| /* |
| Copyright (c) 2013 Cisco and/or its affiliates. |
| |
| * Licensed under the Apache License, Version 2.0 (the "License"); |
| * you may not use this file except in compliance with the License. |
| * You may obtain a copy of the License at: |
| * |
| * http://www.apache.org/licenses/LICENSE-2.0 |
| * |
| * Unless required by applicable law or agreed to in writing, software |
| * distributed under the License is distributed on an "AS IS" BASIS, |
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| * See the License for the specific language governing permissions and |
| * limitations under the License. |
| */ |
| |
| #include <vppinfra/pfhash.h> |
| #include <vppinfra/format.h> |
| |
| /* This is incredibly handy when debugging */ |
| u32 vl (void *v) __attribute__ ((weak)); |
| u32 |
| vl (void *v) |
| { |
| return vec_len (v); |
| } |
| |
| #if defined(CLIB_HAVE_VEC128) && ! defined (__ALTIVEC__) |
| |
| typedef struct |
| { |
| u8 *key[16]; |
| u64 value; |
| } pfhash_show_t; |
| |
| static int |
| sh_compare (pfhash_show_t * sh0, pfhash_show_t * sh1) |
| { |
| return ((i32) (sh0->value) - ((i32) sh1->value)); |
| } |
| |
| u8 * |
| format_pfhash (u8 * s, va_list * args) |
| { |
| pfhash_t *p = va_arg (*args, pfhash_t *); |
| int verbose = va_arg (*args, int); |
| |
| if (p == 0 || p->overflow_hash == 0 || p->buckets == 0) |
| { |
| s = format (s, "*** uninitialized ***"); |
| return s; |
| } |
| |
| s = format (s, "Prefetch hash '%s'\n", p->name); |
| s = |
| format (s, " %d buckets, %u bucket overflows, %.1f%% bucket overflow \n", |
| vec_len (p->buckets), p->overflow_count, |
| 100.0 * ((f64) p->overflow_count) / ((f64) vec_len (p->buckets))); |
| if (p->nitems) |
| s = |
| format (s, |
| " %u items, %u items in overflow, %.1f%% items in overflow\n", |
| p->nitems, p->nitems_in_overflow, |
| 100.0 * ((f64) p->nitems_in_overflow) / ((f64) p->nitems)); |
| |
| if (verbose) |
| { |
| pfhash_show_t *shs = 0, *sh; |
| hash_pair_t *hp; |
| int i, j; |
| |
| for (i = 0; i < vec_len (p->buckets); i++) |
| { |
| pfhash_kv_t *kv; |
| pfhash_kv_16_t *kv16; |
| pfhash_kv_8_t *kv8; |
| pfhash_kv_8v8_t *kv8v8; |
| pfhash_kv_4_t *kv4; |
| |
| if (p->buckets[i] == 0 || p->buckets[i] == PFHASH_BUCKET_OVERFLOW) |
| continue; |
| |
| kv = pool_elt_at_index (p->kvp, p->buckets[i]); |
| |
| switch (p->key_size) |
| { |
| case 16: |
| kv16 = &kv->kv16; |
| for (j = 0; j < 3; j++) |
| { |
| if (kv16->values[j] != (u32) ~ 0) |
| { |
| vec_add2 (shs, sh, 1); |
| clib_memcpy (sh->key, &kv16->kb.k_u32x4[j], |
| p->key_size); |
| sh->value = kv16->values[j]; |
| } |
| } |
| break; |
| case 8: |
| if (p->value_size == 4) |
| { |
