_to_t4, const unsigned char *t1, const unsigned char *t3) {
+    //  t4 = T_sk.t1 * T_sk.t3 - T_sk.t2
+    unsigned char temp[_V1_BYTE + 32];
+    unsigned char *t4 = t2_to_t4;
+    for (unsigned int i = 0; i < _O2; i++) { /// t3 width
+        gfmat_prod(temp, t1, _V1_BYTE, _O1, t3);
+        gf256v_add(t4, temp, _V1_BYTE);
+        t4 += _V1_BYTE;
+        t3 += _O1_BYTE;
+    }
+}
+
+static void obsfucate_l1_polys(unsigned char *l1_polys, const unsigned char *l2_polys, unsigned int n_terms, const unsigned char *s1) {
+    unsigned char temp[_O1_BYTE + 32];
+    while (n_terms--) {
+        gfmat_prod(temp, s1, _O1_BYTE, _O2, l2_polys);
+        gf256v_add(l1_polys, temp, _O1_BYTE);
+        l1_polys += _O1_BYTE;
+        l2_polys += _O2_BYTE;
+    }
+}
+
+///////////////////  Classic //////////////////////////////////
+
+
+/////////////////////   Cyclic   //////////////////////////////////
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_generate_keypair_cyclic(cpk_t *pk, sk_t *sk, const unsigned char *pk_seed, const unsigned char *sk_seed) {
+    memcpy(pk->pk_seed, pk_seed, LEN_PKSEED);
+    memcpy(sk->sk_seed, sk_seed, LEN_SKSEED);
+
+    // prng for sk
+    prng_t prng;
+    prng_t *prng0 = &prng;
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_set(prng0, sk_seed, LEN_SKSEED);
+    generate_S_T(sk->s1, prng0); // S,T:  only a part of sk
+
+    unsigned char t2[sizeof(sk->t4)];
+    memcpy(t2, sk->t4, _V1_BYTE * _O2);   // temporarily store t2
+    calculate_t4(sk->t4, sk->t1, sk->t3); // t2 <- t4
+
+    // prng for pk
+    sk_t inst_Qs;
+    sk_t *Qs = &inst_Qs;
+    prng_t *prng1 = &prng;
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_set(prng1, pk_seed, LEN_PKSEED);
+    generate_B1_B2(Qs->l1_F1, prng1); // generating l1_Q1, l1_Q2, l2_Q1, l2_Q2, l2_Q3, l2_Q5, l2_Q6
+    obsfucate_l1_polys(Qs->l1_F1, Qs->l2_F1, N_TRIANGLE_TERMS(_V1), sk->s1);
+    obsfucate_l1_polys(Qs->l1_F2, Qs->l2_F2, _V1 * _O1, sk->s1);
+    // so far, the Qs contains l1_F1, l1_F2, l2_F1, l2_F2, l2_F3, l2_F5, l2_F6.
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_calculate_F_from_Q(sk, Qs, sk); // calcuate the rest parts of secret key from Qs and S,T
+
+    unsigned char t4[sizeof(sk->t4)];
+    memcpy(t4, sk->t4, _V1_BYTE * _O2);                      // temporarily store t4
+    memcpy(sk->t4, t2, _V1_BYTE * _O2);                      // restore t2
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_calculate_Q_from_F_cyclic(pk, sk, sk); // calculate the rest parts of public key: l1_Q3, l1_Q5, l1_Q6, l1_Q9, l2_Q9
+    memcpy(sk->t4, t4, _V1_BYTE * _O2);                      // restore t4
+
+    obsfucate_l1_polys(pk->l1_Q3, Qs->l2_F3, _V1 * _O2, sk->s1);
+    obsfucate_l1_polys(pk->l1_Q5, Qs->l2_F5, N_TRIANGLE_TERMS(_O1), sk->s1);
+    obsfucate_l1_polys(pk->l1_Q6, Qs->l2_F6, _O1 * _O2, sk->s1);
+    obsfucate_l1_polys(pk->l1_Q9, pk->l2_Q9, N_TRIANGLE_TERMS(_O2), sk->s1);
+
+    // clean
+    memset(&prng, 0, sizeof(prng_t));
+}
+
+
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_cpk_to_pk(pk_t *rpk, const cpk_t *cpk) {
+    // procedure:  cpk_t --> extcpk_t  --> pk_t
+
+    // convert from cpk_t to extcpk_t
+    ext_cpk_t pk;
+
+    // setup prng
+    prng_t prng0;
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_set(&prng0, cpk->pk_seed, LEN_SKSEED);
+
+    // generating parts of key with prng
+    generate_l1_F12(pk.l1_Q1, &prng0);
+    // copying parts of key from input. l1_Q3, l1_Q5, l1_Q6, l1_Q9
+    memcpy(pk.l1_Q3, cpk->l1_Q3, _O1_BYTE * (_V1 * _O2 + N_TRIANGLE_TERMS(_O1) + _O1 * _O2 + N_TRIANGLE_TERMS(_O2)));
+
+    // generating parts of key with prng
+    generate_l2_F12356(pk.l2_Q1, &prng0);
+    // copying parts of key from input: l2_Q9
+    memcpy(pk.l2_Q9, cpk->l2_Q9, _O2_BYTE * N_TRIANGLE_TERMS(_O2));
+
+    // convert from extcpk_t to pk_t
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_extcpk_to_pk(rpk, &pk);
+}

crypto_sign/rainbowIIIc-cyclic/clean/rainbow_keypair.h

@@ -0,0 +1,94 @@
+#ifndef _RAINBOW_KEYPAIR_H_
+#define _RAINBOW_KEYPAIR_H_
+/// @file rainbow_keypair.h
+/// @brief Formats of key pairs and functions for generating key pairs.
+/// Formats of key pairs and functions for generating key pairs.
+///
+
+#include "rainbow_config.h"
+
+#define N_TRIANGLE_TERMS(n_var) ((n_var) * ((n_var) + 1) / 2)
+
+/// @brief public key for classic rainbow
+///
+///  public key for classic rainbow
+///
+typedef struct rainbow_publickey {
+    unsigned char pk[(_PUB_M_BYTE)*N_TRIANGLE_TERMS(_PUB_N)];
+} pk_t;
+
+/// @brief secret key for classic rainbow
+///
+/// secret key for classic rainbow
+///
+typedef struct rainbow_secretkey {
+    ///
+    /// seed for generating secret key.
+    /// Generating S, T, and F for classic rainbow.
+    /// Generating S and T only for cyclic rainbow.
+    unsigned char sk_seed[LEN_SKSEED];
+
+    unsigned char s1[_O1_BYTE * _O2]; ///< part of S map
+    unsigned char t1[_V1_BYTE * _O1]; ///< part of T map
+    unsigned char t4[_V1_BYTE * _O2]; ///< part of T map
+    unsigned char t3[_O1_BYTE * _O2]; ///< part of T map
+
+    unsigned char l1_F1[_O1_BYTE * N_TRIANGLE_TERMS(_V1)]; ///< part of C-map, F1, Layer1
+    unsigned char l1_F2[_O1_BYTE * _V1 * _O1];             ///< part of C-map, F2, Layer1
+
+    unsigned char l2_F1[_O2_BYTE * N_TRIANGLE_TERMS(_V1)]; ///< part of C-map, F1, Layer2
+    unsigned char l2_F2[_O2_BYTE * _V1 * _O1];             ///< part of C-map, F2, Layer2
+
+    unsigned char l2_F3[_O2_BYTE * _V1 * _O2];             ///< part of C-map, F3, Layer2
+    unsigned char l2_F5[_O2_BYTE * N_TRIANGLE_TERMS(_O1)]; ///< part of C-map, F5, Layer2
+    unsigned char l2_F6[_O2_BYTE * _O1 * _O2];             ///< part of C-map, F6, Layer2
+} sk_t;
+
+/// @brief public key for cyclic rainbow
+///
+///  public key for cyclic rainbow
+///
+typedef struct rainbow_publickey_cyclic {
+    unsigned char pk_seed[LEN_PKSEED]; ///< seed for generating l1_Q1,l1_Q2,l2_Q1,l2_Q2,l2_Q3,l2_Q5,l2_Q6
+
+    unsigned char l1_Q3[_O1_BYTE * _V1 * _O2];             ///< Q3, layer1
+    unsigned char l1_Q5[_O1_BYTE * N_TRIANGLE_TERMS(_O1)]; ///< Q5, layer1
+    unsigned char l1_Q6[_O1_BYTE * _O1 * _O2];             ///< Q6, layer1
+    unsigned char l1_Q9[_O1_BYTE * N_TRIANGLE_TERMS(_O2)]; ///< Q9, layer1
+
+    unsigned char l2_Q9[_O2_BYTE * N_TRIANGLE_TERMS(_O2)]; ///< Q9, layer2
+} cpk_t;
+
+/// @brief compressed secret key for cyclic rainbow
+///
+/// compressed secret key for cyclic rainbow
+///
+typedef struct rainbow_secretkey_cyclic {
+    unsigned char pk_seed[LEN_PKSEED]; ///< seed for generating a part of public key.
+    unsigned char sk_seed[LEN_SKSEED]; ///< seed for generating a part of secret key.
+} csk_t;
+
+
+///
+/// @brief Generate key pairs for cyclic rainbow.
+///
+/// @param[out] pk        - the public key.
+/// @param[out] sk        - the secret key.
+/// @param[in]  pk_seed   - seed for generating parts of public key.
+/// @param[in]  sk_seed   - seed for generating secret key.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_generate_keypair_cyclic(cpk_t *pk, sk_t *sk, const unsigned char *pk_seed, const unsigned char *sk_seed);
+
+
+
+////////////////////////////////////
+
+///
+/// @brief converting formats of public keys : from cyclic version to classic key
+///
+/// @param[out] pk       - the classic public key.
+/// @param[in]  cpk      - the cyclic  public key.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_cpk_to_pk(pk_t *pk, const cpk_t *cpk);
+
+#endif //  _RAINBOW_KEYPAIR_H_

