numam-dpdk/lib/librte_acl/acl_run_altivec.h
Jerin Jacob c0583d98a9 eal: introduce macro for always inline
Different drivers use internal macros like force_inline for compiler
always inline feature.
Standardizing it through __rte_always_inline macro.

Verified the change by comparing the output binary file.
No difference found in the output binary file with this change.

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
2017-06-06 17:21:55 +02:00

330 lines
11 KiB
C

/*
* BSD LICENSE
*
* Copyright (C) IBM Corporation 2016.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
* * Neither the name of IBM Corporation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "acl_run.h"
#include "acl_vect.h"
struct _altivec_acl_const {
rte_xmm_t xmm_shuffle_input;
rte_xmm_t xmm_index_mask;
rte_xmm_t xmm_ones_16;
rte_xmm_t range_base;
} altivec_acl_const __attribute__((aligned(RTE_CACHE_LINE_SIZE))) = {
{
.u32 = {0x00000000, 0x04040404, 0x08080808, 0x0c0c0c0c}
},
{
.u32 = {RTE_ACL_NODE_INDEX, RTE_ACL_NODE_INDEX,
RTE_ACL_NODE_INDEX, RTE_ACL_NODE_INDEX}
},
{
.u16 = {1, 1, 1, 1, 1, 1, 1, 1}
},
{
.u32 = {0xffffff00, 0xffffff04, 0xffffff08, 0xffffff0c}
},
};
/*
* Resolve priority for multiple results (altivec version).
* This consists comparing the priority of the current traversal with the
* running set of results for the packet.
* For each result, keep a running array of the result (rule number) and
* its priority for each category.
*/
static inline void
resolve_priority_altivec(uint64_t transition, int n,
const struct rte_acl_ctx *ctx, struct parms *parms,
const struct rte_acl_match_results *p, uint32_t categories)
{
uint32_t x;
xmm_t results, priority, results1, priority1;
vector bool int selector;
xmm_t *saved_results, *saved_priority;
for (x = 0; x < categories; x += RTE_ACL_RESULTS_MULTIPLIER) {
saved_results = (xmm_t *)(&parms[n].cmplt->results[x]);
saved_priority =
(xmm_t *)(&parms[n].cmplt->priority[x]);
/* get results and priorities for completed trie */
results = *(const xmm_t *)&p[transition].results[x];
priority = *(const xmm_t *)&p[transition].priority[x];
/* if this is not the first completed trie */
if (parms[n].cmplt->count != ctx->num_tries) {
/* get running best results and their priorities */
results1 = *saved_results;
priority1 = *saved_priority;
/* select results that are highest priority */
selector = vec_cmpgt(priority1, priority);
results = vec_sel(results, results1, selector);
priority = vec_sel(priority, priority1,
selector);
}
/* save running best results and their priorities */
*saved_results = results;
*saved_priority = priority;
}
}
/*
* Check for any match in 4 transitions
*/
static __rte_always_inline uint32_t
check_any_match_x4(uint64_t val[])
{
return (val[0] | val[1] | val[2] | val[3]) & RTE_ACL_NODE_MATCH;
}
static __rte_always_inline void
acl_match_check_x4(int slot, const struct rte_acl_ctx *ctx, struct parms *parms,
struct acl_flow_data *flows, uint64_t transitions[])
{
while (check_any_match_x4(transitions)) {
transitions[0] = acl_match_check(transitions[0], slot, ctx,
parms, flows, resolve_priority_altivec);
transitions[1] = acl_match_check(transitions[1], slot + 1, ctx,