| kv8 = &kv->kv8; |
| for (j = 0; j < 5; j++) |
| { |
| if (kv8->values[j] != (u32) ~ 0) |
| { |
| vec_add2 (shs, sh, 1); |
| clib_memcpy (sh->key, &kv8->kb.k_u64[j], |
| p->key_size); |
| sh->value = kv8->values[j]; |
| } |
| } |
| } |
| else |
| { |
| kv8v8 = &kv->kv8v8; |
| for (j = 0; j < 4; j++) |
| { |
| if (kv8v8->values[j] != (u64) ~ 0) |
| { |
| vec_add2 (shs, sh, 1); |
| clib_memcpy (sh->key, &kv8v8->kb.k_u64[j], |
| p->key_size); |
| sh->value = kv8v8->values[j]; |
| } |
| } |
| |
| } |
| break; |
| case 4: |
| kv4 = &kv->kv4; |
| for (j = 0; j < 8; j++) |
| { |
| if (kv4->values[j] != (u32) ~ 0) |
| { |
| vec_add2 (shs, sh, 1); |
| clib_memcpy (sh->key, &kv4->kb.kb[j], p->key_size); |
| sh->value = kv4->values[j]; |
| } |
| } |
| break; |
| } |
| } |
| |
| /* *INDENT-OFF* */ |
| hash_foreach_pair (hp, p->overflow_hash, |
| ({ |
| vec_add2 (shs, sh, 1); |
| clib_memcpy (sh->key, (u8 *)hp->key, p->key_size); |
| sh->value = hp->value[0]; |
| })); |
| /* *INDENT-ON* */ |
| |
| vec_sort_with_function (shs, sh_compare); |
| |
| for (i = 0; i < vec_len (shs); i++) |
| { |
| sh = vec_elt_at_index (shs, i); |
| s = format (s, " %U value %u\n", format_hex_bytes, sh->key, |
| p->key_size, sh->value); |
| } |
| vec_free (shs); |
| } |
| return s; |
| } |
| |
| |
| void abort (void); |
| |
| void |
| pfhash_init (pfhash_t * p, char *name, u32 key_size, u32 value_size, |
| u32 nbuckets) |
| { |
| pfhash_kv_t *kv; |
| memset (p, 0, sizeof (*p)); |
| u32 key_bytes; |
| |
| switch (key_size) |
| { |
| case 4: |
| key_bytes = 4; |
| break; |
| case 8: |
| key_bytes = 8; |
| break; |
| case 16: |
| key_bytes = 16; |
| break; |
| default: |
| ASSERT (0); |
| abort (); |
| } |
| |
| switch (value_size) |
| { |
| case 4: |
| case 8: |
| break; |
| default: |
| ASSERT (0); |
| abort (); |
| } |
| |
| |
| p->name = format (0, "%s", name); |
| vec_add1 (p->name, 0); |
| p->overflow_hash = hash_create_mem (0, key_bytes, sizeof (uword)); |
| |
| nbuckets = 1 << (max_log2 (nbuckets)); |
| |
| /* This sets the entire bucket array to zero */ |
| vec_validate (p->buckets, nbuckets - 1); |
| p->key_size = key_size; |
| p->value_size = value_size; |
| |
| /* |
| * Unset buckets implicitly point at the 0th pool elt. |
| * All search routines will return ~0 if they go there. |
| */ |
| pool_get_aligned (p->kvp, kv, 16); |
| memset (kv, 0xff, sizeof (*kv)); |
| } |
| |
| static pfhash_kv_16_t * |
| pfhash_get_kv_16 (pfhash_t * p, u32 bucket_contents, |
| u32x4 * key, u32 * match_index) |
| { |
| u32x4 diff[3]; |
| u32 is_equal[3]; |
| pfhash_kv_16_t *kv = 0; |
| |
| *match_index = (u32) ~ 0; |
| |
| kv = &p->kvp[bucket_contents].kv16; |
| |