crypto_sign/rainbowIIIc-cyclic/clean/rainbow_keypair_computation.c

@@ -0,0 +1,213 @@
+/// @file rainbow_keypair_computation.c
+/// @brief Implementations for functions in rainbow_keypair_computation.h
+///
+
+#include "rainbow_keypair_computation.h"
+#include "blas.h"
+#include "blas_comm.h"
+#include "rainbow_blas.h"
+#include "rainbow_keypair.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_extcpk_to_pk(pk_t *pk, const ext_cpk_t *cpk) {
+    const unsigned char *idx_l1 = cpk->l1_Q1;
+    const unsigned char *idx_l2 = cpk->l2_Q1;
+    for (unsigned int i = 0; i < _V1; i++) {
+        for (unsigned int j = i; j < _V1; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q2;
+    idx_l2 = cpk->l2_Q2;
+    for (unsigned int i = 0; i < _V1; i++) {
+        for (unsigned int j = _V1; j < _V1 + _O1; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q3;
+    idx_l2 = cpk->l2_Q3;
+    for (unsigned int i = 0; i < _V1; i++) {
+        for (unsigned int j = _V1 + _O1; j < _PUB_N; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q5;
+    idx_l2 = cpk->l2_Q5;
+    for (unsigned int i = _V1; i < _V1 + _O1; i++) {
+        for (unsigned int j = i; j < _V1 + _O1; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q6;
+    idx_l2 = cpk->l2_Q6;
+    for (unsigned int i = _V1; i < _V1 + _O1; i++) {
+        for (unsigned int j = _V1 + _O1; j < _PUB_N; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+    idx_l1 = cpk->l1_Q9;
+    idx_l2 = cpk->l2_Q9;
+    for (unsigned int i = _V1 + _O1; i < _PUB_N; i++) {
+        for (unsigned int j = i; j < _PUB_N; j++) {
+            unsigned int pub_idx = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_idx_of_trimat(i, j, _PUB_N);
+            memcpy(&pk->pk[_PUB_M_BYTE * pub_idx], idx_l1, _O1_BYTE);
+            memcpy((&pk->pk[_PUB_M_BYTE * pub_idx]) + _O1_BYTE, idx_l2, _O2_BYTE);
+            idx_l1 += _O1_BYTE;
+            idx_l2 += _O2_BYTE;
+        }
+    }
+}
+
+static void calculate_F_from_Q_ref(sk_t *Fs, const sk_t *Qs, sk_t *Ts) {
+    // Layer 1
+    // F_sk.l1_F1s[i] = Q_pk.l1_F1s[i]
+    memcpy(Fs->l1_F1, Qs->l1_F1, _O1_BYTE * N_TRIANGLE_TERMS(_V1));
+
+    // F_sk.l1_F2s[i] = ( Q_pk.l1_F1s[i] + Q_pk.l1_F1s[i].transpose() ) * T_sk.t1 + Q_pk.l1_F2s[i]
+    memcpy(Fs->l1_F2, Qs->l1_F2, _O1_BYTE * _V1 * _O1);
+    batch_2trimat_madd(Fs->l1_F2, Qs->l1_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O1_BYTE);
+
+    /*
+          Layer 2
+          computations:
+
+          F_sk.l2_F1s[i] = Q_pk.l2_F1s[i]
+
+          Q1_T1 = Q_pk.l2_F1s[i]*T_sk.t1
+          F_sk.l2_F2s[i] =              Q1_T1 + Q_pk.l2_F2s[i]     + Q_pk.l2_F1s[i].transpose() * T_sk.t1
+          F_sk.l2_F5s[i] = UT( t1_tr* ( Q1_T1 + Q_pk.l2_F2s[i] ) ) + Q_pk.l2_F5s[i]
+
+          Q1_Q1T_T4 =  (Q_pk.l2_F1s[i] + Q_pk.l2_F1s[i].transpose()) * t4
+          #Q1_Q1T_T4 =  Q1_Q1T * t4
+          Q2_T3 = Q_pk.l2_F2s[i]*T_sk.t3
+          F_sk.l2_F3s[i] =           Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i]
+          F_sk.l2_F6s[i] = t1_tr * ( Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i] )
+                          +  Q_pk.l2_F2s[i].transpose() * t4
+                          + (Q_pk.l2_F5s[i] + Q_pk.l2_F5s[i].transpose())*T_sk.t3   + Q_pk.l2_F6s[i]
+
+      */
+    memcpy(Fs->l2_F1, Qs->l2_F1, _O2_BYTE * N_TRIANGLE_TERMS(_V1)); // F_sk.l2_F1s[i] = Q_pk.l2_F1s[i]
+
+    // F_sk.l2_F2s[i] =              Q1_T1 + Q_pk.l2_F2s[i]     + Q_pk.l2_F1s[i].transpose() * T_sk.t1
+    // F_sk.l2_F5s[i] = UT( t1_tr* ( Q1_T1 + Q_pk.l2_F2s[i] ) ) + Q_pk.l2_F5s[i]
+    memcpy(Fs->l2_F2, Qs->l2_F2, _O2_BYTE * _V1 * _O1);
+    batch_trimat_madd(Fs->l2_F2, Qs->l2_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O2_BYTE); // Q1_T1+ Q2
+
+    unsigned char tempQ[_O1 * _O1 * _O2_BYTE + 32];
+    memset(tempQ, 0, _O1 * _O1 * _O2_BYTE);
+    batch_matTr_madd(tempQ, Ts->t1, _V1, _V1_BYTE, _O1, Fs->l2_F2, _O1, _O2_BYTE); // t1_tr*(Q1_T1+Q2)
+    memcpy(Fs->l2_F5, Qs->l2_F5, _O2_BYTE * N_TRIANGLE_TERMS(_O1));                // F5
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_UpperTrianglize(Fs->l2_F5, tempQ, _O1, _O2_BYTE);            // UT( ... )
+
+    batch_trimatTr_madd(Fs->l2_F2, Qs->l2_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O2_BYTE); // F2 = Q1_T1 + Q2 + Q1^tr*t1
+
+    // Q1_Q1T_T4 =  (Q_pk.l2_F1s[i] + Q_pk.l2_F1s[i].transpose()) * t4
+    // Q2_T3 = Q_pk.l2_F2s[i]*T_sk.t3
+    // F_sk.l2_F3s[i] =           Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i]
+    memcpy(Fs->l2_F3, Qs->l2_F3, _V1 * _O2 * _O2_BYTE);
+    batch_2trimat_madd(Fs->l2_F3, Qs->l2_F1, Ts->t4, _V1, _V1_BYTE, _O2, _O2_BYTE);  // Q1_Q1T_T4
+    batch_mat_madd(Fs->l2_F3, Qs->l2_F2, _V1, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE); // Q2_T3
+
+    // F_sk.l2_F6s[i] = t1_tr * ( Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i] )
+    //                +  Q_pk.l2_F2s[i].transpose() * t4
+    //                + (Q_pk.l2_F5s[i] + Q_pk.l2_F5s[i].transpose())*T_sk.t3   + Q_pk.l2_F6s[i]
+    memcpy(Fs->l2_F6, Qs->l2_F6, _O1 * _O2 * _O2_BYTE);
+    batch_matTr_madd(Fs->l2_F6, Ts->t1, _V1, _V1_BYTE, _O1, Fs->l2_F3, _O2, _O2_BYTE); // t1_tr * ( Q1_Q1T_T4 + Q2_T3 + Q_pk.l2_F3s[i] )
+    batch_2trimat_madd(Fs->l2_F6, Qs->l2_F5, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE);    // (Q_pk.l2_F5s[i] + Q_pk.l2_F5s[i].transpose())*T_sk.t3
+    batch_bmatTr_madd(Fs->l2_F6, Qs->l2_F2, _O1, Ts->t4, _V1, _V1_BYTE, _O2, _O2_BYTE);
+}
+
+static void calculate_Q_from_F_cyclic_ref(cpk_t *Qs, const sk_t *Fs, const sk_t *Ts) {
+    // Layer 1: Computing Q5, Q3, Q6, Q9
+
+    //  Q_pk.l1_F5s[i] = UT( T1tr* (F1 * T1 + F2) )
+    const unsigned char *t2 = Ts->t4;
+    sk_t tempQ;
+    memcpy(tempQ.l1_F2, Fs->l1_F2, _O1_BYTE * _V1 * _O1);
+    batch_trimat_madd(tempQ.l1_F2, Fs->l1_F1, Ts->t1, _V1, _V1_BYTE, _O1, _O1_BYTE); // F1*T1 + F2
+    memset(tempQ.l2_F1, 0, sizeof(tempQ.l2_F1));
+    memset(tempQ.l2_F2, 0, sizeof(tempQ.l2_F2));
+    batch_matTr_madd(tempQ.l2_F1, Ts->t1, _V1, _V1_BYTE, _O1, tempQ.l1_F2, _O1, _O1_BYTE); // T1tr*(F1*T1 + F2)
+    memset(Qs->l1_Q5, 0, _O1_BYTE * N_TRIANGLE_TERMS(_O1));
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_UpperTrianglize(Qs->l1_Q5, tempQ.l2_F1, _O1, _O1_BYTE); // UT( ... )   // Q5
+
+    /*
+          F1_T2     = F1 * t2
+          F2_T3     = F2 * t3
+          F1_F1T_T2 + F2_T3 = F1_T2 + F2_T3 + F1tr * t2
+          Q_pk.l1_F3s[i] =         F1_F1T_T2 + F2_T3
+          Q_pk.l1_F6s[i] = T1tr* ( F1_F1T_T2 + F2_T3 ) + F2tr * t2
+          Q_pk.l1_F9s[i] = UT( T2tr* ( F1_T2 + F2_T3 ) )
+      */
+    memset(Qs->l1_Q3, 0, _O1_BYTE * _V1 * _O2);
+    memset(Qs->l1_Q6, 0, _O1_BYTE * _O1 * _O2);
+    memset(Qs->l1_Q9, 0, _O1_BYTE * N_TRIANGLE_TERMS(_O2));
+
+    batch_trimat_madd(Qs->l1_Q3, Fs->l1_F1, t2, _V1, _V1_BYTE, _O2, _O1_BYTE);       // F1*T2
+    batch_mat_madd(Qs->l1_Q3, Fs->l1_F2, _V1, Ts->t3, _O1, _O1_BYTE, _O2, _O1_BYTE); // F1_T2 + F2_T3
+
+    memset(tempQ.l1_F2, 0, _O1_BYTE * _V1 * _O2);                                    // should be F3. assuming: _O1 >= _O2
+    batch_matTr_madd(tempQ.l1_F2, t2, _V1, _V1_BYTE, _O2, Qs->l1_Q3, _O2, _O1_BYTE); // T2tr * ( F1_T2 + F2_T3 )
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_UpperTrianglize(Qs->l1_Q9, tempQ.l1_F2, _O2, _O1_BYTE);        // Q9
+
+    batch_trimatTr_madd(Qs->l1_Q3, Fs->l1_F1, t2, _V1, _V1_BYTE, _O2, _O1_BYTE); // F1_F1T_T2 + F2_T3  // Q3
+
+    batch_bmatTr_madd(Qs->l1_Q6, Fs->l1_F2, _O1, t2, _V1, _V1_BYTE, _O2, _O1_BYTE);    // F2tr*T2
+    batch_matTr_madd(Qs->l1_Q6, Ts->t1, _V1, _V1_BYTE, _O1, Qs->l1_Q3, _O2, _O1_BYTE); // Q6
+    /*
+          Layer 2
+          Computing Q9:
+
+          F1_T2     = F1 * t2
+          F2_T3     = F2 * t3
+          Q9 = UT( T2tr*( F1*T2 + F2*T3 + F3 )  +  T3tr*( F5*T3 + F6 ) )
+      */
+    sk_t tempQ2;
+    memcpy(tempQ2.l2_F3, Fs->l2_F3, _O2_BYTE * _V1 * _O2);                              /// F3 actually.
+    batch_trimat_madd(tempQ2.l2_F3, Fs->l2_F1, t2, _V1, _V1_BYTE, _O2, _O2_BYTE);       // F1*T2 + F3
+    batch_mat_madd(tempQ2.l2_F3, Fs->l2_F2, _V1, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE); // F1_T2 + F2_T3 + F3
+
+    memset(tempQ.l2_F3, 0, _O2_BYTE * _V1 * _O2);
+    batch_matTr_madd(tempQ.l2_F3, t2, _V1, _V1_BYTE, _O2, tempQ2.l2_F3, _O2, _O2_BYTE); // T2tr * ( ..... )
+
+    memcpy(tempQ.l2_F6, Fs->l2_F6, _O2_BYTE * _O1 * _O2);
+    batch_trimat_madd(tempQ.l2_F6, Fs->l2_F5, Ts->t3, _O1, _O1_BYTE, _O2, _O2_BYTE); // F5*T3 + F6
+
+    batch_matTr_madd(tempQ.l2_F3, Ts->t3, _O1, _O1_BYTE, _O2, tempQ.l2_F6, _O2, _O2_BYTE); // T2tr*( ..... ) + T3tr*( ..... )
+    memset(Qs->l2_Q9, 0, _O2_BYTE * N_TRIANGLE_TERMS(_O2));
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_UpperTrianglize(Qs->l2_Q9, tempQ.l2_F3, _O2, _O2_BYTE); // Q9
+}
+
+// Choosing implementations depends on the macros: _BLAS_SSE_ and _BLAS_AVX2_
+#define calculate_F_from_Q_impl calculate_F_from_Q_ref
+#define calculate_Q_from_F_cyclic_impl calculate_Q_from_F_cyclic_ref
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_calculate_F_from_Q(sk_t *Fs, const sk_t *Qs, sk_t *Ts) {
+    calculate_F_from_Q_impl(Fs, Qs, Ts);
+}
+
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_calculate_Q_from_F_cyclic(cpk_t *Qs, const sk_t *Fs, const sk_t *Ts) {
+    calculate_Q_from_F_cyclic_impl(Qs, Fs, Ts);
+}

