diff options
author | Damjan Marion <damarion@cisco.com> | 2021-11-28 22:57:15 +0100 |
---|---|---|
committer | Ole Tr�an <otroan@employees.org> | 2021-12-14 11:06:49 +0000 |
commit | 2e5921b08b167793eacf3085a98344ad00afb5ac (patch) | |
tree | 5804720826747d06a2af943a7169a0029ef41a77 /src/vppinfra | |
parent | f622bf4279466eae6874dead7a56b59669783490 (diff) |
vppinfra: toeplitz hash
Type: feature
Change-Id: I0ccf280c4b769ee624cfb780146b847a0e8e4942
Signed-off-by: Damjan Marion <dmarion@me.com>
Diffstat (limited to 'src/vppinfra')
-rw-r--r-- | src/vppinfra/CMakeLists.txt | 3 | ||||
-rw-r--r-- | src/vppinfra/vector/test/toeplitz.c | 336 | ||||
-rw-r--r-- | src/vppinfra/vector/toeplitz.c | 122 | ||||
-rw-r--r-- | src/vppinfra/vector/toeplitz.h | 218 | ||||
-rw-r--r-- | src/vppinfra/vector_avx512.h | 25 |
5 files changed, 704 insertions, 0 deletions
diff --git a/src/vppinfra/CMakeLists.txt b/src/vppinfra/CMakeLists.txt index 14eb81fb1dd..1a43f25b2c6 100644 --- a/src/vppinfra/CMakeLists.txt +++ b/src/vppinfra/CMakeLists.txt @@ -99,6 +99,7 @@ set(VPPINFRA_SRCS valloc.c vec.c vector.c + vector/toeplitz.c cJSON.c ) @@ -196,6 +197,7 @@ set(VPPINFRA_HEADERS vector/index_to_ptr.h vector/ip_csum.h vector/mask_compare.h + vector/toeplitz.h vector.h vector_neon.h vector_sse42.h @@ -280,6 +282,7 @@ set(test_files vector/test/mask_compare.c vector/test/memcpy_x86_64.c vector/test/sha2.c + vector/test/toeplitz.c ) add_vpp_executable(test_vector_funcs diff --git a/src/vppinfra/vector/test/toeplitz.c b/src/vppinfra/vector/test/toeplitz.c new file mode 100644 index 00000000000..0efc4239a13 --- /dev/null +++ b/src/vppinfra/vector/test/toeplitz.c @@ -0,0 +1,336 @@ +/* SPDX-License-Identifier: Apache-2.0 + * Copyright(c) 2021 Cisco Systems, Inc. + */ + +#include <vppinfra/format.h> +#include <vppinfra/vector/test/test.h> +#include <vppinfra/vector/toeplitz.h> + +/* secret key and test cases taken from: + * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/verifying-the-rss-hash-calculation + */ + +typedef struct +{ + u32 sip, dip; + u16 sport, dport; +} __clib_packed ip4_key_t; + +typedef struct +{ + ip4_key_t key; + u32 hash_2t, hash_4t; +} ip4_test_t; + +typedef struct +{ + u16 sip[8], dip[8]; + u16 sport, dport; +} __clib_packed ip6_key_t; + +typedef struct +{ + ip6_key_t key; + u32 hash_2t, hash_4t; +} ip6_test_t; + +#define N_IP4_TESTS 5 +#define N_IP6_TESTS 3 +#define N_LENGTH_TESTS 240 + +#ifndef CLIB_MARCH_VARIANT +#define _IP4(a, b, c, d) ((d) << 24 | (c) << 16 | (b) << 8 | (a)) +#define _IP6(a, b, c, d, e, f, g, h) \ + { \ + (u16) ((a) << 8) | (u8) ((a) >> 8), (u16) ((b) << 8) | (u8) ((b) >> 8), \ + (u16) ((c) << 8) | (u8) ((c) >> 8), (u16) ((d) << 8) | (u8) ((d) >> 8), \ + (u16) ((e) << 8) | (u8) ((e) >> 8), (u16) ((f) << 8) | (u8) ((f) >> 8), \ + (u16) ((g) << 8) | (u8) ((g) >> 8), (u16) ((h) << 8) | (u8) ((h) >> 8), \ + } +#define _PORT(a) ((a) >> 8 | (((a) &0xff) << 8)) + +const ip4_test_t ip4_tests[N_IP4_TESTS] = { + /* ipv4 tests */ + { + .key.sip = _IP4 (66, 9, 149, 