SHUFFLE32_SWAP64 = 0x4e,
};
-static const rte_xmm_t mm_type_quad_range = {
- .u32 = {
- RTE_ACL_NODE_QRANGE,
- RTE_ACL_NODE_QRANGE,
- RTE_ACL_NODE_QRANGE,
- RTE_ACL_NODE_QRANGE,
- },
-};
-
-static const rte_xmm_t mm_type_quad_range64 = {
- .u32 = {
- RTE_ACL_NODE_QRANGE,
- RTE_ACL_NODE_QRANGE,
- 0,
- 0,
- },
-};
-
static const rte_xmm_t mm_shuffle_input = {
.u32 = {0x00000000, 0x04040404, 0x08080808, 0x0c0c0c0c},
};
.u16 = {1, 1, 1, 1, 1, 1, 1, 1},
};
-static const rte_xmm_t mm_bytes = {
- .u32 = {UINT8_MAX, UINT8_MAX, UINT8_MAX, UINT8_MAX},
-};
-
-static const rte_xmm_t mm_bytes64 = {
- .u32 = {UINT8_MAX, UINT8_MAX, 0, 0},
-};
-
static const rte_xmm_t mm_match_mask = {
.u32 = {
RTE_ACL_NODE_MATCH,
* Extract transitions from an XMM register and check for any matches
*/
static void
-acl_process_matches(xmm_t *indicies, int slot, const struct rte_acl_ctx *ctx,
+acl_process_matches(xmm_t *indices, int slot, const struct rte_acl_ctx *ctx,
struct parms *parms, struct acl_flow_data *flows)
{
uint64_t transition1, transition2;
/* extract transition from low 64 bits. */
- transition1 = MM_CVT64(*indicies);
+ transition1 = MM_CVT64(*indices);
/* extract transition from high 64 bits. */
- *indicies = MM_SHUFFLE32(*indicies, SHUFFLE32_SWAP64);
- transition2 = MM_CVT64(*indicies);
+ *indices = MM_SHUFFLE32(*indices, SHUFFLE32_SWAP64);
+ transition2 = MM_CVT64(*indices);
transition1 = acl_match_check(transition1, slot, ctx,
parms, flows, resolve_priority_sse);
transition2 = acl_match_check(transition2, slot + 1, ctx,
parms, flows, resolve_priority_sse);
- /* update indicies with new transitions. */
- *indicies = MM_SET64(transition2, transition1);
+ /* update indices with new transitions. */
+ *indices = MM_SET64(transition2, transition1);
}
/*
*/
static inline void
acl_match_check_x2(int slot, const struct rte_acl_ctx *ctx, struct parms *parms,
- struct acl_flow_data *flows, xmm_t *indicies, xmm_t match_mask)
+ struct acl_flow_data *flows, xmm_t *indices, xmm_t match_mask)
{
xmm_t temp;
- temp = MM_AND(match_mask, *indicies);
+ temp = MM_AND(match_mask, *indices);
while (!MM_TESTZ(temp, temp)) {
- acl_process_matches(indicies, slot, ctx, parms, flows);
- temp = MM_AND(match_mask, *indicies);
+ acl_process_matches(indices, slot, ctx, parms, flows);
+ temp = MM_AND(match_mask, *indices);
}
}
*/
static inline void
acl_match_check_x4(int slot, const struct rte_acl_ctx *ctx, struct parms *parms,
- struct acl_flow_data *flows, xmm_t *indicies1, xmm_t *indicies2,
+ struct acl_flow_data *flows, xmm_t *indices1, xmm_t *indices2,
xmm_t match_mask)
{
xmm_t temp;
/* put low 32 bits of each transition into one register */
- temp = (xmm_t)MM_SHUFFLEPS((__m128)*indicies1, (__m128)*indicies2,
+ temp = (xmm_t)MM_SHUFFLEPS((__m128)*indices1, (__m128)*indices2,
0x88);
/* test for match node */
temp = MM_AND(match_mask, temp);
while (!MM_TESTZ(temp, temp)) {
- acl_process_matches(indicies1, slot, ctx, parms, flows);
- acl_process_matches(indicies2, slot + 2, ctx, parms, flows);
+ acl_process_matches(indices1, slot, ctx, parms, flows);
+ acl_process_matches(indices2, slot + 2, ctx, parms, flows);
- temp = (xmm_t)MM_SHUFFLEPS((__m128)*indicies1,
- (__m128)*indicies2,
+ temp = (xmm_t)MM_SHUFFLEPS((__m128)*indices1,
+ (__m128)*indices2,
0x88);
temp = MM_AND(match_mask, temp);
}
*/
static inline xmm_t
acl_calc_addr(xmm_t index_mask, xmm_t next_input, xmm_t shuffle_input,
- xmm_t ones_16, xmm_t bytes, xmm_t type_quad_range,
- xmm_t *indicies1, xmm_t *indicies2)
+ xmm_t ones_16, xmm_t indices1, xmm_t indices2)
{
- xmm_t addr, node_types, temp;
+ xmm_t addr, node_types, range, temp;
+ xmm_t dfa_msk, dfa_ofs, quad_ofs;
+ xmm_t in, r, t;
+
+ const xmm_t range_base = _mm_set_epi32(0xffffff0c, 0xffffff08,
+ 0xffffff04, 0xffffff00);
/*
* Note that no transition is done for a match
* it reaches a match.