crypto_sign/rainbowIIIc-cyclic/clean/rainbow_keypair_computation.h

@@ -0,0 +1,71 @@
+#ifndef _RAINBOW_KEYPAIR_COMP_H_
+#define _RAINBOW_KEYPAIR_COMP_H_
+/// @file rainbow_keypair_computation.h
+/// @brief Functions for calculating pk/sk while generating keys.
+///
+/// Defining an internal structure of public key.
+/// Functions for calculating pk/sk for key generation.
+///
+
+#include "rainbow_keypair.h"
+
+/// @brief The (internal use) public key for rainbow
+///
+/// The (internal use) public key for rainbow. The public
+/// polynomials are divided into l1_Q1, l1_Q2, ... l1_Q9,
+/// l2_Q1, .... , l2_Q9.
+///
+typedef struct rainbow_extend_publickey {
+    unsigned char l1_Q1[_O1_BYTE * N_TRIANGLE_TERMS(_V1)];
+    unsigned char l1_Q2[_O1_BYTE * _V1 * _O1];
+    unsigned char l1_Q3[_O1_BYTE * _V1 * _O2];
+    unsigned char l1_Q5[_O1_BYTE * N_TRIANGLE_TERMS(_O1)];
+    unsigned char l1_Q6[_O1_BYTE * _O1 * _O2];
+    unsigned char l1_Q9[_O1_BYTE * N_TRIANGLE_TERMS(_O2)];
+
+    unsigned char l2_Q1[_O2_BYTE * N_TRIANGLE_TERMS(_V1)];
+    unsigned char l2_Q2[_O2_BYTE * _V1 * _O1];
+    unsigned char l2_Q3[_O2_BYTE * _V1 * _O2];
+    unsigned char l2_Q5[_O2_BYTE * N_TRIANGLE_TERMS(_O1)];
+    unsigned char l2_Q6[_O2_BYTE * _O1 * _O2];
+    unsigned char l2_Q9[_O2_BYTE * N_TRIANGLE_TERMS(_O2)];
+} ext_cpk_t;
+
+///
+/// @brief converting formats of public keys : from ext_cpk_t version to pk_t
+///
+/// @param[out] pk       - the classic public key.
+/// @param[in]  cpk      - the internel public key.
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_extcpk_to_pk(pk_t *pk, const ext_cpk_t *cpk);
+/////////////////////////////////////////////////
+
+///
+/// @brief Computing public key from secret key
+///
+/// @param[out] Qs       - the public key
+/// @param[in]  Fs       - parts of the secret key: l1_F1, l1_F2, l2_F1, l2_F2, l2_F3, l2_F5, l2_F6
+/// @param[in]  Ts       - parts of the secret key: T1, T4, T3
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_calculate_Q_from_F(ext_cpk_t *Qs, const sk_t *Fs, const sk_t *Ts);
+
+
+///
+/// @brief Computing parts of the sk from parts of pk and sk
+///
+/// @param[out] Fs       - parts of the sk: l1_F1, l1_F2, l2_F1, l2_F2, l2_F3, l2_F5, l2_F6
+/// @param[in]  Qs       - parts of the pk: l1_Q1, l1_Q2, l2_Q1, l2_Q2, l2_Q3, l2_Q5, l2_Q6
+/// @param[in]  Ts       - parts of the sk: T1, T4, T3
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_calculate_F_from_Q(sk_t *Fs, const sk_t *Qs, sk_t *Ts);
+
+///
+/// @brief Computing parts of the pk from the secret key
+///
+/// @param[out] Qs       - parts of the pk: l1_Q3, l1_Q5, l2_Q6, l1_Q9, l2_Q9
+/// @param[in]  Fs       - parts of the sk: l1_F1, l1_F2, l2_F1, l2_F2, l2_F3, l2_F5, l2_F6
+/// @param[in]  Ts       - parts of the sk: T1, T4, T3
+///
+void PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_calculate_Q_from_F_cyclic(cpk_t *Qs, const sk_t *Fs, const sk_t *Ts);
+
+#endif // _RAINBOW_KEYPAIR_COMP_H_

crypto_sign/rainbowIIIc-cyclic/clean/sign.c

@@ -0,0 +1,76 @@
+///  @file  sign.c
+///  @brief the implementations for functions in api.h
+///
+///
+
+#include "api.h"
+#include "rainbow.h"
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+#include "randombytes.h"
+#include "utils_hash.h"
+#include <stdlib.h>
+#include <string.h>
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign_keypair(unsigned char *pk, unsigned char *sk) {
+    unsigned char sk_seed[LEN_SKSEED] = {0};
+    randombytes(sk_seed, LEN_SKSEED);
+
+    unsigned char pk_seed[LEN_PKSEED] = {0};
+    randombytes(pk_seed, LEN_PKSEED);
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_generate_keypair_cyclic((cpk_t *)pk, (sk_t *)sk, pk_seed, sk_seed);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign(unsigned char *sm, size_t *smlen, const unsigned char *m, size_t mlen, const unsigned char *sk) {
+    unsigned char digest[_HASH_LEN];
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(digest, _HASH_LEN, m, mlen);
+
+    memcpy(sm, m, mlen);
+    smlen[0] = mlen + _SIGNATURE_BYTE;
+
+    return PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_sign(sm + mlen, (const sk_t *)sk, digest);
+}
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign_open(unsigned char *m, size_t *mlen, const unsigned char *sm, size_t smlen, const unsigned char *pk) {
+    int rc;
+    if (_SIGNATURE_BYTE > smlen) {
+        rc = -1;
+    } else {
+        *mlen = smlen - _SIGNATURE_BYTE;
+
+        unsigned char digest[_HASH_LEN];
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(digest, _HASH_LEN, sm, *mlen);
+
+        rc = PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_verify_cyclic(digest, sm + mlen[0], (const cpk_t *)pk);
+    }
+    if (!rc) {
+        memcpy(m, sm, smlen - _SIGNATURE_BYTE);
+    } else { // bad signature
+        *mlen = (size_t) -1;
+        memset(m, 0, smlen);
+    }
+    return rc;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign_signature(
+    uint8_t *sig, size_t *siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *sk) {
+    unsigned char digest[_HASH_LEN];
+
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(digest, _HASH_LEN, m, mlen);
+    *siglen = _SIGNATURE_BYTE;
+    return PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_sign(sig, (const sk_t *)sk, digest);
+}
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_crypto_sign_verify(
+    const uint8_t *sig, size_t siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *pk) {
+    if (siglen != _SIGNATURE_BYTE) {
+        return -1;
+    }
+    unsigned char digest[_HASH_LEN];
+    PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(digest, _HASH_LEN, m, mlen);
+    return PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_rainbow_verify_cyclic(digest, sig, (const cpk_t *)pk);
+}

crypto_sign/rainbowIIIc-cyclic/clean/utils_hash.c

@@ -0,0 +1,50 @@
+/// @file utils_hash.c
+/// @brief the adapter for SHA2 families.
+///
+///
+
+#include "utils_hash.h"
+#include "rainbow_config.h"
+#include "sha2.h"
+
+static inline int _hash(unsigned char *digest, const unsigned char *m, size_t mlen) {
+    sha384(digest, m, mlen);
+    return 0;
+}
+
+static inline int expand_hash(unsigned char *digest, size_t n_digest, const unsigned char *hash) {
+    if (_HASH_LEN >= n_digest) {
+        for (size_t i = 0; i < n_digest; i++) {
+            digest[i] = hash[i];
+        }
+        return 0;
+    }
+    for (size_t i = 0; i < _HASH_LEN; i++) {
+        digest[i] = hash[i];
+    }
+    n_digest -= _HASH_LEN;
+
+    while (_HASH_LEN <= n_digest) {
+        _hash(digest + _HASH_LEN, digest, _HASH_LEN);
+
+        n_digest -= _HASH_LEN;
+        digest += _HASH_LEN;
+    }
+    unsigned char temp[_HASH_LEN];
+    if (n_digest) {
+        _hash(temp, digest, _HASH_LEN);
+        for (size_t i = 0; i < n_digest; i++) {
+            digest[_HASH_LEN + i] = temp[i];
+        }
+    }
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(unsigned char *digest,
+        size_t len_digest,
+        const unsigned char *m,
+        size_t mlen) {
+    unsigned char buf[_HASH_LEN];
+    _hash(buf, m, mlen);
+    return expand_hash(digest, len_digest, buf);
+}

crypto_sign/rainbowIIIc-cyclic/clean/utils_hash.h

@@ -0,0 +1,11 @@
+#ifndef _UTILS_HASH_H_
+#define _UTILS_HASH_H_
+/// @file utils_hash.h
+/// @brief the interface for adapting hash functions.
+///
+
+#include <stddef.h>
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(unsigned char *digest, size_t len_digest, const unsigned char *m, size_t mlen);
+
+#endif // _UTILS_HASH_H_

crypto_sign/rainbowIIIc-cyclic/clean/utils_prng.c

@@ -0,0 +1,95 @@
+/// @file utils_prng.c
+/// @brief The implementation of PRNG related functions.
+///
+
+#include "utils_prng.h"
+#include "aes.h"
+#include "randombytes.h"
+#include "utils_hash.h"
+#include <stdlib.h>
+#include <string.h>
+
+static void prng_update(const unsigned char *provided_data,
+                        unsigned char *Key,
+                        unsigned char *V) {
+    unsigned char temp[48];
+    aes256ctx ctx;
+    aes256_keyexp(&ctx, Key);
+    for (int i = 0; i < 3; i++) {
+        //increment V
+        for (int j = 15; j >= 0; j--) {
+            if (V[j] == 0xff) {
+                V[j] = 0x00;
+            } else {
+                V[j]++;
+                break;
+            }
+        }
+        aes256_ecb(temp + 16 * i, V, 1, &ctx);
+    }
+    if (provided_data != NULL) {
+        for (int i = 0; i < 48; i++) {
+            temp[i] ^= provided_data[i];
+        }
+    }
+    memcpy(Key, temp, 32);
+    memcpy(V, temp + 32, 16);
+}
+static void randombytes_init_with_state(prng_t *state,
+                                        unsigned char *entropy_input_48bytes) {
+    memset(state->Key, 0x00, 32);
+    memset(state->V, 0x00, 16);
+    prng_update(entropy_input_48bytes, state->Key, state->V);
+}
+
+static int randombytes_with_state(prng_t *state,
+                                  unsigned char *x,
+                                  size_t xlen) {
+
+    unsigned char block[16];
+    int i = 0;
+
+    aes256ctx ctx;
+    aes256_keyexp(&ctx, state->Key);
+
+    while (xlen > 0) {
+        //increment V
+        for (int j = 15; j >= 0; j--) {
+            if (state->V[j] == 0xff) {
+                state->V[j] = 0x00;
+            } else {
+                state->V[j]++;
+                break;
+            }
+        }
+        aes256_ecb(block, state->V, 1, &ctx);
+        if (xlen > 15) {
+            memcpy(x + i, block, 16);
+            i += 16;
+            xlen -= 16;
+        } else {
+            memcpy(x + i, block, xlen);
+            xlen = 0;
+        }
+    }
+    prng_update(NULL, state->Key, state->V);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_set(prng_t *ctx, const void *prng_seed, unsigned long prng_seedlen) {
+    unsigned char seed[48];
+    if (prng_seedlen >= 48) {
+        memcpy(seed, prng_seed, 48);
+    } else {
+        memcpy(seed, prng_seed, prng_seedlen);
+        PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_hash_msg(seed + prng_seedlen, 48 - (unsigned)prng_seedlen, (const unsigned char *)prng_seed, prng_seedlen);
+    }
+
+    randombytes_init_with_state(ctx, seed);
+
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng_t *ctx, unsigned char *out, unsigned long outlen) {
+    return randombytes_with_state(ctx, out, outlen);
+}

