DPDK patches and discussions
 help / color / mirror / Atom feed
From: 孙越池 <sunyuechi@iscas.ac.cn>
To: "Medvedkin, Vladimir" <vladimir.medvedkin@intel.com>, dev@dpdk.org
Subject: Re: Re: [PATCH v4 2/3] lib/lpm: R-V V rte_lpm_lookupx4
Date: Wed, 11 Jun 2025 23:04:19 +0800 (GMT+08:00)	[thread overview]
Message-ID: <5b1f501.423af.1975f85459b.Coremail.sunyuechi@iscas.ac.cn> (raw)
In-Reply-To: <c2f7c2a6-d91e-46da-8cca-01df5cd313e2@intel.com>

Thanks, they have been updated in v5.


&gt; -----原始邮件-----
&gt; 发件人: "Medvedkin, Vladimir" <vladimir.medvedkin@intel.com>
&gt; 发送时间: 2025-06-11 19:52:57 (星期三)
&gt; 收件人: uk7b@foxmail.com, dev@dpdk.org
&gt; 抄送: "Sun Yuechi" <sunyuechi@iscas.ac.cn>, "Thomas Monjalon" <thomas@monjalon.net>, "Bruce Richardson" <bruce.richardson@intel.com>, "Stanislaw Kardach" <stanislaw.kardach@gmail.com>
&gt; 主题: Re: [PATCH v4 2/3] lib/lpm: R-V V rte_lpm_lookupx4
&gt; 
&gt; Hi Sun,
&gt; 
&gt; You did not address my previous comments regarding commit message. You 
&gt; can put everything you've wrote in this commit as a note and add 
&gt; meaningful description about what commit generally does, like (please 
&gt; correct if needed):
&gt; 
&gt; "Implement LPM lookupx4 routine for RISC-V architecture using RISC-V 
&gt; Vector Extension instruction set"
&gt; 
&gt; Everything else (performance tests, implementation thoughts and 
&gt; considerations, etc.) should be in the patch notes. For more information 
&gt; on what "patch notes" are, you may want refer to Git documentation [1].
&gt; 
&gt; [1] https://git-scm.com/docs/git-notes
&gt; 
&gt; On 05/06/2025 11:58, uk7b@foxmail.com wrote:
&gt; 
&gt; &gt; From: Sun Yuechi <sunyuechi@iscas.ac.cn>
&gt; &gt;
&gt; &gt; The initialization of vtbl_entry is not fully vectorized here because
&gt; &gt; doing so would require __riscv_vluxei32_v_u32m1, which is slower
&gt; &gt; than the scalar approach in this small-scale scenario.
&gt; &gt;
&gt; &gt; - Test: app/test/lpm_perf_autotest
&gt; &gt; - Platform: Banana Pi(BPI-F3)
&gt; &gt; - SoC: Spacemit X60 (8 cores with Vector extension)
&gt; &gt; - CPU Frequency: up to 1.6 GHz
&gt; &gt; - Cache: 256 KiB L1d ×8, 256 KiB L1i ×8, 1 MiB L2 ×2
&gt; &gt; - Memory: 16 GiB
&gt; &gt; - Kernel: Linux 6.6.36
&gt; &gt; - Compiler: GCC 14.2.0 (with RVV intrinsic support)
&gt; &gt;
&gt; &gt; Test results(LPM LookupX4):
&gt; &gt;      scalar: 5.7 cycles
&gt; &gt;      rvv:    4.6 cycles
&gt; &gt;
&gt; &gt; Signed-off-by: Sun Yuechi <sunyuechi@iscas.ac.cn>
&gt; &gt; ---
&gt; &gt;   MAINTAINERS           |  2 ++
&gt; &gt;   lib/lpm/meson.build   |  1 +
&gt; &gt;   lib/lpm/rte_lpm.h     |  2 ++
&gt; &gt;   lib/lpm/rte_lpm_rvv.h | 62 +++++++++++++++++++++++++++++++++++++++++++
&gt; &gt;   4 files changed, 67 insertions(+)
&gt; &gt;   create mode 100644 lib/lpm/rte_lpm_rvv.h
&gt; &gt;
&gt; &gt; diff --git a/MAINTAINERS b/MAINTAINERS
&gt; &gt; index 3e16789250..0f207ac129 100644
&gt; &gt; --- a/MAINTAINERS
&gt; &gt; +++ b/MAINTAINERS
&gt; &gt; @@ -340,6 +340,8 @@ M: Stanislaw Kardach <stanislaw.kardach@gmail.com>
&gt; &gt;   F: config/riscv/
&gt; &gt;   F: doc/guides/linux_gsg/cross_build_dpdk_for_riscv.rst
&gt; &gt;   F: lib/eal/riscv/
&gt; &gt; +M: sunyuechi <sunyuechi@iscas.ac.cn>
&gt; &gt; +F: lib/**/*rvv*
&gt; &gt;   