*/
- /* Shuffle low 32 into temp and high 32 into indicies2 */
- temp = (xmm_t)MM_SHUFFLEPS((__m128)*indicies1, (__m128)*indicies2,
- 0x88);
- *indicies2 = (xmm_t)MM_SHUFFLEPS((__m128)*indicies1,
- (__m128)*indicies2, 0xdd);
+ /* Shuffle low 32 into temp and high 32 into indices2 */
+ temp = (xmm_t)MM_SHUFFLEPS((__m128)indices1, (__m128)indices2, 0x88);
+ range = (xmm_t)MM_SHUFFLEPS((__m128)indices1, (__m128)indices2, 0xdd);
+
+ t = MM_XOR(index_mask, index_mask);
+
+ /* shuffle input byte to all 4 positions of 32 bit value */
+ in = MM_SHUFFLE8(next_input, shuffle_input);
/* Calc node type and node addr */
node_types = MM_ANDNOT(index_mask, temp);
*/
/* mask for DFA type (0) nodes */
- temp = MM_CMPEQ32(node_types, MM_XOR(node_types, node_types));
+ dfa_msk = MM_CMPEQ32(node_types, t);
- /* add input byte to DFA position */
- temp = MM_AND(temp, bytes);
- temp = MM_AND(temp, next_input);
- addr = MM_ADD32(addr, temp);
+ r = _mm_srli_epi32(in, 30);
+ r = _mm_add_epi8(r, range_base);
- /*
- * Calc addr for Range nodes -> range_index + range(input)
- */
- node_types = MM_CMPEQ32(node_types, type_quad_range);
+ t = _mm_srli_epi32(in, 24);
+ r = _mm_shuffle_epi8(range, r);
+
+ dfa_ofs = _mm_sub_epi32(t, r);
/*
* Calculate number of range boundaries that are less than the
* input value. Range boundaries for each node are in signed 8 bit,
- * ordered from -128 to 127 in the indicies2 register.
+ * ordered from -128 to 127 in the indices2 register.
* This is effectively a popcnt of bytes that are greater than the
* input byte.
*/
- /* shuffle input byte to all 4 positions of 32 bit value */
- temp = MM_SHUFFLE8(next_input, shuffle_input);
-
/* check ranges */
- temp = MM_CMPGT8(temp, *indicies2);
+ temp = MM_CMPGT8(in, range);
/* convert -1 to 1 (bytes greater than input byte */
temp = MM_SIGN8(temp, temp);
temp = MM_MADD8(temp, temp);
/* horizontal add pairs of words into dwords */
- temp = MM_MADD16(temp, ones_16);
+ quad_ofs = MM_MADD16(temp, ones_16);
/* mask to range type nodes */
- temp = MM_AND(temp, node_types);
+ temp = _mm_blendv_epi8(quad_ofs, dfa_ofs, dfa_msk);
/* add index into node position */
return MM_ADD32(addr, temp);
*/
static inline xmm_t
transition4(xmm_t index_mask, xmm_t next_input, xmm_t shuffle_input,
- xmm_t ones_16, xmm_t bytes, xmm_t type_quad_range,
- const uint64_t *trans, xmm_t *indicies1, xmm_t *indicies2)
+ xmm_t ones_16, const uint64_t *trans,
+ xmm_t *indices1, xmm_t *indices2)
{
xmm_t addr;
uint64_t trans0, trans2;
/* Calculate the address (array index) for all 4 transitions. */
addr = acl_calc_addr(index_mask, next_input, shuffle_input, ones_16,
- bytes, type_quad_range, indicies1, indicies2);
+ *indices1, *indices2);
/* Gather 64 bit transitions and pack back into 2 registers. */
/* {x2, x1, x2, x3} -> {x1, x1, x2, x3} */
addr = MM_SHUFFLE32(addr, SHUFFLE32_SLOT1);
- *indicies1 = MM_SET64(trans[MM_CVT32(addr)], trans0);
+ *indices1 = MM_SET64(trans[MM_CVT32(addr)], trans0);
/* get slot 3 */
/* {x1, x1, x2, x3} -> {x3, x1, x2, x3} */
addr = MM_SHUFFLE32(addr, SHUFFLE32_SLOT3);
- *indicies2 = MM_SET64(trans[MM_CVT32(addr)], trans2);
+ *indices2 = MM_SET64(trans[MM_CVT32(addr)], trans2);
return MM_SRL32(next_input, 8);
}
struct completion cmplt[MAX_SEARCHES_SSE8];
struct parms parms[MAX_SEARCHES_SSE8];
xmm_t input0, input1;
- xmm_t indicies1, indicies2, indicies3, indicies4;