crypto_sign/rainbowIIIc-cyclic/clean/utils_prng.h

@@ -0,0 +1,18 @@
+#ifndef _UTILS_PRNG_H_
+#define _UTILS_PRNG_H_
+/// @file utils_prng.h
+/// @brief the interface for adapting PRNG functions.
+///
+///
+
+#include "randombytes.h"
+
+typedef struct {
+    unsigned char Key[32];
+    unsigned char V[16];
+} prng_t;
+
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_set(prng_t *ctx, const void *prng_seed, unsigned long prng_seedlen);
+int PQCLEAN_RAINBOWIIICCYCLIC_CLEAN_prng_gen(prng_t *ctx, unsigned char *out, unsigned long outlen);
+
+#endif // _UTILS_PRNG_H_

crypto_sign/rainbowIa-classic/META.yml

@@ -0,0 +1,18 @@
+name: Rainbow-Ia-classic
+type: signature
+claimed-nist-level: 1
+length-public-key: 148992
+length-secret-key: 92960
+length-signature: 64
+nistkat-sha256: b75c6fcda2100e2f6f56e9b97c4cbdda4b533116ab217f24f12e08788eb37fd0
+testvectors-sha256: edc48db3f93a66c0aa497fbbdba0bad173e3ab9cd0e3f651004b3e94d2187b75
+principal-submitters:
+  - Jintai Ding
+auxiliary-submitters:
+  - Ming-Shing Chen
+  - Albrecht Petzoldt
+  - Dieter Schmidt
+  - Bo-Yin Yang
+implementations:
+    - name: clean
+      version: https://github.com/fast-crypto-lab/rainbow-submission-round2/commit/af826fcb78f6af51a02d0352cff28a9690467bfd

crypto_sign/rainbowIa-classic/clean/LICENSE

@@ -0,0 +1,8 @@
+`Software implementation of Rainbow for NIST R2 submission' by Ming-Shing Chen
+
+To the extent possible under law, the person who associated CC0 with
+`Software implementation of Rainbow for NIST R2 submission' has waived all copyright and related or neighboring rights
+to `Software implementation of Rainbow for NIST R2 submission'.
+
+You should have received a copy of the CC0 legalcode along with this
+work.  If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.

crypto_sign/rainbowIa-classic/clean/Makefile

@@ -0,0 +1,20 @@
+# This Makefile can be used with GNU Make or BSD Make
+
+LIB=librainbowIa-classic_clean.a
+
+HEADERS = api.h blas_comm.h blas.h blas_u32.h gf.h parallel_matrix_op.h rainbow_blas.h rainbow_config.h rainbow.h rainbow_keypair_computation.h rainbow_keypair.h utils_hash.h utils_prng.h 
+OBJECTS =  blas_comm.o parallel_matrix_op.o rainbow.o rainbow_keypair.o rainbow_keypair_computation.o sign.o utils_hash.o utils_prng.o blas_u32.o gf.o
+
+CFLAGS=-O3 -Wall -Wconversion -Wextra -Wpedantic -Wvla -Werror -Wmissing-prototypes -Wredundant-decls -std=c99 -I../../../common $(EXTRAFLAGS)
+
+all: $(LIB)
+
+%.o: %.c $(HEADERS)
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(LIB): $(OBJECTS)
+	$(AR) -r $@ $(OBJECTS)
+
+clean:
+	$(RM) $(OBJECTS)
+	$(RM) $(LIB)

crypto_sign/rainbowIa-classic/clean/Makefile.Microsoft_nmake

@@ -0,0 +1,19 @@
+# This Makefile can be used with Microsoft Visual Studio's nmake using the command:
+#    nmake /f Makefile.Microsoft_nmake
+
+LIBRARY=librainbowIa-classic_clean.lib
+OBJECTS =  blas_comm.obj parallel_matrix_op.obj rainbow.obj rainbow_keypair.obj rainbow_keypair_computation.obj sign.obj utils_hash.obj utils_prng.obj blas_u32.obj gf.obj
+
+CFLAGS=/nologo /I ..\..\..\common /W4 /WX
+
+all: $(LIBRARY)
+
+# Make sure objects are recompiled if headers change.
+$(OBJECTS): *.h
+
+$(LIBRARY): $(OBJECTS)
+	LIB.EXE /NOLOGO /WX /OUT:$@ $**
+
+clean:
+    -DEL $(OBJECTS)
+    -DEL $(LIBRARY)

crypto_sign/rainbowIa-classic/clean/api.h

@@ -0,0 +1,32 @@
+#ifndef PQCLEAN_RAINBOWIACLASSIC_CLEAN_API_H
+#define PQCLEAN_RAINBOWIACLASSIC_CLEAN_API_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+#define PQCLEAN_RAINBOWIACLASSIC_CLEAN_CRYPTO_SECRETKEYBYTES 92960
+#define PQCLEAN_RAINBOWIACLASSIC_CLEAN_CRYPTO_PUBLICKEYBYTES 148992
+#define PQCLEAN_RAINBOWIACLASSIC_CLEAN_CRYPTO_BYTES 64
+#define PQCLEAN_RAINBOWIACLASSIC_CLEAN_CRYPTO_ALGNAME "RAINBOW(16,32,32,32) - classic"
+
+int PQCLEAN_RAINBOWIACLASSIC_CLEAN_crypto_sign_keypair(uint8_t *pk, uint8_t *sk);
+
+
+int PQCLEAN_RAINBOWIACLASSIC_CLEAN_crypto_sign_signature(
+    uint8_t *sig, size_t *siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *sk);
+
+int PQCLEAN_RAINBOWIACLASSIC_CLEAN_crypto_sign_verify(
+    const uint8_t *sig, size_t siglen,
+    const uint8_t *m, size_t mlen, const uint8_t *pk);
+
+int PQCLEAN_RAINBOWIACLASSIC_CLEAN_crypto_sign(uint8_t *sm, size_t *smlen,
+        const uint8_t *m, size_t mlen,
+        const uint8_t *sk);
+
+int PQCLEAN_RAINBOWIACLASSIC_CLEAN_crypto_sign_open(uint8_t *m, size_t *mlen,
+        const uint8_t *sm, size_t smlen,
+        const uint8_t *pk);
+
+
+#endif

crypto_sign/rainbowIa-classic/clean/blas.h

@@ -0,0 +1,20 @@
+#ifndef _BLAS_H_
+#define _BLAS_H_
+/// @file blas.h
+/// @brief Defining the implementations for linear algebra functions depending on the machine architecture.
+///
+
+#include "blas_comm.h"
+#include "blas_u32.h"
+#include "rainbow_config.h"
+
+#define gf256v_predicated_add PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_predicated_add_u32
+#define gf256v_add PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_add_u32
+
+
+#define gf16v_mul_scalar PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_scalar_u32
+#define gf16v_madd PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_madd_u32
+#define gf16v_dot PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_dot_u32
+
+
+#endif // _BLAS_H_

crypto_sign/rainbowIa-classic/clean/blas_comm.c

@@ -0,0 +1,152 @@
+/// @file blas_comm.c
+/// @brief The standard implementations for blas_comm.h
+///
+
+#include "blas_comm.h"
+#include "blas.h"
+#include "gf.h"
+#include "rainbow_config.h"
+
+#include <stdint.h>
+#include <string.h>
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(uint8_t *b, unsigned int _num_byte) {
+    for (size_t i = 0; i < _num_byte; i++) {
+        b[i] = 0;
+    }
+}
+
+/// @brief get an element from GF(16) vector .
+///
+/// @param[in]  a         - the input vector a.
+/// @param[in]  i         - the index in the vector a.
+/// @return  the value of the element.
+///
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(const uint8_t *a, unsigned int i) {
+    uint8_t r = a[i >> 1];
+    uint8_t r0 = r & 0xf;
+    uint8_t r1 = r >> 4;
+    uint8_t m = (uint8_t)(-((int8_t)i & 1));
+    return (uint8_t)((r1 & m) | ((~m) & r0));
+}
+
+/// @brief set an element for a GF(16) vector .
+///
+/// @param[in,out]   a   - the vector a.
+/// @param[in]  i        - the index in the vector a.
+/// @param[in]  v        - the value for the i-th element in vector a.
+/// @return  the value of the element.
+///
+static uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_set_ele(uint8_t *a, unsigned int i, uint8_t v) {
+    uint8_t m = (uint8_t)(0xf ^ (-((int8_t)i & 1)));                      ///  1--> 0xf0 , 0--> 0x0f
+    uint8_t ai_remaining = (uint8_t)(a[i >> 1] & (~m));                   /// erase
+    a[i >> 1] = (uint8_t)(ai_remaining | (m & (v << 4)) | (m & v & 0xf)); /// set
+    return v;
+}
+
+static void gf16mat_prod_ref(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b) {
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(c, n_A_vec_byte);
+    for (unsigned int i = 0; i < n_A_width; i++) {
+        uint8_t bb = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(b, i);
+        gf16v_madd(c, matA, bb, n_A_vec_byte);
+        matA += n_A_vec_byte;
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_mul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int len_vec) {
+    unsigned int n_vec_byte = (len_vec + 1) / 2;
+    for (unsigned int k = 0; k < len_vec; k++) {
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(c, n_vec_byte);
+        const uint8_t *bk = b + n_vec_byte * k;
+        for (unsigned int i = 0; i < len_vec; i++) {
+            uint8_t bb = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(bk, i);
+            gf16v_madd(c, a + n_vec_byte * i, bb, n_vec_byte);
+        }
+        c += n_vec_byte;
+    }
+}
+
+static unsigned int gf16mat_gauss_elim_ref(uint8_t *mat, unsigned int h, unsigned int w) {
+    unsigned int n_w_byte = (w + 1) / 2;
+    unsigned int r8 = 1;
+    for (unsigned int i = 0; i < h; i++) {
+        unsigned int offset_byte = i >> 1;
+        uint8_t *ai = mat + n_w_byte * i;
+        for (unsigned int j = i + 1; j < h; j++) {
+            uint8_t *aj = mat + n_w_byte * j;
+            gf256v_predicated_add(ai + offset_byte, !PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16_is_nonzero(PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(ai, i)), aj + offset_byte, n_w_byte - offset_byte);
+        }
+        uint8_t pivot = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(ai, i);
+        r8 &= PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16_is_nonzero(pivot);
+        pivot = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16_inv(pivot);
+        offset_byte = (i + 1) >> 1;
+        gf16v_mul_scalar(ai + offset_byte, pivot, n_w_byte - offset_byte);
+        for (unsigned int j = 0; j < h; j++) {
+            if (i == j) {
+                continue;
+            }
+            uint8_t *aj = mat + n_w_byte * j;
+            gf16v_madd(aj + offset_byte, ai + offset_byte, PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(aj, i), n_w_byte - offset_byte);
+        }
+    }
+    return r8;
+}
+
+static unsigned int gf16mat_solve_linear_eq_ref(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n) {
+    uint8_t mat[64 * 33];
+    unsigned int n_byte = (n + 1) >> 1;
+    for (unsigned int i = 0; i < n; i++) {
+        memcpy(mat + i * (n_byte + 1), inp_mat + i * n_byte, n_byte);
+        mat[i * (n_byte + 1) + n_byte] = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(c_terms, i);
+    }
+    unsigned int r8 = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_gauss_elim(mat, n, n + 2);
+    for (unsigned int i = 0; i < n; i++) {
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_set_ele(sol, i, mat[i * (n_byte + 1) + n_byte]);
+    }
+    return r8;
+}
+
+static inline void gf16mat_submat(uint8_t *mat2, unsigned int w2, unsigned int st, const uint8_t *mat, unsigned int w, unsigned int h) {
+    unsigned int n_byte_w1 = (w + 1) / 2;
+    unsigned int n_byte_w2 = (w2 + 1) / 2;
+    unsigned int st_2 = st / 2;
+    for (unsigned int i = 0; i < h; i++) {
+        for (unsigned int j = 0; j < n_byte_w2; j++) {
+            mat2[i * n_byte_w2 + j] = mat[i * n_byte_w1 + st_2 + j];
+        }
+    }
+}
+
+unsigned int PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_inv(uint8_t *inv_a, const uint8_t *a, unsigned int H, uint8_t *buffer) {
+    unsigned int n_w_byte = (H + 1) / 2;
+
+    uint8_t *aa = buffer;
+    for (unsigned int i = 0; i < H; i++) {
+        uint8_t *ai = aa + i * 2 * n_w_byte;
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(ai, 2 * n_w_byte);
+        gf256v_add(ai, a + i * n_w_byte, n_w_byte);
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_set_ele(ai + n_w_byte, i, 1);
+    }
+    unsigned int r8 = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_gauss_elim(aa, H, 2 * H);
+    gf16mat_submat(inv_a, H, H, aa, 2 * H, H);
+    return r8;
+}
+
+// choosing the implementations depends on the macros _BLAS_AVX2_ and _BLAS_SSE
+
+#define gf16mat_prod_impl gf16mat_prod_ref
+#define gf16mat_gauss_elim_impl gf16mat_gauss_elim_ref
+#define gf16mat_solve_linear_eq_impl gf16mat_solve_linear_eq_ref
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_prod(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b) {
+    gf16mat_prod_impl(c, matA, n_A_vec_byte, n_A_width, b);
+}
+
+unsigned int PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_gauss_elim(uint8_t *mat, unsigned int h, unsigned int w) {
+    return gf16mat_gauss_elim_impl(mat, h, w);
+}
+
+unsigned int PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_solve_linear_eq(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n) {
+    return gf16mat_solve_linear_eq_impl(sol, inp_mat, c_terms, n);
+}
+