187), + .key.dip = _IP4 (161, 142, 100, 80), + .key.sport = _PORT (2794), + .key.dport = _PORT (1766), + .hash_2t = 0x323e8fc2, + .hash_4t = 0x51ccc178, + }, + { + .key.sip = _IP4 (199, 92, 111, 2), + .key.dip = _IP4 (65, 69, 140, 83), + .key.sport = _PORT (14230), + .key.dport = _PORT (4739), + .hash_2t = 0xd718262a, + .hash_4t = 0xc626b0ea, + }, + { + .key.sip = _IP4 (24, 19, 198, 95), + .key.dip = _IP4 (12, 22, 207, 184), + .key.sport = _PORT (12898), + .key.dport = _PORT (38024), + .hash_2t = 0xd2d0a5de, + .hash_4t = 0x5c2b394a, + }, + { + .key.sip = _IP4 (38, 27, 205, 30), + .key.dip = _IP4 (209, 142, 163, 6), + .key.sport = _PORT (48228), + .key.dport = _PORT (2217), + .hash_2t = 0x82989176, + .hash_4t = 0xafc7327f, + }, + { + .key.sip = _IP4 (153, 39, 163, 191), + .key.dip = _IP4 (202, 188, 127, 2), + .key.sport = _PORT (44251), + .key.dport = _PORT (1303), + .hash_2t = 0x5d1809c5, + .hash_4t = 0x10e828a2, + } +}; + +const ip6_test_t ip6_tests[N_IP6_TESTS] = { + { + .key.sip = _IP6 (0x3ffe, 0x2501, 0x200, 0x1fff, 0, 0, 0, 7), + .key.dip = _IP6 (0x3ffe, 0x2501, 0x200, 3, 0, 0, 0, 1), + .key.sport = _PORT (2794), + .key.dport = _PORT (1766), + .hash_2t = 0x2cc18cd5, + .hash_4t = 0x40207d3d, + }, + { + .key.sip = _IP6 (0x3ffe, 0x501, 8, 0, 0x260, 0x97ff, 0xfe40, 0xefab), + .key.dip = _IP6 (0xff02, 0, 0, 0, 0, 0, 0, 1), + .key.sport = _PORT (14230), + .key.dport = _PORT (4739), + .hash_2t = 0x0f0c461c, + .hash_4t = 0xdde51bbf, + }, + { + .key.sip = _IP6 (0x3ffe, 0x1900, 0x4545, 3, 0x200, 0xf8ff, 0xfe21, 0x67cf), + .key.dip = _IP6 (0xfe80, 0, 0, 0, 0x200, 0xf8ff, 0xfe21, 0x67cf), + .key.sport = _PORT (44251), + .key.dport = _PORT (38024), + .hash_2t = 0x4b61e985, + .hash_4t = 0x02d1feef, + } +}; + +const u32 length_test_hashes[N_LENGTH_TESTS] = { + 0x00000000, 0x00000000, 0x2b6d12ad, 0x9de4446e, 0x061f00bf, 0xad7ed8f7, + 0x4bc7b068, 0x231fc545, 0xdbd97a33, 0xcdab29e7, 0x2d665c0c, 0x31e28ed7, + 0x14e19218, 0x5aa89f0f, 0xd47de07f, 0x355ec712, 0x7e1cbfc0, 0xf84de19d, + 0xbcf66bd3, 0x104086c6, 0x71900b34, 0xcd2f9819, 0xeae68ebb, 0x54d63b4c, + 0x5f865a2c, 0x9d6ded08, 0xe00b0912, 0x3fcf07a6, 0x3bd9ca93, 0x3f4f3bbb, + 0xd0b82624, 0xa28a08e1, 0xa585969f, 0x0c8f4a71, 0x5dce7bdd, 0x4fcf2a6d, + 0x91c89ae9, 0xbef8a24d, 0x8e3d30fe, 0xc8027848, 0xc1e7e513, 0xa12bd3d9, + 0x46700bb4, 0xc6339dab, 0x970805ad, 0xfcb50ac8, 0xc6db4f44, 0x792e2987, + 0xacfb7836, 0xa25ec529, 0x957d7beb, 0x6732809a, 0x891836ed, 0xeefb83b2, + 0xca96b40b, 0x93fd5abd, 0x9076f922, 0x59adb4eb, 0x9705aafb, 0x282719b1, + 0xdda9cb8a, 0x3f499131, 0x47491130, 0x30ef0759, 0xad1cf855, 0x428aa312, + 0x4200240a, 0x71a72857, 0x16b30c36, 0x10cca9a3, 0x166f091e, 0x30e00560, + 0x8acd20ba, 0xfa633d76, 0x0fe32eb7, 0xdcc0122f, 0x20aa8ab0, 0x62b2a9af, + 0x7a6c80a6, 0x27e87268, 0x95b797a8, 0x25d18ccd, 0x68a7fb00, 0xc54bcdad, + 0x3bd0e717, 0xf0df54c9, 0x780daadf, 0x7b435605, 0x150c1e10, 0x8a892e54, + 0x9d27cb25, 0xe23383a5, 0x57aac408, 0x83b8abf8, 0x560f33af, 0xd5cb3307, + 0x79ae8edc, 0x9b127665, 0x320f18bd, 0x385d636b, 0xbd1b2dbf, 