| diff[0] = u32x4_sub (kv->kb.k_u32x4[0], key[0]); |
| diff[1] = u32x4_sub (kv->kb.k_u32x4[1], key[0]); |
| diff[2] = u32x4_sub (kv->kb.k_u32x4[2], key[0]); |
| |
| is_equal[0] = u32x4_zero_byte_mask (diff[0]) == 0xffff; |
| is_equal[1] = u32x4_zero_byte_mask (diff[1]) == 0xffff; |
| is_equal[2] = u32x4_zero_byte_mask (diff[2]) == 0xffff; |
| |
| if (is_equal[0]) |
| *match_index = 0; |
| if (is_equal[1]) |
| *match_index = 1; |
| if (is_equal[2]) |
| *match_index = 2; |
| |
| return kv; |
| } |
| |
| static pfhash_kv_8_t * |
| pfhash_get_kv_8 (pfhash_t * p, u32 bucket_contents, |
| u64 * key, u32 * match_index) |
| { |
| pfhash_kv_8_t *kv; |
| |
| *match_index = (u32) ~ 0; |
| |
| kv = &p->kvp[bucket_contents].kv8; |
| |
| if (kv->kb.k_u64[0] == key[0]) |
| *match_index = 0; |
| if (kv->kb.k_u64[1] == key[0]) |
| *match_index = 1; |
| if (kv->kb.k_u64[2] == key[0]) |
| *match_index = 2; |
| if (kv->kb.k_u64[3] == key[0]) |
| *match_index = 3; |
| if (kv->kb.k_u64[4] == key[0]) |
| *match_index = 4; |
| |
| return kv; |
| } |
| |
| static pfhash_kv_8v8_t * |
| pfhash_get_kv_8v8 (pfhash_t * p, |
| u32 bucket_contents, u64 * key, u32 * match_index) |
| { |
| pfhash_kv_8v8_t *kv; |
| |
| *match_index = (u32) ~ 0; |
| |
| kv = &p->kvp[bucket_contents].kv8v8; |
| |
| if (kv->kb.k_u64[0] == key[0]) |
| *match_index = 0; |
| if (kv->kb.k_u64[1] == key[0]) |
| *match_index = 1; |
| if (kv->kb.k_u64[2] == key[0]) |
| *match_index = 2; |
| if (kv->kb.k_u64[3] == key[0]) |
| *match_index = 3; |
| |
| return kv; |
| } |
| |
| static pfhash_kv_4_t * |
| pfhash_get_kv_4 (pfhash_t * p, u32 bucket_contents, |
| u32 * key, u32 * match_index) |
| { |
| u32x4 vector_key; |
| u32x4 is_equal[2]; |
| u32 zbm[2], winner_index; |
| pfhash_kv_4_t *kv; |
| |
| *match_index = (u32) ~ 0; |
| |
| kv = &p->kvp[bucket_contents].kv4; |
| |
| vector_key = u32x4_splat (key[0]); |
| |
| is_equal[0] = u32x4_is_equal (kv->kb.k_u32x4[0], vector_key); |
| is_equal[1] = u32x4_is_equal (kv->kb.k_u32x4[1], vector_key); |
| zbm[0] = ~u32x4_zero_byte_mask (is_equal[0]) & 0xFFFF; |
| zbm[1] = ~u32x4_zero_byte_mask (is_equal[1]) & 0xFFFF; |
| |
| if (PREDICT_FALSE ((zbm[0] == 0) && (zbm[1] == 0))) |
| return kv; |
| |
| winner_index = min_log2 (zbm[0]) >> 2; |
| winner_index = zbm[1] ? (4 + (min_log2 (zbm[1]) >> 2)) : winner_index; |
| |
| *match_index = winner_index; |
| return kv; |
| } |
| |
| static pfhash_kv_t * |
| pfhash_get_internal (pfhash_t * p, u32 bucket_contents, |
| void *key, u32 * match_index) |
| { |
| pfhash_kv_t *kv = 0; |
| |
| switch (p->key_size) |
| { |
| case 16: |
| kv = |
| (pfhash_kv_t *) pfhash_get_kv_16 (p, bucket_contents, key, |