crypto_sign/rainbowIa-classic/clean/blas_comm.h

@@ -0,0 +1,74 @@
+#ifndef _BLAS_COMM_H_
+#define _BLAS_COMM_H_
+/// @file blas_comm.h
+/// @brief Common functions for linear algebra.
+///
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+/// @brief set a vector to 0.
+///
+/// @param[in,out]   b      - the vector b.
+/// @param[in]  _num_byte   - number of bytes for the vector b.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(uint8_t *b, unsigned int _num_byte);
+
+/// @brief get an element from GF(16) vector .
+///
+/// @param[in]  a         - the input vector a.
+/// @param[in]  i         - the index in the vector a.
+/// @return  the value of the element.
+///
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(const uint8_t *a, unsigned int i);
+
+/// @brief matrix-matrix multiplication:  c = a * b , in GF(16)
+///
+/// @param[out]  c         - the output matrix c
+/// @param[in]   c         - a matrix a.
+/// @param[in]   b         - a matrix b.
+/// @param[in]   len_vec   - the length of column vectors.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_mul(uint8_t *c, const uint8_t *a, const uint8_t *b, unsigned int len_vec);
+
+/// @brief Gauss elimination for a matrix, in GF(16)
+///
+/// @param[in,out]  mat    - the matrix.
+/// @param[in]   h         - the height of the matrix.
+/// @param[in]   w         - the width of the matrix.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_gauss_elim(uint8_t *mat, unsigned int h, unsigned int w);
+
+/// @brief Solving linear equations, in GF(16)
+///
+/// @param[out]  sol       - the solutions.
+/// @param[in]   inp_mat   - the matrix parts of input equations.
+/// @param[in]   c_terms   - the constant terms of the input equations.
+/// @param[in]   n         - the number of equations.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_solve_linear_eq(uint8_t *sol, const uint8_t *inp_mat, const uint8_t *c_terms, unsigned int n);
+
+/// @brief Computing the inverse matrix, in GF(16)
+///
+/// @param[out]  inv_a     - the output of matrix a.
+/// @param[in]   a         - a matrix a.
+/// @param[in]   H         - height of matrix a, i.e., matrix a is an HxH matrix.
+/// @param[in]   buffer    - The buffer for computations. it has to be as large as 2 input matrixes.
+/// @return   1(true) if success. 0(false) if the matrix is singular.
+///
+unsigned int PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_inv(uint8_t *inv_a, const uint8_t *a, unsigned int H, uint8_t *buffer);
+
+/// @brief matrix-vector multiplication:  c = matA * b , in GF(16)
+///
+/// @param[out]  c         - the output vector c
+/// @param[in]   matA      - a column-major matrix A.
+/// @param[in]   n_A_vec_byte  - the size of column vectors in bytes.
+/// @param[in]   n_A_width   - the width of matrix A.
+/// @param[in]   b          - the vector b.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_prod(uint8_t *c, const uint8_t *matA, unsigned int n_A_vec_byte, unsigned int n_A_width, const uint8_t *b);
+
+
+#endif // _BLAS_COMM_H_

crypto_sign/rainbowIa-classic/clean/blas_u32.c

@@ -0,0 +1,115 @@
+#include "blas_u32.h"
+#include "gf.h"
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_predicated_add_u32(uint8_t *accu_b, uint8_t predicate, const uint8_t *a, unsigned int _num_byte) {
+    uint32_t pr_u32 = ((uint32_t)0) - ((uint32_t)predicate);
+    uint8_t pr_u8 = pr_u32 & 0xff;
+
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *b_u32 = (uint32_t *)accu_b;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        b_u32[i] ^= (a_u32[i] & pr_u32);
+    }
+
+    a += (n_u32 << 2);
+    accu_b += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_b[i] ^= (a[i] & pr_u8);
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_add_u32(uint8_t *accu_b, const uint8_t *a, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *b_u32 = (uint32_t *)accu_b;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        b_u32[i] ^= a_u32[i];
+    }
+
+    a += (n_u32 << 2);
+    accu_b += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_b[i] ^= a[i];
+    }
+}
+
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_scalar_u32(uint8_t *a, uint8_t gf16_b, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *a_u32 = (uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        a_u32[i] = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32(a_u32[i], gf16_b);
+    }
+
+    union tmp_32 {
+        uint8_t u8[4];
+        uint32_t u32;
+    } t;
+    t.u32 = 0;
+    a += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        t.u8[i] = a[i];
+    }
+    t.u32 = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32(t.u32, gf16_b);
+    for (unsigned int i = 0; i < rem; i++) {
+        a[i] = t.u8[i];
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_madd_u32(uint8_t *accu_c, const uint8_t *a, uint8_t gf16_b, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    uint32_t *c_u32 = (uint32_t *)accu_c;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        c_u32[i] ^= PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32(a_u32[i], gf16_b);
+    }
+
+    union tmp_32 {
+        uint8_t u8[4];
+        uint32_t u32;
+    } t;
+    t.u32 = 0;
+    accu_c += (n_u32 << 2);
+    a += (n_u32 << 2);
+    unsigned int rem = _num_byte & 3;
+    for (unsigned int i = 0; i < rem; i++) {
+        t.u8[i] = a[i];
+    }
+    t.u32 = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32(t.u32, gf16_b);
+    for (unsigned int i = 0; i < rem; i++) {
+        accu_c[i] ^= t.u8[i];
+    }
+}
+
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_dot_u32(const uint8_t *a, const uint8_t *b, unsigned int _num_byte) {
+    unsigned int n_u32 = _num_byte >> 2;
+    const uint32_t *a_u32 = (const uint32_t *)a;
+    const uint32_t *b_u32 = (const uint32_t *)b;
+    uint32_t r = 0;
+    for (unsigned int i = 0; i < n_u32; i++) {
+        r ^= PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32_u32(a_u32[i], b_u32[i]);
+    }
+
+    unsigned int rem = _num_byte & 3;
+    if (rem) {
+        union tmp_32 {
+            uint8_t u8[4];
+            uint32_t u32;
+        } ta, tb;
+        ta.u32 = 0;
+        tb.u32 = 0;
+        for (unsigned int i = 0; i < rem; i++) {
+            ta.u8[i] = a[(n_u32 << 2) + i];
+        }
+        for (unsigned int i = 0; i < rem; i++) {
+            tb.u8[i] = b[(n_u32 << 2) + i];
+        }
+        r ^= PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32_u32(ta.u32, tb.u32);
+    }
+    return PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_reduce_u32(r);
+}
+

crypto_sign/rainbowIa-classic/clean/blas_u32.h

@@ -0,0 +1,19 @@
+#ifndef _BLAS_U32_H_
+#define _BLAS_U32_H_
+/// @file blas_u32.h
+/// @brief Inlined functions for implementing basic linear algebra functions for uint32 arch.
+///
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_predicated_add_u32(uint8_t *accu_b, uint8_t predicate, const uint8_t *a, unsigned int _num_byte);
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_add_u32(uint8_t *accu_b, const uint8_t *a, unsigned int _num_byte);
+
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_madd_u32(uint8_t *accu_c, const uint8_t *a, uint8_t gf16_b, unsigned int _num_byte);
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_scalar_u32(uint8_t *a, uint8_t gf16_b, unsigned int _num_byte);
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_dot_u32(const uint8_t *a, const uint8_t *b, unsigned int _num_byte);
+
+
+#endif // _BLAS_U32_H_

crypto_sign/rainbowIa-classic/clean/gf.c

@@ -0,0 +1,124 @@
+#include "gf.h"
+
+//// gf4 := gf2[x]/x^2+x+1
+static inline uint8_t gf4_mul_2(uint8_t a) {
+    uint8_t r = (uint8_t)(a << 1);
+    r ^= (uint8_t)((a >> 1) * 7);
+    return r;
+}
+
+static inline uint8_t gf4_mul(uint8_t a, uint8_t b) {
+    uint8_t r = (uint8_t)(a * (b & 1));
+    return r ^ (uint8_t)(gf4_mul_2(a) * (b >> 1));
+}
+
+static inline uint8_t gf4_squ(uint8_t a) {
+    return a ^ (a >> 1);
+}
+
+static inline uint32_t gf4v_mul_2_u32(uint32_t a) {
+    uint32_t bit0 = a & 0x55555555;
+    uint32_t bit1 = a & 0xaaaaaaaa;
+    return (bit0 << 1) ^ bit1 ^ (bit1 >> 1);
+}
+
+static inline uint32_t gf4v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t bit0_b = ((uint32_t)0) - ((uint32_t)(b & 1));
+    uint32_t bit1_b = ((uint32_t)0) - ((uint32_t)((b >> 1) & 1));
+    return (a & bit0_b) ^ (bit1_b & gf4v_mul_2_u32(a));
+}
+
+//// gf16 := gf4[y]/y^2+y+x
+static inline uint8_t gf16_mul(uint8_t a, uint8_t b) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = (a >> 2);
+    uint8_t b0 = b & 3;
+    uint8_t b1 = (b >> 2);
+    uint8_t a0b0 = gf4_mul(a0, b0);
+    uint8_t a1b1 = gf4_mul(a1, b1);
+    uint8_t a0b1_a1b0 = gf4_mul(a0 ^ a1, b0 ^ b1) ^ a0b0 ^ a1b1;
+    uint8_t a1b1_x2 = gf4_mul_2(a1b1);
+    return (uint8_t)((a0b1_a1b0 ^ a1b1) << 2 ^ a0b0 ^ a1b1_x2);
+}
+
+static inline uint8_t gf16_squ(uint8_t a) {
+    uint8_t a0 = a & 3;
+    uint8_t a1 = (a >> 2);
+    a1 = gf4_squ(a1);
+    uint8_t a1squ_x2 = gf4_mul_2(a1);
+    return (uint8_t)((a1 << 2) ^ a1squ_x2 ^ gf4_squ(a0));
+}
+
+// gf16 := gf4[y]/y^2+y+x
+uint32_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32(uint32_t a, uint8_t b) {
+    uint32_t axb0 = gf4v_mul_u32(a, b);
+    uint32_t axb1 = gf4v_mul_u32(a, b >> 2);
+    uint32_t a0b1 = (axb1 << 2) & 0xcccccccc;
+    uint32_t a1b1 = axb1 & 0xcccccccc;
+    uint32_t a1b1_2 = a1b1 >> 2;
+
+    return axb0 ^ a0b1 ^ a1b1 ^ gf4v_mul_2_u32(a1b1_2);
+}
+
+
+static inline uint32_t _gf4v_mul_u32_u32(uint32_t a0, uint32_t a1, uint32_t b0, uint32_t b1) {
+    uint32_t c0 = a0 & b0;
+    uint32_t c2 = a1 & b1;
+    uint32_t c1_ = (a0 ^ a1) & (b0 ^ b1);
+    return ((c1_ ^ c0) << 1) ^ c0 ^ c2;
+}
+
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16_is_nonzero(uint8_t a) {
+    unsigned int a4 = a & 0xf;
+    unsigned int r = ((unsigned int)0) - a4;
+    r >>= 4;
+    return r & 1;
+}
+
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16_inv(uint8_t a) {
+    uint8_t a2 = gf16_squ(a);
+    uint8_t a4 = gf16_squ(a2);
+    uint8_t a8 = gf16_squ(a4);
+    uint8_t a6 = gf16_mul(a4, a2);
+    return gf16_mul(a8, a6);
+}
+
+static inline uint32_t _gf16v_mul_u32_u32(uint32_t a0, uint32_t a1, uint32_t a2, uint32_t a3, uint32_t b0, uint32_t b1, uint32_t b2, uint32_t b3) {
+    uint32_t c0 = _gf4v_mul_u32_u32(a0, a1, b0, b1);
+    uint32_t c1_ = _gf4v_mul_u32_u32(a0 ^ a2, a1 ^ a3, b0 ^ b2, b1 ^ b3);
+
+    uint32_t c2_0 = a2 & b2;
+    uint32_t c2_2 = a3 & b3;
+    uint32_t c2_1_ = (a2 ^ a3) & (b2 ^ b3);
+    uint32_t c2_r0 = c2_0 ^ c2_2;
+    uint32_t c2_r1 = c2_0 ^ c2_1_;
+    // GF(4) x2: (bit0<<1)^bit1^(bit1>>1);
+    return ((c1_ ^ c0) << 2) ^ c0 ^ (c2_r0 << 1) ^ c2_r1 ^ (c2_r1 << 1);
+}
+
+uint32_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32_u32(uint32_t a, uint32_t b) {
+    uint32_t a0 = a & 0x11111111;
+    uint32_t a1 = (a >> 1) & 0x11111111;
+    uint32_t a2 = (a >> 2) & 0x11111111;
+    uint32_t a3 = (a >> 3) & 0x11111111;
+    uint32_t b0 = b & 0x11111111;
+    uint32_t b1 = (b >> 1) & 0x11111111;
+    uint32_t b2 = (b >> 2) & 0x11111111;
+    uint32_t b3 = (b >> 3) & 0x11111111;
+
+    return _gf16v_mul_u32_u32(a0, a1, a2, a3, b0, b1, b2, b3);
+}
+
+static inline uint8_t gf256v_reduce_u32(uint32_t a) {
+    // https://godbolt.org/z/7hirMb
+    uint16_t *aa = (uint16_t *)(&a);
+    uint16_t r = aa[0] ^ aa[1];
+    uint8_t *rr = (uint8_t *)(&r);
+    return rr[0] ^ rr[1];
+}
+
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_reduce_u32(uint32_t a) {
+    uint8_t r256 = gf256v_reduce_u32(a);
+    return (uint8_t)((r256 & 0xf) ^ (r256 >> 4));
+}
+

