rhys: gimli optimized for avr

Co-authored-by: Rhys Weatherley <rhys.weatherley@gmail.com>

gimli/Implementations/crypto_aead/gimli24v1/rhys-avr/aead-common.c

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "aead-common.h"
+
+int aead_check_tag
+    (unsigned char *plaintext, unsigned long long plaintext_len,
+     const unsigned char *tag1, const unsigned char *tag2,
+     unsigned size)
+{
+    /* Set "accum" to -1 if the tags match, or 0 if they don't match */
+    int accum = 0;
+    while (size > 0) {
+        accum |= (*tag1++ ^ *tag2++);
+        --size;
+    }
+    accum = (accum - 1) >> 8;
+
+    /* Destroy the plaintext if the tag match failed */
+    while (plaintext_len > 0) {
+        *plaintext++ &= accum;
+        --plaintext_len;
+    }
+
+    /* If "accum" is 0, return -1, otherwise return 0 */
+    return ~accum;
+}
+
+int aead_check_tag_precheck
+    (unsigned char *plaintext, unsigned long long plaintext_len,
+     const unsigned char *tag1, const unsigned char *tag2,
+     unsigned size, int precheck)
+{
+    /* Set "accum" to -1 if the tags match, or 0 if they don't match */
+    int accum = 0;
+    while (size > 0) {
+        accum |= (*tag1++ ^ *tag2++);
+        --size;
+    }
+    accum = ((accum - 1) >> 8) & precheck;
+
+    /* Destroy the plaintext if the tag match failed */
+    while (plaintext_len > 0) {
+        *plaintext++ &= accum;
+        --plaintext_len;
+    }
+
+    /* If "accum" is 0, return -1, otherwise return 0 */
+    return ~accum;
+}

gimli/Implementations/crypto_aead/gimli24v1/rhys-avr/aead-common.h

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef LWCRYPTO_AEAD_COMMON_H
+#define LWCRYPTO_AEAD_COMMON_H
+
+#include <stddef.h>
+
+/**
+ * \file aead-common.h
+ * \brief Definitions that are common across AEAD schemes.
+ *
+ * AEAD stands for "Authenticated Encryption with Associated Data".
+ * It is a standard API pattern for securely encrypting and
+ * authenticating packets of data.
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * \brief Encrypts and authenticates a packet with an AEAD scheme.
+ *
+ * \param c Buffer to receive the output.
+ * \param clen On exit, set to the length of the output which includes
+ * the ciphertext and the authentication tag.
+ * \param m Buffer that contains the plaintext message to encrypt.
+ * \param mlen Length of the plaintext message in bytes.
+ * \param ad Buffer that contains associated data to authenticate
+ * along with the packet but which does not need to be encrypted.
+ * \param adlen Length of the associated data in bytes.
+ * \param nsec Secret nonce - normally not used by AEAD schemes.
+ * \param npub Points to the public nonce for the packet.
+ * \param k Points to the key to use to encrypt the packet.
+ *
+ * \return 0 on success, or a negative value if there was an error in
+ * the parameters.
+ */
+typedef int (*aead_cipher_encrypt_t)
+    (unsigned char *c, unsigned long long *clen,
+     const unsigned char *m, unsigned long long mlen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *nsec,
+     const unsigned char *npub,
+     const unsigned char *k);
+
+/**
+ * \brief Decrypts and authenticates a packet with an AEAD scheme.
+ *
+ * \param m Buffer to receive the plaintext message on output.
+ * \param mlen Receives the length of the plaintext message on output.
+ * \param nsec Secret nonce - normally not used by AEAD schemes.
+ * \param c Buffer that contains the ciphertext and authentication
+ * tag to decrypt.
+ * \param clen Length of the input data in bytes, which includes the
+ * ciphertext and the authentication tag.
+ * \param ad Buffer that contains associated data to authenticate
+ * along with the packet but which does not need to be encrypted.
+ * \param adlen Length of the associated data in bytes.
+ * \param npub Points to the public nonce for the packet.
+ * \param k Points to the key to use to decrypt the packet.
+ *
+ * \return 0 on success, -1 if the authentication tag was incorrect,
+ * or some other negative number if there was an error in the parameters.
+ */
+typedef int (*aead_cipher_decrypt_t)
+    (unsigned char *m, unsigned long long *mlen,
+     unsigned char *nsec,
+     const unsigned char *c, unsigned long long clen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *npub,
+     const unsigned char *k);
+
+/**
+ * \brief Hashes a block of input data.
+ *
+ * \param out Buffer to receive the hash output.
+ * \param in Points to the input data to be hashed.
+ * \param inlen Length of the input data in bytes.
+ *
+ * \return Returns zero on success or -1 if there was an error in the
+ * parameters.
+ */
+typedef int (*aead_hash_t)
+    (unsigned char *out, const unsigned char *in, unsigned long long inlen);
+
+/**
+ * \brief Initializes the state for a hashing operation.
+ *
+ * \param state Hash state to be initialized.
+ */
+typedef void (*aead_hash_init_t)(void *state);
+
+/**
+ * \brief Updates a hash state with more input data.
+ *
+ * \param state Hash state to be updated.
+ * \param in Points to the input data to be incorporated into the state.
+ * \param inlen Length of the input data to be incorporated into the state.
+ */
+typedef void (*aead_hash_update_t)
+    (void *state, const unsigned char *in, unsigned long long inlen);
+
+/**
+ * \brief Returns the final hash value from a hashing operation.
+ *
+ * \param Hash state to be finalized.
+ * \param out Points to the output buffer to receive the hash value.
+ */
+typedef void (*aead_hash_finalize_t)(void *state, unsigned char *out);
+
+/**
+ * \brief Aborbs more input data into an XOF state.
+ *
+ * \param state XOF state to be updated.
+ * \param in Points to the input data to be absorbed into the state.
+ * \param inlen Length of the input data to be absorbed into the state.
+ *
+ * \sa ascon_xof_init(), ascon_xof_squeeze()
+ */
+typedef void (*aead_xof_absorb_t)
+    (void *state, const unsigned char *in, unsigned long long inlen);
+
+/**
+ * \brief Squeezes output data from an XOF state.
+ *
+ * \param state XOF state to squeeze the output data from.
+ * \param out Points to the output buffer to receive the squeezed data.
+ * \param outlen Number of bytes of data to squeeze out of the state.
+ */
+typedef void (*aead_xof_squeeze_t)
+    (void *state, unsigned char *out, unsigned long long outlen);
+
+/**
+ * \brief No special AEAD features.
+ */
+#define AEAD_FLAG_NONE          0x0000
+
+/**
+ * \brief The natural byte order of the AEAD cipher is little-endian.
+ *
+ * If this flag is not present, then the natural byte order of the
+ * AEAD cipher should be assumed to be big-endian.
+ *
+ * The natural byte order may be useful when formatting packet sequence
+ * numbers as nonces.  The application needs to know whether the sequence
+ * number should be packed into the leading or trailing bytes of the nonce.
+ */
+#define AEAD_FLAG_LITTLE_ENDIAN 0x0001
+
+/**
+ * \brief Meta-information about an AEAD cipher.
+ */
+typedef struct
+{
+    const char *name;               /**< Name of the cipher */
+    unsigned key_len;               /**< Length of the key in bytes */
+    unsigned nonce_len;             /**< Length of the nonce in bytes */
+    unsigned tag_len;               /**< Length of the tag in bytes */
+    unsigned flags;                 /**< Flags for extra features */
+    aead_cipher_encrypt_t encrypt;  /**< AEAD encryption function */
+    aead_cipher_decrypt_t decrypt;  /**< AEAD decryption function */
+
+} aead_cipher_t;
+
+/**
+ * \brief Meta-information about a hash algorithm that is related to an AEAD.
+ *
+ * Regular hash algorithms should provide the "hash", "init", "update",
+ * and "finalize" functions.  Extensible Output Functions (XOF's) should
+ * proivde the "hash", "init", "absorb", and "squeeze" functions.
+ */
+typedef struct
+{
+    const char *name;           /**< Name of the hash algorithm */
+    size_t state_size;          /**< Size of the incremental state structure */
+    unsigned hash_len;          /**< Length of the hash in bytes */
+    unsigned flags;             /**< Flags for extra features */
+    aead_hash_t hash;           /**< All in one hashing function */
+    aead_hash_init_t init;      /**< Incremental hash/XOF init function */
+    aead_hash_update_t update;  /**< Incremental hash update function */
+    aead_hash_finalize_t finalize; /**< Incremental hash finalize function */
+    aead_xof_absorb_t absorb;   /**< Incremental XOF absorb function */
+    aead_xof_squeeze_t squeeze; /**< Incremental XOF squeeze function */
+
+} aead_hash_algorithm_t;
+
+/**
+ * \brief Check an authentication tag in constant time.
+ *
+ * \param plaintext Points to the plaintext data.
+ * \param plaintext_len Length of the plaintext in bytes.
+ * \param tag1 First tag to compare.
+ * \param tag2 Second tag to compare.
+ * \param tag_len Length of the tags in bytes.
+ *
+ * \return Returns -1 if the tag check failed or 0 if the check succeeded.
+ *
+ * If the tag check fails, then the \a plaintext will also be zeroed to
+ * prevent it from being used accidentally by the application when the
+ * ciphertext was invalid.
+ */
+int aead_check_tag
+    (unsigned char *plaintext, unsigned long long plaintext_len,
+     const unsigned char *tag1, const unsigned char *tag2,
+     unsigned tag_len);
+
+/**
+ * \brief Check an authentication tag in constant time with a previous check.
+ *
+ * \param plaintext Points to the plaintext data.
+ * \param plaintext_len Length of the plaintext in bytes.
+ * \param tag1 First tag to compare.
+ * \param tag2 Second tag to compare.
+ * \param tag_len Length of the tags in bytes.
+ * \param precheck Set to -1 if previous check succeeded or 0 if it failed.
+ *
+ * \return Returns -1 if the tag check failed or 0 if the check succeeded.
+ *
+ * If the tag check fails, then the \a plaintext will also be zeroed to
+ * prevent it from being used accidentally by the application when the
+ * ciphertext was invalid.
+ *
+ * This version can be used to incorporate other information about the
+ * correctness of the plaintext into the final result.
+ */
+int aead_check_tag_precheck
+    (unsigned char *plaintext, unsigned long long plaintext_len,
+     const unsigned char *tag1, const unsigned char *tag2,
+     unsigned tag_len, int precheck);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

gimli/Implementations/crypto_aead/gimli24v1/rhys-avr/api.h

+#define CRYPTO_KEYBYTES 32
+#define CRYPTO_NSECBYTES 0 
+#define CRYPTO_NPUBBYTES 16 
+#define CRYPTO_ABYTES 16 
+#define CRYPTO_NOOVERLAP 1 

gimli/Implementations/crypto_aead/gimli24v1/rhys-avr/encrypt.c

+
+#include "gimli24.h"
+
+int crypto_aead_encrypt
+    (unsigned char *c, unsigned long long *clen,
+     const unsigned char *m, unsigned long long mlen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *nsec,
+     const unsigned char *npub,
+     const unsigned char *k)
+{
+    return gimli24_aead_encrypt
+        (c, clen, m, mlen, ad, adlen, nsec, npub, k);
+}
+
+int crypto_aead_decrypt
+    (unsigned char *m, unsigned long long *mlen,
+     unsigned char *nsec,
+     const unsigned char *c, unsigned long long clen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *npub,
+     const unsigned char *k)
+{
+    return gimli24_aead_decrypt
+        (m, mlen, nsec, c, clen, ad, adlen, npub, k);
+}

