DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Yao, Lei A" <lei.a.yao@intel.com>
To: "Wang, Zhihong" <zhihong.wang@intel.com>, "dev@dpdk.org" <dev@dpdk.org>
Cc: "yuanhan.liu@linux.intel.com" <yuanhan.liu@linux.intel.com>,
	"thomas.monjalon@6wind.com" <thomas.monjalon@6wind.com>
Subject: Re: [dpdk-dev] [PATCH v2] eal: optimize aligned rte_memcpy
Date: Thu, 8 Dec 2016 00:55:39 +0000	[thread overview]
Message-ID: <2DBBFF226F7CF64BAFCA79B681719D9537F45981@shsmsx102.ccr.corp.intel.com> (raw)
In-Reply-To: <1481074266-4461-1-git-send-email-zhihong.wang@intel.com>

Tested-by: Lei Yao <lei.a.yao@intel.com>
- Apply patch to v16.11
I have tested the loopback performance for this patch on 3 following settings:
CPU: IVB
Ubutnu16.04
Kernal:  4.4.0
gcc : 5.4.0

CPU: HSW
Fedora 21
Kernal: 4.1.13
gcc: 4.9.2

CPU:BDW
Ubutnu16.04
Kernal:  4.4.0
gcc : 5.4.0

I can see 10%~20% performance gain for different packet size on mergeable path. Only on IVB + gcc5.4.0, slight performance drop(~4%) on vector path for packet size 128 ,260. It's may related to gcc version as this performance drop not see with gcc 6+.


-----Original Message-----
From: Wang, Zhihong 
Sent: Wednesday, December 7, 2016 9:31 AM
To: dev@dpdk.org
Cc: yuanhan.liu@linux.intel.com; thomas.monjalon@6wind.com; Yao, Lei A <lei.a.yao@intel.com>; Wang, Zhihong <zhihong.wang@intel.com>
Subject: [PATCH v2] eal: optimize aligned rte_memcpy

This patch optimizes rte_memcpy for well aligned cases, where both
dst and src addr are aligned to maximum MOV width. It introduces a
dedicated function called rte_memcpy_aligned to handle the aligned
cases with simplified instruction stream. The existing rte_memcpy
is renamed as rte_memcpy_generic. The selection between them 2 is
done at the entry of rte_memcpy.

The existing rte_memcpy is for generic cases, it handles unaligned
copies and make store aligned, it even makes load aligned for micro
architectures like Ivy Bridge. However alignment handling comes at
a price: It adds extra load/store instructions, which can cause
complications sometime.

DPDK Vhost memcpy with Mergeable Rx Buffer feature as an example:
The copy is aligned, and remote, and there is header write along
which is also remote. In this case the memcpy instruction stream
should be simplified, to reduce extra load/store, therefore reduce
the probability of load/store buffer full caused pipeline stall, to
let the actual memcpy instructions be issued and let H/W prefetcher
goes to work as early as possible.

This patch is tested on Ivy Bridge, Haswell and Skylake, it provides
up to 20% gain for Virtio Vhost PVP traffic, with packet size ranging
from 64 to 1500 bytes.

The test can also be conducted without NIC, by setting loopback
traffic between Virtio and Vhost. For example, modify the macro
TXONLY_DEF_PACKET_LEN to the requested packet size in testpmd.h,
rebuild and start testpmd in both host and guest, then "start" on
one side and "start tx_first 32" on the other.


Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
---
 .../common/include/arch/x86/rte_memcpy.h           | 81 +++++++++++++++++++++-
 1 file changed, 78 insertions(+), 3 deletions(-)

diff --git a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
index b3bfc23..b9785e8 100644
--- a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
+++ b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
@@ -69,6 +69,8 @@ rte_memcpy(void *dst, const void *src, size_t n) __attribute__((always_inline));
 
 #ifdef RTE_MACHINE_CPUFLAG_AVX512F
 
+#define ALIGNMENT_MASK 0x3F
+
 /**
  * AVX512 implementation below
  */
@@ -189,7 +191,7 @@ rte_mov512blocks(uint8_t *dst, const uint8_t *src, size_t n)
 }
 
 static inline void *
-rte_memcpy(void *dst, const void *src, size_t n)
+rte_memcpy_generic(void *dst, const void *src, size_t n)
 {
 	uintptr_t dstu = (uintptr_t)dst;
 	uintptr_t srcu = (uintptr_t)src;
@@ -308,6 +310,8 @@ COPY_BLOCK_128_BACK63:
 