crypto_sign/rainbowIa-classic/clean/gf.h

@@ -0,0 +1,20 @@
+#ifndef _GF16_H_
+#define _GF16_H_
+
+#include "rainbow_config.h"
+#include <stdint.h>
+
+/// @file gf16.h
+/// @brief Library for arithmetics in GF(16) and GF(256)
+///
+
+uint32_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32(uint32_t a, uint8_t b);
+
+
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16_is_nonzero(uint8_t a);
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16_inv(uint8_t a);
+uint32_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_u32_u32(uint32_t a, uint32_t b);
+uint8_t PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_reduce_u32(uint32_t a);
+
+
+#endif // _GF16_H_

crypto_sign/rainbowIa-classic/clean/parallel_matrix_op.c

@@ -0,0 +1,182 @@
+///  @file parallel_matrix_op.c
+///  @brief the standard implementations for functions in parallel_matrix_op.h
+///
+///  the standard implementations for functions in parallel_matrix_op.h
+///
+
+#include "parallel_matrix_op.h"
+#include "blas.h"
+#include "blas_comm.h"
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle(UT) matrix.
+///
+/// @param[in]  i_row     - the i-th row in an upper-triangle matrix.
+/// @param[in]  j_col     - the j-th column in an upper-triangle matrix.
+/// @param[in]  dim       - the dimension of the upper-triangle matrix, i.e., an dim x dim matrix.
+/// @return  the corresponding index in an array storage.
+///
+unsigned int PQCLEAN_RAINBOWIACLASSIC_CLEAN_idx_of_trimat(unsigned int i_row, unsigned int j_col, unsigned int dim) {
+    return (dim + dim - i_row + 1) * i_row / 2 + j_col - i_row;
+}
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle or lower-triangle matrix.
+///
+/// @param[in]  i_row     - the i-th row in a triangle matrix.
+/// @param[in]  j_col     - the j-th column in a triangle matrix.
+/// @param[in]  dim       - the dimension of the triangle matrix, i.e., an dim x dim matrix.
+/// @return    the corresponding index in an array storage.
+///
+static inline unsigned int idx_of_2trimat(unsigned int i_row, unsigned int j_col, unsigned int n_var) {
+    if (i_row > j_col) {
+        return PQCLEAN_RAINBOWIACLASSIC_CLEAN_idx_of_trimat(j_col, i_row, n_var);
+    }
+    return PQCLEAN_RAINBOWIACLASSIC_CLEAN_idx_of_trimat(i_row, j_col, n_var);
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_UpperTrianglize(unsigned char *btriC, const unsigned char *bA, unsigned int Awidth, unsigned int size_batch) {
+    unsigned char *runningC = btriC;
+    unsigned int Aheight = Awidth;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < i; j++) {
+            unsigned int idx = PQCLEAN_RAINBOWIACLASSIC_CLEAN_idx_of_trimat(j, i, Aheight);
+            gf256v_add(btriC + idx * size_batch, bA + size_batch * (i * Awidth + j), size_batch);
+        }
+        gf256v_add(runningC, bA + size_batch * (i * Awidth + i), size_batch * (Aheight - i));
+        runningC += size_batch * (Aheight - i);
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Awidth = Bheight;
+    unsigned int Aheight = Awidth;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (k < i) {
+                    continue;
+                }
+                gf16v_madd(bC, &btriA[(k - i) * size_batch], PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+        btriA += (Aheight - i) * size_batch;
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_trimatTr_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Aheight = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (i < k) {
+                    continue;
+                }
+                gf16v_madd(bC, &btriA[size_batch * (PQCLEAN_RAINBOWIACLASSIC_CLEAN_idx_of_trimat(k, i, Aheight))], PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_2trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Aheight = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                if (i == k) {
+                    continue;
+                }
+                gf16v_madd(bC, &btriA[size_batch * (idx_of_2trimat(i, k, Aheight))], PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_matTr_madd_gf16(unsigned char *bC, const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Atr_height = Awidth;
+    unsigned int Atr_width = Aheight;
+    for (unsigned int i = 0; i < Atr_height; i++) {
+        for (unsigned int j = 0; j < Atr_width; j++) {
+            gf16v_madd(bC, &bB[j * Bwidth * size_batch], PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(&A_to_tr[size_Acolvec * i], j), size_batch * Bwidth);
+        }
+        bC += size_batch * Bwidth;
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_bmatTr_madd_gf16(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    const unsigned char *bA = bA_to_tr;
+    unsigned int Aheight = Awidth_before_tr;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                gf16v_madd(bC, &bA[size_batch * (i + k * Aheight)], PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_mat_madd_gf16(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch) {
+    unsigned int Awidth = Bheight;
+    for (unsigned int i = 0; i < Aheight; i++) {
+        for (unsigned int j = 0; j < Bwidth; j++) {
+            for (unsigned int k = 0; k < Bheight; k++) {
+                gf16v_madd(bC, &bA[k * size_batch], PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(&B[j * size_Bcolvec], k), size_batch);
+            }
+            bC += size_batch;
+        }
+        bA += (Awidth) * size_batch;
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_quad_recmat_eval_gf16(unsigned char *z, const unsigned char *y, unsigned int dim_y, const unsigned char *mat,
+        const unsigned char *x, unsigned int dim_x, unsigned int size_batch) {
+    unsigned char tmp[128];
+
+    unsigned char _x[128];
+    for (unsigned int i = 0; i < dim_x; i++) {
+        _x[i] = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(x, i);
+    }
+    unsigned char _y[128];
+    for (unsigned int i = 0; i < dim_y; i++) {
+        _y[i] = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(y, i);
+    }
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(z, size_batch);
+    for (unsigned int i = 0; i < dim_y; i++) {
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(tmp, size_batch);
+        for (unsigned int j = 0; j < dim_x; j++) {
+            gf16v_madd(tmp, mat, _x[j], size_batch);
+            mat += size_batch;
+        }
+        gf16v_madd(z, tmp, _y[i], size_batch);
+    }
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_quad_trimat_eval_gf16(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch) {
+    unsigned char tmp[256];
+
+    unsigned char _x[256];
+    for (unsigned int i = 0; i < dim; i++) {
+        _x[i] = PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele(x, i);
+    }
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(y, size_batch);
+    for (unsigned int i = 0; i < dim; i++) {
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(tmp, size_batch);
+        for (unsigned int j = i; j < dim; j++) {
+            gf16v_madd(tmp, trimat, _x[j], size_batch);
+            trimat += size_batch;
+        }
+        gf16v_madd(y, tmp, _x[i], size_batch);
+    }
+}

