From: Konstantin Ananyev <konstantin.ananyev@intel.com>
To: dev@dpdk.org
Subject: [dpdk-dev] [PATCH v2 17/17] libte_acl: remove unused macros.
Date: Mon, 12 Jan 2015 19:16:21 +0000 [thread overview]
Message-ID: <1421090181-17150-18-git-send-email-konstantin.ananyev@intel.com> (raw)
In-Reply-To: <1421090181-17150-1-git-send-email-konstantin.ananyev@intel.com>
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
---
lib/librte_acl/acl.h | 39 ++++++++++++++++++++++++++++++++++++++-
lib/librte_acl/acl_run.h | 1 -
2 files changed, 38 insertions(+), 2 deletions(-)
diff --git a/lib/librte_acl/acl.h b/lib/librte_acl/acl.h
index 61b849a..e65e079 100644
--- a/lib/librte_acl/acl.h
+++ b/lib/librte_acl/acl.h
@@ -62,13 +62,50 @@ struct rte_acl_bitset {
#define RTE_ACL_NODE_DFA (0 << RTE_ACL_TYPE_SHIFT)
#define RTE_ACL_NODE_SINGLE (1U << RTE_ACL_TYPE_SHIFT)
-#define RTE_ACL_NODE_QEXACT (2U << RTE_ACL_TYPE_SHIFT)
#define RTE_ACL_NODE_QRANGE (3U << RTE_ACL_TYPE_SHIFT)
#define RTE_ACL_NODE_MATCH (4U << RTE_ACL_TYPE_SHIFT)
#define RTE_ACL_NODE_TYPE (7U << RTE_ACL_TYPE_SHIFT)
#define RTE_ACL_NODE_UNDEFINED UINT32_MAX
/*
+ * ACL RT structure is a set of multibit tries (with stride == 8)
+ * represented by an array of transitions. The next position is calculated
+ * based on the current position and the input byte.
+ * Each transition is 64 bit value with the following format:
+ * | node_type_specific : 32 | node_type : 3 | node_addr : 29 |
+ * For all node types except RTE_ACL_NODE_MATCH, node_addr is an index
+ * to the start of the node in the transtions array.
+ * Few different node types are used:
+ * RTE_ACL_NODE_MATCH:
+ * node_addr value is and index into an array that contains the return value
+ * and its priority for each category.
+ * Upper 32 bits of the transtion value are not used for that node type.
+ * RTE_ACL_NODE_QRANGE:
+ * that node consist of up to 5 transitions.
+ * Upper 32 bits are interpreted as 4 signed character values which
+ * are ordered from smallest(INT8_MIN) to largest (INT8_MAX).
+ * These values define 5 ranges:
+ * INT8_MIN <= range[0] <= ((int8_t *)&transition)[4]
+ * ((int8_t *)&transition)[4] < range[1] <= ((int8_t *)&transition)[5]
+ * ((int8_t *)&transition)[5] < range[2] <= ((int8_t *)&transition)[6]
+ * ((int8_t *)&transition)[6] < range[3] <= ((int8_t *)&transition)[7]
+ * ((int8_t *)&transition)[7] < range[4] <= INT8_MAX
+ * So for input byte value within range[i] i-th transition within that node
+ * will be used.
+ * RTE_ACL_NODE_SINGLE:
+ * always transitions to the same node regardless of the input value.
+ * RTE_ACL_NODE_DFA:
+ * that node consits of up to 256 transitions.
+ * In attempt to conserve space all transitions are divided into 4 consecutive
+ * groups, by 64 transitions per group:
+ * group64[i] contains transitions[i * 64, .. i * 64 + 63].
+ * Upper 32 bits are interpreted as 4 unsigned character values one per group,
+ * which contain index to the start of the given group within the node.
+ * So to calculate transition index within the node for given input byte value:
+ * input_byte - ((uint8_t *)&transition)[4 + input_byte / 64].
+ */
+
+/*
* Structure of a node is a set of ptrs and each ptr has a bit map
* of values associated with this transition.
*/
diff --git a/lib/librte_acl/acl_run.h b/lib/librte_acl/acl_run.h
index 850bc81..b2fc42c 100644
--- a/lib/librte_acl/acl_run.h
+++ b/lib/librte_acl/acl_run.h
@@ -40,7 +40,6 @@
#define MAX_SEARCHES_AVX16 16
#define MAX_SEARCHES_SSE8 8
#define MAX_SEARCHES_SSE4 4
-#define MAX_SEARCHES_SSE2 2
#define MAX_SEARCHES_SCALAR 2
#define GET_NEXT_4BYTES(prm, idx) \
--
1.8.5.3
next prev parent reply other threads:[~2015-01-12 19:16 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-01-12 19:16 [dpdk-dev] [PATCH v2 00/17] ACL: New AVX2 classify method and several other enhancements Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 01/17] fix fix compilation issues with RTE_LIBRTE_ACL_STANDALONE=y Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 02/17] app/test: few small fixes fot test_acl.c Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 03/17] librte_acl: make data_indexes long enough to survive idle transitions Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 04/17] librte_acl: remove build phase heuristsic with negative perfomance effect Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 05/17] librte_acl: fix a bug at build phase that can cause matches beeing overwirtten Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 06/17] librte_acl: introduce DFA nodes compression (group64) for identical entries Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 07/17] librte_acl: build/gen phase - simplify the way match nodes are allocated Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 08/17] librte_acl: make scalar RT code to be more similar to vector one Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 09/17] librte_acl: a bit of RT code deduplication Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 10/17] EAL: introduce rte_ymm and relatives in rte_common_vect.h Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 11/17] librte_acl: add AVX2 as new rte_acl_classify() method Konstantin Ananyev
2015-01-19 17:22 ` Thomas Monjalon
2015-01-20 10:56 ` Ananyev, Konstantin
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 12/17] test-acl: add ability to manually select RT method Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 13/17] librte_acl: Remove search_sse_2 and relatives Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 14/17] libter_acl: move lo/hi dwords shuffle out from calc_addr Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 15/17] libte_acl: make calc_addr a define to deduplicate the code Konstantin Ananyev
2015-01-12 19:16 ` [dpdk-dev] [PATCH v2 16/17] libte_acl: introduce max_size into rte_acl_config Konstantin Ananyev
2015-01-12 19:16 ` Konstantin Ananyev [this message]
2015-01-19 17:17 ` [dpdk-dev] [PATCH v2 17/17] libte_acl: remove unused macros Thomas Monjalon
2015-01-20 10:09 ` Ananyev, Konstantin
2015-01-20 10:48 ` Jim Thompson
[not found] ` <2601191342CEEE43887BDE71AB977258213DE0BB@irsmsx105.ger.corp.intel.com>
2015-01-20 11:11 ` Ananyev, Konstantin
2015-01-20 12:26 ` Thomas Monjalon
2015-01-14 18:39 ` [dpdk-dev] [PATCH v2 00/17] ACL: New AVX2 classify method and several other enhancements Neil Horman
2015-01-19 17:16 ` Thomas Monjalon
2015-01-19 18:39 ` Neil Horman
2015-01-20 10:11 ` Ananyev, Konstantin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1421090181-17150-18-git-send-email-konstantin.ananyev@intel.com \
--to=konstantin.ananyev@intel.com \
--cc=dev@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).