gimli/Implementations/crypto_aead/gimli24v1/rhys-avr/gimli24.c

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "gimli24.h"
+#include "internal-gimli24.h"
+#include <string.h>
+
+aead_cipher_t const gimli24_cipher = {
+    "GIMLI-24",
+    GIMLI24_KEY_SIZE,
+    GIMLI24_NONCE_SIZE,
+    GIMLI24_TAG_SIZE,
+    AEAD_FLAG_LITTLE_ENDIAN,
+    gimli24_aead_encrypt,
+    gimli24_aead_decrypt
+};
+
+aead_hash_algorithm_t const gimli24_hash_algorithm = {
+    "GIMLI-24-HASH",
+    sizeof(gimli24_hash_state_t),
+    GIMLI24_HASH_SIZE,
+    AEAD_FLAG_LITTLE_ENDIAN,
+    gimli24_hash,
+    (aead_hash_init_t)gimli24_hash_init,
+    (aead_hash_update_t)gimli24_hash_absorb,
+    (aead_hash_finalize_t)gimli24_hash_finalize,
+    (aead_xof_absorb_t)gimli24_hash_absorb,
+    (aead_xof_squeeze_t)gimli24_hash_squeeze
+};
+
+/**
+ * \brief Number of bytes of input or output data to process per block.
+ */
+#define GIMLI24_BLOCK_SIZE 16
+
+/**
+ * \brief Structure of the GIMLI-24 state as both an array of words
+ * and an array of bytes.
+ */
+typedef union
+{
+    uint32_t words[12];     /**< Words in the state */
+    uint8_t bytes[48];      /**< Bytes in the state */
+
+} gimli24_state_t;
+
+/**
+ * \brief Absorbs data into a GIMLI-24 state.
+ *
+ * \param state The state to absorb the data into.
+ * \param data Points to the data to be absorbed.
+ * \param len Length of the data to be absorbed.
+ */
+static void gimli24_absorb
+    (gimli24_state_t *state, const unsigned char *data, unsigned long long len)
+{
+    unsigned temp;
+    while (len >= GIMLI24_BLOCK_SIZE) {
+        lw_xor_block(state->bytes, data, GIMLI24_BLOCK_SIZE);
+        gimli24_permute(state->words);
+        data += GIMLI24_BLOCK_SIZE;
+        len -= GIMLI24_BLOCK_SIZE;
+    }
+    temp = (unsigned)len;
+    lw_xor_block(state->bytes, data, temp);
+    state->bytes[temp] ^= 0x01; /* Padding */
+    state->bytes[47] ^= 0x01;
+    gimli24_permute(state->words);
+}
+
+/**
+ * \brief Encrypts a block of data with a GIMLI-24 state.
+ *
+ * \param state The state to encrypt with.
+ * \param dest Points to the destination buffer.
+ * \param src Points to the source buffer.
+ * \param len Length of the data to encrypt from \a src into \a dest.
+ */
+static void gimli24_encrypt
+    (gimli24_state_t *state, unsigned char *dest,
+     const unsigned char *src, unsigned long long len)
+{
+    unsigned temp;
+    while (len >= GIMLI24_BLOCK_SIZE) {
+        lw_xor_block_2_dest(dest, state->bytes, src, GIMLI24_BLOCK_SIZE);
+        gimli24_permute(state->words);
+        dest += GIMLI24_BLOCK_SIZE;
+        src += GIMLI24_BLOCK_SIZE;
+        len -= GIMLI24_BLOCK_SIZE;
+    }
+    temp = (unsigned)len;
+    lw_xor_block_2_dest(dest, state->bytes, src, temp);
+    state->bytes[temp] ^= 0x01; /* Padding */
+    state->bytes[47] ^= 0x01;
+    gimli24_permute(state->words);
+}
+
+/**
+ * \brief Decrypts a block of data with a GIMLI-24 state.
+ *
+ * \param state The state to decrypt with.
+ * \param dest Points to the destination buffer.
+ * \param src Points to the source buffer.
+ * \param len Length of the data to decrypt from \a src into \a dest.
+ */
+static void gimli24_decrypt
+    (gimli24_state_t *state, unsigned char *dest,
+     const unsigned char *src, unsigned long long len)
+{
+    unsigned temp;
+    while (len >= GIMLI24_BLOCK_SIZE) {
+        lw_xor_block_swap(dest, state->bytes, src, GIMLI24_BLOCK_SIZE);
+        gimli24_permute(state->words);
+        dest += GIMLI24_BLOCK_SIZE;
+        src += GIMLI24_BLOCK_SIZE;
+        len -= GIMLI24_BLOCK_SIZE;
+    }
+    temp = (unsigned)len;
+    lw_xor_block_swap(dest, state->bytes, src, temp);
+    state->bytes[temp] ^= 0x01; /* Padding */
+    state->bytes[47] ^= 0x01;
+    gimli24_permute(state->words);
+}
+
+int gimli24_aead_encrypt
+    (unsigned char *c, unsigned long long *clen,
+     const unsigned char *m, unsigned long long mlen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *nsec,
+     const unsigned char *npub,
+     const unsigned char *k)
+{
+    gimli24_state_t state;
+    (void)nsec;
+
+    /* Set the length of the returned ciphertext */
+    *clen = mlen + GIMLI24_TAG_SIZE;
+
+    /* Format the initial GIMLI state from the nonce and the key */
+    memcpy(state.words, npub, GIMLI24_NONCE_SIZE);
+    memcpy(state.words + 4, k, GIMLI24_KEY_SIZE);
+
+    /* Permute the initial state */
+    gimli24_permute(state.words);
+
+    /* Absorb the associated data */
+    gimli24_absorb(&state, ad, adlen);
+
+    /* Encrypt the plaintext to produce the ciphertext */
+    gimli24_encrypt(&state, c, m, mlen);
+
+    /* Generate the authentication tag at the end of the ciphertext */
+    memcpy(c + mlen, state.bytes, GIMLI24_TAG_SIZE);
+    return 0;
+}
+
+int gimli24_aead_decrypt
+    (unsigned char *m, unsigned long long *mlen,
+     unsigned char *nsec,
+     const unsigned char *c, unsigned long long clen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *npub,
+     const unsigned char *k)
+{
+    gimli24_state_t state;
+    (void)nsec;
+
+    /* Validate the ciphertext length and set the return "mlen" value */
+    if (clen < GIMLI24_TAG_SIZE)
+        return -1;
+    *mlen = clen - GIMLI24_TAG_SIZE;
+
+    /* Format the initial GIMLI state from the nonce and the key */
+    memcpy(state.words, npub, GIMLI24_NONCE_SIZE);
+    memcpy(state.words + 4, k, GIMLI24_KEY_SIZE);
+
+    /* Permute the initial state */
+    gimli24_permute(state.words);
+
+    /* Absorb the associated data */
+    gimli24_absorb(&state, ad, adlen);
+
+    /* Decrypt the ciphertext to produce the plaintext */
+    gimli24_decrypt(&state, m, c, *mlen);
+
+    /* Check the authentication tag at the end of the packet */
+    return aead_check_tag
+        (m, *mlen, state.bytes, c + *mlen, GIMLI24_TAG_SIZE);
+}
+
+int gimli24_hash
+    (unsigned char *out, const unsigned char *in, unsigned long long inlen)
+{
+    gimli24_state_t state;
+
+    /* Initialize the hash state to all zeroes */
+    memset(&state, 0, sizeof(state));
+
+    /* Absorb the input */
+    gimli24_absorb(&state, in, inlen);
+
+    /* Generate the output hash */
+    memcpy(out, state.bytes, GIMLI24_HASH_SIZE / 2);
+    gimli24_permute(state.words);
+    memcpy(out + GIMLI24_HASH_SIZE / 2, state.bytes, GIMLI24_HASH_SIZE / 2);
+    return 0;
+}
+
+void gimli24_hash_init(gimli24_hash_state_t *state)
+{
+    memset(state, 0, sizeof(gimli24_hash_state_t));
+}
+
+#define GIMLI24_XOF_RATE 16
+#define gimli24_xof_permute() \
+    gimli24_permute((uint32_t *)(state->s.state))
+
+void gimli24_hash_absorb
+    (gimli24_hash_state_t *state, const unsigned char *in,
+     unsigned long long inlen)
+{
+    unsigned temp;
+
+    if (state->s.mode) {
+        /* We were squeezing output - go back to the absorb phase */
+        state->s.mode = 0;
+        state->s.count = 0;
+        gimli24_xof_permute();
+    }
+
+    /* Handle the partial left-over block from last time */
+    if (state->s.count) {
+        temp = GIMLI24_XOF_RATE - state->s.count;
+        if (temp > inlen) {
+            temp = (unsigned)inlen;
+            lw_xor_block(state->s.state + state->s.count, in, temp);
+            state->s.count += temp;
+            return;
+        }
+        lw_xor_block(state->s.state + state->s.count, in, temp);
+        state->s.count = 0;
+        in += temp;
+        inlen -= temp;
+        gimli24_xof_permute();
+    }
+
+    /* Process full blocks that are aligned at state->s.count == 0 */
+    while (inlen >= GIMLI24_XOF_RATE) {
+        lw_xor_block(state->s.state, in, GIMLI24_XOF_RATE);
+        in += GIMLI24_XOF_RATE;
+        inlen -= GIMLI24_XOF_RATE;
+        gimli24_xof_permute();
+    }
+
+    /* Process the left-over block at the end of the input */
+    temp = (unsigned)inlen;
+    lw_xor_block(state->s.state, in, temp);
+    state->s.count = temp;
+}
+
+void gimli24_hash_squeeze
+    (gimli24_hash_state_t *state, unsigned char *out,
+     unsigned long long outlen)
+{
+    unsigned temp;
+
+    /* Pad the final input block if we were still in the absorb phase */
+    if (!state->s.mode) {
+        state->s.state[state->s.count] ^= 0x01;
+        state->s.state[47] ^= 0x01;
+        state->s.count = 0;
+        state->s.mode = 1;
+    }
+
+    /* Handle left-over partial blocks from last time */
+    if (state->s.count) {
+        temp = GIMLI24_XOF_RATE - state->s.count;
+        if (temp > outlen) {
+            temp = (unsigned)outlen;
+            memcpy(out, state->s.state + state->s.count, temp);
+            state->s.count += temp;
+            return;
+        }
+        memcpy(out, state->s.state + state->s.count, temp);
+        out += temp;
+        outlen -= temp;
+        state->s.count = 0;
+    }
+
+    /* Handle full blocks */
+    while (outlen >= GIMLI24_XOF_RATE) {
+        gimli24_xof_permute();
+        memcpy(out, state->s.state, GIMLI24_XOF_RATE);
+        out += GIMLI24_XOF_RATE;
+        outlen -= GIMLI24_XOF_RATE;
+    }
+
+    /* Handle the left-over block */
+    if (outlen > 0) {
+        temp = (unsigned)outlen;
+        gimli24_xof_permute();
+        memcpy(out, state->s.state, temp);
+        state->s.count = temp;
+    }
+}
+
+void gimli24_hash_finalize
+    (gimli24_hash_state_t *state, unsigned char *out)
+{
+    gimli24_hash_squeeze(state, out, GIMLI24_HASH_SIZE);
+}

gimli/Implementations/crypto_aead/gimli24v1/rhys-avr/gimli24.h

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef LWCRYPTO_GIMLI24_H
+#define LWCRYPTO_GIMLI24_H
+
+#include "aead-common.h"
+
+/**
+ * \file gimli24.h
+ * \brief Gimli authenticated encryption algorithm.
+ *
+ * GIMLI-24-CIPHER has a 256-bit key, a 128-bit nonce, and a 128-bit tag.
+ * It is the spiritual successor to the widely used ChaCha20 and has a
+ * similar design.
+ *
+ * This library also includes an implementation of the hash algorithm
+ * GIMLI-24-HASH in both regular hashing and XOF modes.
+ *
+ * References: https://gimli.cr.yp.to/
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * \brief Size of the key for GIMLI-24.
+ */
+#define GIMLI24_KEY_SIZE 32
+
+/**
+ * \brief Size of the nonce for GIMLI-24.
+ */
+#define GIMLI24_NONCE_SIZE 16
+
+/**
+ * \brief Size of the authentication tag for GIMLI-24.
+ */
+#define GIMLI24_TAG_SIZE 16
+
+/**
+ * \brief Size of the hash output for GIMLI-24.
+ */
+#define GIMLI24_HASH_SIZE 32
+
+/**
+ * \brief State information for GIMLI-24-HASH incremental modes.
+ */
+typedef union
+{
+    struct {
+        unsigned char state[48]; /**< Current hash state */
+        unsigned char count;     /**< Number of bytes in the current block */
+        unsigned char mode;      /**< Hash mode: 0 for absorb, 1 for squeeze */
+    } s;                         /**< State */
+    unsigned long long align;    /**< For alignment of this structure */
+
+} gimli24_hash_state_t;
+
+/**
+ * \brief Meta-information block for the GIMLI-24 cipher.
+ */
+extern aead_cipher_t const gimli24_cipher;
+
+/**
+ * \brief Meta-information block for the GIMLI-24-HASH algorithm.
+ *
+ * This meta-information block can also be used in XOF mode.
+ */
+extern aead_hash_algorithm_t const gimli24_hash_algorithm;
+
+/**
+ * \brief Encrypts and authenticates a packet with GIMLI-24 using the
+ * full AEAD mode.
+ *
+ * \param c Buffer to receive the output.
+ * \param clen On exit, set to the length of the output which includes
+ * the ciphertext and the 16 byte authentication tag.
+ * \param m Buffer that contains the plaintext message to encrypt.
+ * \param mlen Length of the plaintext message in bytes.
+ * \param ad Buffer that contains associated data to authenticate
+ * along with the packet but which does not need to be encrypted.
+ * \param adlen Length of the associated data in bytes.
+ * \param nsec Secret nonce - not used by this algorithm.
+ * \param npub Points to the public nonce for the packet which must
+ * be 16 bytes in length.
+ * \param k Points to the 32 bytes of the key to use to encrypt the packet.
+ *
+ * \return 0 on success, or a negative value if there was an error in
+ * the parameters.
+ *
+ * \sa gimli24_aead_decrypt()
+ */
+int gimli24_aead_encrypt
+    (unsigned char *c, unsigned long long *clen,
+     const unsigned char *m, unsigned long long mlen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *nsec,
+     const unsigned char *npub,
+     const unsigned char *k);
+
+/**
+ * \brief Decrypts and authenticates a packet with GIMLI-24 using the
+ * full AEAD mode.
+ *
+ * \param m Buffer to receive the plaintext message on output.
+ * \param mlen Receives the length of the plaintext message on output.
+ * \param nsec Secret nonce - not used by this algorithm.
+ * \param c Buffer that contains the ciphertext and authentication
+ * tag to decrypt.
+ * \param clen Length of the input data in bytes, which includes the
+ * ciphertext and the 16 byte authentication tag.
+ * \param ad Buffer that contains associated data to authenticate
+ * along with the packet but which does not need to be encrypted.
+ * \param adlen Length of the associated data in bytes.
+ * \param npub Points to the public nonce for the packet which must
+ * be 16 bytes in length.
+ * \param k Points to the 32 bytes of the key to use to decrypt the packet.
+ *
+ * \return 0 on success, -1 if the authentication tag was incorrect,
+ * or some other negative number if there was an error in the parameters.
+ *
+ * \sa gimli24_aead_encrypt()
+ */
+int gimli24_aead_decrypt
+    (unsigned char *m, unsigned long long *mlen,
+     unsigned char *nsec,
+     const unsigned char *c, unsigned long long clen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *npub,
+     const unsigned char *k);
+
+/**
+ * \brief Hashes a block of input data with GIMLI-24 to generate a hash value.
+ *
+ * \param out Buffer to receive the hash output which must be at least
+ * GIMLI24_HASH_SIZE bytes in length.
+ * \param in Points to the input data to be hashed.
+ * \param inlen Length of the input data in bytes.
+ *
+ * \return Returns zero on success or -1 if there was an error in the
+ * parameters.
+ */
+int gimli24_hash
+    (unsigned char *out, const unsigned char *in, unsigned long long inlen);
+
+/**
+ * \brief Initializes the state for a GIMLI-24-HASH hashing operation.
+ *
+ * \param state Hash state to be initialized.
+ *
+ * \sa gimli24_hash_absorb(), gimli24_hash_squeeze(), gimli24_hash()
+ */
+void gimli24_hash_init(gimli24_hash_state_t *state);
+
+/**
+ * \brief Aborbs more input data into a GIMLI-24-HASH state.
+ *
+ * \param state Hash state to be updated.
+ * \param in Points to the input data to be absorbed into the state.
+ * \param inlen Length of the input data to be absorbed into the state.
+ *
+ * \sa gimli24_hash_init(), gimli24_hash_squeeze()
+ */
+void gimli24_hash_absorb
+    (gimli24_hash_state_t *state, const unsigned char *in,
+     unsigned long long inlen);
+
+/**
+ * \brief Squeezes output data from an GIMLI-24-HASH state.
+ *
+ * \param state Hash state to squeeze the output data from.
+ * \param out Points to the output buffer to receive the squeezed data.
+ * \param outlen Number of bytes of data to squeeze out of the state.
+ *
+ * \sa gimli24_hash_init(), gimli24_hash_absorb()
+ */
+void gimli24_hash_squeeze
+    (gimli24_hash_state_t *state, unsigned char *out,
+     unsigned long long outlen);
+
+/**
+ * \brief Returns the final hash value from a GIMLI-24-HASH hashing operation.
+ *
+ * \param state Hash state to be finalized.
+ * \param out Points to the output buffer to receive the hash value.
+ *
+ * \note This is a wrapper around gimli24_hash_squeeze() for a fixed length
+ * of GIMLI24_HASH_SIZE bytes.
+ *
+ * \sa gimli24_hash_init(), gimli24_hash_absorb()
+ */
+void gimli24_hash_finalize
+    (gimli24_hash_state_t *state, unsigned char *out);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

gimli/Implementations/crypto_aead/gimli24v1/rhys-avr/internal-gimli24.h

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef LW_INTERNAL_GIMLI24_H
+#define LW_INTERNAL_GIMLI24_H
+
+#include "internal-util.h"
+
+/**
+ * \file internal-gimli24.h
+ * \brief Internal implementation of the GIMLI-24 permutation.
+ *
+ * References: https://gimli.cr.yp.to/
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * \brief Permutes the GIMLI-24 state.
+ *
+ * \param state The GIMLI-24 state to be permuted.
+ *
+ * The input and output \a state will be in little-endian byte order.
+ */
+void gimli24_permute(uint32_t state[12]);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