crypto_sign/rainbowIa-classic/clean/parallel_matrix_op.h

@@ -0,0 +1,260 @@
+#ifndef _P_MATRIX_OP_H_
+#define _P_MATRIX_OP_H_
+///  @file  parallel_matrix_op.h
+///  @brief Librarys for operations of batched matrixes.
+///
+///
+
+////////////////  Section:  triangle matrix <-> rectangle matrix   ///////////////////////////////////
+
+///
+/// @brief  Calculate the corresponding index in an array for an upper-triangle(UT) matrix.
+///
+/// @param[in]  i_row     - the i-th row in an upper-triangle matrix.
+/// @param[in]  j_col     - the j-th column in an upper-triangle matrix.
+/// @param[in]  dim       - the dimension of the upper-triangle matrix, i.e., an dim x dim matrix.
+/// @return  the corresponding index in an array storage.
+///
+unsigned int PQCLEAN_RAINBOWIACLASSIC_CLEAN_idx_of_trimat(unsigned int i_row, unsigned int j_col, unsigned int dim);
+
+///
+/// @brief  Upper trianglize a rectangle matrix to the corresponding upper-trangle matrix.
+///
+/// @param[out]  btriC    - the batched upper-trianglized matrix C.
+/// @param[in]   bA       - a batched retangle matrix A.
+/// @param[in]   bwidth   - the width of the batched matrix A, i.e., A is a Awidth x Awidth matrix.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_UpperTrianglize(unsigned char *btriC, const unsigned char *bA, unsigned int Awidth, unsigned int size_batch);
+
+////////////////////  Section:  matrix multiplications  ///////////////////////////////
+
+///
+/// @brief  bC += btriA * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA^Tr * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. A will be transposed while multiplying.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_trimatTr_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += btriA^Tr * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A, which will be transposed while multiplying.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_trimatTr_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC +=  (btriA + btriA^Tr) *B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. The operand for multiplication is (btriA + btriA^Tr).
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_2trimat_madd_gf16(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC +=  (btriA + btriA^Tr) *B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   btriA      - a batched UT matrix A. The operand for multiplication is (btriA + btriA^Tr).
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_2trimat_madd_gf256(unsigned char *bC, const unsigned char *btriA,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += A^Tr * bB  , in GF(16)
+///
+/// @param[out]  bC           - the batched matrix C.
+/// @param[in]   A_to_tr      - a column-major matrix A. The operand for multiplication is A^Tr.
+/// @param[in]   Aheight      - the height of A.
+/// @param[in]   size_Acolvec    - the size of a column vector in A.
+/// @param[in]   Awidth           - the width of A.
+/// @param[in]   bB          - a batched matrix B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_matTr_madd_gf16(unsigned char *bC,
+        const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += A^Tr * bB  , in GF(256)
+///
+/// @param[out]  bC           - the batched matrix C.
+/// @param[in]   A_to_tr      - a column-major matrix A. The operand for multiplication is A^Tr.
+/// @param[in]   Aheight      - the height of A.
+/// @param[in]   size_Acolvec    - the size of a column vector in A.
+/// @param[in]   Awidth           - the width of A.
+/// @param[in]   bB          - a batched matrix B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_matTr_madd_gf256(unsigned char *bC,
+        const unsigned char *A_to_tr, unsigned int Aheight, unsigned int size_Acolvec, unsigned int Awidth,
+        const unsigned char *bB, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA^Tr * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA_to_tr   - a batched matrix A. The operand for multiplication is (bA^Tr).
+/// @param[in]   Awidth_befor_tr     - the width of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_bmatTr_madd_gf16(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA^Tr * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA_to_tr   - a batched matrix A. The operand for multiplication is (bA^Tr).
+/// @param[in]   Awidth_befor_tr     - the width of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_bmatTr_madd_gf256(unsigned char *bC, const unsigned char *bA_to_tr, unsigned int Awidth_before_tr,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA * B  , in GF(16)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA         - a batched matrix A.
+/// @param[in]   Aheigh     - the height of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_mat_madd_gf16(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+///
+/// @brief  bC += bA * B  , in GF(256)
+///
+/// @param[out]  bC         - the batched matrix C.
+/// @param[in]   bA         - a batched matrix A.
+/// @param[in]   Aheigh     - the height of A.
+/// @param[in]   B          - a column-major matrix B.
+/// @param[in]   Bheight          - the height of B.
+/// @param[in]   size_Bcolvec     - the size of the column vector in B.
+/// @param[in]   Bwidth           - the width of B.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_mat_madd_gf256(unsigned char *bC, const unsigned char *bA, unsigned int Aheight,
+        const unsigned char *B, unsigned int Bheight, unsigned int size_Bcolvec, unsigned int Bwidth, unsigned int size_batch);
+
+////////////////////  Section: "quadratric" matrix evaluation  ///////////////////////////////
+
+///
+/// @brief  y =  x^Tr * trimat * x  , in GF(16)
+///
+/// @param[out]  y          - the returned batched element y.
+/// @param[in]   trimat     - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim        - the dimension of matrix trimat (and x).
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_quad_trimat_eval_gf16(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch);
+
+///
+/// @brief  y =  x^Tr * trimat * x  , in GF(256)
+///
+/// @param[out]  y          - the returned batched element y.
+/// @param[in]   trimat     - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim        - the dimension of matrix trimat (and x).
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_quad_trimat_eval_gf256(unsigned char *y, const unsigned char *trimat, const unsigned char *x, unsigned int dim, unsigned int size_batch);
+
+///
+/// @brief  z =  y^Tr * mat * x  , in GF(16)
+///
+/// @param[out]  z          - the returned batched element z.
+/// @param[in]   y          - an input vector y.
+/// @param[in]   dim_y      - the length of y.
+/// @param[in]   mat        - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim_x        - the length of x.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_quad_recmat_eval_gf16(unsigned char *z, const unsigned char *y, unsigned int dim_y,
+        const unsigned char *mat, const unsigned char *x, unsigned int dim_x, unsigned int size_batch);
+
+///
+/// @brief  z =  y^Tr * mat * x  , in GF(256)
+///
+/// @param[out]  z          - the returned batched element z.
+/// @param[in]   y          - an input vector y.
+/// @param[in]   dim_y      - the length of y.
+/// @param[in]   mat        - a batched matrix.
+/// @param[in]   x          - an input vector x.
+/// @param[in]   dim_x        - the length of x.
+/// @param[in]   size_batch - number of the batched elements in the corresponding position of the matrix.
+///
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_quad_recmat_eval_gf256(unsigned char *z, const unsigned char *y, unsigned int dim_y,
+        const unsigned char *mat, const unsigned char *x, unsigned int dim_x, unsigned int size_batch);
+
+#endif // _P_MATRIX_OP_H_

crypto_sign/rainbowIa-classic/clean/rainbow.c

@@ -0,0 +1,169 @@
+/// @file rainbow.c
+/// @brief The standard implementations for functions in rainbow.h
+///
+
+#include "rainbow.h"
+#include "blas.h"
+#include "rainbow_blas.h"
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+#include "utils_hash.h"
+#include "utils_prng.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#define MAX_ATTEMPT_FRMAT 128
+#define _MAX_O ((_O1 > _O2) ? _O1 : _O2)
+#define _MAX_O_BYTE ((_O1_BYTE > _O2_BYTE) ? _O1_BYTE : _O2_BYTE)
+
+int PQCLEAN_RAINBOWIACLASSIC_CLEAN_rainbow_sign(uint8_t *signature, const sk_t *sk, const uint8_t *_digest) {
+    uint8_t mat_l1[_O1 * _O1_BYTE];
+    uint8_t mat_l2[_O2 * _O2_BYTE];
+    uint8_t mat_buffer[2 * _MAX_O * _MAX_O_BYTE];
+
+    // setup PRNG
+    prng_t prng_sign;
+    uint8_t prng_preseed[LEN_SKSEED + _HASH_LEN];
+    memcpy(prng_preseed, sk->sk_seed, LEN_SKSEED);
+    memcpy(prng_preseed + LEN_SKSEED, _digest, _HASH_LEN); // prng_preseed = sk_seed || digest
+    uint8_t prng_seed[_HASH_LEN];
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_hash_msg(prng_seed, _HASH_LEN, prng_preseed, _HASH_LEN + LEN_SKSEED);
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_set(&prng_sign, prng_seed, _HASH_LEN); // seed = H( sk_seed || digest )
+    for (unsigned int i = 0; i < LEN_SKSEED + _HASH_LEN; i++) {
+        prng_preseed[i] ^= prng_preseed[i]; // clean
+    }
+    for (unsigned int i = 0; i < _HASH_LEN; i++) {
+        prng_seed[i] ^= prng_seed[i]; // clean
+    }
+
+    // roll vinegars.
+    uint8_t vinegar[_V1_BYTE];
+    unsigned int n_attempt = 0;
+    unsigned int l1_succ = 0;
+    while (!l1_succ) {
+        if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+            break;
+        }
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(&prng_sign, vinegar, _V1_BYTE);   // generating vinegars
+        gfmat_prod(mat_l1, sk->l1_F2, _O1 * _O1_BYTE, _V1, vinegar); // generating the linear equations for layer 1
+        l1_succ = gfmat_inv(mat_l1, mat_l1, _O1, mat_buffer);        // check if the linear equation solvable
+        n_attempt++;
+    }
+
+    // Given the vinegars, pre-compute variables needed for layer 2
+    uint8_t r_l1_F1[_O1_BYTE] = {0};
+    uint8_t r_l2_F1[_O2_BYTE] = {0};
+    batch_quad_trimat_eval(r_l1_F1, sk->l1_F1, vinegar, _V1, _O1_BYTE);
+    batch_quad_trimat_eval(r_l2_F1, sk->l2_F1, vinegar, _V1, _O2_BYTE);
+    uint8_t mat_l2_F3[_O2 * _O2_BYTE];
+    uint8_t mat_l2_F2[_O1 * _O2_BYTE];
+    gfmat_prod(mat_l2_F3, sk->l2_F3, _O2 * _O2_BYTE, _V1, vinegar);
+    gfmat_prod(mat_l2_F2, sk->l2_F2, _O1 * _O2_BYTE, _V1, vinegar);
+
+    // Some local variables.
+    uint8_t _z[_PUB_M_BYTE];
+    uint8_t y[_PUB_M_BYTE];
+    uint8_t *x_v1 = vinegar;
+    uint8_t x_o1[_O1_BYTE];
+    uint8_t x_o2[_O1_BYTE];
+
+    uint8_t digest_salt[_HASH_LEN + _SALT_BYTE];
+    memcpy(digest_salt, _digest, _HASH_LEN);
+    uint8_t *salt = digest_salt + _HASH_LEN;
+
+    uint8_t temp_o[_MAX_O_BYTE + 32] = {0};
+    unsigned int succ = 0;
+    while (!succ) {
+        if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+            break;
+        }
+        // The computation:  H(digest||salt)  -->   z   --S-->   y  --C-map-->   x   --T-->   w
+
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(&prng_sign, salt, _SALT_BYTE);                         // roll the salt
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_hash_msg(_z, _PUB_M_BYTE, digest_salt, _HASH_LEN + _SALT_BYTE); // H(digest||salt)
+
+        //  y = S^-1 * z
+        memcpy(y, _z, _PUB_M_BYTE); // identity part of S
+        gfmat_prod(temp_o, sk->s1, _O1_BYTE, _O2, _z + _O1_BYTE);
+        gf256v_add(y, temp_o, _O1_BYTE);
+
+        // Central Map:
+        // layer 1: calculate x_o1
+        memcpy(temp_o, r_l1_F1, _O1_BYTE);
+        gf256v_add(temp_o, y, _O1_BYTE);
+        gfmat_prod(x_o1, mat_l1, _O1_BYTE, _O1, temp_o);
+
+        // layer 2: calculate x_o2
+        PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf256v_set_zero(temp_o, _O2_BYTE);
+        gfmat_prod(temp_o, mat_l2_F2, _O2_BYTE, _O1, x_o1);             // F2
+        batch_quad_trimat_eval(mat_l2, sk->l2_F5, x_o1, _O1, _O2_BYTE); // F5
+        gf256v_add(temp_o, mat_l2, _O2_BYTE);
+        gf256v_add(temp_o, r_l2_F1, _O2_BYTE); // F1
+        gf256v_add(temp_o, y + _O1_BYTE, _O2_BYTE);
+
+        // generate the linear equations of the 2nd layer
+        gfmat_prod(mat_l2, sk->l2_F6, _O2 * _O2_BYTE, _O1, x_o1); // F6
+        gf256v_add(mat_l2, mat_l2_F3, _O2 * _O2_BYTE);            // F3
+        succ = gfmat_inv(mat_l2, mat_l2, _O2, mat_buffer);
+        gfmat_prod(x_o2, mat_l2, _O2_BYTE, _O2, temp_o); // solve l2 eqs
+
+        n_attempt++;
+    };
+    //  w = T^-1 * y
+    uint8_t w[_PUB_N_BYTE];
+    // identity part of T.
+    memcpy(w, x_v1, _V1_BYTE);
+    memcpy(w + _V1_BYTE, x_o1, _O1_BYTE);
+    memcpy(w + _V2_BYTE, x_o2, _O2_BYTE);
+    // Computing the t1 part.
+    gfmat_prod(y, sk->t1, _V1_BYTE, _O1, x_o1);
+    gf256v_add(w, y, _V1_BYTE);
+    // Computing the t4 part.
+    gfmat_prod(y, sk->t4, _V1_BYTE, _O2, x_o2);
+    gf256v_add(w, y, _V1_BYTE);
+    // Computing the t3 part.
+    gfmat_prod(y, sk->t3, _O1_BYTE, _O2, x_o2);
+    gf256v_add(w + _V1_BYTE, y, _O1_BYTE);
+
+    memset(signature, 0, _SIGNATURE_BYTE); // set the output 0
+    // clean
+    memset(&prng_sign, 0, sizeof(prng_t));
+    memset(vinegar, 0, _V1_BYTE);
+    memset(r_l1_F1, 0, _O1_BYTE);
+    memset(r_l2_F1, 0, _O2_BYTE);
+    memset(_z, 0, _PUB_M_BYTE);
+    memset(y, 0, _PUB_M_BYTE);
+    memset(x_o1, 0, _O1_BYTE);
+    memset(x_o2, 0, _O2_BYTE);
+    memset(temp_o, 0, sizeof(temp_o));
+
+    // return: copy w and salt to the signature.
+    if (MAX_ATTEMPT_FRMAT <= n_attempt) {
+        return -1;
+    }
+    gf256v_add(signature, w, _PUB_N_BYTE);
+    gf256v_add(signature + _PUB_N_BYTE, salt, _SALT_BYTE);
+    return 0;
+}
+
+int PQCLEAN_RAINBOWIACLASSIC_CLEAN_rainbow_verify(const uint8_t *digest, const uint8_t *signature, const pk_t *pk) {
+    unsigned char digest_ck[_PUB_M_BYTE];
+    // public_map( digest_ck , pk , signature ); Evaluating the quadratic public polynomials.
+    batch_quad_trimat_eval(digest_ck, pk->pk, signature, _PUB_N, _PUB_M_BYTE);
+
+    unsigned char correct[_PUB_M_BYTE];
+    unsigned char digest_salt[_HASH_LEN + _SALT_BYTE];
+    memcpy(digest_salt, digest, _HASH_LEN);
+    memcpy(digest_salt + _HASH_LEN, signature + _PUB_N_BYTE, _SALT_BYTE);
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_hash_msg(correct, _PUB_M_BYTE, digest_salt, _HASH_LEN + _SALT_BYTE); // H( digest || salt )
+
+    // check consistancy.
+    unsigned char cc = 0;
+    for (unsigned int i = 0; i < _PUB_M_BYTE; i++) {
+        cc |= (digest_ck[i] ^ correct[i]);
+    }
+    return (0 == cc) ? 0 : -1;
+}
+
+

