mirror_zfs/module/icp/algs/blake3/blake3_x86-64.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or https://opensource.org/licenses/CDDL-1.0.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */

/*
 * Copyright (c) 2021-2022 Tino Reichardt <milky-zfs@mcmilk.de>
 */

#include "blake3_impl.h"

#if defined(__aarch64__) || \
	(defined(__x86_64) && defined(HAVE_SSE2)) || \
	(defined(__PPC64__) && defined(__LITTLE_ENDIAN__))

extern void zfs_blake3_compress_in_place_sse2(uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags);

extern void zfs_blake3_compress_xof_sse2(const uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags, uint8_t out[64]);

extern void zfs_blake3_hash_many_sse2(const uint8_t * const *inputs,
    size_t num_inputs, size_t blocks, const uint32_t key[8],
    uint64_t counter, boolean_t increment_counter, uint8_t flags,
    uint8_t flags_start, uint8_t flags_end, uint8_t *out);

static void blake3_compress_in_place_sse2(uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags) {
	kfpu_begin();
	zfs_blake3_compress_in_place_sse2(cv, block, block_len, counter,
	    flags);
	kfpu_end();
}

static void blake3_compress_xof_sse2(const uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags, uint8_t out[64]) {
	kfpu_begin();
	zfs_blake3_compress_xof_sse2(cv, block, block_len, counter, flags,
	    out);
	kfpu_end();
}

static void blake3_hash_many_sse2(const uint8_t * const *inputs,
    size_t num_inputs, size_t blocks, const uint32_t key[8],
    uint64_t counter, boolean_t increment_counter, uint8_t flags,
    uint8_t flags_start, uint8_t flags_end, uint8_t *out) {
	kfpu_begin();
	zfs_blake3_hash_many_sse2(inputs, num_inputs, blocks, key, counter,
	    increment_counter, flags, flags_start, flags_end, out);
	kfpu_end();
}

static boolean_t blake3_is_sse2_supported(void)
{
#if defined(__x86_64)
	return (kfpu_allowed() && zfs_sse2_available());
#elif defined(__PPC64__) && defined(__linux__)
	return (kfpu_allowed() && zfs_vsx_available());
#else
	return (kfpu_allowed());
#endif
}

const blake3_ops_t blake3_sse2_impl = {
	.compress_in_place = blake3_compress_in_place_sse2,
	.compress_xof = blake3_compress_xof_sse2,
	.hash_many = blake3_hash_many_sse2,
	.is_supported = blake3_is_sse2_supported,
	.degree = 4,
	.name = "sse2"
};
#endif

#if defined(__aarch64__) || \
	(defined(__x86_64) && defined(HAVE_SSE2)) || \
	(defined(__PPC64__) && defined(__LITTLE_ENDIAN__))

extern void zfs_blake3_compress_in_place_sse41(uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags);

extern void zfs_blake3_compress_xof_sse41(const uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags, uint8_t out[64]);

extern void zfs_blake3_hash_many_sse41(const uint8_t * const *inputs,
    size_t num_inputs, size_t blocks, const uint32_t key[8],
    uint64_t counter, boolean_t increment_counter, uint8_t flags,
    uint8_t flags_start, uint8_t flags_end, uint8_t *out);

static void blake3_compress_in_place_sse41(uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags) {
	kfpu_begin();
	zfs_blake3_compress_in_place_sse41(cv, block, block_len, counter,
	    flags);
	kfpu_end();
}

static void blake3_compress_xof_sse41(const uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags, uint8_t out[64]) {
	kfpu_begin();
	zfs_blake3_compress_xof_sse41(cv, block, block_len, counter, flags,
	    out);
	kfpu_end();
}

static void blake3_hash_many_sse41(const uint8_t * const *inputs,
    size_t num_inputs, size_t blocks, const uint32_t key[8],
    uint64_t counter, boolean_t increment_counter, uint8_t flags,
    uint8_t flags_start, uint8_t flags_end, uint8_t *out) {
	kfpu_begin();
	zfs_blake3_hash_many_sse41(inputs, num_inputs, blocks, key, counter,
	    increment_counter, flags, flags_start, flags_end, out);
	kfpu_end();
}