gimli/Implementations/crypto_aead/gimli24v1/rhys-avr/internal-util.h

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef LW_INTERNAL_UTIL_H
+#define LW_INTERNAL_UTIL_H
+
+#include <stdint.h>
+
+/* Figure out how to inline functions using this C compiler */
+#if defined(__STDC__) && __STDC_VERSION__ >= 199901L
+#define STATIC_INLINE static inline
+#elif defined(__GNUC__) || defined(__clang__)
+#define STATIC_INLINE static __inline__
+#else
+#define STATIC_INLINE static
+#endif
+
+/* Try to figure out whether the CPU is little-endian or big-endian.
+ * May need to modify this to include new compiler-specific defines.
+ * Alternatively, define __LITTLE_ENDIAN__ or __BIG_ENDIAN__ in your
+ * compiler flags when you compile this library */
+#if defined(__x86_64) || defined(__x86_64__) || \
+    defined(__i386) || defined(__i386__) || \
+    defined(__AVR__) || defined(__arm) || defined(__arm__) || \
+    defined(_M_AMD64) || defined(_M_X64) || defined(_M_IX86) || \
+    defined(_M_IA64) || defined(_M_ARM) || defined(_M_ARM_FP) || \
+    (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == 1234) || \
+    defined(__LITTLE_ENDIAN__)
+#define LW_UTIL_LITTLE_ENDIAN 1
+#elif (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == 4321) || \
+    defined(__BIG_ENDIAN__)
+/* Big endian */
+#else
+#error "Cannot determine the endianess of this platform"
+#endif
+
+/* Helper macros to load and store values while converting endian-ness */
+
+/* Load a big-endian 32-bit word from a byte buffer */
+#define be_load_word32(ptr) \
+    ((((uint32_t)((ptr)[0])) << 24) | \
+     (((uint32_t)((ptr)[1])) << 16) | \
+     (((uint32_t)((ptr)[2])) << 8) | \
+      ((uint32_t)((ptr)[3])))
+
+/* Store a big-endian 32-bit word into a byte buffer */
+#define be_store_word32(ptr, x) \
+    do { \
+        uint32_t _x = (x); \
+        (ptr)[0] = (uint8_t)(_x >> 24); \
+        (ptr)[1] = (uint8_t)(_x >> 16); \
+        (ptr)[2] = (uint8_t)(_x >> 8); \
+        (ptr)[3] = (uint8_t)_x; \
+    } while (0)
+
+/* Load a little-endian 32-bit word from a byte buffer */
+#define le_load_word32(ptr) \
+    ((((uint32_t)((ptr)[3])) << 24) | \
+     (((uint32_t)((ptr)[2])) << 16) | \
+     (((uint32_t)((ptr)[1])) << 8) | \
+      ((uint32_t)((ptr)[0])))
+
+/* Store a little-endian 32-bit word into a byte buffer */
+#define le_store_word32(ptr, x) \
+    do { \
+        uint32_t _x = (x); \
+        (ptr)[0] = (uint8_t)_x; \
+        (ptr)[1] = (uint8_t)(_x >> 8); \
+        (ptr)[2] = (uint8_t)(_x >> 16); \
+        (ptr)[3] = (uint8_t)(_x >> 24); \
+    } while (0)
+
+/* Load a big-endian 64-bit word from a byte buffer */
+#define be_load_word64(ptr) \
+    ((((uint64_t)((ptr)[0])) << 56) | \
+     (((uint64_t)((ptr)[1])) << 48) | \
+     (((uint64_t)((ptr)[2])) << 40) | \
+     (((uint64_t)((ptr)[3])) << 32) | \
+     (((uint64_t)((ptr)[4])) << 24) | \
+     (((uint64_t)((ptr)[5])) << 16) | \
+     (((uint64_t)((ptr)[6])) << 8) | \
+      ((uint64_t)((ptr)[7])))
+
+/* Store a big-endian 64-bit word into a byte buffer */
+#define be_store_word64(ptr, x) \
+    do { \
+        uint64_t _x = (x); \
+        (ptr)[0] = (uint8_t)(_x >> 56); \
+        (ptr)[1] = (uint8_t)(_x >> 48); \
+        (ptr)[2] = (uint8_t)(_x >> 40); \
+        (ptr)[3] = (uint8_t)(_x >> 32); \
+        (ptr)[4] = (uint8_t)(_x >> 24); \
+        (ptr)[5] = (uint8_t)(_x >> 16); \
+        (ptr)[6] = (uint8_t)(_x >> 8); \
+        (ptr)[7] = (uint8_t)_x; \
+    } while (0)
+
+/* Load a little-endian 64-bit word from a byte buffer */
+#define le_load_word64(ptr) \
+    ((((uint64_t)((ptr)[7])) << 56) | \
+     (((uint64_t)((ptr)[6])) << 48) | \
+     (((uint64_t)((ptr)[5])) << 40) | \
+     (((uint64_t)((ptr)[4])) << 32) | \
+     (((uint64_t)((ptr)[3])) << 24) | \
+     (((uint64_t)((ptr)[2])) << 16) | \
+     (((uint64_t)((ptr)[1])) << 8) | \
+      ((uint64_t)((ptr)[0])))
+
+/* Store a little-endian 64-bit word into a byte buffer */
+#define le_store_word64(ptr, x) \
+    do { \
+        uint64_t _x = (x); \
+        (ptr)[0] = (uint8_t)_x; \
+        (ptr)[1] = (uint8_t)(_x >> 8); \
+        (ptr)[2] = (uint8_t)(_x >> 16); \
+        (ptr)[3] = (uint8_t)(_x >> 24); \
+        (ptr)[4] = (uint8_t)(_x >> 32); \
+        (ptr)[5] = (uint8_t)(_x >> 40); \
+        (ptr)[6] = (uint8_t)(_x >> 48); \
+        (ptr)[7] = (uint8_t)(_x >> 56); \
+    } while (0)
+
+/* Load a big-endian 16-bit word from a byte buffer */
+#define be_load_word16(ptr) \
+    ((((uint16_t)((ptr)[0])) << 8) | \
+      ((uint16_t)((ptr)[1])))
+
+/* Store a big-endian 16-bit word into a byte buffer */
+#define be_store_word16(ptr, x) \
+    do { \
+        uint16_t _x = (x); \
+        (ptr)[0] = (uint8_t)(_x >> 8); \
+        (ptr)[1] = (uint8_t)_x; \
+    } while (0)
+
+/* Load a little-endian 16-bit word from a byte buffer */
+#define le_load_word16(ptr) \
+    ((((uint16_t)((ptr)[1])) << 8) | \
+      ((uint16_t)((ptr)[0])))
+
+/* Store a little-endian 16-bit word into a byte buffer */
+#define le_store_word16(ptr, x) \
+    do { \
+        uint16_t _x = (x); \
+        (ptr)[0] = (uint8_t)_x; \
+        (ptr)[1] = (uint8_t)(_x >> 8); \
+    } while (0)
+
+/* XOR a source byte buffer against a destination */
+#define lw_xor_block(dest, src, len) \
+    do { \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src = (src); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            *_dest++ ^= *_src++; \
+            --_len; \
+        } \
+    } while (0)
+
+/* XOR two source byte buffers and put the result in a destination buffer */
+#define lw_xor_block_2_src(dest, src1, src2, len) \
+    do { \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src1 = (src1); \
+        const unsigned char *_src2 = (src2); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            *_dest++ = *_src1++ ^ *_src2++; \
+            --_len; \
+        } \
+    } while (0)
+
+/* XOR a source byte buffer against a destination and write to another
+ * destination at the same time */
+#define lw_xor_block_2_dest(dest2, dest, src, len) \
+    do { \
+        unsigned char *_dest2 = (dest2); \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src = (src); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            *_dest2++ = (*_dest++ ^= *_src++); \
+            --_len; \
+        } \
+    } while (0)
+
+/* XOR two byte buffers and write to a destination which at the same
+ * time copying the contents of src2 to dest2 */
+#define lw_xor_block_copy_src(dest2, dest, src1, src2, len) \
+    do { \
+        unsigned char *_dest2 = (dest2); \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src1 = (src1); \
+        const unsigned char *_src2 = (src2); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            unsigned char _temp = *_src2++; \
+            *_dest2++ = _temp; \
+            *_dest++ = *_src1++ ^ _temp; \
+            --_len; \
+        } \
+    } while (0)
+
+/* XOR a source byte buffer against a destination and write to another
+ * destination at the same time.  This version swaps the source value
+ * into the "dest" buffer */
+#define lw_xor_block_swap(dest2, dest, src, len) \
+    do { \
+        unsigned char *_dest2 = (dest2); \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src = (src); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            unsigned char _temp = *_src++; \
+            *_dest2++ = *_dest ^ _temp; \
+            *_dest++ = _temp; \
+            --_len; \
+        } \
+    } while (0)
+
+/* Rotation functions need to be optimised for best performance on AVR.
+ * The most efficient rotations are where the number of bits is 1 or a
+ * multiple of 8, so we compose the efficient rotations to produce all
+ * other rotation counts of interest. */
+
+#if defined(__AVR__)
+#define LW_CRYPTO_ROTATE32_COMPOSED 1
+#else
+#define LW_CRYPTO_ROTATE32_COMPOSED 0
+#endif
+
+/* Rotation macros for 32-bit arguments */
+
+/* Generic left rotate */
+#define leftRotate(a, bits) \
+    (__extension__ ({ \
+        uint32_t _temp = (a); \
+        (_temp << (bits)) | (_temp >> (32 - (bits))); \
+    }))
+
+/* Generic right rotate */
+#define rightRotate(a, bits) \
+    (__extension__ ({ \
+        uint32_t _temp = (a); \
+        (_temp >> (bits)) | (_temp << (32 - (bits))); \
+    }))
+
+#if !LW_CRYPTO_ROTATE32_COMPOSED
+
+/* Left rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define leftRotate1(a)  (leftRotate((a), 1))
+#define leftRotate2(a)  (leftRotate((a), 2))
+#define leftRotate3(a)  (leftRotate((a), 3))
+#define leftRotate4(a)  (leftRotate((a), 4))
+#define leftRotate5(a)  (leftRotate((a), 5))
+#define leftRotate6(a)  (leftRotate((a), 6))
+#define leftRotate7(a)  (leftRotate((a), 7))
+#define leftRotate8(a)  (leftRotate((a), 8))
+#define leftRotate9(a)  (leftRotate((a), 9))
+#define leftRotate10(a) (leftRotate((a), 10))
+#define leftRotate11(a) (leftRotate((a), 11))
+#define leftRotate12(a) (leftRotate((a), 12))
+#define leftRotate13(a) (leftRotate((a), 13))
+#define leftRotate14(a) (leftRotate((a), 14))
+#define leftRotate15(a) (leftRotate((a), 15))
+#define leftRotate16(a) (leftRotate((a), 16))
+#define leftRotate17(a) (leftRotate((a), 17))
+#define leftRotate18(a) (leftRotate((a), 18))
+#define leftRotate19(a) (leftRotate((a), 19))
+#define leftRotate20(a) (leftRotate((a), 20))
+#define leftRotate21(a) (leftRotate((a), 21))
+#define leftRotate22(a) (leftRotate((a), 22))
+#define leftRotate23(a) (leftRotate((a), 23))
+#define leftRotate24(a) (leftRotate((a), 24))
+#define leftRotate25(a) (leftRotate((a), 25))
+#define leftRotate26(a) (leftRotate((a), 26))
+#define leftRotate27(a) (leftRotate((a), 27))
+#define leftRotate28(a) (leftRotate((a), 28))
+#define leftRotate29(a) (leftRotate((a), 29))
+#define leftRotate30(a) (leftRotate((a), 30))
+#define leftRotate31(a) (leftRotate((a), 31))
+
+/* Right rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define rightRotate1(a)  (rightRotate((a), 1))
+#define rightRotate2(a)  (rightRotate((a), 2))
+#define rightRotate3(a)  (rightRotate((a), 3))
+#define rightRotate4(a)  (rightRotate((a), 4))
+#define rightRotate5(a)  (rightRotate((a), 5))
+#define rightRotate6(a)  (rightRotate((a), 6))
+#define rightRotate7(a)  (rightRotate((a), 7))
+#define rightRotate8(a)  (rightRotate((a), 8))
+#define rightRotate9(a)  (rightRotate((a), 9))
+#define rightRotate10(a) (rightRotate((a), 10))
+#define rightRotate11(a) (rightRotate((a), 11))
+#define rightRotate12(a) (rightRotate((a), 12))
+#define rightRotate13(a) (rightRotate((a), 13))
+#define rightRotate14(a) (rightRotate((a), 14))
+#define rightRotate15(a) (rightRotate((a), 15))
+#define rightRotate16(a) (rightRotate((a), 16))
+#define rightRotate17(a) (rightRotate((a), 17))
+#define rightRotate18(a) (rightRotate((a), 18))
+#define rightRotate19(a) (rightRotate((a), 19))
+#define rightRotate20(a) (rightRotate((a), 20))
+#define rightRotate21(a) (rightRotate((a), 21))
+#define rightRotate22(a) (rightRotate((a), 22))
+#define rightRotate23(a) (rightRotate((a), 23))
+#define rightRotate24(a) (rightRotate((a), 24))
+#define rightRotate25(a) (rightRotate((a), 25))
+#define rightRotate26(a) (rightRotate((a), 26))
+#define rightRotate27(a) (rightRotate((a), 27))
+#define rightRotate28(a) (rightRotate((a), 28))
+#define rightRotate29(a) (rightRotate((a), 29))
+#define rightRotate30(a) (rightRotate((a), 30))
+#define rightRotate31(a) (rightRotate((a), 31))
+
+#else /* LW_CRYPTO_ROTATE32_COMPOSED */
+
+/* Composed rotation macros where 1 and 8 are fast, but others are slow */
+
+/* Left rotate by 1 */
+#define leftRotate1(a)  (leftRotate((a), 1))
+
+/* Left rotate by 2 */
+#define leftRotate2(a)  (leftRotate(leftRotate((a), 1), 1))
+
+/* Left rotate by 3 */
+#define leftRotate3(a)  (leftRotate(leftRotate(leftRotate((a), 1), 1), 1))
+
+/* Left rotate by 4 */
+#define leftRotate4(a)  (leftRotate(leftRotate(leftRotate(leftRotate((a), 1), 1), 1), 1))
+
+/* Left rotate by 5: Rotate left by 8, then right by 3 */
+#define leftRotate5(a)  (rightRotate(rightRotate(rightRotate(leftRotate((a), 8), 1), 1), 1))
+
+/* Left rotate by 6: Rotate left by 8, then right by 2 */
+#define leftRotate6(a)  (rightRotate(rightRotate(leftRotate((a), 8), 1), 1))
+
+/* Left rotate by 7: Rotate left by 8, then right by 1 */
+#define leftRotate7(a)  (rightRotate(leftRotate((a), 8), 1))
+
+/* Left rotate by 8 */
+#define leftRotate8(a)  (leftRotate((a), 8))
+
+/* Left rotate by 9: Rotate left by 8, then left by 1 */
+#define leftRotate9(a)  (leftRotate(leftRotate((a), 8), 1))
+
+/* Left rotate by 10: Rotate left by 8, then left by 2 */
+#define leftRotate10(a) (leftRotate(leftRotate(leftRotate((a), 8), 1), 1))
+
+/* Left rotate by 11: Rotate left by 8, then left by 3 */
+#define leftRotate11(a) (leftRotate(leftRotate(leftRotate(leftRotate((a), 8), 1), 1), 1))
+
+/* Left rotate by 12: Rotate left by 16, then right by 4 */
+#define leftRotate12(a) (rightRotate(rightRotate(rightRotate(rightRotate(leftRotate((a), 16), 1), 1), 1), 1))
+
+/* Left rotate by 13: Rotate left by 16, then right by 3 */
+#define leftRotate13(a) (rightRotate(rightRotate(rightRotate(leftRotate((a), 16), 1), 1), 1))
+
+/* Left rotate by 14: Rotate left by 16, then right by 2 */
+#define leftRotate14(a) (rightRotate(rightRotate(leftRotate((a), 16), 1), 1))
+
+/* Left rotate by 15: Rotate left by 16, then right by 1 */
+#define leftRotate15(a) (rightRotate(leftRotate((a), 16), 1))
+
+/* Left rotate by 16 */
+#define leftRotate16(a) (leftRotate((a), 16))
+
+/* Left rotate by 17: Rotate left by 16, then left by 1 */
+#define leftRotate17(a) (leftRotate(leftRotate((a), 16), 1))
+
+/* Left rotate by 18: Rotate left by 16, then left by 2 */
+#define leftRotate18(a) (leftRotate(leftRotate(leftRotate((a), 16), 1), 1))
+
+/* Left rotate by 19: Rotate left by 16, then left by 3 */
+#define leftRotate19(a) (leftRotate(leftRotate(leftRotate(leftRotate((a), 16), 1), 1), 1))
+
+/* Left rotate by 20: Rotate left by 16, then left by 4 */
+#define leftRotate20(a) (leftRotate(leftRotate(leftRotate(leftRotate(leftRotate((a), 16), 1), 1), 1), 1))
+
+/* Left rotate by 21: Rotate left by 24, then right by 3 */
+#define leftRotate21(a) (rightRotate(rightRotate(rightRotate(leftRotate((a), 24), 1), 1), 1))
+
+/* Left rotate by 22: Rotate left by 24, then right by 2 */
+#define leftRotate22(a) (rightRotate(rightRotate(leftRotate((a), 24), 1), 1))
+
+/* Left rotate by 23: Rotate left by 24, then right by 1 */
+#define leftRotate23(a) (rightRotate(leftRotate((a), 24), 1))
+
+/* Left rotate by 24 */
+#define leftRotate24(a) (leftRotate((a), 24))
+
+/* Left rotate by 25: Rotate left by 24, then left by 1 */
+#define leftRotate25(a) (leftRotate(leftRotate((a), 24), 1))
+
+/* Left rotate by 26: Rotate left by 24, then left by 2 */
+#define leftRotate26(a) (leftRotate(leftRotate(leftRotate((a), 24), 1), 1))
+
+/* Left rotate by 27: Rotate left by 24, then left by 3 */
+#define leftRotate27(a) (leftRotate(leftRotate(leftRotate(leftRotate((a), 24), 1), 1), 1))
+
+/* Left rotate by 28: Rotate right by 4 */
+#define leftRotate28(a) (rightRotate(rightRotate(rightRotate(rightRotate((a), 1), 1), 1), 1))
+
+/* Left rotate by 29: Rotate right by 3 */
+#define leftRotate29(a) (rightRotate(rightRotate(rightRotate((a), 1), 1), 1))
+
+/* Left rotate by 30: Rotate right by 2 */
+#define leftRotate30(a) (rightRotate(rightRotate((a), 1), 1))
+
+/* Left rotate by 31: Rotate right by 1 */
+#define leftRotate31(a) (rightRotate((a), 1))
+
+/* Define the 32-bit right rotations in terms of left rotations */
+#define rightRotate1(a)  (leftRotate31((a)))
+#define rightRotate2(a)  (leftRotate30((a)))
+#define rightRotate3(a)  (leftRotate29((a)))
+#define rightRotate4(a)  (leftRotate28((a)))
+#define rightRotate5(a)  (leftRotate27((a)))
+#define rightRotate6(a)  (leftRotate26((a)))
+#define rightRotate7(a)  (leftRotate25((a)))
+#define rightRotate8(a)  (leftRotate24((a)))
+#define rightRotate9(a)  (leftRotate23((a)))
+#define rightRotate10(a) (leftRotate22((a)))
+#define rightRotate11(a) (leftRotate21((a)))
+#define rightRotate12(a) (leftRotate20((a)))
+#define rightRotate13(a) (leftRotate19((a)))
+#define rightRotate14(a) (leftRotate18((a)))
+#define rightRotate15(a) (leftRotate17((a)))
+#define rightRotate16(a) (leftRotate16((a)))
+#define rightRotate17(a) (leftRotate15((a)))
+#define rightRotate18(a) (leftRotate14((a)))
+#define rightRotate19(a) (leftRotate13((a)))
+#define rightRotate20(a) (leftRotate12((a)))
+#define rightRotate21(a) (leftRotate11((a)))
+#define rightRotate22(a) (leftRotate10((a)))
+#define rightRotate23(a) (leftRotate9((a)))
+#define rightRotate24(a) (leftRotate8((a)))
+#define rightRotate25(a) (leftRotate7((a)))
+#define rightRotate26(a) (leftRotate6((a)))
+#define rightRotate27(a) (leftRotate5((a)))
+#define rightRotate28(a) (leftRotate4((a)))
+#define rightRotate29(a) (leftRotate3((a)))
+#define rightRotate30(a) (leftRotate2((a)))
+#define rightRotate31(a) (leftRotate1((a)))
+
+#endif /* LW_CRYPTO_ROTATE32_COMPOSED */
+
+/* Rotation macros for 64-bit arguments */
+
+/* Generic left rotate */
+#define leftRotate_64(a, bits) \
+    (__extension__ ({ \
+        uint64_t _temp = (a); \
+        (_temp << (bits)) | (_temp >> (64 - (bits))); \
+    }))
+
+/* Generic right rotate */
+#define rightRotate_64(a, bits) \
+    (__extension__ ({ \
+        uint64_t _temp = (a); \
+        (_temp >> (bits)) | (_temp << (64 - (bits))); \
+    }))
+
+/* Left rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define leftRotate1_64(a)  (leftRotate_64((a), 1))
+#define leftRotate2_64(a)  (leftRotate_64((a), 2))
+#define leftRotate3_64(a)  (leftRotate_64((a), 3))
+#define leftRotate4_64(a)  (leftRotate_64((a), 4))
+#define leftRotate5_64(a)  (leftRotate_64((a), 5))
+#define leftRotate6_64(a)  (leftRotate_64((a), 6))
+#define leftRotate7_64(a)  (leftRotate_64((a), 7))
+#define leftRotate8_64(a)  (leftRotate_64((a), 8))
+#define leftRotate9_64(a)  (leftRotate_64((a), 9))
+#define leftRotate10_64(a) (leftRotate_64((a), 10))
+#define leftRotate11_64(a) (leftRotate_64((a), 11))
+#define leftRotate12_64(a) (leftRotate_64((a), 12))
+#define leftRotate13_64(a) (leftRotate_64((a), 13))
+#define leftRotate14_64(a) (leftRotate_64((a), 14))
+#define leftRotate15_64(a) (leftRotate_64((a), 15))
+#define leftRotate16_64(a) (leftRotate_64((a), 16))
+#define leftRotate17_64(a) (leftRotate_64((a), 17))
+#define leftRotate18_64(a) (leftRotate_64((a), 18))
+#define leftRotate19_64(a) (leftRotate_64((a), 19))
+#define leftRotate20_64(a) (leftRotate_64((a), 20))
+#define leftRotate21_64(a) (leftRotate_64((a), 21))
+#define leftRotate22_64(a) (leftRotate_64((a), 22))
+#define leftRotate23_64(a) (leftRotate_64((a), 23))
+#define leftRotate24_64(a) (leftRotate_64((a), 24))
+#define leftRotate25_64(a) (leftRotate_64((a), 25))
+#define leftRotate26_64(a) (leftRotate_64((a), 26))
+#define leftRotate27_64(a) (leftRotate_64((a), 27))
+#define leftRotate28_64(a) (leftRotate_64((a), 28))
+#define leftRotate29_64(a) (leftRotate_64((a), 29))
+#define leftRotate30_64(a) (leftRotate_64((a), 30))
+#define leftRotate31_64(a) (leftRotate_64((a), 31))
+#define leftRotate32_64(a) (leftRotate_64((a), 32))
+#define leftRotate33_64(a) (leftRotate_64((a), 33))
+#define leftRotate34_64(a) (leftRotate_64((a), 34))
+#define leftRotate35_64(a) (leftRotate_64((a), 35))
+#define leftRotate36_64(a) (leftRotate_64((a), 36))
+#define leftRotate37_64(a) (leftRotate_64((a), 37))
+#define leftRotate38_64(a) (leftRotate_64((a), 38))
+#define leftRotate39_64(a) (leftRotate_64((a), 39))
+#define leftRotate40_64(a) (leftRotate_64((a), 40))
+#define leftRotate41_64(a) (leftRotate_64((a), 41))
+#define leftRotate42_64(a) (leftRotate_64((a), 42))
+#define leftRotate43_64(a) (leftRotate_64((a), 43))
+#define leftRotate44_64(a) (leftRotate_64((a), 44))
+#define leftRotate45_64(a) (leftRotate_64((a), 45))
+#define leftRotate46_64(a) (leftRotate_64((a), 46))
+#define leftRotate47_64(a) (leftRotate_64((a), 47))
+#define leftRotate48_64(a) (leftRotate_64((a), 48))
+#define leftRotate49_64(a) (leftRotate_64((a), 49))
+#define leftRotate50_64(a) (leftRotate_64((a), 50))
+#define leftRotate51_64(a) (leftRotate_64((a), 51))
+#define leftRotate52_64(a) (leftRotate_64((a), 52))
+#define leftRotate53_64(a) (leftRotate_64((a), 53))
+#define leftRotate54_64(a) (leftRotate_64((a), 54))
+#define leftRotate55_64(a) (leftRotate_64((a), 55))
+#define leftRotate56_64(a) (leftRotate_64((a), 56))
+#define leftRotate57_64(a) (leftRotate_64((a), 57))
+#define leftRotate58_64(a) (leftRotate_64((a), 58))
+#define leftRotate59_64(a) (leftRotate_64((a), 59))
+#define leftRotate60_64(a) (leftRotate_64((a), 60))
+#define leftRotate61_64(a) (leftRotate_64((a), 61))
+#define leftRotate62_64(a) (leftRotate_64((a), 62))
+#define leftRotate63_64(a) (leftRotate_64((a), 63))
+
+/* Right rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define rightRotate1_64(a)  (rightRotate_64((a), 1))
+#define rightRotate2_64(a)  (rightRotate_64((a), 2))
+#define rightRotate3_64(a)  (rightRotate_64((a), 3))
+#define rightRotate4_64(a)  (rightRotate_64((a), 4))
+#define rightRotate5_64(a)  (rightRotate_64((a), 5))
+#define rightRotate6_64(a)  (rightRotate_64((a), 6))
+#define rightRotate7_64(a)  (rightRotate_64((a), 7))
+#define rightRotate8_64(a)  (rightRotate_64((a), 8))
+#define rightRotate9_64(a)  (rightRotate_64((a), 9))
+#define rightRotate10_64(a) (rightRotate_64((a), 10))
+#define rightRotate11_64(a) (rightRotate_64((a), 11))
+#define rightRotate12_64(a) (rightRotate_64((a), 12))
+#define rightRotate13_64(a) (rightRotate_64((a), 13))
+#define rightRotate14_64(a) (rightRotate_64((a), 14))
+#define rightRotate15_64(a) (rightRotate_64((a), 15))
+#define rightRotate16_64(a) (rightRotate_64((a), 16))
+#define rightRotate17_64(a) (rightRotate_64((a), 17))
+#define rightRotate18_64(a) (rightRotate_64((a), 18))
+#define rightRotate19_64(a) (rightRotate_64((a), 19))
+#define rightRotate20_64(a) (rightRotate_64((a), 20))
+#define rightRotate21_64(a) (rightRotate_64((a), 21))
+#define rightRotate22_64(a) (rightRotate_64((a), 22))
+#define rightRotate23_64(a) (rightRotate_64((a), 23))
+#define rightRotate24_64(a) (rightRotate_64((a), 24))
+#define rightRotate25_64(a) (rightRotate_64((a), 25))
+#define rightRotate26_64(a) (rightRotate_64((a), 26))
+#define rightRotate27_64(a) (rightRotate_64((a), 27))
+#define rightRotate28_64(a) (rightRotate_64((a), 28))
+#define rightRotate29_64(a) (rightRotate_64((a), 29))
+#define rightRotate30_64(a) (rightRotate_64((a), 30))
+#define rightRotate31_64(a) (rightRotate_64((a), 31))
+#define rightRotate32_64(a) (rightRotate_64((a), 32))
+#define rightRotate33_64(a) (rightRotate_64((a), 33))
+#define rightRotate34_64(a) (rightRotate_64((a), 34))
+#define rightRotate35_64(a) (rightRotate_64((a), 35))
+#define rightRotate36_64(a) (rightRotate_64((a), 36))
+#define rightRotate37_64(a) (rightRotate_64((a), 37))
+#define rightRotate38_64(a) (rightRotate_64((a), 38))
+#define rightRotate39_64(a) (rightRotate_64((a), 39))
+#define rightRotate40_64(a) (rightRotate_64((a), 40))
+#define rightRotate41_64(a) (rightRotate_64((a), 41))
+#define rightRotate42_64(a) (rightRotate_64((a), 42))
+#define rightRotate43_64(a) (rightRotate_64((a), 43))
+#define rightRotate44_64(a) (rightRotate_64((a), 44))
+#define rightRotate45_64(a) (rightRotate_64((a), 45))
+#define rightRotate46_64(a) (rightRotate_64((a), 46))
+#define rightRotate47_64(a) (rightRotate_64((a), 47))
+#define rightRotate48_64(a) (rightRotate_64((a), 48))
+#define rightRotate49_64(a) (rightRotate_64((a), 49))
+#define rightRotate50_64(a) (rightRotate_64((a), 50))
+#define rightRotate51_64(a) (rightRotate_64((a), 51))
+#define rightRotate52_64(a) (rightRotate_64((a), 52))
+#define rightRotate53_64(a) (rightRotate_64((a), 53))
+#define rightRotate54_64(a) (rightRotate_64((a), 54))
+#define rightRotate55_64(a) (rightRotate_64((a), 55))
+#define rightRotate56_64(a) (rightRotate_64((a), 56))
+#define rightRotate57_64(a) (rightRotate_64((a), 57))
+#define rightRotate58_64(a) (rightRotate_64((a), 58))
+#define rightRotate59_64(a) (rightRotate_64((a), 59))
+#define rightRotate60_64(a) (rightRotate_64((a), 60))
+#define rightRotate61_64(a) (rightRotate_64((a), 61))
+#define rightRotate62_64(a) (rightRotate_64((a), 62))
+#define rightRotate63_64(a) (rightRotate_64((a), 63))
+
+/* Rotate a 16-bit value left by a number of bits */
+#define leftRotate_16(a, bits) \
+    (__extension__ ({ \
+        uint16_t _temp = (a); \
+        (_temp << (bits)) | (_temp >> (16 - (bits))); \
+    }))
+
+/* Rotate a 16-bit value right by a number of bits */
+#define rightRotate_16(a, bits) \
+    (__extension__ ({ \
+        uint16_t _temp = (a); \
+        (_temp >> (bits)) | (_temp << (16 - (bits))); \
+    }))
+
+/* Left rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define leftRotate1_16(a)  (leftRotate_16((a), 1))
+#define leftRotate2_16(a)  (leftRotate_16((a), 2))
+#define leftRotate3_16(a)  (leftRotate_16((a), 3))
+#define leftRotate4_16(a)  (leftRotate_16((a), 4))
+#define leftRotate5_16(a)  (leftRotate_16((a), 5))
+#define leftRotate6_16(a)  (leftRotate_16((a), 6))
+#define leftRotate7_16(a)  (leftRotate_16((a), 7))
+#define leftRotate8_16(a)  (leftRotate_16((a), 8))
+#define leftRotate9_16(a)  (leftRotate_16((a), 9))
+#define leftRotate10_16(a) (leftRotate_16((a), 10))
+#define leftRotate11_16(a) (leftRotate_16((a), 11))
+#define leftRotate12_16(a) (leftRotate_16((a), 12))
+#define leftRotate13_16(a) (leftRotate_16((a), 13))
+#define leftRotate14_16(a) (leftRotate_16((a), 14))
+#define leftRotate15_16(a) (leftRotate_16((a), 15))
+
+/* Right rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define rightRotate1_16(a)  (rightRotate_16((a), 1))
+#define rightRotate2_16(a)  (rightRotate_16((a), 2))
+#define rightRotate3_16(a)  (rightRotate_16((a), 3))
+#define rightRotate4_16(a)  (rightRotate_16((a), 4))
+#define rightRotate5_16(a)  (rightRotate_16((a), 5))
+#define rightRotate6_16(a)  (rightRotate_16((a), 6))
+#define rightRotate7_16(a)  (rightRotate_16((a), 7))
+#define rightRotate8_16(a)  (rightRotate_16((a), 8))
+#define rightRotate9_16(a)  (rightRotate_16((a), 9))
+#define rightRotate10_16(a) (rightRotate_16((a), 10))
+#define rightRotate11_16(a) (rightRotate_16((a), 11))
+#define rightRotate12_16(a) (rightRotate_16((a), 12))
+#define rightRotate13_16(a) (rightRotate_16((a), 13))
+#define rightRotate14_16(a) (rightRotate_16((a), 14))
+#define rightRotate15_16(a) (rightRotate_16((a), 15))
+
+/* Rotate an 8-bit value left by a number of bits */
+#define leftRotate_8(a, bits) \
+    (__extension__ ({ \
+        uint8_t _temp = (a); \
+        (_temp << (bits)) | (_temp >> (8 - (bits))); \
+    }))
+
+/* Rotate an 8-bit value right by a number of bits */
+#define rightRotate_8(a, bits) \
+    (__extension__ ({ \
+        uint8_t _temp = (a); \
+        (_temp >> (bits)) | (_temp << (8 - (bits))); \
+    }))
+
+/* Left rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define leftRotate1_8(a)  (leftRotate_8((a), 1))
+#define leftRotate2_8(a)  (leftRotate_8((a), 2))
+#define leftRotate3_8(a)  (leftRotate_8((a), 3))
+#define leftRotate4_8(a)  (leftRotate_8((a), 4))
+#define leftRotate5_8(a)  (leftRotate_8((a), 5))
+#define leftRotate6_8(a)  (leftRotate_8((a), 6))
+#define leftRotate7_8(a)  (leftRotate_8((a), 7))
+
+/* Right rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define rightRotate1_8(a)  (rightRotate_8((a), 1))
+#define rightRotate2_8(a)  (rightRotate_8((a), 2))
+#define rightRotate3_8(a)  (rightRotate_8((a), 3))
+#define rightRotate4_8(a)  (rightRotate_8((a), 4))
+#define rightRotate5_8(a)  (rightRotate_8((a), 5))
+#define rightRotate6_8(a)  (rightRotate_8((a), 6))
+#define rightRotate7_8(a)  (rightRotate_8((a), 7))
+
+#endif