crypto_sign/rainbowIa-classic/clean/rainbow.h

@@ -0,0 +1,33 @@
+#ifndef _RAINBOW_H_
+#define _RAINBOW_H_
+/// @file rainbow.h
+/// @brief APIs for rainbow.
+///
+
+#include "rainbow_config.h"
+#include "rainbow_keypair.h"
+
+#include <stdint.h>
+
+///
+/// @brief Signing function for classical secret key.
+///
+/// @param[out] signature - the signature.
+/// @param[in]  sk        - the secret key.
+/// @param[in]  digest    - the digest.
+///
+int PQCLEAN_RAINBOWIACLASSIC_CLEAN_rainbow_sign(uint8_t *signature, const sk_t *sk, const uint8_t *digest);
+
+///
+/// @brief Verifying function.
+///
+/// @param[in]  digest    - the digest.
+/// @param[in]  signature - the signature.
+/// @param[in]  pk        - the public key.
+/// @return 0 for successful verified. -1 for failed verification.
+///
+int PQCLEAN_RAINBOWIACLASSIC_CLEAN_rainbow_verify(const uint8_t *digest, const uint8_t *signature, const pk_t *pk);
+
+
+
+#endif // _RAINBOW_H_

crypto_sign/rainbowIa-classic/clean/rainbow_blas.h

@@ -0,0 +1,31 @@
+#ifndef _RAINBOW_BLAS_H_
+#define _RAINBOW_BLAS_H_
+/// @file rainbow_blas.h
+/// @brief Defining the functions used in rainbow.c acconding to the definitions in rainbow_config.h
+///
+///  Defining the functions used in rainbow.c acconding to the definitions in rainbow_config.h
+
+#include "blas.h"
+#include "parallel_matrix_op.h"
+#include "rainbow_config.h"
+
+
+#define gfv_get_ele PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_get_ele
+#define gfv_mul_scalar PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_mul_scalar
+#define gfv_madd PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16v_madd
+
+#define gfmat_prod PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_prod
+#define gfmat_inv PQCLEAN_RAINBOWIACLASSIC_CLEAN_gf16mat_inv
+
+#define batch_trimat_madd PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_trimat_madd_gf16
+#define batch_trimatTr_madd PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_trimatTr_madd_gf16
+#define batch_2trimat_madd PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_2trimat_madd_gf16
+#define batch_matTr_madd PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_matTr_madd_gf16
+#define batch_bmatTr_madd PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_bmatTr_madd_gf16
+#define batch_mat_madd PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_mat_madd_gf16
+
+#define batch_quad_trimat_eval PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_quad_trimat_eval_gf16
+#define batch_quad_recmat_eval PQCLEAN_RAINBOWIACLASSIC_CLEAN_batch_quad_recmat_eval_gf16
+
+
+#endif // _RAINBOW_BLAS_H_

crypto_sign/rainbowIa-classic/clean/rainbow_config.h

@@ -0,0 +1,47 @@
+#ifndef _H_RAINBOW_CONFIG_H_
+#define _H_RAINBOW_CONFIG_H_
+
+/// @file rainbow_config.h
+/// @brief Defining the parameters of the Rainbow and the corresponding constants.
+///
+
+#define _USE_GF16
+#define _GFSIZE 16
+#define _V1 32
+#define _O1 32
+#define _O2 32
+#define _HASH_LEN 32
+
+
+#define _V2 ((_V1) + (_O1))
+
+/// size of N, in # of gf elements.
+#define _PUB_N (_V1 + _O1 + _O2)
+
+/// size of M, in # gf elements.
+#define _PUB_M (_O1 + _O2)
+
+/// size of variables, in # bytes.
+
+// GF16
+#define _V1_BYTE (_V1 / 2)
+#define _V2_BYTE (_V2 / 2)
+#define _O1_BYTE (_O1 / 2)
+#define _O2_BYTE (_O2 / 2)
+#define _PUB_N_BYTE (_PUB_N / 2)
+#define _PUB_M_BYTE (_PUB_M / 2)
+
+
+/// length of seed for public key, in # bytes
+#define LEN_PKSEED 32
+
+/// length of seed for secret key, in # bytes
+#define LEN_SKSEED 32
+
+/// length of salt for a signature, in # bytes
+#define _SALT_BYTE 16
+
+/// length of a signature
+#define _SIGNATURE_BYTE (_PUB_N_BYTE + _SALT_BYTE)
+
+#endif //  _H_RAINBOW_CONFIG_H_

crypto_sign/rainbowIa-classic/clean/rainbow_keypair.c

@@ -0,0 +1,126 @@
+/// @file rainbow_keypair.c
+/// @brief implementations of functions in rainbow_keypair.h
+///
+
+#include "rainbow_keypair.h"
+#include "blas.h"
+#include "blas_comm.h"
+#include "rainbow_blas.h"
+#include "rainbow_keypair_computation.h"
+#include "utils_prng.h"
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+static void generate_S_T(unsigned char *s_and_t, prng_t *prng0) {
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, s_and_t, _O1_BYTE * _O2); // S1
+    s_and_t += _O1_BYTE * _O2;
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, s_and_t, _V1_BYTE * _O1); // T1
+    s_and_t += _V1_BYTE * _O1;
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, s_and_t, _V1_BYTE * _O2); // T2
+    s_and_t += _V1_BYTE * _O2;
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, s_and_t, _O1_BYTE * _O2); // T3
+}
+
+static unsigned int generate_l1_F12(unsigned char *sk, prng_t *prng0) {
+    unsigned int n_byte_generated = 0;
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, sk, _O1_BYTE * N_TRIANGLE_TERMS(_V1)); // l1_F1
+    sk += _O1_BYTE * N_TRIANGLE_TERMS(_V1);
+    n_byte_generated += _O1_BYTE * N_TRIANGLE_TERMS(_V1);
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, sk, _O1_BYTE * _V1 * _O1); // l1_F2
+    n_byte_generated += _O1_BYTE * _V1 * _O1;
+    return n_byte_generated;
+}
+
+static unsigned int generate_l2_F12356(unsigned char *sk, prng_t *prng0) {
+    unsigned int n_byte_generated = 0;
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * N_TRIANGLE_TERMS(_V1)); // l2_F1
+    sk += _O2_BYTE * N_TRIANGLE_TERMS(_V1);
+    n_byte_generated += _O2_BYTE * N_TRIANGLE_TERMS(_V1);
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _V1 * _O1); // l2_F2
+    sk += _O2_BYTE * _V1 * _O1;
+    n_byte_generated += _O2_BYTE * _V1 * _O1;
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _V1 * _O2); // l2_F3
+    sk += _O2_BYTE * _V1 * _O1;
+    n_byte_generated += _O2_BYTE * _V1 * _O1;
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * N_TRIANGLE_TERMS(_O1)); // l2_F5
+    sk += _O2_BYTE * N_TRIANGLE_TERMS(_O1);
+    n_byte_generated += _O2_BYTE * N_TRIANGLE_TERMS(_O1);
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_gen(prng0, sk, _O2_BYTE * _O1 * _O2); // l2_F6
+    n_byte_generated += _O2_BYTE * _O1 * _O2;
+
+    return n_byte_generated;
+}
+
+static void generate_B1_B2(unsigned char *sk, prng_t *prng0) {
+    sk += generate_l1_F12(sk, prng0);
+    generate_l2_F12356(sk, prng0);
+}
+
+static void calculate_t4(unsigned char *t2_to_t4, const unsigned char *t1, const unsigned char *t3) {
+    //  t4 = T_sk.t1 * T_sk.t3 - T_sk.t2
+    unsigned char temp[_V1_BYTE + 32];
+    unsigned char *t4 = t2_to_t4;
+    for (unsigned int i = 0; i < _O2; i++) { /// t3 width
+        gfmat_prod(temp, t1, _V1_BYTE, _O1, t3);
+        gf256v_add(t4, temp, _V1_BYTE);
+        t4 += _V1_BYTE;
+        t3 += _O1_BYTE;
+    }
+}
+
+static void obsfucate_l1_polys(unsigned char *l1_polys, const unsigned char *l2_polys, unsigned int n_terms, const unsigned char *s1) {
+    unsigned char temp[_O1_BYTE + 32];
+    while (n_terms--) {
+        gfmat_prod(temp, s1, _O1_BYTE, _O2, l2_polys);
+        gf256v_add(l1_polys, temp, _O1_BYTE);
+        l1_polys += _O1_BYTE;
+        l2_polys += _O2_BYTE;
+    }
+}
+
+///////////////////  Classic //////////////////////////////////
+
+static void _generate_secretkey(sk_t *sk, const unsigned char *sk_seed) {
+    memcpy(sk->sk_seed, sk_seed, LEN_SKSEED);
+
+    // set up prng
+    prng_t prng0;
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_prng_set(&prng0, sk_seed, LEN_SKSEED);
+
+    // generating secret key with prng.
+    generate_S_T(sk->s1, &prng0);
+    generate_B1_B2(sk->l1_F1, &prng0);
+
+    // clean prng
+    memset(&prng0, 0, sizeof(prng_t));
+}
+
+void PQCLEAN_RAINBOWIACLASSIC_CLEAN_generate_keypair(pk_t *rpk, sk_t *sk, const unsigned char *sk_seed) {
+    _generate_secretkey(sk, sk_seed);
+
+    // set up a temporary structure ext_cpk_t for calculating public key.
+    ext_cpk_t pk;
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_calculate_Q_from_F(&pk, sk, sk); // compute the public key in ext_cpk_t format.
+    calculate_t4(sk->t4, sk->t1, sk->t3);
+
+    obsfucate_l1_polys(pk.l1_Q1, pk.l2_Q1, N_TRIANGLE_TERMS(_V1), sk->s1);
+    obsfucate_l1_polys(pk.l1_Q2, pk.l2_Q2, _V1 * _O1, sk->s1);
+    obsfucate_l1_polys(pk.l1_Q3, pk.l2_Q3, _V1 * _O2, sk->s1);
+    obsfucate_l1_polys(pk.l1_Q5, pk.l2_Q5, N_TRIANGLE_TERMS(_O1), sk->s1);
+    obsfucate_l1_polys(pk.l1_Q6, pk.l2_Q6, _O1 * _O2, sk->s1);
+    obsfucate_l1_polys(pk.l1_Q9, pk.l2_Q9, N_TRIANGLE_TERMS(_O2), sk->s1);
+    // so far, the pk contains the full pk but in ext_cpk_t format.
+
+    PQCLEAN_RAINBOWIACLASSIC_CLEAN_extcpk_to_pk(rpk, &pk); // convert the public key from ext_cpk_t to pk_t.
+}
+
+
+