| match_index); |
| break; |
| case 8: |
| if (p->value_size == 4) |
| kv = (pfhash_kv_t *) pfhash_get_kv_8 (p, bucket_contents, |
| key, match_index); |
| else |
| kv = (pfhash_kv_t *) pfhash_get_kv_8v8 (p, bucket_contents, |
| key, match_index); |
| break; |
| case 4: |
| kv = |
| (pfhash_kv_t *) pfhash_get_kv_4 (p, bucket_contents, key, |
| match_index); |
| break; |
| default: |
| ASSERT (0); |
| } |
| return kv; |
| } |
| |
| u64 |
| pfhash_get (pfhash_t * p, u32 bucket, void *key) |
| { |
| pfhash_kv_t *kv; |
| u32 match_index = ~0; |
| pfhash_kv_16_t *kv16; |
| pfhash_kv_8_t *kv8; |
| pfhash_kv_8v8_t *kv8v8; |
| pfhash_kv_4_t *kv4; |
| |
| u32 bucket_contents = pfhash_read_bucket_prefetch_kv (p, bucket); |
| |
| if (bucket_contents == PFHASH_BUCKET_OVERFLOW) |
| { |
| uword *hp; |
| |
| hp = hash_get_mem (p->overflow_hash, key); |
| if (hp) |
| return hp[0]; |
| return (u64) ~ 0; |
| } |
| |
| kv = pfhash_get_internal (p, bucket_contents, key, &match_index); |
| if (match_index == (u32) ~ 0) |
| return (u64) ~ 0; |
| |
| kv16 = (void *) kv; |
| kv8 = (void *) kv; |
| kv4 = (void *) kv; |
| kv8v8 = (void *) kv; |
| |
| switch (p->key_size) |
| { |
| case 16: |
| return (kv16->values[match_index] == (u32) ~ 0) |
| ? (u64) ~ 0 : (u64) kv16->values[match_index]; |
| case 8: |
| if (p->value_size == 4) |
| return (kv8->values[match_index] == (u32) ~ 0) |
| ? (u64) ~ 0 : (u64) kv8->values[match_index]; |
| else |
| return kv8v8->values[match_index]; |
| case 4: |
| return (kv4->values[match_index] == (u32) ~ 0) |
| ? (u64) ~ 0 : (u64) kv4->values[match_index]; |
| default: |
| ASSERT (0); |
| } |
| return (u64) ~ 0; |
| } |
| |
| void |
| pfhash_set (pfhash_t * p, u32 bucket, void *key, void *value) |
| { |
| u32 bucket_contents = pfhash_read_bucket_prefetch_kv (p, bucket); |
| u32 match_index = (u32) ~ 0; |
| pfhash_kv_t *kv; |
| pfhash_kv_16_t *kv16; |
| pfhash_kv_8_t *kv8; |
| pfhash_kv_8v8_t *kv8v8; |
| pfhash_kv_4_t *kv4; |
| int i; |
| u8 *kcopy; |
| |
| if (bucket_contents == PFHASH_BUCKET_OVERFLOW) |
| { |
| hash_pair_t *hp; |
| hp = hash_get_pair_mem (p->overflow_hash, key); |
| if (hp) |
| { |
| clib_warning ("replace value 0x%08x with value 0x%08x", |
| hp->value[0], (u64) value); |
| hp->value[0] = (u64) value; |
| return; |
| } |
| kcopy = clib_mem_alloc (p->key_size); |
| clib_memcpy (kcopy, key, p->key_size); |
| hash_set_mem (p->overflow_hash, kcopy, value); |
| p->nitems++; |
| p->nitems_in_overflow++; |
| return; |
| } |
| |
| if (bucket_contents == 0) |
| { |
| pool_get_aligned (p->kvp, kv, 16); |
| memset (kv, 0xff, sizeof (*kv)); |
| p->buckets[bucket] = kv - p->kvp; |
| } |
| else |