gimli/Implementations/crypto_hash/gimli24v1/rhys-avr/aead-common.c

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "aead-common.h"
+
+int aead_check_tag
+    (unsigned char *plaintext, unsigned long long plaintext_len,
+     const unsigned char *tag1, const unsigned char *tag2,
+     unsigned size)
+{
+    /* Set "accum" to -1 if the tags match, or 0 if they don't match */
+    int accum = 0;
+    while (size > 0) {
+        accum |= (*tag1++ ^ *tag2++);
+        --size;
+    }
+    accum = (accum - 1) >> 8;
+
+    /* Destroy the plaintext if the tag match failed */
+    while (plaintext_len > 0) {
+        *plaintext++ &= accum;
+        --plaintext_len;
+    }
+
+    /* If "accum" is 0, return -1, otherwise return 0 */
+    return ~accum;
+}
+
+int aead_check_tag_precheck
+    (unsigned char *plaintext, unsigned long long plaintext_len,
+     const unsigned char *tag1, const unsigned char *tag2,
+     unsigned size, int precheck)
+{
+    /* Set "accum" to -1 if the tags match, or 0 if they don't match */
+    int accum = 0;
+    while (size > 0) {
+        accum |= (*tag1++ ^ *tag2++);
+        --size;
+    }
+    accum = ((accum - 1) >> 8) & precheck;
+
+    /* Destroy the plaintext if the tag match failed */
+    while (plaintext_len > 0) {
+        *plaintext++ &= accum;
+        --plaintext_len;
+    }
+
+    /* If "accum" is 0, return -1, otherwise return 0 */
+    return ~accum;
+}