&gt; &gt;   Intel x86
&gt; &gt;   M: Bruce Richardson <bruce.richardson@intel.com>
&gt; &gt; diff --git a/lib/lpm/meson.build b/lib/lpm/meson.build
&gt; &gt; index fae4f79fb9..09133061e5 100644
&gt; &gt; --- a/lib/lpm/meson.build
&gt; &gt; +++ b/lib/lpm/meson.build
&gt; &gt; @@ -17,6 +17,7 @@ indirect_headers += files(
&gt; &gt;           'rte_lpm_scalar.h',
&gt; &gt;           'rte_lpm_sse.h',
&gt; &gt;           'rte_lpm_sve.h',
&gt; &gt; +        'rte_lpm_rvv.h',
&gt; &gt;   )
&gt; &gt;   deps += ['hash']
&gt; &gt;   deps += ['rcu']
&gt; &gt; diff --git a/lib/lpm/rte_lpm.h b/lib/lpm/rte_lpm.h
&gt; &gt; index 7df64f06b1..b06517206f 100644
&gt; &gt; --- a/lib/lpm/rte_lpm.h
&gt; &gt; +++ b/lib/lpm/rte_lpm.h
&gt; &gt; @@ -408,6 +408,8 @@ rte_lpm_lookupx4(const struct rte_lpm *lpm, xmm_t ip, uint32_t hop[4],
&gt; &gt;   #include "rte_lpm_altivec.h"
&gt; &gt;   #elif defined(RTE_ARCH_X86)
&gt; &gt;   #include "rte_lpm_sse.h"
&gt; &gt; +#elif defined(RTE_ARCH_RISCV) &amp;&amp; defined(RTE_RISCV_FEATURE_V)
&gt; &gt; +#include "rte_lpm_rvv.h"
&gt; &gt;   #else
&gt; &gt;   #include "rte_lpm_scalar.h"
&gt; &gt;   #endif
&gt; &gt; diff --git a/lib/lpm/rte_lpm_rvv.h b/lib/lpm/rte_lpm_rvv.h
&gt; &gt; new file mode 100644
&gt; &gt; index 0000000000..5f48fb2b32
&gt; &gt; --- /dev/null
&gt; &gt; +++ b/lib/lpm/rte_lpm_rvv.h
&gt; &gt; @@ -0,0 +1,62 @@
&gt; &gt; +/* SPDX-License-Identifier: BSD-3-Clause
&gt; &gt; + * Copyright (c) 2025 Institute of Software Chinese Academy of Sciences (ISCAS).
&gt; &gt; + */
&gt; &gt; +
&gt; &gt; +#ifndef _RTE_LPM_RVV_H_
&gt; &gt; +#define _RTE_LPM_RVV_H_
&gt; &gt; +
&gt; &gt; +#include <rte_vect.h>
&gt; &gt; +
&gt; &gt; +#include <rte_cpuflags.h>
&gt; &gt; +#include <riscv_vector.h>
&gt; &gt; +
&gt; &gt; +#ifdef __cplusplus
&gt; &gt; +extern "C" {
&gt; &gt; +#endif
&gt; &gt; +
&gt; &gt; +#define RTE_LPM_LOOKUP_SUCCESS 0x01000000
&gt; &gt; +#define RTE_LPM_VALID_EXT_ENTRY_BITMASK 0x03000000
&gt; &gt; +
&gt; &gt; +static inline void rte_lpm_lookupx4(
&gt; &gt; +	const struct rte_lpm *lpm, xmm_t ip, uint32_t hop[4], uint32_t defv)
&gt; &gt; +{
&gt; &gt; +	size_t vl = 4;
&gt; &gt; +
&gt; &gt; +	const uint32_t *tbl24_p = (const uint32_t *)lpm-&gt;tbl24;
&gt; &gt; +	uint32_t tbl_entries[4] = {
&gt; &gt; +		tbl24_p[((uint32_t)ip[0]) &gt;&gt; 8],
&gt; &gt; +		tbl24_p[((uint32_t)ip[1]) &gt;&gt; 8],
&gt; &gt; +		tbl24_p[((uint32_t)ip[2]) &gt;&gt; 8],
&gt; &gt; +		tbl24_p[((uint32_t)ip[3]) &gt;&gt; 8],
&gt; &gt; +	};
&gt; &gt; +	vuint32m1_t vtbl_entry = __riscv_vle32_v_u32m1(tbl_entries, vl);
&gt; &gt; +
&gt; &gt; +	vbool32_t mask = __riscv_vmseq_vx_u32m1_b32(
&gt; &gt; +	    __riscv_vand_vx_u32m1(vtbl_entry, RTE_LPM_VALID_EXT_ENTRY_BITMASK, vl),
&gt; &gt; +	    RTE_LPM_VALID_EXT_ENTRY_BITMASK, vl);
&gt; &gt; +
&gt; &gt; +	vuint32m1_t vtbl8_index = __riscv_vsll_vx_u32m1(
&gt; &gt; +	    __riscv_vadd_vv_u32m1(
&gt; &gt; +		__riscv_vsll_vx_u32m1(__riscv_vand_vx_u32m1(vtbl_entry, 0x00FFFFFF, vl), 8, vl),
&gt; &gt; +		__riscv_vand_vx_u32m1(
&gt; &gt; +		    __riscv_vle32_v_u32m1((const uint32_t *)&amp;ip, vl), 0x000000FF, vl),
&gt; &gt; +		vl),
&gt; &gt; +	    2, vl);
&gt; &gt; +
&gt; &gt; +	vtbl_entry = __riscv_vluxei32_v_u32m1_mu(
&gt; &gt; +	    mask, vtbl_entry, (const uint32_t *)(lpm-&gt;tbl8), vtbl8_index, vl);