0x97679888, + 0x738894a4, 0xeba2afb0, 0xfa7c2d50, 0xb6741aa1, 0x28922bba, 0x7783242b, + 0xa694cca2, 0xa32781c0, 0x696cd670, 0xa714d72f, 0xea34d35a, 0xc5aed81e, + 0x0438433a, 0xc1939ab2, 0xb51c123a, 0x121426b9, 0x1add93ba, 0x50c56b6a, + 0x7e90902a, 0xae3abd85, 0x2f7a0088, 0xb45cf6f9, 0x80070094, 0x8bd46467, + 0xdfd1b762, 0x0bb25856, 0x48eefe84, 0x0989dbb9, 0xfc32472b, 0x965fec6b, + 0x5a256bd0, 0x6df7127a, 0x7856d0d6, 0xedc82bd3, 0x1b563b96, 0xc73eace7, + 0xba4c0a93, 0xdfd6dd97, 0x923c41db, 0x14926ca6, 0x22e52ab1, 0x22852a66, + 0x79606b9c, 0xb0f22b23, 0xb46354ba, 0x9c3cd931, 0x03a92bd6, 0x84000834, + 0x5425df65, 0xf4dd3fc9, 0x391cc873, 0xa560b52e, 0x828037d9, 0x31323dd5, + 0x5c6e3147, 0x28e21f85, 0xa431eb51, 0xf468c4a3, 0x9bea1d2e, 0x43d9109c, + 0x5bb9b081, 0xe0825675, 0xc9c92591, 0xd29fc812, 0x03136bc9, 0x5e005a1f, + 0x6d821ed8, 0x3f0bfcc4, 0x24774162, 0x893bde94, 0x6475efea, 0x6711538e, + 0xc4755f6d, 0x9425ebe2, 0xacf471b4, 0xb947ab0c, 0x1f78c455, 0x372b3ed7, + 0xb3ec24d7, 0x18c4459f, 0xa8ff3695, 0xe4aa2b85, 0x8a52ad7e, 0xe05e8177, + 0x7aa348ed, 0x3e4ac6aa, 0x17dcf8a5, 0x93b933b0, 0x8f7413ec, 0xc77bfe61, + 0xfdb72874, 0x4370f138, 0xdf3462ad, 0xc8970a59, 0xb4a9fed8, 0xa2ddc39b, + 0xd61db62a, 0x95c5fc1b, 0x7b22e6e0, 0x1969702c, 0x7992aebb, 0x59d7c225, + 0x0e16db0b, 0x9f2afc21, 0x246cf66b, 0xb3d6569d, 0x29c532d7, 0xe155747a, + 0xe38d7872, 0xea704969, 0xb69095b0, 0x1b198efd, 0x55daab76, 0xa2a377b6, + 0xb31aa2fa, 0x48b73c41, 0xf0cc501a, 0x9c9ca831, 0x1b591b99, 0xb2d8d22f, + 0xab4b5f69, 0x4fe00e71, 0xdf5480bd, 0x982540d7, 0x7f34ea4f, 0xd7be66e1, + 0x9d2ab1ba, 0x1ba62e12, 0xee3fb36c, 0xf28d7c5a, 0x756311eb, 0xc68567f2, + 0x7b6ea177, 0xc398d9f3 +}; + +#else +extern const ip4_test_t ip4_tests[N_IP4_TESTS]; +extern const ip6_test_t ip6_tests[N_IP6_TESTS]; +extern const u32 length_test_hashes[N_LENGTH_TESTS]; +#endif + +__test_funct_fn u32 +wrapper (clib_toeplitz_hash_key_t *k, u8 *data, u32 n_bytes) +{ + return clib_toeplitz_hash (k, data, n_bytes); +} + +static clib_error_t * +test_clib_toeplitz_hash (clib_error_t *err) +{ + u32 r; + int n_key_copies, bigkey_len, bigdata_len; + u8 *bigkey, *bigdata; + clib_toeplitz_hash_key_t *k; + + k = clib_toeplitz_hash_key_init (0, 0); + + for (int i = 0; i < N_IP4_TESTS; i++) + { + r = wrapper (k, (u8 *) &ip4_tests[i].key, 8); + if (ip4_tests[i].hash_2t != r) + return clib_error_return (err, + "wrong IPv4 2 tuple hash for test %u, " + "calculated 0x%08x expected 0x%08x", + i, ip4_tests[i].hash_2t, r); + + r = wrapper (k, (u8 *) &ip4_tests[i].key, 12); + if (ip4_tests[i].hash_4t != r) + return clib_error_return (err, + "wrong IPv4 4 tuple hash for test %u, " + "calculated 0x%08x expected 0x%08x", + i, ip4_tests[i].hash_4t, r); + } + + for (int i = 0; i < N_IP6_TESTS; i++) + { + r = wrapper (k, (u8 *) &ip6_tests[i].key, 32); + if (ip6_tests[i].hash_2t != r) + return clib_error_return (err, + "wrong IPv6 2 tuple hash for test %u, " + "calculated 0x%08x expected 0x%08x", + i, ip6_tests[i].hash_2t, r); + + r = wrapper (k, (u8 *) &ip6_tests[i].key, 36); + if (ip6_tests[i].hash_4t != r) + return clib_error_return (err, + "wrong IPv6 4 tuple hash for test %u, " + "calculated 0x%08x expected 0x%08x", + i, ip6_tests[i].hash_4t, r); + } + + n_key_copies = 6; + bigkey_len = k->key_length * n_key_copies; + bigdata_len = bigkey_len - 4; + bigkey = clib_mem_alloc (bigkey_len); + bigdata = clib_mem_alloc (bigdata_len); + u32 key_len = k->key_length; + + for (int i = 0; i < n_key_copies; i++) + clib_memcpy (bigkey + i * key_len, k->data, key_len); + + for (int i = 0; i < bigdata_len; i++) + bigdata[i] = (u8) i; + + clib_toeplitz_hash_key_free (k); + k = clib_toeplitz_hash_key_init (bigkey, n_key_copies * key_len); + + for (int i = 0; i < N_LENGTH_TESTS - 4; i++) + { + r = wrapper (k, bigdata, i); + if (length_test_hashes[i] != r) + { + err = clib_error_return (err, + "wrong length test hash for length %u, " + "calculated 0x%08x expected 0x%08x " + "xor 0x%08x", + i, r, length_test_hashes[i], + r ^ length_test_hashes[i]); + goto done; + } + } + +done: + clib_toeplitz_hash_key_free (k); + clib_mem_free (bigkey); + clib_mem_free (bigdata); + return err; +} + +void __test_perf_fn +perftest_fixed_12byte (int fd, test_perf_t *tp) +{ + u32 n = tp->n_ops; + u8 *data = test_mem_alloc_and_splat (12, n, (void *) &ip4_tests[0].key); + u8 *res = test_mem_alloc (4 * n); + clib_toeplitz_hash_key_t *k = clib_toeplitz_hash_key_init (0, 0); + + test_perf_event_enable (fd); + for (int i = 0; i < n; i++) + ((u32 *) res)[i] = clib_toeplitz_hash (k, data + i * 12, 12); + test_perf_event_disable (fd); + + clib_toeplitz_hash_key_free (k); + test_mem_free (data); + test_mem_free (res); +} + +void __test_perf_fn +perftest_fixed_36byte (int fd, test_perf_t *tp) +{ + u32 n = tp->n_ops; + u8 *data = test_mem_alloc_and_splat (36, n, (void *) &ip6_tests[0].key); + u8 *res = test_mem_alloc (4 * n); + clib_toeplitz_hash_key_t *k = clib_toeplitz_hash_key_init (0, 0); + + test_perf_event_enable (fd); + for (int i = 0; i < n; i++) + ((u32 *) res)[i] = clib_toeplitz_hash (k, data + i * 36, 36); + test_perf_event_disable (fd); + + clib_toeplitz_hash_key_free (k); + test_mem_free (data); + test_mem_free (res); +} + +void __test_perf_fn +perftest_variable_size (int fd, test_perf_t *tp) +{ + u32 key_len, n_keys, n = tp->n_ops; + u8 *key, *data = test_mem_alloc (n); + u32 *res = test_mem_alloc (sizeof (u32)); + clib_toeplitz_hash_key_t *k = clib_toeplitz_hash_key_init (0, 0); + + k = clib_toeplitz_hash_key_init (0, 0); + key_len = k->key_length; + n_keys = ((n + 4) / k->key_length) + 1; + key = test_mem_alloc_and_splat (n_keys, key_len, k->data); + clib_toeplitz_hash_key_free (k); + k = clib_toeplitz_hash_key_init (key, key_len * n_keys); + + test_perf_event_enable (fd); + res[0] = clib_toeplitz_hash (k, data, n); + test_perf_event_disable (fd); + + clib_toeplitz_hash_key_free (k); + test_mem_free (data); + test_mem_free (res); + test_mem_free (key); +} + +REGISTER_TEST (clib_toeplitz_hash) = { + .name = "clib_toeplitz_hash", + .fn = test_clib_toeplitz_hash, + .perf_tests = PERF_TESTS ({ .name = "fixed_12", + .op_name = "12B Tuple", + .n_ops = 1024, + .fn = perftest_fixed_12byte }, + { .name = "fixed_36", + .op_name = "36B Tuple", + .n_ops = 1024, + .fn = perftest_fixed_36byte }, + { .name = "variable_size", + .op_name = "Byte", + .n_ops = 16384, + .fn = perftest_variable_size }), +}; diff --git a/src/vppinfra/vector/toeplitz.c b/src/vppinfra/vector/toeplitz.c new file mode 100644 index 00000000000..b0140d51e6f --- /dev/null +++ b/src/vppinfra/vector/toeplitz.c @@ -0,0 +1,122 @@ +/* SPDX-License-Identifier: Apache-2.0 + * Copyright(c) 2021 Cisco Systems, Inc. + */ + +#include <vppinfra/clib.h> +#include <vppinfra/mem.h> +#include <vppinfra/vector/toeplitz.h> + +static u8 default_key[40] = { + 0x6d, 0x5a, 0x56, 0xda, 0x25, 0x5b, 0x0e, 0xc2, 0x41, 0x67, + 0x25, 0x3d, 0x43, 0xa3, 0x8f, 0xb0, 0xd0, 0xca, 0x2b, 0xcb, + 0xae, 0x7b, 0x30, 0xb4, 0x77, 0xcb, 0x2d, 0xa3, 0x80, 0x30, + 0xf2, 0x0c, 0x6a, 0x42, 0xb7, 0x3b, 0xbe, 0xac, 0x01, 0xfa, +}; + +#ifdef __x86_64__ +static_always_inline void +clib_toeplitz_hash_key_expand_8 (u64x2 kv, u64x8u *m) +{ + u64x8 kv4, a, b, shift = { 0, 1, 2, 3, 4, 5, 6, 7 }; + + kv4 = (u64x8){ kv[0], kv[1], kv[0], kv[1], kv[0], kv[1], kv[0], kv[1] }; + + /* clang-format off */ + /* create 8 byte-swapped copies of the bytes 0 - 7 */ + a = (u64x8) u8x64_shuffle (kv4, + 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0, + 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0, + 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0, + 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0, + 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0, + 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0, + 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0, + 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0); + /* create 8 byte-swapped copies of the bytes 4 - 11 */ + b = (u64x8) u8x64_shuffle (kv4, + 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, + 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, + 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, + 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, + 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, + 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, + 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, + 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4); + /* clang-format on */ + + /* shift each 64-bit element for 0 - 7 bits */ + a <<= shift; + b <<= shift; + + /* clang-format off */ + /* construct eight 8x8 bit matrix used by gf2p8affine */ + * m = (u64x8) u8x64_shuffle2 (a, b, + 0x07, 0x0f, 0x17, 0x1f, 0x27, 0x2f, 0x37, 0x3f, + 0x06, 0x0e, 0x16, 0x1e, 0x26, 0x2e, 0x36, 0x3e, + 0x05, 0x0d, 0x15, 0x1d, 0x25, 0x2d, 0x35, 0x3d, + 0x04, 0x0c, 0x14, 0x1c, 0x24, 0x2c, 0x34, 0x3c, + 0x47, 0x4f, 0x57, 0x5f, 0x67, 0x6f, 0x77, 0x7f, + 0x46, 0x4e, 0x56, 0x5e, 0x66, 0x6e, 0x76, 0x7e, + 0x45, 0x4d, 0x55, 0x5d, 0x65, 0x6d, 0x75, 0x7d, + 0x44, 0x4c, 0x54, 0x5c, 0x64, 0x6c, 0x74, 0x7c); + /* clang-format