gimli/Implementations/crypto_hash/gimli24v1/rhys-avr/aead-common.h

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef LWCRYPTO_AEAD_COMMON_H
+#define LWCRYPTO_AEAD_COMMON_H
+
+#include <stddef.h>
+
+/**
+ * \file aead-common.h
+ * \brief Definitions that are common across AEAD schemes.
+ *
+ * AEAD stands for "Authenticated Encryption with Associated Data".
+ * It is a standard API pattern for securely encrypting and
+ * authenticating packets of data.
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * \brief Encrypts and authenticates a packet with an AEAD scheme.
+ *
+ * \param c Buffer to receive the output.
+ * \param clen On exit, set to the length of the output which includes
+ * the ciphertext and the authentication tag.
+ * \param m Buffer that contains the plaintext message to encrypt.
+ * \param mlen Length of the plaintext message in bytes.
+ * \param ad Buffer that contains associated data to authenticate
+ * along with the packet but which does not need to be encrypted.
+ * \param adlen Length of the associated data in bytes.
+ * \param nsec Secret nonce - normally not used by AEAD schemes.
+ * \param npub Points to the public nonce for the packet.
+ * \param k Points to the key to use to encrypt the packet.
+ *
+ * \return 0 on success, or a negative value if there was an error in
+ * the parameters.
+ */
+typedef int (*aead_cipher_encrypt_t)
+    (unsigned char *c, unsigned long long *clen,
+     const unsigned char *m, unsigned long long mlen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *nsec,
+     const unsigned char *npub,
+     const unsigned char *k);
+
+/**
+ * \brief Decrypts and authenticates a packet with an AEAD scheme.
+ *
+ * \param m Buffer to receive the plaintext message on output.
+ * \param mlen Receives the length of the plaintext message on output.
+ * \param nsec Secret nonce - normally not used by AEAD schemes.
+ * \param c Buffer that contains the ciphertext and authentication
+ * tag to decrypt.
+ * \param clen Length of the input data in bytes, which includes the
+ * ciphertext and the authentication tag.
+ * \param ad Buffer that contains associated data to authenticate
+ * along with the packet but which does not need to be encrypted.
+ * \param adlen Length of the associated data in bytes.
+ * \param npub Points to the public nonce for the packet.
+ * \param k Points to the key to use to decrypt the packet.
+ *
+ * \return 0 on success, -1 if the authentication tag was incorrect,
+ * or some other negative number if there was an error in the parameters.
+ */
+typedef int (*aead_cipher_decrypt_t)
+    (unsigned char *m, unsigned long long *mlen,
+     unsigned char *nsec,
+     const unsigned char *c, unsigned long long clen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *npub,
+     const unsigned char *k);
+
+/**
+ * \brief Hashes a block of input data.
+ *
+ * \param out Buffer to receive the hash output.
+ * \param in Points to the input data to be hashed.
+ * \param inlen Length of the input data in bytes.
+ *
+ * \return Returns zero on success or -1 if there was an error in the
+ * parameters.
+ */
+typedef int (*aead_hash_t)
+    (unsigned char *out, const unsigned char *in, unsigned long long inlen);
+
+/**
+ * \brief Initializes the state for a hashing operation.
+ *
+ * \param state Hash state to be initialized.
+ */
+typedef void (*aead_hash_init_t)(void *state);
+
+/**
+ * \brief Updates a hash state with more input data.
+ *
+ * \param state Hash state to be updated.
+ * \param in Points to the input data to be incorporated into the state.
+ * \param inlen Length of the input data to be incorporated into the state.
+ */
+typedef void (*aead_hash_update_t)
+    (void *state, const unsigned char *in, unsigned long long inlen);
+
+/**
+ * \brief Returns the final hash value from a hashing operation.
+ *
+ * \param Hash state to be finalized.
+ * \param out Points to the output buffer to receive the hash value.
+ */
+typedef void (*aead_hash_finalize_t)(void *state, unsigned char *out);
+
+/**
+ * \brief Aborbs more input data into an XOF state.
+ *
+ * \param state XOF state to be updated.
+ * \param in Points to the input data to be absorbed into the state.
+ * \param inlen Length of the input data to be absorbed into the state.
+ *
+ * \sa ascon_xof_init(), ascon_xof_squeeze()
+ */
+typedef void (*aead_xof_absorb_t)
+    (void *state, const unsigned char *in, unsigned long long inlen);
+
+/**
+ * \brief Squeezes output data from an XOF state.
+ *
+ * \param state XOF state to squeeze the output data from.
+ * \param out Points to the output buffer to receive the squeezed data.
+ * \param outlen Number of bytes of data to squeeze out of the state.
+ */
+typedef void (*aead_xof_squeeze_t)
+    (void *state, unsigned char *out, unsigned long long outlen);
+
+/**
+ * \brief No special AEAD features.
+ */
+#define AEAD_FLAG_NONE          0x0000
+
+/**
+ * \brief The natural byte order of the AEAD cipher is little-endian.
+ *
+ * If this flag is not present, then the natural byte order of the
+ * AEAD cipher should be assumed to be big-endian.
+ *
+ * The natural byte order may be useful when formatting packet sequence
+ * numbers as nonces.  The application needs to know whether the sequence
+ * number should be packed into the leading or trailing bytes of the nonce.
+ */
+#define AEAD_FLAG_LITTLE_ENDIAN 0x0001
+
+/**
+ * \brief Meta-information about an AEAD cipher.
+ */
+typedef struct
+{
+    const char *name;               /**< Name of the cipher */
+    unsigned key_len;               /**< Length of the key in bytes */
+    unsigned nonce_len;             /**< Length of the nonce in bytes */
+    unsigned tag_len;               /**< Length of the tag in bytes */
+    unsigned flags;                 /**< Flags for extra features */
+    aead_cipher_encrypt_t encrypt;  /**< AEAD encryption function */
+    aead_cipher_decrypt_t decrypt;  /**< AEAD decryption function */
+
+} aead_cipher_t;
+
+/**
+ * \brief Meta-information about a hash algorithm that is related to an AEAD.
+ *
+ * Regular hash algorithms should provide the "hash", "init", "update",
+ * and "finalize" functions.  Extensible Output Functions (XOF's) should
+ * proivde the "hash", "init", "absorb", and "squeeze" functions.
+ */
+typedef struct
+{
+    const char *name;           /**< Name of the hash algorithm */
+    size_t state_size;          /**< Size of the incremental state structure */
+    unsigned hash_len;          /**< Length of the hash in bytes */
+    unsigned flags;             /**< Flags for extra features */
+    aead_hash_t hash;           /**< All in one hashing function */
+    aead_hash_init_t init;      /**< Incremental hash/XOF init function */
+    aead_hash_update_t update;  /**< Incremental hash update function */
+    aead_hash_finalize_t finalize; /**< Incremental hash finalize function */
+    aead_xof_absorb_t absorb;   /**< Incremental XOF absorb function */
+    aead_xof_squeeze_t squeeze; /**< Incremental XOF squeeze function */
+
+} aead_hash_algorithm_t;
+
+/**
+ * \brief Check an authentication tag in constant time.
+ *
+ * \param plaintext Points to the plaintext data.
+ * \param plaintext_len Length of the plaintext in bytes.
+ * \param tag1 First tag to compare.
+ * \param tag2 Second tag to compare.
+ * \param tag_len Length of the tags in bytes.
+ *
+ * \return Returns -1 if the tag check failed or 0 if the check succeeded.
+ *
+ * If the tag check fails, then the \a plaintext will also be zeroed to
+ * prevent it from being used accidentally by the application when the
+ * ciphertext was invalid.
+ */
+int aead_check_tag
+    (unsigned char *plaintext, unsigned long long plaintext_len,
+     const unsigned char *tag1, const unsigned char *tag2,
+     unsigned tag_len);
+
+/**
+ * \brief Check an authentication tag in constant time with a previous check.
+ *
+ * \param plaintext Points to the plaintext data.
+ * \param plaintext_len Length of the plaintext in bytes.
+ * \param tag1 First tag to compare.
+ * \param tag2 Second tag to compare.
+ * \param tag_len Length of the tags in bytes.
+ * \param precheck Set to -1 if previous check succeeded or 0 if it failed.
+ *
+ * \return Returns -1 if the tag check failed or 0 if the check succeeded.
+ *
+ * If the tag check fails, then the \a plaintext will also be zeroed to
+ * prevent it from being used accidentally by the application when the
+ * ciphertext was invalid.
+ *
+ * This version can be used to incorporate other information about the
+ * correctness of the plaintext into the final result.
+ */
+int aead_check_tag_precheck
+    (unsigned char *plaintext, unsigned long long plaintext_len,
+     const unsigned char *tag1, const unsigned char *tag2,
+     unsigned tag_len, int precheck);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

gimli/Implementations/crypto_hash/gimli24v1/rhys-avr/api.h

+#define CRYPTO_BYTES 32

gimli/Implementations/crypto_hash/gimli24v1/rhys-avr/gimli24.c

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "gimli24.h"
+#include "internal-gimli24.h"
+#include <string.h>
+
+aead_cipher_t const gimli24_cipher = {
+    "GIMLI-24",
+    GIMLI24_KEY_SIZE,
+    GIMLI24_NONCE_SIZE,
+    GIMLI24_TAG_SIZE,
+    AEAD_FLAG_LITTLE_ENDIAN,
+    gimli24_aead_encrypt,
+    gimli24_aead_decrypt
+};
+
+aead_hash_algorithm_t const gimli24_hash_algorithm = {
+    "GIMLI-24-HASH",
+    sizeof(gimli24_hash_state_t),
+    GIMLI24_HASH_SIZE,
+    AEAD_FLAG_LITTLE_ENDIAN,
+    gimli24_hash,
+    (aead_hash_init_t)gimli24_hash_init,
+    (aead_hash_update_t)gimli24_hash_absorb,
+    (aead_hash_finalize_t)gimli24_hash_finalize,
+    (aead_xof_absorb_t)gimli24_hash_absorb,
+    (aead_xof_squeeze_t)gimli24_hash_squeeze
+};
+
+/**
+ * \brief Number of bytes of input or output data to process per block.
+ */
+#define GIMLI24_BLOCK_SIZE 16
+
+/**
+ * \brief Structure of the GIMLI-24 state as both an array of words
+ * and an array of bytes.
+ */
+typedef union
+{
+    uint32_t words[12];     /**< Words in the state */
+    uint8_t bytes[48];      /**< Bytes in the state */
+
+} gimli24_state_t;
+
+/**
+ * \brief Absorbs data into a GIMLI-24 state.
+ *
+ * \param state The state to absorb the data into.
+ * \param data Points to the data to be absorbed.
+ * \param len Length of the data to be absorbed.
+ */
+static void gimli24_absorb
+    (gimli24_state_t *state, const unsigned char *data, unsigned long long len)
+{
+    unsigned temp;
+    while (len >= GIMLI24_BLOCK_SIZE) {
+        lw_xor_block(state->bytes, data, GIMLI24_BLOCK_SIZE);
+        gimli24_permute(state->words);
+        data += GIMLI24_BLOCK_SIZE;
+        len -= GIMLI24_BLOCK_SIZE;
+    }
+    temp = (unsigned)len;
+    lw_xor_block(state->bytes, data, temp);
+    state->bytes[temp] ^= 0x01; /* Padding */
+    state->bytes[47] ^= 0x01;
+    gimli24_permute(state->words);
+}
+
+/**
+ * \brief Encrypts a block of data with a GIMLI-24 state.
+ *
+ * \param state The state to encrypt with.
+ * \param dest Points to the destination buffer.
+ * \param src Points to the source buffer.
+ * \param len Length of the data to encrypt from \a src into \a dest.
+ */
+static void gimli24_encrypt
+    (gimli24_state_t *state, unsigned char *dest,
+     const unsigned char *src, unsigned long long len)
+{
+    unsigned temp;
+    while (len >= GIMLI24_BLOCK_SIZE) {
+        lw_xor_block_2_dest(dest, state->bytes, src, GIMLI24_BLOCK_SIZE);
+        gimli24_permute(state->words);
+        dest += GIMLI24_BLOCK_SIZE;
+        src += GIMLI24_BLOCK_SIZE;
+        len -= GIMLI24_BLOCK_SIZE;
+    }
+    temp = (unsigned)len;
+    lw_xor_block_2_dest(dest, state->bytes, src, temp);
+    state->bytes[temp] ^= 0x01; /* Padding */
+    state->bytes[47] ^= 0x01;
+    gimli24_permute(state->words);
+}
+
+/**
+ * \brief Decrypts a block of data with a GIMLI-24 state.
+ *
+ * \param state The state to decrypt with.
+ * \param dest Points to the destination buffer.
+ * \param src Points to the source buffer.
+ * \param len Length of the data to decrypt from \a src into \a dest.
+ */
+static void gimli24_decrypt
+    (gimli24_state_t *state, unsigned char *dest,
+     const unsigned char *src, unsigned long long len)
+{
+    unsigned temp;
+    while (len >= GIMLI24_BLOCK_SIZE) {
+        lw_xor_block_swap(dest, state->bytes, src, GIMLI24_BLOCK_SIZE);
+        gimli24_permute(state->words);
+        dest += GIMLI24_BLOCK_SIZE;
+        src += GIMLI24_BLOCK_SIZE;
+        len -= GIMLI24_BLOCK_SIZE;
+    }
+    temp = (unsigned)len;
+    lw_xor_block_swap(dest, state->bytes, src, temp);
+    state->bytes[temp] ^= 0x01; /* Padding */
+    state->bytes[47] ^= 0x01;
+    gimli24_permute(state->words);
+}
+
+int gimli24_aead_encrypt
+    (unsigned char *c, unsigned long long *clen,
+     const unsigned char *m, unsigned long long mlen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *nsec,
+     const unsigned char *npub,
+     const unsigned char *k)
+{
+    gimli24_state_t state;
+    (void)nsec;
+
+    /* Set the length of the returned ciphertext */
+    *clen = mlen + GIMLI24_TAG_SIZE;
+
+    /* Format the initial GIMLI state from the nonce and the key */
+    memcpy(state.words, npub, GIMLI24_NONCE_SIZE);
+    memcpy(state.words + 4, k, GIMLI24_KEY_SIZE);
+
+    /* Permute the initial state */
+    gimli24_permute(state.words);
+
+    /* Absorb the associated data */
+    gimli24_absorb(&state, ad, adlen);
+
+    /* Encrypt the plaintext to produce the ciphertext */
+    gimli24_encrypt(&state, c, m, mlen);
+
+    /* Generate the authentication tag at the end of the ciphertext */
+    memcpy(c + mlen, state.bytes, GIMLI24_TAG_SIZE);
+    return 0;
+}
+
+int gimli24_aead_decrypt
+    (unsigned char *m, unsigned long long *mlen,
+     unsigned char *nsec,
+     const unsigned char *c, unsigned long long clen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *npub,
+     const unsigned char *k)
+{
+    gimli24_state_t state;
+    (void)nsec;
+
+    /* Validate the ciphertext length and set the return "mlen" value */
+    if (clen < GIMLI24_TAG_SIZE)
+        return -1;
+    *mlen = clen - GIMLI24_TAG_SIZE;
+
+    /* Format the initial GIMLI state from the nonce and the key */
+    memcpy(state.words, npub, GIMLI24_NONCE_SIZE);
+    memcpy(state.words + 4, k, GIMLI24_KEY_SIZE);
+
+    /* Permute the initial state */
+    gimli24_permute(state.words);
+
+    /* Absorb the associated data */
+    gimli24_absorb(&state, ad, adlen);
+
+    /* Decrypt the ciphertext to produce the plaintext */
+    gimli24_decrypt(&state, m, c, *mlen);
+
+    /* Check the authentication tag at the end of the packet */
+    return aead_check_tag
+        (m, *mlen, state.bytes, c + *mlen, GIMLI24_TAG_SIZE);
+}
+
+int gimli24_hash
+    (unsigned char *out, const unsigned char *in, unsigned long long inlen)
+{
+    gimli24_state_t state;
+
+    /* Initialize the hash state to all zeroes */
+    memset(&state, 0, sizeof(state));
+
+    /* Absorb the input */
+    gimli24_absorb(&state, in, inlen);
+
+    /* Generate the output hash */
+    memcpy(out, state.bytes, GIMLI24_HASH_SIZE / 2);
+    gimli24_permute(state.words);
+    memcpy(out + GIMLI24_HASH_SIZE / 2, state.bytes, GIMLI24_HASH_SIZE / 2);
+    return 0;
+}
+
+void gimli24_hash_init(gimli24_hash_state_t *state)
+{
+    memset(state, 0, sizeof(gimli24_hash_state_t));
+}
+
+#define GIMLI24_XOF_RATE 16
+#define gimli24_xof_permute() \
+    gimli24_permute((uint32_t *)(state->s.state))
+
+void gimli24_hash_absorb
+    (gimli24_hash_state_t *state, const unsigned char *in,
+     unsigned long long inlen)
+{
+    unsigned temp;
+
+    if (state->s.mode) {
+        /* We were squeezing output - go back to the absorb phase */
+        state->s.mode = 0;
+        state->s.count = 0;
+        gimli24_xof_permute();
+    }
+
+    /* Handle the partial left-over block from last time */
+    if (state->s.count) {
+        temp = GIMLI24_XOF_RATE - state->s.count;
+        if (temp > inlen) {
+            temp = (unsigned)inlen;
+            lw_xor_block(state->s.state + state->s.count, in, temp);
+            state->s.count += temp;
+            return;
+        }
+        lw_xor_block(state->s.state + state->s.count, in, temp);
+        state->s.count = 0;
+        in += temp;
+        inlen -= temp;
+        gimli24_xof_permute();
+    }
+
+    /* Process full blocks that are aligned at state->s.count == 0 */
+    while (inlen >= GIMLI24_XOF_RATE) {
+        lw_xor_block(state->s.state, in, GIMLI24_XOF_RATE);
+        in += GIMLI24_XOF_RATE;
+        inlen -= GIMLI24_XOF_RATE;
+        gimli24_xof_permute();
+    }
+
+    /* Process the left-over block at the end of the input */
+    temp = (unsigned)inlen;
+    lw_xor_block(state->s.state, in, temp);
+    state->s.count = temp;
+}
+
+void gimli24_hash_squeeze
+    (gimli24_hash_state_t *state, unsigned char *out,
+     unsigned long long outlen)
+{
+    unsigned temp;
+
+    /* Pad the final input block if we were still in the absorb phase */
+    if (!state->s.mode) {
+        state->s.state[state->s.count] ^= 0x01;
+        state->s.state[47] ^= 0x01;
+        state->s.count = 0;
+        state->s.mode = 1;
+    }
+
+    /* Handle left-over partial blocks from last time */
+    if (state->s.count) {
+        temp = GIMLI24_XOF_RATE - state->s.count;
+        if (temp > outlen) {
+            temp = (unsigned)outlen;
+            memcpy(out, state->s.state + state->s.count, temp);
+            state->s.count += temp;
+            return;
+        }
+        memcpy(out, state->s.state + state->s.count, temp);
+        out += temp;
+        outlen -= temp;
+        state->s.count = 0;
+    }
+
+    /* Handle full blocks */
+    while (outlen >= GIMLI24_XOF_RATE) {
+        gimli24_xof_permute();
+        memcpy(out, state->s.state, GIMLI24_XOF_RATE);
+        out += GIMLI24_XOF_RATE;
+        outlen -= GIMLI24_XOF_RATE;
+    }
+
+    /* Handle the left-over block */
+    if (outlen > 0) {
+        temp = (unsigned)outlen;
+        gimli24_xof_permute();
+        memcpy(out, state->s.state, temp);
+        state->s.count = temp;
+    }
+}
+
+void gimli24_hash_finalize
+    (gimli24_hash_state_t *state, unsigned char *out)
+{
+    gimli24_hash_squeeze(state, out, GIMLI24_HASH_SIZE);
+}