parms, flows, resolve_priority_altivec);
transitions[2] = acl_match_check(transitions[2], slot + 2, ctx,
parms, flows, resolve_priority_altivec);
transitions[3] = acl_match_check(transitions[3], slot + 3, ctx,
parms, flows, resolve_priority_altivec);
}
}
/*
* Process 4 transitions (in 2 XMM registers) in parallel
*/
static inline __attribute__((optimize("O2"))) xmm_t
transition4(xmm_t next_input, const uint64_t *trans,
xmm_t *indices1, xmm_t *indices2)
{
xmm_t addr, tr_lo, tr_hi;
xmm_t in, node_type, r, t;
xmm_t dfa_ofs, quad_ofs;
xmm_t *index_mask, *tp;
vector bool int dfa_msk;
vector signed char zeroes = {};
union {
uint64_t d64[2];
uint32_t d32[4];
} v;
/* Move low 32 into tr_lo and high 32 into tr_hi */
tr_lo = (xmm_t){(*indices1)[0], (*indices1)[2],
(*indices2)[0], (*indices2)[2]};
tr_hi = (xmm_t){(*indices1)[1], (*indices1)[3],
(*indices2)[1], (*indices2)[3]};
/* Calculate the address (array index) for all 4 transitions. */
index_mask = (xmm_t *)&altivec_acl_const.xmm_index_mask.u32;
t = vec_xor(*index_mask, *index_mask);
in = vec_perm(next_input, (xmm_t){},
*(vector unsigned char *)&altivec_acl_const.xmm_shuffle_input);
/* Calc node type and node addr */
node_type = vec_and(vec_nor(*index_mask, *index_mask), tr_lo);
addr = vec_and(tr_lo, *index_mask);
/* mask for DFA type(0) nodes */
dfa_msk = vec_cmpeq(node_type, t);
/* DFA calculations. */
r = vec_sr(in, (vector unsigned int){30, 30, 30, 30});
tp = (xmm_t *)&altivec_acl_const.range_base.u32;
r = vec_add(r, *tp);
t = vec_sr(in, (vector unsigned int){24, 24, 24, 24});
r = vec_perm(tr_hi, (xmm_t){(uint16_t)0 << 16},
(vector unsigned char)r);
dfa_ofs = vec_sub(t, r);
/* QUAD/SINGLE caluclations. */
t = (xmm_t)vec_cmpgt((vector signed char)in, (vector signed char)tr_hi);
t = (xmm_t)vec_sel(
vec_sel(
(vector signed char)vec_sub(
zeroes, (vector signed char)t),
(vector signed char)t,
vec_cmpgt((vector signed char)t, zeroes)),
zeroes,
vec_cmpeq((vector signed char)t, zeroes));
t = (xmm_t)vec_msum((vector signed char)t,
(vector unsigned char)t, (xmm_t){});
quad_ofs = (xmm_t)vec_msum((vector signed short)t,
*(vector signed short *)&altivec_acl_const.xmm_ones_16.u16,
(xmm_t){});
/* blend DFA and QUAD/SINGLE. */
t = vec_sel(quad_ofs, dfa_ofs, dfa_msk);
/* calculate address for next transitions. */
addr = vec_add(addr, t);
v.d64[0] = (uint64_t)trans[addr[0]];
v.d64[1] = (uint64_t)trans[addr[1]];
*indices1 = (xmm_t){v.d32[0], v.d32[1], v.d32[2], v.d32[3]};
v.d64[0] = (uint64_t)trans[addr[2]];
v.d64[1] = (uint64_t)trans[addr[3]];
*indices2 = (xmm_t){v.d32[0], v.d32[1], v.d32[2], v.d32[3]};
return vec_sr(next_input,
(vector unsigned int){CHAR_BIT, CHAR_BIT, CHAR_BIT, CHAR_BIT});
}
/*
* Execute trie traversal with 8 traversals in parallel
*/
static inline int
search_altivec_8(const struct rte_acl_ctx *ctx, const uint8_t **data,
uint32_t *results, uint32_t total_packets, uint32_t categories)
{
int n;
struct acl_flow_data flows;
uint64_t index_array[MAX_SEARCHES_ALTIVEC8];
struct completion cmplt[MAX_SEARCHES_ALTIVEC8];
struct parms parms[MAX_SEARCHES_ALTIVEC8];
xmm_t input0, input1;
acl_set_flow(&flows, cmplt, RTE_DIM(cmplt), data, results,