on */ +} + +void +clib_toeplitz_hash_key_expand (u64 *matrixes, u8 *key, int size) +{ + u64x8u *m = (u64x8u *) matrixes; + u64x2 kv, zero = {}; + + while (size >= 8) + { + kv = *(u64x2u *) key; + clib_toeplitz_hash_key_expand_8 (kv, m); + key += 8; + m++; + size -= 8; + } + + kv = u64x2_shuffle2 (kv, zero, 1, 2); + clib_toeplitz_hash_key_expand_8 (kv, m); +} +#endif + +__clib_export clib_toeplitz_hash_key_t * +clib_toeplitz_hash_key_init (u8 *key, u32 keylen) +{ + clib_toeplitz_hash_key_t *k; + u32 size, gfni_size = 0; + + if (key == 0) + { + key = default_key; + keylen = sizeof (default_key); + } + + size = + round_pow2 (sizeof (clib_toeplitz_hash_key_t) + round_pow2 (keylen, 16), + CLIB_CACHE_LINE_BYTES); +#ifdef __x86_64__ + gfni_size = round_pow2 ((keylen + 1) * 8, CLIB_CACHE_LINE_BYTES); +#endif + + k = clib_mem_alloc_aligned (size + gfni_size, CLIB_CACHE_LINE_BYTES); + clib_memset_u8 (k, 0, size + gfni_size); + k->key_length = keylen; + k->gfni_offset = size; + clib_memcpy_fast (k->data, key, keylen); + +#ifdef __x86_64__ + clib_toeplitz_hash_key_expand ((u64 *) ((u8 *) k + k->gfni_offset), k->data, + k->key_length); +#endif + + return k; +} + +__clib_export void +clib_toeplitz_hash_key_free (clib_toeplitz_hash_key_t *k) +{ + clib_mem_free (k); +} diff --git a/src/vppinfra/vector/toeplitz.h b/src/vppinfra/vector/toeplitz.h new file mode 100644 index 00000000000..a199a4d68a6 --- /dev/null +++ b/src/vppinfra/vector/toeplitz.h @@ -0,0 +1,218 @@ +/* SPDX-License-Identifier: Apache-2.0 + * Copyright(c) 2021 Cisco Systems, Inc. + */ + +#ifndef included_vector_toeplitz_h +#define included_vector_toeplitz_h +#include <vppinfra/clib.h> + +typedef struct +{ + u16 key_length; + u16 gfni_offset; + u8 data[]; +} clib_toeplitz_hash_key_t; + +clib_toeplitz_hash_key_t *clib_toeplitz_hash_key_init (u8 *key, u32 keylen); +void clib_toeplitz_hash_key_free (clib_toeplitz_hash_key_t *k); + +#if defined(__GFNI__) && defined(__AVX512F__) + +#define u64x8_gf2p8_affine(d, m, imm) \ + (u64x8) _mm512_gf2p8affine_epi64_epi8 ((__m512i) (d), (__m512i) (m), imm) + +#endif + +#ifdef CLIB_HAVE_VEC256 +static_always_inline u32x8 +toeplitz_hash_one_x8 (u32x8 hash, u64x4 v4, u8 data, u8 off) +{ + u32x8 v8 = u32x8_shuffle2 (v4 << (off * 8), v4 << (off * 8 + 4), + /*uppper 32 bits of each u64 in reverse order */ + 15, 13, 11, 9, 7, 5, 3, 1); + +#ifdef CLIB_HAVE_VEC256_MASK_BITWISE_OPS + return u32x8_mask_xor (hash, v8, data); +#else + static const u32x8 bits = { 1, 2, 4, 8, 16, 32, 64, 128 }; + return hash ^ (((u32x8_splat (data) & bits) != u32x8_zero ()) & v8); +#endif +} +#endif + +static_always_inline u32 +clib_toeplitz_hash (clib_toeplitz_hash_key_t *k, u8 *data, int n_bytes) +{ + u8 *key = k->data; + /* key must be 4 bytes longer than data */ + ASSERT (k->key_length - n_bytes >= 4); + +#if defined(__GFNI__) && defined(__AVX512F__) + u8x64 a, b, dv; + u64x8 xor_sum_x8 = {}; + u64x8u *m = (u64x8u *) ((u8 *) k + k->gfni_offset); + + u8x64 idx = { 0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x02, 