+ xmm_t indices1, indices2, indices3, indices4;
acl_set_flow(&flows, cmplt, RTE_DIM(cmplt), data, results,
total_packets, categories, ctx->trans_table);
}
/*
- * indicies1 contains index_array[0,1]
- * indicies2 contains index_array[2,3]
- * indicies3 contains index_array[4,5]
- * indicies4 contains index_array[6,7]
+ * indices1 contains index_array[0,1]
+ * indices2 contains index_array[2,3]
+ * indices3 contains index_array[4,5]
+ * indices4 contains index_array[6,7]
*/
- indicies1 = MM_LOADU((xmm_t *) &index_array[0]);
- indicies2 = MM_LOADU((xmm_t *) &index_array[2]);
+ indices1 = MM_LOADU((xmm_t *) &index_array[0]);
+ indices2 = MM_LOADU((xmm_t *) &index_array[2]);
- indicies3 = MM_LOADU((xmm_t *) &index_array[4]);
- indicies4 = MM_LOADU((xmm_t *) &index_array[6]);
+ indices3 = MM_LOADU((xmm_t *) &index_array[4]);
+ indices4 = MM_LOADU((xmm_t *) &index_array[6]);
/* Check for any matches. */
acl_match_check_x4(0, ctx, parms, &flows,
- &indicies1, &indicies2, mm_match_mask.m);
+ &indices1, &indices2, mm_match_mask.m);
acl_match_check_x4(4, ctx, parms, &flows,
- &indicies3, &indicies4, mm_match_mask.m);
+ &indices3, &indices4, mm_match_mask.m);
while (flows.started > 0) {
input0 = transition4(mm_index_mask.m, input0,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies1, &indicies2);
+ flows.trans, &indices1, &indices2);
input1 = transition4(mm_index_mask.m, input1,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies3, &indicies4);
+ flows.trans, &indices3, &indices4);
input0 = transition4(mm_index_mask.m, input0,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies1, &indicies2);
+ flows.trans, &indices1, &indices2);
input1 = transition4(mm_index_mask.m, input1,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies3, &indicies4);
+ flows.trans, &indices3, &indices4);
input0 = transition4(mm_index_mask.m, input0,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies1, &indicies2);
+ flows.trans, &indices1, &indices2);
input1 = transition4(mm_index_mask.m, input1,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies3, &indicies4);
+ flows.trans, &indices3, &indices4);
input0 = transition4(mm_index_mask.m, input0,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies1, &indicies2);
+ flows.trans, &indices1, &indices2);
input1 = transition4(mm_index_mask.m, input1,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies3, &indicies4);
+ flows.trans, &indices3, &indices4);
/* Check for any matches. */
acl_match_check_x4(0, ctx, parms, &flows,
- &indicies1, &indicies2, mm_match_mask.m);
+ &indices1, &indices2, mm_match_mask.m);
acl_match_check_x4(4, ctx, parms, &flows,
- &indicies3, &indicies4, mm_match_mask.m);
+ &indices3, &indices4, mm_match_mask.m);
}
return 0;
uint64_t index_array[MAX_SEARCHES_SSE4];
struct completion cmplt[MAX_SEARCHES_SSE4];
struct parms parms[MAX_SEARCHES_SSE4];
- xmm_t input, indicies1, indicies2;
+ xmm_t input, indices1, indices2;
acl_set_flow(&flows, cmplt, RTE_DIM(cmplt), data, results,
total_packets, categories, ctx->trans_table);
index_array[n] = acl_start_next_trie(&flows, parms, n, ctx);
}
- indicies1 = MM_LOADU((xmm_t *) &index_array[0]);
- indicies2 = MM_LOADU((xmm_t *) &index_array[2]);
+ indices1 = MM_LOADU((xmm_t *) &index_array[0]);
+ indices2 = MM_LOADU((xmm_t *) &index_array[2]);
/* Check for any matches. */