 #elif defined RTE_MACHINE_CPUFLAG_AVX2
 
+#define ALIGNMENT_MASK 0x1F
+
 /**
  * AVX2 implementation below
  */
@@ -387,7 +391,7 @@ rte_mov128blocks(uint8_t *dst, const uint8_t *src, size_t n)
 }
 
 static inline void *
-rte_memcpy(void *dst, const void *src, size_t n)
+rte_memcpy_generic(void *dst, const void *src, size_t n)
 {
 	uintptr_t dstu = (uintptr_t)dst;
 	uintptr_t srcu = (uintptr_t)src;
@@ -499,6 +503,8 @@ COPY_BLOCK_128_BACK31:
 
 #else /* RTE_MACHINE_CPUFLAG */
 
+#define ALIGNMENT_MASK 0x0F
+
 /**
  * SSE & AVX implementation below
  */
@@ -677,7 +683,7 @@ __extension__ ({                                                      \
 })
 
 static inline void *
-rte_memcpy(void *dst, const void *src, size_t n)
+rte_memcpy_generic(void *dst, const void *src, size_t n)
 {
 	__m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8;
 	uintptr_t dstu = (uintptr_t)dst;
@@ -821,6 +827,75 @@ COPY_BLOCK_64_BACK15:
 
 #endif /* RTE_MACHINE_CPUFLAG */
 
+static inline void *
+rte_memcpy_aligned(void *dst, const void *src, size_t n)
+{
+	void *ret = dst;
+
+	/* Copy size <= 16 bytes */
+	if (n < 16) {
+		if (n & 0x01) {
+			*(uint8_t *)dst = *(const uint8_t *)src;
+			src = (const uint8_t *)src + 1;
+			dst = (uint8_t *)dst + 1;
+		}
+		if (n & 0x02) {
+			*(uint16_t *)dst = *(const uint16_t *)src;
+			src = (const uint16_t *)src + 1;
+			dst = (uint16_t *)dst + 1;
+		}
+		if (n & 0x04) {
+			*(uint32_t *)dst = *(const uint32_t *)src;
+			src = (const uint32_t *)src + 1;
+			dst = (uint32_t *)dst + 1;
+		}
+		if (n & 0x08)
+			*(uint64_t *)dst = *(const uint64_t *)src;
+
+		return ret;
+	}
+
+	/* Copy 16 <= size <= 32 bytes */
+	if (n <= 32) {
+		rte_mov16((uint8_t *)dst, (const uint8_t *)src);
+		rte_mov16((uint8_t *)dst - 16 + n,
+				(const uint8_t *)src - 16 + n);
+
+		return ret;
+	}
+
+	/* Copy 32 < size <= 64 bytes */
+	if (n <= 64) {
+		rte_mov32((uint8_t *)dst, (const uint8_t *)src);
+		rte_mov32((uint8_t *)dst - 32 + n,
+				(const uint8_t *)src - 32 + n);
+
+		return ret;
+	}
+
+	/* Copy 64 bytes blocks */
+	for (; n >= 64; n -= 64) {
+		rte_mov64((uint8_t *)dst, (const uint8_t *)src);
+		dst = (uint8_t *)dst + 64;
+		src = (const uint8_t *)src + 64;
+	}
+
+	/* Copy whatever left */
+	rte_mov64((uint8_t *)dst - 64 + n,
+			(const uint8_t *)src - 64 + n);
+
+	return ret;
+}
+
+static inline void *
+rte_memcpy(void *dst, const void *src, size_t n)
+{
+	if (!(((uintptr_t)dst | (uintptr_t)src) & ALIGNMENT_MASK))
+		return rte_memcpy_aligned(dst, src, n);
+	else
+		return rte_memcpy_generic(dst, src, n);
+}
+
 #ifdef __cplusplus
 }
 #endif
-- 
2.7.4

  reply	other threads:[~2016-12-08  0:55 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-12-02  1:19 [dpdk-dev] [PATCH] vhost: optimize vhost memcpy Zhihong Wang
2016-12-02 10:30 ` Thomas Monjalon
2016-12-05  8:27 ` Yuanhan Liu
2016-12-05 10:27   ` Wang, Zhihong
2016-12-05 10:37     ` Yuanhan Liu
2016-12-07  6:11       ` Wang, Zhihong
2016-12-07  1:31 ` [dpdk-dev] [PATCH v2] eal: optimize aligned rte_memcpy Zhihong Wang
2016-12-08  0:55   ` Yao, Lei A [this message]
2016-12-08  2:18   ` Yuanhan Liu
2017-01-17 15:08     ` Thomas Monjalon

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=2DBBFF226F7CF64BAFCA79B681719D9537F45981@shsmsx102.ccr.corp.intel.com \
    --to=lei.a.yao@intel.com \
    --cc=dev@dpdk.org \
    --cc=thomas.monjalon@6wind.com \
    --cc=yuanhan.liu@linux.intel.com \
    --cc=zhihong.wang@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).