0x03, 0x01, 0x02, + 0x03, 0x04, 0x01, 0x02, 0x03, 0x04, 0x02, 0x03, 0x04, 0x05, + 0x02, 0x03, 0x04, 0x05, 0x03, 0x04, 0x05, 0x06, 0x03, 0x04, + 0x05, 0x06, 0x04, 0x05, 0x06, 0x07, 0x04, 0x05, 0x06, 0x07, + 0x05, 0x06, 0x07, 0x08, 0x05, 0x06, 0x07, 0x08, 0x06, 0x07, + 0x08, 0x09, 0x06, 0x07, 0x08, 0x09, 0x07, 0x08, 0x09, 0x0a, + 0x07, 0x08, 0x09, 0x0a }; + + /* move data ptr backwards for 3 byte so mask load "prepends" three zeros */ + data -= 3; + n_bytes += 3; + + if (n_bytes < 64) + { + dv = u8x64_mask_load_zero ((u8 *) data, pow2_mask (n_bytes - 3) << 3); + goto last8; + } + + dv = u8x64_mask_load_zero ((u8 *) data, -1ULL << 3); +next56: + a = u8x64_permute (idx, dv); + b = u8x64_permute (idx, (u8x64) u64x8_align_right (dv, dv, 1)); + xor_sum_x8 = u64x8_xor3 (xor_sum_x8, u64x8_gf2p8_affine (a, m[0], 0), + u64x8_gf2p8_affine (b, m[1], 0)); + + a = u8x64_permute (idx, (u8x64) u64x8_align_right (dv, dv, 2)); + b = u8x64_permute (idx, (u8x64) u64x8_align_right (dv, dv, 3)); + xor_sum_x8 = u64x8_xor3 (xor_sum_x8, u64x8_gf2p8_affine (a, m[2], 0), + u64x8_gf2p8_affine (b, m[3], 0)); + + a = u8x64_permute (idx, (u8x64) u64x8_align_right (dv, dv, 4)); + b = u8x64_permute (idx, (u8x64) u64x8_align_right (dv, dv, 5)); + xor_sum_x8 = u64x8_xor3 (xor_sum_x8, u64x8_gf2p8_affine (a, m[4], 0), + u64x8_gf2p8_affine (b, m[5], 0)); + + a = u8x64_permute (idx, (u8x64) u64x8_align_right (dv, dv, 6)); + xor_sum_x8 ^= u64x8_gf2p8_affine (a, m[6], 0); + n_bytes -= 56; + data += 56; + m += 7; + + if (n_bytes >= 64) + { + dv = *(u8x64u *) data; + goto next56; + } + + if (n_bytes == 0) + goto done; + + dv = u8x64_mask_load_zero ((u8 *) data, pow2_mask (n_bytes)); +last8: + a = u8x64_permute (idx, dv); + xor_sum_x8 ^= u64x8_gf2p8_affine (a, m[0], 0); + n_bytes -= 8; + + if (n_bytes > 0) + { + m += 1; + dv = (u8x64) u64x8_align_right (u64x8_zero (), dv, 1); + goto last8; + } + +done: + /* horizontal xor */ + xor_sum_x8 ^= u64x8_align_right (xor_sum_x8, xor_sum_x8, 4); + xor_sum_x8 ^= u64x8_align_right (xor_sum_x8, xor_sum_x8, 2); + return xor_sum_x8[0] ^ xor_sum_x8[1]; +#elif defined(CLIB_HAVE_VEC256) + u64x4 v4, shift = { 0, 1, 2, 3 }; + u32x8 hash8 = {}; + u32x4 hash4; + + while (n_bytes >= 4) + { + v4 = u64x4_splat (clib_net_to_host_u64 (*(u64u *) key)) << shift; + + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[0], 0); + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[1], 1); + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[2], 2); + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[3], 3); + + data += 4; + key += 4; + n_bytes -= 4; + } + + if (n_bytes) + { + u64 v = (u64) clib_net_to_host_u32 ((u64) (*(u32u *) key)) << 32; + v |= (u64) key[4] << 24; + + if (n_bytes == 3) + { + v |= (u64) key[5] << 16; + v |= (u64) key[6] << 8; + v4 = u64x4_splat (v) << shift; + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[0], 0); + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[1], 1); + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[2], 