gimli/Implementations/crypto_hash/gimli24v1/rhys-avr/gimli24.h

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef LWCRYPTO_GIMLI24_H
+#define LWCRYPTO_GIMLI24_H
+
+#include "aead-common.h"
+
+/**
+ * \file gimli24.h
+ * \brief Gimli authenticated encryption algorithm.
+ *
+ * GIMLI-24-CIPHER has a 256-bit key, a 128-bit nonce, and a 128-bit tag.
+ * It is the spiritual successor to the widely used ChaCha20 and has a
+ * similar design.
+ *
+ * This library also includes an implementation of the hash algorithm
+ * GIMLI-24-HASH in both regular hashing and XOF modes.
+ *
+ * References: https://gimli.cr.yp.to/
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * \brief Size of the key for GIMLI-24.
+ */
+#define GIMLI24_KEY_SIZE 32
+
+/**
+ * \brief Size of the nonce for GIMLI-24.
+ */
+#define GIMLI24_NONCE_SIZE 16
+
+/**
+ * \brief Size of the authentication tag for GIMLI-24.
+ */
+#define GIMLI24_TAG_SIZE 16
+
+/**
+ * \brief Size of the hash output for GIMLI-24.
+ */
+#define GIMLI24_HASH_SIZE 32
+
+/**
+ * \brief State information for GIMLI-24-HASH incremental modes.
+ */
+typedef union
+{
+    struct {
+        unsigned char state[48]; /**< Current hash state */
+        unsigned char count;     /**< Number of bytes in the current block */
+        unsigned char mode;      /**< Hash mode: 0 for absorb, 1 for squeeze */
+    } s;                         /**< State */
+    unsigned long long align;    /**< For alignment of this structure */
+
+} gimli24_hash_state_t;
+
+/**
+ * \brief Meta-information block for the GIMLI-24 cipher.
+ */
+extern aead_cipher_t const gimli24_cipher;
+
+/**
+ * \brief Meta-information block for the GIMLI-24-HASH algorithm.
+ *
+ * This meta-information block can also be used in XOF mode.
+ */
+extern aead_hash_algorithm_t const gimli24_hash_algorithm;
+
+/**
+ * \brief Encrypts and authenticates a packet with GIMLI-24 using the
+ * full AEAD mode.
+ *
+ * \param c Buffer to receive the output.
+ * \param clen On exit, set to the length of the output which includes
+ * the ciphertext and the 16 byte authentication tag.
+ * \param m Buffer that contains the plaintext message to encrypt.
+ * \param mlen Length of the plaintext message in bytes.
+ * \param ad Buffer that contains associated data to authenticate
+ * along with the packet but which does not need to be encrypted.
+ * \param adlen Length of the associated data in bytes.
+ * \param nsec Secret nonce - not used by this algorithm.
+ * \param npub Points to the public nonce for the packet which must
+ * be 16 bytes in length.
+ * \param k Points to the 32 bytes of the key to use to encrypt the packet.
+ *
+ * \return 0 on success, or a negative value if there was an error in
+ * the parameters.
+ *
+ * \sa gimli24_aead_decrypt()
+ */
+int gimli24_aead_encrypt
+    (unsigned char *c, unsigned long long *clen,
+     const unsigned char *m, unsigned long long mlen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *nsec,
+     const unsigned char *npub,
+     const unsigned char *k);
+
+/**
+ * \brief Decrypts and authenticates a packet with GIMLI-24 using the
+ * full AEAD mode.
+ *
+ * \param m Buffer to receive the plaintext message on output.
+ * \param mlen Receives the length of the plaintext message on output.
+ * \param nsec Secret nonce - not used by this algorithm.
+ * \param c Buffer that contains the ciphertext and authentication
+ * tag to decrypt.
+ * \param clen Length of the input data in bytes, which includes the
+ * ciphertext and the 16 byte authentication tag.
+ * \param ad Buffer that contains associated data to authenticate
+ * along with the packet but which does not need to be encrypted.
+ * \param adlen Length of the associated data in bytes.
+ * \param npub Points to the public nonce for the packet which must
+ * be 16 bytes in length.
+ * \param k Points to the 32 bytes of the key to use to decrypt the packet.
+ *
+ * \return 0 on success, -1 if the authentication tag was incorrect,
+ * or some other negative number if there was an error in the parameters.
+ *
+ * \sa gimli24_aead_encrypt()
+ */
+int gimli24_aead_decrypt
+    (unsigned char *m, unsigned long long *mlen,
+     unsigned char *nsec,
+     const unsigned char *c, unsigned long long clen,
+     const unsigned char *ad, unsigned long long adlen,
+     const unsigned char *npub,
+     const unsigned char *k);
+
+/**
+ * \brief Hashes a block of input data with GIMLI-24 to generate a hash value.
+ *
+ * \param out Buffer to receive the hash output which must be at least
+ * GIMLI24_HASH_SIZE bytes in length.
+ * \param in Points to the input data to be hashed.
+ * \param inlen Length of the input data in bytes.
+ *
+ * \return Returns zero on success or -1 if there was an error in the
+ * parameters.
+ */
+int gimli24_hash
+    (unsigned char *out, const unsigned char *in, unsigned long long inlen);
+
+/**
+ * \brief Initializes the state for a GIMLI-24-HASH hashing operation.
+ *
+ * \param state Hash state to be initialized.
+ *
+ * \sa gimli24_hash_absorb(), gimli24_hash_squeeze(), gimli24_hash()
+ */
+void gimli24_hash_init(gimli24_hash_state_t *state);
+
+/**
+ * \brief Aborbs more input data into a GIMLI-24-HASH state.
+ *
+ * \param state Hash state to be updated.
+ * \param in Points to the input data to be absorbed into the state.
+ * \param inlen Length of the input data to be absorbed into the state.
+ *
+ * \sa gimli24_hash_init(), gimli24_hash_squeeze()
+ */
+void gimli24_hash_absorb
+    (gimli24_hash_state_t *state, const unsigned char *in,
+     unsigned long long inlen);
+
+/**
+ * \brief Squeezes output data from an GIMLI-24-HASH state.
+ *
+ * \param state Hash state to squeeze the output data from.
+ * \param out Points to the output buffer to receive the squeezed data.
+ * \param outlen Number of bytes of data to squeeze out of the state.
+ *
+ * \sa gimli24_hash_init(), gimli24_hash_absorb()
+ */
+void gimli24_hash_squeeze
+    (gimli24_hash_state_t *state, unsigned char *out,
+     unsigned long long outlen);
+
+/**
+ * \brief Returns the final hash value from a GIMLI-24-HASH hashing operation.
+ *
+ * \param state Hash state to be finalized.
+ * \param out Points to the output buffer to receive the hash value.
+ *
+ * \note This is a wrapper around gimli24_hash_squeeze() for a fixed length
+ * of GIMLI24_HASH_SIZE bytes.
+ *
+ * \sa gimli24_hash_init(), gimli24_hash_absorb()
+ */
+void gimli24_hash_finalize
+    (gimli24_hash_state_t *state, unsigned char *out);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

gimli/Implementations/crypto_hash/gimli24v1/rhys-avr/hash.c

+
+#include "gimli24.h"
+
+int crypto_hash
+    (unsigned char *out, const unsigned char *in, unsigned long long inlen)
+{
+    return gimli24_hash(out, in, inlen);
+}

gimli/Implementations/crypto_hash/gimli24v1/rhys-avr/internal-gimli24.c

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "internal-gimli24.h"
+
+#if !defined(__AVR__)
+
+/* Apply the SP-box to a specific column in the state array */
+#define GIMLI24_SP(s0, s4, s8) \
+    do { \
+        x = leftRotate24(s0); \
+        y = leftRotate9(s4); \
+        s4 = y ^ x ^ ((x | s8) << 1); \
+        s0 = s8 ^ y ^ ((x & y) << 3); \
+        s8 = x ^ (s8 << 1) ^ ((y & s8) << 2); \
+    } while (0)
+
+void gimli24_permute(uint32_t state[12])
+{
+    uint32_t s0, s1, s2, s3, s4,  s5;
+    uint32_t s6, s7, s8, s9, s10, s11;
+    uint32_t x, y;
+    unsigned round;
+
+    /* Load the state into local variables and convert from little-endian */
+#if defined(LW_UTIL_LITTLE_ENDIAN)
+    s0  = state[0];
+    s1  = state[1];
+    s2  = state[2];
+    s3  = state[3];
+    s4  = state[4];
+    s5  = state[5];
+    s6  = state[6];
+    s7  = state[7];
+    s8  = state[8];
+    s9  = state[9];
+    s10 = state[10];
+    s11 = state[11];
+#else
+    s0  = le_load_word32((const unsigned char *)(&(state[0])));
+    s1  = le_load_word32((const unsigned char *)(&(state[1])));
+    s2  = le_load_word32((const unsigned char *)(&(state[2])));
+    s3  = le_load_word32((const unsigned char *)(&(state[3])));
+    s4  = le_load_word32((const unsigned char *)(&(state[4])));
+    s5  = le_load_word32((const unsigned char *)(&(state[5])));
+    s6  = le_load_word32((const unsigned char *)(&(state[6])));
+    s7  = le_load_word32((const unsigned char *)(&(state[7])));
+    s8  = le_load_word32((const unsigned char *)(&(state[8])));
+    s9  = le_load_word32((const unsigned char *)(&(state[9])));
+    s10 = le_load_word32((const unsigned char *)(&(state[10])));
+    s11 = le_load_word32((const unsigned char *)(&(state[11])));
+#endif
+
+    /* Unroll and perform the rounds 4 at a time */
+    for (round = 24; round > 0; round -= 4) {
+        /* Round 0: SP-box, small swap, add round constant */
+        GIMLI24_SP(s0, s4, s8);
+        GIMLI24_SP(s1, s5, s9);
+        GIMLI24_SP(s2, s6, s10);
+        GIMLI24_SP(s3, s7, s11);
+        x = s0;
+        y = s2;
+        s0 = s1 ^ 0x9e377900U ^ round;
+        s1 = x;
+        s2 = s3;
+        s3 = y;
+
+        /* Round 1: SP-box only */
+        GIMLI24_SP(s0, s4, s8);
+        GIMLI24_SP(s1, s5, s9);
+        GIMLI24_SP(s2, s6, s10);
+        GIMLI24_SP(s3, s7, s11);
+
+        /* Round 2: SP-box, big swap */
+        GIMLI24_SP(s0, s4, s8);
+        GIMLI24_SP(s1, s5, s9);
+        GIMLI24_SP(s2, s6, s10);
+        GIMLI24_SP(s3, s7, s11);
+        x = s0;
+        y = s1;
+        s0 = s2;
+        s1 = s3;
+        s2 = x;
+        s3 = y;
+
+        /* Round 3: SP-box only */
+        GIMLI24_SP(s0, s4, s8);
+        GIMLI24_SP(s1, s5, s9);
+        GIMLI24_SP(s2, s6, s10);
+        GIMLI24_SP(s3, s7, s11);
+    }
+
+    /* Convert state to little-endian if the platform is not little-endian */
+#if defined(LW_UTIL_LITTLE_ENDIAN)
+    state[0]  = s0;
+    state[1]  = s1;
+    state[2]  = s2;
+    state[3]  = s3;
+    state[4]  = s4;
+    state[5]  = s5;
+    state[6]  = s6;
+    state[7]  = s7;
+    state[8]  = s8;
+    state[9]  = s9;
+    state[10] = s10;
+    state[11] = s11;
+#else
+    le_store_word32(((unsigned char *)(&(state[0]))),  s0);
+    le_store_word32(((unsigned char *)(&(state[1]))),  s1);
+    le_store_word32(((unsigned char *)(&(state[2]))),  s2);
+    le_store_word32(((unsigned char *)(&(state[3]))),  s3);
+    le_store_word32(((unsigned char *)(&(state[4]))),  s4);
+    le_store_word32(((unsigned char *)(&(state[5]))),  s5);
+    le_store_word32(((unsigned char *)(&(state[6]))),  s6);
+    le_store_word32(((unsigned char *)(&(state[7]))),  s7);
+    le_store_word32(((unsigned char *)(&(state[8]))),  s8);
+    le_store_word32(((unsigned char *)(&(state[9]))),  s9);
+    le_store_word32(((unsigned char *)(&(state[10]))), s10);
+    le_store_word32(((unsigned char *)(&(state[11]))), s11);
+#endif
+}
+
+#endif /* !__AVR__ */

gimli/Implementations/crypto_hash/gimli24v1/rhys-avr/internal-gimli24.h

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef LW_INTERNAL_GIMLI24_H
+#define LW_INTERNAL_GIMLI24_H
+
+#include "internal-util.h"
+
+/**
+ * \file internal-gimli24.h
+ * \brief Internal implementation of the GIMLI-24 permutation.
+ *
+ * References: https://gimli.cr.yp.to/
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * \brief Permutes the GIMLI-24 state.
+ *
+ * \param state The GIMLI-24 state to be permuted.
+ *
+ * The input and output \a state will be in little-endian byte order.
+ */
+void gimli24_permute(uint32_t state[12]);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

