[dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Hiroshi Shimamoto]

From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>

To reduce instruction cache miss, add branch condition hints into
recv/xmit functions. This improves a bit performance.

We can see performance improvements with memnic-tester.
Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
 size |  before  |  after
   64 | 5.54Mpps | 5.55Mpps
  128 | 5.46Mpps | 5.44Mpps
  256 | 5.21Mpps | 5.22Mpps
  512 | 4.50Mpps | 4.52Mpps
 1024 | 3.71Mpps | 3.73Mpps
 1280 | 3.21Mpps | 3.22Mpps
 1518 | 2.92Mpps | 2.93Mpps

Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
---
 pmd/pmd_memnic.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
index 7fc3093..875d3ea 100644
--- a/pmd/pmd_memnic.c
+++ b/pmd/pmd_memnic.c
@@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
 	int idx, next;
 	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
 
-	if (!adapter->nic->hdr.valid)
+	if (unlikely(!adapter->nic->hdr.valid))
 		return 0;
 
 	pkts = bytes = errs = 0;
 	idx = adapter->up_idx;
 	for (nr = 0; nr < nb_pkts; nr++) {
 		p = &data->packets[idx];
-		if (p->status != MEMNIC_PKT_ST_FILLED)
+		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
 			break;
 		/* prefetch the next area */
 		next = idx;
-		if (++next >= MEMNIC_NR_PACKET)
+		if (unlikely(++next >= MEMNIC_NR_PACKET))
 			next = 0;
 		rte_prefetch0(&data->packets[next]);
-		if (p->len > framesz) {
+		if (unlikely(p->len > framesz)) {
 			errs++;
 			goto drop;
 		}
 		mb = rte_pktmbuf_alloc(adapter->mp);
-		if (!mb)
+		if (unlikely(!mb))
 			break;
 
 		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
@@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
 	uint64_t pkts, bytes, errs;
 	uint32_t framesz = adapter->framesz;
 
-	if (!adapter->nic->hdr.valid)
+	if (unlikely(!adapter->nic->hdr.valid))
 		return 0;
 
 	pkts = bytes = errs = 0;
@@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
 		struct rte_mbuf *sg;
 		void *ptr;
 
-		if (pkt_len > framesz) {
+		if (unlikely(pkt_len > framesz)) {
 			errs++;
 			break;
 		}
@@ -379,7 +379,7 @@ retry:
 			goto retry;
 		}
 
-		if (idx != ACCESS_ONCE(adapter->down_idx)) {
+		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
 			/*
 			 * host freed this and got false positive,
 			 * need to recover the status and retry.
@@ -388,7 +388,7 @@ retry:
 			goto retry;
 		}
 
-		if (++idx >= MEMNIC_NR_PACKET)
+		if (unlikely(++idx >= MEMNIC_NR_PACKET))
 			idx = 0;
 		adapter->down_idx = idx;
 
-- 
1.8.3.1

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Xie, Huawei]

The patch is ok. For the commit message, is it better 
"to reduce branch mispredication"?