&gt; &gt; +
&gt; &gt; +	vuint32m1_t vnext_hop = __riscv_vand_vx_u32m1(vtbl_entry, 0x00FFFFFF, vl);
&gt; &gt; +	mask = __riscv_vmseq_vx_u32m1_b32(
&gt; &gt; +	    __riscv_vand_vx_u32m1(vtbl_entry, RTE_LPM_LOOKUP_SUCCESS, vl), 0, vl);
&gt; &gt; +
&gt; &gt; +	vnext_hop = __riscv_vmerge_vxm_u32m1(vnext_hop, defv, mask, vl);
&gt; &gt; +
&gt; &gt; +	__riscv_vse32_v_u32m1(hop, vnext_hop, vl);
&gt; &gt; +}
&gt; &gt; +
&gt; &gt; +#ifdef __cplusplus
&gt; &gt; +}
&gt; &gt; +#endif
&gt; &gt; +
&gt; &gt; +#endif /* _RTE_LPM_RVV_H_ */
&gt; 
&gt; -- 
&gt; Regards,
&gt; Vladimir
</riscv_vector.h></rte_cpuflags.h></rte_vect.h></bruce.richardson@intel.com></sunyuechi@iscas.ac.cn></stanislaw.kardach@gmail.com></sunyuechi@iscas.ac.cn></sunyuechi@iscas.ac.cn></stanislaw.kardach@gmail.com></bruce.richardson@intel.com></thomas@monjalon.net></sunyuechi@iscas.ac.cn></vladimir.medvedkin@intel.com>

  reply	other threads:[~2025-06-11 15:04 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-05-28 16:57 [PATCH 1/3] config/riscv: detect V extension uk7b
2025-06-04 11:49 ` [PATCH v2 0/3] Add RISC-V V extension detection and LPM optimization uk7b
2025-06-04 11:49 ` [PATCH v2 1/3] config/riscv: detect V extension uk7b
2025-06-04 19:54   ` Stephen Hemminger
2025-06-04 11:49 ` [PATCH v2 2/3] lib/lpm: R-V V rte_lpm_lookupx4 uk7b
2025-06-04 11:49 ` [PATCH v2 3/3] riscv: override machine_args only when default uk7b
2025-06-04 13:07 ` [PATCH v2 0/3] Add RISC-V V extension detection and LPM optimization uk7b
2025-06-04 13:16   ` 孙越池
2025-06-04 13:07 ` [PATCH v2 1/3] config/riscv: detect V extension uk7b
2025-06-04 13:07 ` [PATCH v2 2/3] lib/lpm: R-V V rte_lpm_lookupx4 uk7b
2025-06-04 13:07 ` [PATCH v2 3/3] riscv: override machine_args only when default uk7b
2025-06-04 15:47 ` [PATCH v3 0/3] Add RISC-V V extension detection and LPM optimization uk7b
     [not found] ` <20250604154720.3078131-1-uk7b@foxmail.com>
2025-06-04 15:47   ` [PATCH v3 1/3] config/riscv: detect V extension uk7b
2025-06-04 15:47   ` [PATCH v3 2/3] lib/lpm: R-V V rte_lpm_lookupx4 uk7b
2025-06-04 15:47   ` [PATCH v3 3/3] riscv: override machine_args only when default uk7b
2025-06-05 10:58 ` [PATCH v4 0/3] [PATCH v4 0/3] Add RISC-V V extension detection and LPM optimization uk7b
2025-06-11 14:59   ` [PATCH v5 " uk7b
     [not found]   ` <20250611145915.854026-1-uk7b@foxmail.com>
2025-06-11 14:59     ` [PATCH v5 1/3] config/riscv: detect V extension uk7b
2025-06-11 14:59     ` [PATCH v5 2/3] lib/lpm: R-V V rte_lpm_lookupx4 uk7b
2025-06-11 14:59     ` [PATCH v5 3/3] riscv: override machine_args only when default uk7b
     [not found] ` <20250605105844.3931758-1-uk7b@foxmail.com>
2025-06-05 10:58   ` [PATCH v4 1/3] config/riscv: detect V extension uk7b
2025-06-05 10:58   ` [PATCH v4 2/3] lib/lpm: R-V V rte_lpm_lookupx4 uk7b
2025-06-11 11:52     ` Medvedkin, Vladimir
2025-06-11 15:04       ` 孙越池 [this message]
2025-06-05 10:58   ` [PATCH v4 3/3] riscv: override machine_args only when default uk7b

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=5b1f501.423af.1975f85459b.Coremail.sunyuechi@iscas.ac.cn \
    --to=sunyuechi@iscas.ac.cn \
    --cc=dev@dpdk.org \
    --cc=vladimir.medvedkin@intel.com \
    /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).