static boolean_t blake3_is_sse41_supported(void)
{
#if defined(__x86_64)
	return (kfpu_allowed() && zfs_sse4_1_available());
#elif defined(__PPC64__) && defined(__linux__)
	return (kfpu_allowed() && zfs_vsx_available());
#else
	return (kfpu_allowed());
#endif
}

const blake3_ops_t blake3_sse41_impl = {
	.compress_in_place = blake3_compress_in_place_sse41,
	.compress_xof = blake3_compress_xof_sse41,
	.hash_many = blake3_hash_many_sse41,
	.is_supported = blake3_is_sse41_supported,
	.degree = 4,
	.name = "sse41"
};
#endif

#if defined(__x86_64) && defined(HAVE_SSE4_1) && defined(HAVE_AVX2)
extern void zfs_blake3_hash_many_avx2(const uint8_t * const *inputs,
    size_t num_inputs, size_t blocks, const uint32_t key[8],
    uint64_t counter, boolean_t increment_counter, uint8_t flags,
    uint8_t flags_start, uint8_t flags_end, uint8_t *out);

static void blake3_hash_many_avx2(const uint8_t * const *inputs,
    size_t num_inputs, size_t blocks, const uint32_t key[8],
    uint64_t counter, boolean_t increment_counter, uint8_t flags,
    uint8_t flags_start, uint8_t flags_end, uint8_t *out) {
	kfpu_begin();
	zfs_blake3_hash_many_avx2(inputs, num_inputs, blocks, key, counter,
	    increment_counter, flags, flags_start, flags_end, out);
	kfpu_end();
}

static boolean_t blake3_is_avx2_supported(void)
{
	return (kfpu_allowed() && zfs_sse4_1_available() &&
	    zfs_avx2_available());
}

const blake3_ops_t blake3_avx2_impl = {
	.compress_in_place = blake3_compress_in_place_sse41,
	.compress_xof = blake3_compress_xof_sse41,
	.hash_many = blake3_hash_many_avx2,
	.is_supported = blake3_is_avx2_supported,
	.degree = 8,
	.name = "avx2"
};
#endif

#if defined(__x86_64) && defined(HAVE_AVX512F) && defined(HAVE_AVX512VL)
extern void zfs_blake3_compress_in_place_avx512(uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags);

extern void zfs_blake3_compress_xof_avx512(const uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags, uint8_t out[64]);

extern void zfs_blake3_hash_many_avx512(const uint8_t * const *inputs,
    size_t num_inputs, size_t blocks, const uint32_t key[8],
    uint64_t counter, boolean_t increment_counter, uint8_t flags,
    uint8_t flags_start, uint8_t flags_end, uint8_t *out);

static void blake3_compress_in_place_avx512(uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags) {
	kfpu_begin();
	zfs_blake3_compress_in_place_avx512(cv, block, block_len, counter,
	    flags);
	kfpu_end();
}

static void blake3_compress_xof_avx512(const uint32_t cv[8],
    const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
    uint64_t counter, uint8_t flags, uint8_t out[64]) {
	kfpu_begin();
	zfs_blake3_compress_xof_avx512(cv, block, block_len, counter, flags,
	    out);
	kfpu_end();
}

static void blake3_hash_many_avx512(const uint8_t * const *inputs,
    size_t num_inputs, size_t blocks, const uint32_t key[8],
    uint64_t counter, boolean_t increment_counter, uint8_t flags,
    uint8_t flags_start, uint8_t flags_end, uint8_t *out) {
	kfpu_begin();
	zfs_blake3_hash_many_avx512(inputs, num_inputs, blocks, key, counter,
	    increment_counter, flags, flags_start, flags_end, out);
	kfpu_end();
}

static boolean_t blake3_is_avx512_supported(void)
{
	return (kfpu_allowed() && zfs_avx512f_available() &&
	    zfs_avx512vl_available());
}

const blake3_ops_t blake3_avx512_impl = {
	.compress_in_place = blake3_compress_in_place_avx512,
	.compress_xof = blake3_compress_xof_avx512,
	.hash_many = blake3_hash_many_avx512,
	.is_supported = blake3_is_avx512_supported,
	.degree = 16,
	.name = "avx512"
};
#endif