total_packets, categories, ctx->trans_table);
for (n = 0; n < MAX_SEARCHES_ALTIVEC8; n++) {
cmplt[n].count = 0;
index_array[n] = acl_start_next_trie(&flows, parms, n, ctx);
}
/* Check for any matches. */
acl_match_check_x4(0, ctx, parms, &flows, (uint64_t *)&index_array[0]);
acl_match_check_x4(4, ctx, parms, &flows, (uint64_t *)&index_array[4]);
while (flows.started > 0) {
/* Gather 4 bytes of input data for each stream. */
input0 = (xmm_t){GET_NEXT_4BYTES(parms, 0),
GET_NEXT_4BYTES(parms, 1),
GET_NEXT_4BYTES(parms, 2),
GET_NEXT_4BYTES(parms, 3)};
input1 = (xmm_t){GET_NEXT_4BYTES(parms, 4),
GET_NEXT_4BYTES(parms, 5),
GET_NEXT_4BYTES(parms, 6),
GET_NEXT_4BYTES(parms, 7)};
/* Process the 4 bytes of input on each stream. */
input0 = transition4(input0, flows.trans,
(xmm_t *)&index_array[0], (xmm_t *)&index_array[2]);
input1 = transition4(input1, flows.trans,
(xmm_t *)&index_array[4], (xmm_t *)&index_array[6]);
input0 = transition4(input0, flows.trans,
(xmm_t *)&index_array[0], (xmm_t *)&index_array[2]);
input1 = transition4(input1, flows.trans,
(xmm_t *)&index_array[4], (xmm_t *)&index_array[6]);
input0 = transition4(input0, flows.trans,
(xmm_t *)&index_array[0], (xmm_t *)&index_array[2]);
input1 = transition4(input1, flows.trans,
(xmm_t *)&index_array[4], (xmm_t *)&index_array[6]);
input0 = transition4(input0, flows.trans,
(xmm_t *)&index_array[0], (xmm_t *)&index_array[2]);
input1 = transition4(input1, flows.trans,
(xmm_t *)&index_array[4], (xmm_t *)&index_array[6]);
/* Check for any matches. */
acl_match_check_x4(0, ctx, parms, &flows,
(uint64_t *)&index_array[0]);
acl_match_check_x4(4, ctx, parms, &flows,
(uint64_t *)&index_array[4]);
}
return 0;
}
/*
* Execute trie traversal with 4 traversals in parallel
*/
static inline int
search_altivec_4(const struct rte_acl_ctx *ctx, const uint8_t **data,
uint32_t *results, int total_packets, uint32_t categories)
{
int n;
struct acl_flow_data flows;
uint64_t index_array[MAX_SEARCHES_ALTIVEC4];
struct completion cmplt[MAX_SEARCHES_ALTIVEC4];
struct parms parms[MAX_SEARCHES_ALTIVEC4];
xmm_t input;
acl_set_flow(&flows, cmplt, RTE_DIM(cmplt), data, results,
total_packets, categories, ctx->trans_table);
for (n = 0; n < MAX_SEARCHES_ALTIVEC4; n++) {
cmplt[n].count = 0;
index_array[n] = acl_start_next_trie(&flows, parms, n, ctx);
}
/* Check for any matches. */
acl_match_check_x4(0, ctx, parms, &flows, index_array);
while (flows.started > 0) {
/* Gather 4 bytes of input data for each stream. */
input = (xmm_t){GET_NEXT_4BYTES(parms, 0),
GET_NEXT_4BYTES(parms, 1),
GET_NEXT_4BYTES(parms, 2),
GET_NEXT_4BYTES(parms, 3)};
/* Process the 4 bytes of input on each stream. */
input = transition4(input, flows.trans,
(xmm_t *)&index_array[0], (xmm_t *)&index_array[2]);
input = transition4(input, flows.trans,
(xmm_t *)&index_array[0], (xmm_t *)&index_array[2]);
input = transition4(input, flows.trans,
(xmm_t *)&index_array[0], (xmm_t *)&index_array[2]);
input = transition4(input, flows.trans,
(xmm_t *)&index_array[0], (xmm_t *)&index_array[2]);
/* Check for any matches. */
acl_match_check_x4(0, ctx, parms, &flows, index_array);
}
return 0;
}