acl_match_check_x4(0, ctx, parms, &flows,
- &indicies1, &indicies2, mm_match_mask.m);
+ &indices1, &indices2, mm_match_mask.m);
while (flows.started > 0) {
/* Process the 4 bytes of input on each stream. */
input = transition4(mm_index_mask.m, input,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies1, &indicies2);
+ flows.trans, &indices1, &indices2);
input = transition4(mm_index_mask.m, input,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies1, &indicies2);
+ flows.trans, &indices1, &indices2);
input = transition4(mm_index_mask.m, input,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies1, &indicies2);
+ flows.trans, &indices1, &indices2);
input = transition4(mm_index_mask.m, input,
mm_shuffle_input.m, mm_ones_16.m,
- mm_bytes.m, mm_type_quad_range.m,
- flows.trans, &indicies1, &indicies2);
+ flows.trans, &indices1, &indices2);
/* Check for any matches. */
acl_match_check_x4(0, ctx, parms, &flows,
- &indicies1, &indicies2, mm_match_mask.m);
+ &indices1, &indices2, mm_match_mask.m);
}
return 0;
static inline xmm_t
transition2(xmm_t index_mask, xmm_t next_input, xmm_t shuffle_input,
- xmm_t ones_16, xmm_t bytes, xmm_t type_quad_range,
- const uint64_t *trans, xmm_t *indicies1)
+ xmm_t ones_16, const uint64_t *trans, xmm_t *indices1)
{
uint64_t t;
- xmm_t addr, indicies2;
+ xmm_t addr, indices2;
- indicies2 = MM_XOR(ones_16, ones_16);
+ indices2 = MM_XOR(ones_16, ones_16);
addr = acl_calc_addr(index_mask, next_input, shuffle_input, ones_16,
- bytes, type_quad_range, indicies1, &indicies2);
+ *indices1, indices2);
/* Gather 64 bit transitions and pack 2 per register. */
/* get slot 1 */
addr = MM_SHUFFLE32(addr, SHUFFLE32_SLOT1);
- *indicies1 = MM_SET64(trans[MM_CVT32(addr)], t);
+ *indices1 = MM_SET64(trans[MM_CVT32(addr)], t);
return MM_SRL32(next_input, 8);
}
uint64_t index_array[MAX_SEARCHES_SSE2];
struct completion cmplt[MAX_SEARCHES_SSE2];
struct parms parms[MAX_SEARCHES_SSE2];
- xmm_t input, indicies;
+ xmm_t input, indices;
acl_set_flow(&flows, cmplt, RTE_DIM(cmplt), data, results,
total_packets, categories, ctx->trans_table);
index_array[n] = acl_start_next_trie(&flows, parms, n, ctx);
}
- indicies = MM_LOADU((xmm_t *) &index_array[0]);
+ indices = MM_LOADU((xmm_t *) &index_array[0]);
/* Check for any matches. */
- acl_match_check_x2(0, ctx, parms, &flows, &indicies, mm_match_mask64.m);
+ acl_match_check_x2(0, ctx, parms, &flows, &indices, mm_match_mask64.m);
while (flows.started > 0) {
input = transition2(mm_index_mask64.m, input,
mm_shuffle_input64.m, mm_ones_16.m,
- mm_bytes64.m, mm_type_quad_range64.m,
- flows.trans, &indicies);
+ flows.trans, &indices);
input = transition2(mm_index_mask64.m, input,
mm_shuffle_input64.m, mm_ones_16.m,
- mm_bytes64.m, mm_type_quad_range64.m,
- flows.trans, &indicies);
+ flows.trans, &indices);
input = transition2(mm_index_mask64.m, input,
mm_shuffle_input64.m, mm_ones_16.m,
- mm_bytes64.m, mm_type_quad_range64.m,
- flows.trans, &indicies);
+ flows.trans, &indices);
input = transition2(mm_index_mask64.m, input,
mm_shuffle_input64.m, mm_ones_16.m,
- mm_bytes64.m, mm_type_quad_range64.m,
- flows.trans, &indicies);
+ flows.trans, &indices);
/* Check for any matches. */
- acl_match_check_x2(0, ctx, parms, &flows, &indicies,
+ acl_match_check_x2(0, ctx, parms, &flows, &indices,
mm_match_mask64.m);
}
git.droids-corp.org / dpdk.git / blobdiff