gimli/Implementations/crypto_hash/gimli24v1/rhys-avr/internal-util.h

+/*
+ * Copyright (C) 2020 Southern Storm Software, Pty Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef LW_INTERNAL_UTIL_H
+#define LW_INTERNAL_UTIL_H
+
+#include <stdint.h>
+
+/* Figure out how to inline functions using this C compiler */
+#if defined(__STDC__) && __STDC_VERSION__ >= 199901L
+#define STATIC_INLINE static inline
+#elif defined(__GNUC__) || defined(__clang__)
+#define STATIC_INLINE static __inline__
+#else
+#define STATIC_INLINE static
+#endif
+
+/* Try to figure out whether the CPU is little-endian or big-endian.
+ * May need to modify this to include new compiler-specific defines.
+ * Alternatively, define __LITTLE_ENDIAN__ or __BIG_ENDIAN__ in your
+ * compiler flags when you compile this library */
+#if defined(__x86_64) || defined(__x86_64__) || \
+    defined(__i386) || defined(__i386__) || \
+    defined(__AVR__) || defined(__arm) || defined(__arm__) || \
+    defined(_M_AMD64) || defined(_M_X64) || defined(_M_IX86) || \
+    defined(_M_IA64) || defined(_M_ARM) || defined(_M_ARM_FP) || \
+    (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == 1234) || \
+    defined(__LITTLE_ENDIAN__)
+#define LW_UTIL_LITTLE_ENDIAN 1
+#elif (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == 4321) || \
+    defined(__BIG_ENDIAN__)
+/* Big endian */
+#else
+#error "Cannot determine the endianess of this platform"
+#endif
+
+/* Helper macros to load and store values while converting endian-ness */
+
+/* Load a big-endian 32-bit word from a byte buffer */
+#define be_load_word32(ptr) \
+    ((((uint32_t)((ptr)[0])) << 24) | \
+     (((uint32_t)((ptr)[1])) << 16) | \
+     (((uint32_t)((ptr)[2])) << 8) | \
+      ((uint32_t)((ptr)[3])))
+
+/* Store a big-endian 32-bit word into a byte buffer */
+#define be_store_word32(ptr, x) \
+    do { \
+        uint32_t _x = (x); \
+        (ptr)[0] = (uint8_t)(_x >> 24); \
+        (ptr)[1] = (uint8_t)(_x >> 16); \
+        (ptr)[2] = (uint8_t)(_x >> 8); \
+        (ptr)[3] = (uint8_t)_x; \
+    } while (0)
+
+/* Load a little-endian 32-bit word from a byte buffer */
+#define le_load_word32(ptr) \
+    ((((uint32_t)((ptr)[3])) << 24) | \
+     (((uint32_t)((ptr)[2])) << 16) | \
+     (((uint32_t)((ptr)[1])) << 8) | \
+      ((uint32_t)((ptr)[0])))
+
+/* Store a little-endian 32-bit word into a byte buffer */
+#define le_store_word32(ptr, x) \
+    do { \
+        uint32_t _x = (x); \
+        (ptr)[0] = (uint8_t)_x; \
+        (ptr)[1] = (uint8_t)(_x >> 8); \
+        (ptr)[2] = (uint8_t)(_x >> 16); \
+        (ptr)[3] = (uint8_t)(_x >> 24); \
+    } while (0)
+
+/* Load a big-endian 64-bit word from a byte buffer */
+#define be_load_word64(ptr) \
+    ((((uint64_t)((ptr)[0])) << 56) | \
+     (((uint64_t)((ptr)[1])) << 48) | \
+     (((uint64_t)((ptr)[2])) << 40) | \
+     (((uint64_t)((ptr)[3])) << 32) | \
+     (((uint64_t)((ptr)[4])) << 24) | \
+     (((uint64_t)((ptr)[5])) << 16) | \
+     (((uint64_t)((ptr)[6])) << 8) | \
+      ((uint64_t)((ptr)[7])))
+
+/* Store a big-endian 64-bit word into a byte buffer */
+#define be_store_word64(ptr, x) \
+    do { \
+        uint64_t _x = (x); \
+        (ptr)[0] = (uint8_t)(_x >> 56); \
+        (ptr)[1] = (uint8_t)(_x >> 48); \
+        (ptr)[2] = (uint8_t)(_x >> 40); \
+        (ptr)[3] = (uint8_t)(_x >> 32); \
+        (ptr)[4] = (uint8_t)(_x >> 24); \
+        (ptr)[5] = (uint8_t)(_x >> 16); \
+        (ptr)[6] = (uint8_t)(_x >> 8); \
+        (ptr)[7] = (uint8_t)_x; \
+    } while (0)
+
+/* Load a little-endian 64-bit word from a byte buffer */
+#define le_load_word64(ptr) \
+    ((((uint64_t)((ptr)[7])) << 56) | \
+     (((uint64_t)((ptr)[6])) << 48) | \
+     (((uint64_t)((ptr)[5])) << 40) | \
+     (((uint64_t)((ptr)[4])) << 32) | \
+     (((uint64_t)((ptr)[3])) << 24) | \
+     (((uint64_t)((ptr)[2])) << 16) | \
+     (((uint64_t)((ptr)[1])) << 8) | \
+      ((uint64_t)((ptr)[0])))
+
+/* Store a little-endian 64-bit word into a byte buffer */
+#define le_store_word64(ptr, x) \
+    do { \
+        uint64_t _x = (x); \
+        (ptr)[0] = (uint8_t)_x; \
+        (ptr)[1] = (uint8_t)(_x >> 8); \
+        (ptr)[2] = (uint8_t)(_x >> 16); \
+        (ptr)[3] = (uint8_t)(_x >> 24); \
+        (ptr)[4] = (uint8_t)(_x >> 32); \
+        (ptr)[5] = (uint8_t)(_x >> 40); \
+        (ptr)[6] = (uint8_t)(_x >> 48); \
+        (ptr)[7] = (uint8_t)(_x >> 56); \
+    } while (0)
+
+/* Load a big-endian 16-bit word from a byte buffer */
+#define be_load_word16(ptr) \
+    ((((uint16_t)((ptr)[0])) << 8) | \
+      ((uint16_t)((ptr)[1])))
+
+/* Store a big-endian 16-bit word into a byte buffer */
+#define be_store_word16(ptr, x) \
+    do { \
+        uint16_t _x = (x); \
+        (ptr)[0] = (uint8_t)(_x >> 8); \
+        (ptr)[1] = (uint8_t)_x; \
+    } while (0)
+
+/* Load a little-endian 16-bit word from a byte buffer */
+#define le_load_word16(ptr) \
+    ((((uint16_t)((ptr)[1])) << 8) | \
+      ((uint16_t)((ptr)[0])))
+
+/* Store a little-endian 16-bit word into a byte buffer */
+#define le_store_word16(ptr, x) \
+    do { \
+        uint16_t _x = (x); \
+        (ptr)[0] = (uint8_t)_x; \
+        (ptr)[1] = (uint8_t)(_x >> 8); \
+    } while (0)
+
+/* XOR a source byte buffer against a destination */
+#define lw_xor_block(dest, src, len) \
+    do { \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src = (src); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            *_dest++ ^= *_src++; \
+            --_len; \
+        } \
+    } while (0)
+
+/* XOR two source byte buffers and put the result in a destination buffer */
+#define lw_xor_block_2_src(dest, src1, src2, len) \
+    do { \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src1 = (src1); \
+        const unsigned char *_src2 = (src2); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            *_dest++ = *_src1++ ^ *_src2++; \
+            --_len; \
+        } \
+    } while (0)
+
+/* XOR a source byte buffer against a destination and write to another
+ * destination at the same time */
+#define lw_xor_block_2_dest(dest2, dest, src, len) \
+    do { \
+        unsigned char *_dest2 = (dest2); \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src = (src); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            *_dest2++ = (*_dest++ ^= *_src++); \
+            --_len; \
+        } \
+    } while (0)
+
+/* XOR two byte buffers and write to a destination which at the same
+ * time copying the contents of src2 to dest2 */
+#define lw_xor_block_copy_src(dest2, dest, src1, src2, len) \
+    do { \
+        unsigned char *_dest2 = (dest2); \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src1 = (src1); \
+        const unsigned char *_src2 = (src2); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            unsigned char _temp = *_src2++; \
+            *_dest2++ = _temp; \
+            *_dest++ = *_src1++ ^ _temp; \
+            --_len; \
+        } \
+    } while (0)
+
+/* XOR a source byte buffer against a destination and write to another
+ * destination at the same time.  This version swaps the source value
+ * into the "dest" buffer */
+#define lw_xor_block_swap(dest2, dest, src, len) \
+    do { \
+        unsigned char *_dest2 = (dest2); \
+        unsigned char *_dest = (dest); \
+        const unsigned char *_src = (src); \
+        unsigned _len = (len); \
+        while (_len > 0) { \
+            unsigned char _temp = *_src++; \
+            *_dest2++ = *_dest ^ _temp; \
+            *_dest++ = _temp; \
+            --_len; \
+        } \
+    } while (0)
+
+/* Rotation functions need to be optimised for best performance on AVR.
+ * The most efficient rotations are where the number of bits is 1 or a
+ * multiple of 8, so we compose the efficient rotations to produce all
+ * other rotation counts of interest. */
+
+#if defined(__AVR__)
+#define LW_CRYPTO_ROTATE32_COMPOSED 1
+#else
+#define LW_CRYPTO_ROTATE32_COMPOSED 0
+#endif
+
+/* Rotation macros for 32-bit arguments */
+
+/* Generic left rotate */
+#define leftRotate(a, bits) \
+    (__extension__ ({ \
+        uint32_t _temp = (a); \
+        (_temp << (bits)) | (_temp >> (32 - (bits))); \
+    }))
+
+/* Generic right rotate */
+#define rightRotate(a, bits) \
+    (__extension__ ({ \
+        uint32_t _temp = (a); \
+        (_temp >> (bits)) | (_temp << (32 - (bits))); \
+    }))
+
+#if !LW_CRYPTO_ROTATE32_COMPOSED
+
+/* Left rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define leftRotate1(a)  (leftRotate((a), 1))
+#define leftRotate2(a)  (leftRotate((a), 2))
+#define leftRotate3(a)  (leftRotate((a), 3))
+#define leftRotate4(a)  (leftRotate((a), 4))
+#define leftRotate5(a)  (leftRotate((a), 5))
+#define leftRotate6(a)  (leftRotate((a), 6))
+#define leftRotate7(a)  (leftRotate((a), 7))
+#define leftRotate8(a)  (leftRotate((a), 8))
+#define leftRotate9(a)  (leftRotate((a), 9))
+#define leftRotate10(a) (leftRotate((a), 10))
+#define leftRotate11(a) (leftRotate((a), 11))
+#define leftRotate12(a) (leftRotate((a), 12))
+#define leftRotate13(a) (leftRotate((a), 13))
+#define leftRotate14(a) (leftRotate((a), 14))
+#define leftRotate15(a) (leftRotate((a), 15))
+#define leftRotate16(a) (leftRotate((a), 16))
+#define leftRotate17(a) (leftRotate((a), 17))
+#define leftRotate18(a) (leftRotate((a), 18))
+#define leftRotate19(a) (leftRotate((a), 19))
+#define leftRotate20(a) (leftRotate((a), 20))
+#define leftRotate21(a) (leftRotate((a), 21))
+#define leftRotate22(a) (leftRotate((a), 22))
+#define leftRotate23(a) (leftRotate((a), 23))
+#define leftRotate24(a) (leftRotate((a), 24))
+#define leftRotate25(a) (leftRotate((a), 25))
+#define leftRotate26(a) (leftRotate((a), 26))
+#define leftRotate27(a) (leftRotate((a), 27))
+#define leftRotate28(a) (leftRotate((a), 28))
+#define leftRotate29(a) (leftRotate((a), 29))
+#define leftRotate30(a) (leftRotate((a), 30))
+#define leftRotate31(a) (leftRotate((a), 31))
+
+/* Right rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define rightRotate1(a)  (rightRotate((a), 1))
+#define rightRotate2(a)  (rightRotate((a), 2))
+#define rightRotate3(a)  (rightRotate((a), 3))
+#define rightRotate4(a)  (rightRotate((a), 4))
+#define rightRotate5(a)  (rightRotate((a), 5))
+#define rightRotate6(a)  (rightRotate((a), 6))
+#define rightRotate7(a)  (rightRotate((a), 7))
+#define rightRotate8(a)  (rightRotate((a), 8))
+#define rightRotate9(a)  (rightRotate((a), 9))
+#define rightRotate10(a) (rightRotate((a), 10))
+#define rightRotate11(a) (rightRotate((a), 11))
+#define rightRotate12(a) (rightRotate((a), 12))
+#define rightRotate13(a) (rightRotate((a), 13))
+#define rightRotate14(a) (rightRotate((a), 14))
+#define rightRotate15(a) (rightRotate((a), 15))
+#define rightRotate16(a) (rightRotate((a), 16))
+#define rightRotate17(a) (rightRotate((a), 17))
+#define rightRotate18(a) (rightRotate((a), 18))
+#define rightRotate19(a) (rightRotate((a), 19))
+#define rightRotate20(a) (rightRotate((a), 20))
+#define rightRotate21(a) (rightRotate((a), 21))
+#define rightRotate22(a) (rightRotate((a), 22))
+#define rightRotate23(a) (rightRotate((a), 23))
+#define rightRotate24(a) (rightRotate((a), 24))
+#define rightRotate25(a) (rightRotate((a), 25))
+#define rightRotate26(a) (rightRotate((a), 26))
+#define rightRotate27(a) (rightRotate((a), 27))
+#define rightRotate28(a) (rightRotate((a), 28))
+#define rightRotate29(a) (rightRotate((a), 29))
+#define rightRotate30(a) (rightRotate((a), 30))
+#define rightRotate31(a) (rightRotate((a), 31))
+
+#else /* LW_CRYPTO_ROTATE32_COMPOSED */
+
+/* Composed rotation macros where 1 and 8 are fast, but others are slow */
+
+/* Left rotate by 1 */
+#define leftRotate1(a)  (leftRotate((a), 1))
+
+/* Left rotate by 2 */
+#define leftRotate2(a)  (leftRotate(leftRotate((a), 1), 1))
+
+/* Left rotate by 3 */
+#define leftRotate3(a)  (leftRotate(leftRotate(leftRotate((a), 1), 1), 1))
+
+/* Left rotate by 4 */
+#define leftRotate4(a)  (leftRotate(leftRotate(leftRotate(leftRotate((a), 1), 1), 1), 1))
+
+/* Left rotate by 5: Rotate left by 8, then right by 3 */
+#define leftRotate5(a)  (rightRotate(rightRotate(rightRotate(leftRotate((a), 8), 1), 1), 1))
+
+/* Left rotate by 6: Rotate left by 8, then right by 2 */
+#define leftRotate6(a)  (rightRotate(rightRotate(leftRotate((a), 8), 1), 1))
+
+/* Left rotate by 7: Rotate left by 8, then right by 1 */
+#define leftRotate7(a)  (rightRotate(leftRotate((a), 8), 1))
+
+/* Left rotate by 8 */
+#define leftRotate8(a)  (leftRotate((a), 8))
+
+/* Left rotate by 9: Rotate left by 8, then left by 1 */
+#define leftRotate9(a)  (leftRotate(leftRotate((a), 8), 1))
+
+/* Left rotate by 10: Rotate left by 8, then left by 2 */
+#define leftRotate10(a) (leftRotate(leftRotate(leftRotate((a), 8), 1), 1))
+
+/* Left rotate by 11: Rotate left by 8, then left by 3 */
+#define leftRotate11(a) (leftRotate(leftRotate(leftRotate(leftRotate((a), 8), 1), 1), 1))
+
+/* Left rotate by 12: Rotate left by 16, then right by 4 */
+#define leftRotate12(a) (rightRotate(rightRotate(rightRotate(rightRotate(leftRotate((a), 16), 1), 1), 1), 1))
+
+/* Left rotate by 13: Rotate left by 16, then right by 3 */
+#define leftRotate13(a) (rightRotate(rightRotate(rightRotate(leftRotate((a), 16), 1), 1), 1))
+
+/* Left rotate by 14: Rotate left by 16, then right by 2 */
+#define leftRotate14(a) (rightRotate(rightRotate(leftRotate((a), 16), 1), 1))
+
+/* Left rotate by 15: Rotate left by 16, then right by 1 */
+#define leftRotate15(a) (rightRotate(leftRotate((a), 16), 1))
+
+/* Left rotate by 16 */
+#define leftRotate16(a) (leftRotate((a), 16))
+
+/* Left rotate by 17: Rotate left by 16, then left by 1 */
+#define leftRotate17(a) (leftRotate(leftRotate((a), 16), 1))
+
+/* Left rotate by 18: Rotate left by 16, then left by 2 */
+#define leftRotate18(a) (leftRotate(leftRotate(leftRotate((a), 16), 1), 1))
+
+/* Left rotate by 19: Rotate left by 16, then left by 3 */
+#define leftRotate19(a) (leftRotate(leftRotate(leftRotate(leftRotate((a), 16), 1), 1), 1))