2); + } + else if (n_bytes == 2) + { + v |= (u64) key[5] << 16; + v4 = u64x4_splat (v) << shift; + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[0], 0); + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[1], 1); + } + else + { + v4 = u64x4_splat (v) << shift; + hash8 = toeplitz_hash_one_x8 (hash8, v4, data[0], 0); + } + } + + hash4 = u32x8_extract_lo (hash8) ^ u32x8_extract_hi (hash8); + hash4 ^= (u32x4) u8x16_align_right (hash4, hash4, 8); + hash4 ^= (u32x4) u8x16_align_right (hash4, hash4, 4); + return hash4[0]; + +#endif + u64 v, hash = 0; + + while (n_bytes >= 4) + { + v = clib_net_to_host_u64 (*(u64u *) key); + + for (u8 bit = 1 << 7, byte = data[0]; bit; bit >>= 1, v <<= 1) + hash ^= byte & bit ? v : 0; + for (u8 bit = 1 << 7, byte = data[1]; bit; bit >>= 1, v <<= 1) + hash ^= byte & bit ? v : 0; + for (u8 bit = 1 << 7, byte = data[2]; bit; bit >>= 1, v <<= 1) + hash ^= byte & bit ? v : 0; + for (u8 bit = 1 << 7, byte = data[3]; bit; bit >>= 1, v <<= 1) + hash ^= byte & bit ? v : 0; + + data += 4; + key += 4; + n_bytes -= 4; + } + + if (n_bytes) + { + v = (u64) clib_net_to_host_u32 ((u64) (*(u32u *) key)) << 32; + v |= (u64) key[4] << 24; + for (u8 bit = 1 << 7, byte = data[0]; bit; bit >>= 1, v <<= 1) + hash ^= byte & bit ? v : 0; + if (n_bytes > 1) + { + v |= (u64) key[5] << 24; + for (u8 bit = 1 << 7, byte = data[1]; bit; bit >>= 1, v <<= 1) + hash ^= byte & bit ? v : 0; + } + if (n_bytes > 2) + { + v |= (u64) key[6] << 24; + for (u8 bit = 1 << 7, byte = data[2]; bit; bit >>= 1, v <<= 1) + hash ^= byte & bit ? v : 0; + } + } + return hash >> 32; +} + +#endif diff --git a/src/vppinfra/vector_avx512.h b/src/vppinfra/vector_avx512.h index 33f40ef7b5a..3505961dca5 100644 --- a/src/vppinfra/vector_avx512.h +++ b/src/vppinfra/vector_avx512.h @@ -98,6 +98,18 @@ _ (i32x16, i16x32, _mm512_packs_epi32) _ (i32x16, u16x32, _mm512_packus_epi32) #undef _ +static_always_inline u64x8 +u64x8_byte_swap (u64x8 v) +{ + u8x64 swap = { + 7, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8, + 7, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8, + 7, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8, + 7, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8, + }; + return (u64x8) _mm512_shuffle_epi8 ((__m512i) v, (__m512i) swap); +} + static_always_inline u32x16 u32x16_byte_swap (u32x16 v) { @@ -184,6 +196,13 @@ u8x64_xor3 (u8x64 a, u8x64 b, u8x64 c) (__m512i) c, 0x96); } +static_always_inline u64x8 +u64x8_xor3 (u64x8 a, u64x8 b, u64x8 c) +{ + return (u64x8) _mm512_ternarylogic_epi32 ((__m512i) a, (__m512i) b, + (__m512i) c, 0x96); +} + static_always_inline u8x64 u8x64_reflect_u8x16 (u8x64 x) { @@ -302,6 +321,12 @@ u8x64_mask_blend (u8x64 a, u8x64 b, u64 mask) return (u8x64) _mm512_mask_blend_epi8 (mask, (__m512i) a, (__m512i) b); } +static_always_inline u8x64 +u8x64_permute (u8x64 v, u8x64 idx) +{ + return (u8x64) _mm512_permutexvar_epi8 ((__m512i) v, (__m512i) idx); +} + #define _(t, m, e, p, it) \ static_always_inline m t##_is_equal_mask (t a, t b) \ { \ |