| kv = pfhash_get_internal (p, bucket_contents, key, &match_index); |
| |
| kv16 = (void *) kv; |
| kv8 = (void *) kv; |
| kv8v8 = (void *) kv; |
| kv4 = (void *) kv; |
| |
| p->nitems++; |
| |
| if (match_index != (u32) ~ 0) |
| { |
| switch (p->key_size) |
| { |
| case 16: |
| kv16->values[match_index] = (u32) (u64) value; |
| return; |
| |
| case 8: |
| if (p->value_size == 4) |
| kv8->values[match_index] = (u32) (u64) value; |
| else |
| kv8v8->values[match_index] = (u64) value; |
| return; |
| |
| case 4: |
| kv4->values[match_index] = (u64) value; |
| return; |
| |
| default: |
| ASSERT (0); |
| } |
| } |
| |
| switch (p->key_size) |
| { |
| case 16: |
| for (i = 0; i < 3; i++) |
| { |
| if (kv16->values[i] == (u32) ~ 0) |
| { |
| clib_memcpy (&kv16->kb.k_u32x4[i], key, p->key_size); |
| kv16->values[i] = (u32) (u64) value; |
| return; |
| } |
| } |
| /* copy bucket contents to overflow hash tbl */ |
| for (i = 0; i < 3; i++) |
| { |
| kcopy = clib_mem_alloc (p->key_size); |
| clib_memcpy (kcopy, &kv16->kb.k_u32x4[i], p->key_size); |
| hash_set_mem (p->overflow_hash, kcopy, kv16->values[i]); |
| p->nitems_in_overflow++; |
| } |
| /* Add new key to overflow */ |
| kcopy = clib_mem_alloc (p->key_size); |
| clib_memcpy (kcopy, key, p->key_size); |
| hash_set_mem (p->overflow_hash, kcopy, value); |
| p->buckets[bucket] = PFHASH_BUCKET_OVERFLOW; |
| p->overflow_count++; |
| p->nitems_in_overflow++; |
| return; |
| |
| case 8: |
| if (p->value_size == 4) |
| { |
| for (i = 0; i < 5; i++) |
| { |
| if (kv8->values[i] == (u32) ~ 0) |
| { |
| clib_memcpy (&kv8->kb.k_u64[i], key, 8); |
| kv8->values[i] = (u32) (u64) value; |
| return; |
| } |
| } |
| /* copy bucket contents to overflow hash tbl */ |
| for (i = 0; i < 5; i++) |
| { |
| kcopy = clib_mem_alloc (p->key_size); |
| clib_memcpy (kcopy, &kv8->kb.k_u64[i], 8); |
| hash_set_mem (p->overflow_hash, kcopy, kv8->values[i]); |
| p->nitems_in_overflow++; |
| } |
| } |
| else |
| { |
| for (i = 0; i < 4; i++) |
| { |
| if (kv8v8->values[i] == (u64) ~ 0) |
| { |
| clib_memcpy (&kv8v8->kb.k_u64[i], key, 8); |
| kv8v8->values[i] = (u64) value; |
| return; |
| } |
| } |
| /* copy bucket contents to overflow hash tbl */ |
| for (i = 0; i < 4; i++) |
| { |
| kcopy = clib_mem_alloc (p->key_size); |
| clib_memcpy (kcopy, &kv8v8->kb.k_u64[i], 8); |
| hash_set_mem (p->overflow_hash, kcopy, kv8v8->values[i]); |
| p->nitems_in_overflow++; |
| } |
| |
| } |
| /* Add new key to overflow */ |
| kcopy = clib_mem_alloc (p->key_size); |
| clib_memcpy (kcopy, key, p->key_size); |
| hash_set_mem (p->overflow_hash, kcopy, value); |
| p->buckets[bucket] = PFHASH_BUCKET_OVERFLOW; |
| p->overflow_count++; |