+
+/* Left rotate by 20: Rotate left by 16, then left by 4 */
+#define leftRotate20(a) (leftRotate(leftRotate(leftRotate(leftRotate(leftRotate((a), 16), 1), 1), 1), 1))
+
+/* Left rotate by 21: Rotate left by 24, then right by 3 */
+#define leftRotate21(a) (rightRotate(rightRotate(rightRotate(leftRotate((a), 24), 1), 1), 1))
+
+/* Left rotate by 22: Rotate left by 24, then right by 2 */
+#define leftRotate22(a) (rightRotate(rightRotate(leftRotate((a), 24), 1), 1))
+
+/* Left rotate by 23: Rotate left by 24, then right by 1 */
+#define leftRotate23(a) (rightRotate(leftRotate((a), 24), 1))
+
+/* Left rotate by 24 */
+#define leftRotate24(a) (leftRotate((a), 24))
+
+/* Left rotate by 25: Rotate left by 24, then left by 1 */
+#define leftRotate25(a) (leftRotate(leftRotate((a), 24), 1))
+
+/* Left rotate by 26: Rotate left by 24, then left by 2 */
+#define leftRotate26(a) (leftRotate(leftRotate(leftRotate((a), 24), 1), 1))
+
+/* Left rotate by 27: Rotate left by 24, then left by 3 */
+#define leftRotate27(a) (leftRotate(leftRotate(leftRotate(leftRotate((a), 24), 1), 1), 1))
+
+/* Left rotate by 28: Rotate right by 4 */
+#define leftRotate28(a) (rightRotate(rightRotate(rightRotate(rightRotate((a), 1), 1), 1), 1))
+
+/* Left rotate by 29: Rotate right by 3 */
+#define leftRotate29(a) (rightRotate(rightRotate(rightRotate((a), 1), 1), 1))
+
+/* Left rotate by 30: Rotate right by 2 */
+#define leftRotate30(a) (rightRotate(rightRotate((a), 1), 1))
+
+/* Left rotate by 31: Rotate right by 1 */
+#define leftRotate31(a) (rightRotate((a), 1))
+
+/* Define the 32-bit right rotations in terms of left rotations */
+#define rightRotate1(a)  (leftRotate31((a)))
+#define rightRotate2(a)  (leftRotate30((a)))
+#define rightRotate3(a)  (leftRotate29((a)))
+#define rightRotate4(a)  (leftRotate28((a)))
+#define rightRotate5(a)  (leftRotate27((a)))
+#define rightRotate6(a)  (leftRotate26((a)))
+#define rightRotate7(a)  (leftRotate25((a)))
+#define rightRotate8(a)  (leftRotate24((a)))
+#define rightRotate9(a)  (leftRotate23((a)))
+#define rightRotate10(a) (leftRotate22((a)))
+#define rightRotate11(a) (leftRotate21((a)))
+#define rightRotate12(a) (leftRotate20((a)))
+#define rightRotate13(a) (leftRotate19((a)))
+#define rightRotate14(a) (leftRotate18((a)))
+#define rightRotate15(a) (leftRotate17((a)))
+#define rightRotate16(a) (leftRotate16((a)))
+#define rightRotate17(a) (leftRotate15((a)))
+#define rightRotate18(a) (leftRotate14((a)))
+#define rightRotate19(a) (leftRotate13((a)))
+#define rightRotate20(a) (leftRotate12((a)))
+#define rightRotate21(a) (leftRotate11((a)))
+#define rightRotate22(a) (leftRotate10((a)))
+#define rightRotate23(a) (leftRotate9((a)))
+#define rightRotate24(a) (leftRotate8((a)))
+#define rightRotate25(a) (leftRotate7((a)))
+#define rightRotate26(a) (leftRotate6((a)))
+#define rightRotate27(a) (leftRotate5((a)))
+#define rightRotate28(a) (leftRotate4((a)))
+#define rightRotate29(a) (leftRotate3((a)))
+#define rightRotate30(a) (leftRotate2((a)))
+#define rightRotate31(a) (leftRotate1((a)))
+
+#endif /* LW_CRYPTO_ROTATE32_COMPOSED */
+
+/* Rotation macros for 64-bit arguments */
+
+/* Generic left rotate */
+#define leftRotate_64(a, bits) \
+    (__extension__ ({ \
+        uint64_t _temp = (a); \
+        (_temp << (bits)) | (_temp >> (64 - (bits))); \
+    }))
+
+/* Generic right rotate */
+#define rightRotate_64(a, bits) \
+    (__extension__ ({ \
+        uint64_t _temp = (a); \
+        (_temp >> (bits)) | (_temp << (64 - (bits))); \
+    }))
+
+/* Left rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define leftRotate1_64(a)  (leftRotate_64((a), 1))
+#define leftRotate2_64(a)  (leftRotate_64((a), 2))
+#define leftRotate3_64(a)  (leftRotate_64((a), 3))
+#define leftRotate4_64(a)  (leftRotate_64((a), 4))
+#define leftRotate5_64(a)  (leftRotate_64((a), 5))
+#define leftRotate6_64(a)  (leftRotate_64((a), 6))
+#define leftRotate7_64(a)  (leftRotate_64((a), 7))
+#define leftRotate8_64(a)  (leftRotate_64((a), 8))
+#define leftRotate9_64(a)  (leftRotate_64((a), 9))
+#define leftRotate10_64(a) (leftRotate_64((a), 10))
+#define leftRotate11_64(a) (leftRotate_64((a), 11))
+#define leftRotate12_64(a) (leftRotate_64((a), 12))
+#define leftRotate13_64(a) (leftRotate_64((a), 13))
+#define leftRotate14_64(a) (leftRotate_64((a), 14))
+#define leftRotate15_64(a) (leftRotate_64((a), 15))
+#define leftRotate16_64(a) (leftRotate_64((a), 16))
+#define leftRotate17_64(a) (leftRotate_64((a), 17))
+#define leftRotate18_64(a) (leftRotate_64((a), 18))
+#define leftRotate19_64(a) (leftRotate_64((a), 19))
+#define leftRotate20_64(a) (leftRotate_64((a), 20))
+#define leftRotate21_64(a) (leftRotate_64((a), 21))
+#define leftRotate22_64(a) (leftRotate_64((a), 22))
+#define leftRotate23_64(a) (leftRotate_64((a), 23))
+#define leftRotate24_64(a) (leftRotate_64((a), 24))
+#define leftRotate25_64(a) (leftRotate_64((a), 25))
+#define leftRotate26_64(a) (leftRotate_64((a), 26))
+#define leftRotate27_64(a) (leftRotate_64((a), 27))
+#define leftRotate28_64(a) (leftRotate_64((a), 28))
+#define leftRotate29_64(a) (leftRotate_64((a), 29))
+#define leftRotate30_64(a) (leftRotate_64((a), 30))
+#define leftRotate31_64(a) (leftRotate_64((a), 31))
+#define leftRotate32_64(a) (leftRotate_64((a), 32))
+#define leftRotate33_64(a) (leftRotate_64((a), 33))
+#define leftRotate34_64(a) (leftRotate_64((a), 34))
+#define leftRotate35_64(a) (leftRotate_64((a), 35))
+#define leftRotate36_64(a) (leftRotate_64((a), 36))
+#define leftRotate37_64(a) (leftRotate_64((a), 37))
+#define leftRotate38_64(a) (leftRotate_64((a), 38))
+#define leftRotate39_64(a) (leftRotate_64((a), 39))
+#define leftRotate40_64(a) (leftRotate_64((a), 40))
+#define leftRotate41_64(a) (leftRotate_64((a), 41))
+#define leftRotate42_64(a) (leftRotate_64((a), 42))
+#define leftRotate43_64(a) (leftRotate_64((a), 43))
+#define leftRotate44_64(a) (leftRotate_64((a), 44))
+#define leftRotate45_64(a) (leftRotate_64((a), 45))
+#define leftRotate46_64(a) (leftRotate_64((a), 46))
+#define leftRotate47_64(a) (leftRotate_64((a), 47))
+#define leftRotate48_64(a) (leftRotate_64((a), 48))
+#define leftRotate49_64(a) (leftRotate_64((a), 49))
+#define leftRotate50_64(a) (leftRotate_64((a), 50))
+#define leftRotate51_64(a) (leftRotate_64((a), 51))
+#define leftRotate52_64(a) (leftRotate_64((a), 52))
+#define leftRotate53_64(a) (leftRotate_64((a), 53))
+#define leftRotate54_64(a) (leftRotate_64((a), 54))
+#define leftRotate55_64(a) (leftRotate_64((a), 55))
+#define leftRotate56_64(a) (leftRotate_64((a), 56))
+#define leftRotate57_64(a) (leftRotate_64((a), 57))
+#define leftRotate58_64(a) (leftRotate_64((a), 58))
+#define leftRotate59_64(a) (leftRotate_64((a), 59))
+#define leftRotate60_64(a) (leftRotate_64((a), 60))
+#define leftRotate61_64(a) (leftRotate_64((a), 61))
+#define leftRotate62_64(a) (leftRotate_64((a), 62))
+#define leftRotate63_64(a) (leftRotate_64((a), 63))
+
+/* Right rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define rightRotate1_64(a)  (rightRotate_64((a), 1))
+#define rightRotate2_64(a)  (rightRotate_64((a), 2))
+#define rightRotate3_64(a)  (rightRotate_64((a), 3))
+#define rightRotate4_64(a)  (rightRotate_64((a), 4))
+#define rightRotate5_64(a)  (rightRotate_64((a), 5))
+#define rightRotate6_64(a)  (rightRotate_64((a), 6))
+#define rightRotate7_64(a)  (rightRotate_64((a), 7))
+#define rightRotate8_64(a)  (rightRotate_64((a), 8))
+#define rightRotate9_64(a)  (rightRotate_64((a), 9))
+#define rightRotate10_64(a) (rightRotate_64((a), 10))
+#define rightRotate11_64(a) (rightRotate_64((a), 11))
+#define rightRotate12_64(a) (rightRotate_64((a), 12))
+#define rightRotate13_64(a) (rightRotate_64((a), 13))
+#define rightRotate14_64(a) (rightRotate_64((a), 14))
+#define rightRotate15_64(a) (rightRotate_64((a), 15))
+#define rightRotate16_64(a) (rightRotate_64((a), 16))
+#define rightRotate17_64(a) (rightRotate_64((a), 17))
+#define rightRotate18_64(a) (rightRotate_64((a), 18))
+#define rightRotate19_64(a) (rightRotate_64((a), 19))
+#define rightRotate20_64(a) (rightRotate_64((a), 20))
+#define rightRotate21_64(a) (rightRotate_64((a), 21))
+#define rightRotate22_64(a) (rightRotate_64((a), 22))
+#define rightRotate23_64(a) (rightRotate_64((a), 23))
+#define rightRotate24_64(a) (rightRotate_64((a), 24))
+#define rightRotate25_64(a) (rightRotate_64((a), 25))
+#define rightRotate26_64(a) (rightRotate_64((a), 26))
+#define rightRotate27_64(a) (rightRotate_64((a), 27))
+#define rightRotate28_64(a) (rightRotate_64((a), 28))
+#define rightRotate29_64(a) (rightRotate_64((a), 29))
+#define rightRotate30_64(a) (rightRotate_64((a), 30))
+#define rightRotate31_64(a) (rightRotate_64((a), 31))
+#define rightRotate32_64(a) (rightRotate_64((a), 32))
+#define rightRotate33_64(a) (rightRotate_64((a), 33))
+#define rightRotate34_64(a) (rightRotate_64((a), 34))
+#define rightRotate35_64(a) (rightRotate_64((a), 35))
+#define rightRotate36_64(a) (rightRotate_64((a), 36))
+#define rightRotate37_64(a) (rightRotate_64((a), 37))
+#define rightRotate38_64(a) (rightRotate_64((a), 38))
+#define rightRotate39_64(a) (rightRotate_64((a), 39))
+#define rightRotate40_64(a) (rightRotate_64((a), 40))
+#define rightRotate41_64(a) (rightRotate_64((a), 41))
+#define rightRotate42_64(a) (rightRotate_64((a), 42))
+#define rightRotate43_64(a) (rightRotate_64((a), 43))
+#define rightRotate44_64(a) (rightRotate_64((a), 44))
+#define rightRotate45_64(a) (rightRotate_64((a), 45))
+#define rightRotate46_64(a) (rightRotate_64((a), 46))
+#define rightRotate47_64(a) (rightRotate_64((a), 47))
+#define rightRotate48_64(a) (rightRotate_64((a), 48))
+#define rightRotate49_64(a) (rightRotate_64((a), 49))
+#define rightRotate50_64(a) (rightRotate_64((a), 50))
+#define rightRotate51_64(a) (rightRotate_64((a), 51))
+#define rightRotate52_64(a) (rightRotate_64((a), 52))
+#define rightRotate53_64(a) (rightRotate_64((a), 53))
+#define rightRotate54_64(a) (rightRotate_64((a), 54))
+#define rightRotate55_64(a) (rightRotate_64((a), 55))
+#define rightRotate56_64(a) (rightRotate_64((a), 56))
+#define rightRotate57_64(a) (rightRotate_64((a), 57))
+#define rightRotate58_64(a) (rightRotate_64((a), 58))
+#define rightRotate59_64(a) (rightRotate_64((a), 59))
+#define rightRotate60_64(a) (rightRotate_64((a), 60))
+#define rightRotate61_64(a) (rightRotate_64((a), 61))
+#define rightRotate62_64(a) (rightRotate_64((a), 62))
+#define rightRotate63_64(a) (rightRotate_64((a), 63))
+
+/* Rotate a 16-bit value left by a number of bits */
+#define leftRotate_16(a, bits) \
+    (__extension__ ({ \
+        uint16_t _temp = (a); \
+        (_temp << (bits)) | (_temp >> (16 - (bits))); \
+    }))
+
+/* Rotate a 16-bit value right by a number of bits */
+#define rightRotate_16(a, bits) \
+    (__extension__ ({ \
+        uint16_t _temp = (a); \
+        (_temp >> (bits)) | (_temp << (16 - (bits))); \
+    }))
+
+/* Left rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define leftRotate1_16(a)  (leftRotate_16((a), 1))
+#define leftRotate2_16(a)  (leftRotate_16((a), 2))
+#define leftRotate3_16(a)  (leftRotate_16((a), 3))
+#define leftRotate4_16(a)  (leftRotate_16((a), 4))
+#define leftRotate5_16(a)  (leftRotate_16((a), 5))
+#define leftRotate6_16(a)  (leftRotate_16((a), 6))
+#define leftRotate7_16(a)  (leftRotate_16((a), 7))
+#define leftRotate8_16(a)  (leftRotate_16((a), 8))
+#define leftRotate9_16(a)  (leftRotate_16((a), 9))
+#define leftRotate10_16(a) (leftRotate_16((a), 10))
+#define leftRotate11_16(a) (leftRotate_16((a), 11))
+#define leftRotate12_16(a) (leftRotate_16((a), 12))
+#define leftRotate13_16(a) (leftRotate_16((a), 13))
+#define leftRotate14_16(a) (leftRotate_16((a), 14))
+#define leftRotate15_16(a) (leftRotate_16((a), 15))
+
+/* Right rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define rightRotate1_16(a)  (rightRotate_16((a), 1))
+#define rightRotate2_16(a)  (rightRotate_16((a), 2))
+#define rightRotate3_16(a)  (rightRotate_16((a), 3))
+#define rightRotate4_16(a)  (rightRotate_16((a), 4))
+#define rightRotate5_16(a)  (rightRotate_16((a), 5))
+#define rightRotate6_16(a)  (rightRotate_16((a), 6))
+#define rightRotate7_16(a)  (rightRotate_16((a), 7))
+#define rightRotate8_16(a)  (rightRotate_16((a), 8))
+#define rightRotate9_16(a)  (rightRotate_16((a), 9))
+#define rightRotate10_16(a) (rightRotate_16((a), 10))
+#define rightRotate11_16(a) (rightRotate_16((a), 11))
+#define rightRotate12_16(a) (rightRotate_16((a), 12))
+#define rightRotate13_16(a) (rightRotate_16((a), 13))
+#define rightRotate14_16(a) (rightRotate_16((a), 14))
+#define rightRotate15_16(a) (rightRotate_16((a), 15))
+
+/* Rotate an 8-bit value left by a number of bits */
+#define leftRotate_8(a, bits) \
+    (__extension__ ({ \
+        uint8_t _temp = (a); \
+        (_temp << (bits)) | (_temp >> (8 - (bits))); \
+    }))
+
+/* Rotate an 8-bit value right by a number of bits */
+#define rightRotate_8(a, bits) \
+    (__extension__ ({ \
+        uint8_t _temp = (a); \
+        (_temp >> (bits)) | (_temp << (8 - (bits))); \
+    }))
+
+/* Left rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define leftRotate1_8(a)  (leftRotate_8((a), 1))
+#define leftRotate2_8(a)  (leftRotate_8((a), 2))
+#define leftRotate3_8(a)  (leftRotate_8((a), 3))
+#define leftRotate4_8(a)  (leftRotate_8((a), 4))
+#define leftRotate5_8(a)  (leftRotate_8((a), 5))
+#define leftRotate6_8(a)  (leftRotate_8((a), 6))
+#define leftRotate7_8(a)  (leftRotate_8((a), 7))
+
+/* Right rotate by a specific number of bits.  These macros may be replaced
+ * with more efficient ones on platforms that lack a barrel shifter */
+#define rightRotate1_8(a)  (rightRotate_8((a), 1))
+#define rightRotate2_8(a)  (rightRotate_8((a), 2))
+#define rightRotate3_8(a)  (rightRotate_8((a), 3))
+#define rightRotate4_8(a)  (rightRotate_8((a), 4))
+#define rightRotate5_8(a)  (rightRotate_8((a), 5))
+#define rightRotate6_8(a)  (rightRotate_8((a), 6))
+#define rightRotate7_8(a)  (rightRotate_8((a), 7))
+
+#endif