| p->nitems_in_overflow++; |
| return; |
| |
| case 4: |
| for (i = 0; i < 8; i++) |
| { |
| if (kv4->values[i] == (u32) ~ 0) |
| { |
| clib_memcpy (&kv4->kb.kb[i], key, 4); |
| kv4->values[i] = (u32) (u64) value; |
| return; |
| } |
| } |
| /* copy bucket contents to overflow hash tbl */ |
| for (i = 0; i < 8; i++) |
| { |
| kcopy = clib_mem_alloc (p->key_size); |
| clib_memcpy (kcopy, &kv4->kb.kb[i], 4); |
| hash_set_mem (p->overflow_hash, kcopy, kv4->values[i]); |
| p->nitems_in_overflow++; |
| } |
| /* Add new key to overflow */ |
| kcopy = clib_mem_alloc (p->key_size); |
| clib_memcpy (kcopy, key, p->key_size); |
| hash_set_mem (p->overflow_hash, kcopy, value); |
| p->buckets[bucket] = PFHASH_BUCKET_OVERFLOW; |
| p->overflow_count++; |
| p->nitems_in_overflow++; |
| return; |
| |
| default: |
| ASSERT (0); |
| } |
| } |
| |
| void |
| pfhash_unset (pfhash_t * p, u32 bucket, void *key) |
| { |
| u32 bucket_contents = pfhash_read_bucket_prefetch_kv (p, bucket); |
| u32 match_index = (u32) ~ 0; |
| pfhash_kv_t *kv; |
| pfhash_kv_16_t *kv16; |
| pfhash_kv_8_t *kv8; |
| pfhash_kv_8v8_t *kv8v8; |
| pfhash_kv_4_t *kv4; |
| void *oldkey; |
| |
| if (bucket_contents == PFHASH_BUCKET_OVERFLOW) |
| { |
| hash_pair_t *hp; |
| hp = hash_get_pair_mem (p->overflow_hash, key); |
| if (hp) |
| { |
| oldkey = (void *) hp->key; |
| hash_unset_mem (p->overflow_hash, key); |
| clib_mem_free (oldkey); |
| p->nitems--; |
| p->nitems_in_overflow--; |
| } |
| return; |
| } |
| |
| kv = pfhash_get_internal (p, bucket_contents, key, &match_index); |
| if (match_index == (u32) ~ 0) |
| return; |
| |
| p->nitems--; |
| |
| kv16 = (void *) kv; |
| kv8 = (void *) kv; |
| kv8v8 = (void *) kv; |
| kv4 = (void *) kv; |
| |
| switch (p->key_size) |
| { |
| case 16: |
| kv16->values[match_index] = (u32) ~ 0; |
| return; |
| |
| case 8: |
| if (p->value_size == 4) |
| kv8->values[match_index] = (u32) ~ 0; |
| else |
| kv8v8->values[match_index] = (u64) ~ 0; |
| return; |
| |
| case 4: |
| kv4->values[match_index] = (u32) ~ 0; |
| return; |
| |
| default: |
| ASSERT (0); |
| } |
| } |
| |
| void |
| pfhash_free (pfhash_t * p) |
| { |
| hash_pair_t *hp; |
| int i; |
| u8 **keys = 0; |
| |
| vec_free (p->name); |
| |
| pool_free (p->kvp); |
| |
| /* *INDENT-OFF* */ |
| hash_foreach_pair (hp, p->overflow_hash, |
| ({ |
| vec_add1 (keys, (u8 *)hp->key); |
| })); |
| /* *INDENT-ON* */ |
| hash_free (p->overflow_hash); |
| for (i = 0; i < vec_len (keys); i++) |
| vec_free (keys[i]); |
| vec_free (keys); |
| } |
| |
| #endif |
| |
| /* |
| * fd.io coding-style-patch-verification: ON |
| * |
| * Local Variables: |
| * eval: (c-set-style "gnu") |
| * End: |
| */ |