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Hiroshi Shimamoto
> Sent: Tuesday, September 30, 2014 7:15 PM
> To: dev@dpdk.org
> Cc: Hayato Momma
> Subject: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> 
> From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> 
> To reduce instruction cache miss, add branch condition hints into
> recv/xmit functions. This improves a bit performance.
> 
> We can see performance improvements with memnic-tester.
> Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
>  size |  before  |  after
>    64 | 5.54Mpps | 5.55Mpps
>   128 | 5.46Mpps | 5.44Mpps
>   256 | 5.21Mpps | 5.22Mpps
>   512 | 4.50Mpps | 4.52Mpps
>  1024 | 3.71Mpps | 3.73Mpps
>  1280 | 3.21Mpps | 3.22Mpps
>  1518 | 2.92Mpps | 2.93Mpps
> 
> Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> ---
>  pmd/pmd_memnic.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
> 
> diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> index 7fc3093..875d3ea 100644
> --- a/pmd/pmd_memnic.c
> +++ b/pmd/pmd_memnic.c
> @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
>  	int idx, next;
>  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> 
> -	if (!adapter->nic->hdr.valid)
> +	if (unlikely(!adapter->nic->hdr.valid))
>  		return 0;
> 
>  	pkts = bytes = errs = 0;
>  	idx = adapter->up_idx;
>  	for (nr = 0; nr < nb_pkts; nr++) {
>  		p = &data->packets[idx];
> -		if (p->status != MEMNIC_PKT_ST_FILLED)
> +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
>  			break;
>  		/* prefetch the next area */
>  		next = idx;
> -		if (++next >= MEMNIC_NR_PACKET)
> +		if (unlikely(++next >= MEMNIC_NR_PACKET))
>  			next = 0;
>  		rte_prefetch0(&data->packets[next]);
> -		if (p->len > framesz) {
> +		if (unlikely(p->len > framesz)) {
>  			errs++;
>  			goto drop;
>  		}
>  		mb = rte_pktmbuf_alloc(adapter->mp);
> -		if (!mb)
> +		if (unlikely(!mb))
>  			break;
> 
>  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
>  	uint64_t pkts, bytes, errs;
>  	uint32_t framesz = adapter->framesz;
> 
> -	if (!adapter->nic->hdr.valid)
> +	if (unlikely(!adapter->nic->hdr.valid))
>  		return 0;
> 
>  	pkts = bytes = errs = 0;
> @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
>  		struct rte_mbuf *sg;
>  		void *ptr;
> 
> -		if (pkt_len > framesz) {
> +		if (unlikely(pkt_len > framesz)) {
>  			errs++;
>  			break;
>  		}
> @@ -379,7 +379,7 @@ retry:
>  			goto retry;
>  		}
> 
> -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
>  			/*
>  			 * host freed this and got false positive,
>  			 * need to recover the status and retry.
> @@ -388,7 +388,7 @@ retry:
>  			goto retry;
>  		}
> 
> -		if (++idx >= MEMNIC_NR_PACKET)
> +		if (unlikely(++idx >= MEMNIC_NR_PACKET))
>  			idx = 0;
>  		adapter->down_idx = idx;
> 
> --
> 1.8.3.1

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Hiroshi Shimamoto]

Hi,

> Subject: RE: [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> 
> The patch is ok. For the commit message, is it better
> "to reduce branch mispredication"?

yes, that seems more suitable to explain the situation.

Thomas, what do you think? Can you replace the message when you apply
this patch?

thanks,
Hiroshi

> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Hiroshi Shimamoto
> > Sent: Tuesday, September 30, 2014 7:15 PM
> > To: dev@dpdk.org
> > Cc: Hayato Momma
> > Subject: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> >
> > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> >
> > To reduce instruction cache miss, add branch condition hints into
> > recv/xmit functions. This improves a bit performance.
> >
> > We can see performance improvements with memnic-tester.
> > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> >  size |  before  |  after
> >    64 | 5.54Mpps | 5.55Mpps
> >   128 | 5.46Mpps | 5.44Mpps
> >   256 | 5.21Mpps | 5.22Mpps
> >   512 | 4.50Mpps | 4.52Mpps
> >  1024 | 3.71Mpps | 3.73Mpps
> >  1280 | 3.21Mpps | 3.22Mpps
> >  1518 | 2.92Mpps | 2.93Mpps
> >
> > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > ---
> >  pmd/pmd_memnic.c | 18 +++++++++---------
> >  1 file changed, 9 insertions(+), 9 deletions(-)
> >
> > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > index 7fc3093..875d3ea 100644
> > --- a/pmd/pmd_memnic.c
> > +++ b/pmd/pmd_memnic.c
> > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> >  	int idx, next;
> >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> >
> > -	if (!adapter->nic->hdr.valid)
> > +	if (unlikely(!adapter->nic->hdr.valid))
> >  		return 0;
> >
> >  	pkts = bytes = errs = 0;
> >  	idx = adapter->up_idx;
> >  	for (nr = 0; nr < nb_pkts; nr++) {
> >  		p = &data->packets[idx];
> > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> >  			break;
> >  		/* prefetch the next area */
> >  		next = idx;
> > -		if (++next >= MEMNIC_NR_PACKET)
> > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> >  			next = 0;
> >  		rte_prefetch0(&data->packets[next]);
> > -		if (p->len > framesz) {
> > +		if (unlikely(p->len > framesz)) {
> >  			errs++;
> >  			goto drop;
> >  		}
> >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > -		if (!mb)
> > +		if (unlikely(!mb))
> >  			break;
> >
> >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> >  	uint64_t pkts, bytes, errs;
> >  	uint32_t framesz = adapter->framesz;
> >
> > -	if (!adapter->nic->hdr.valid)
> > +	if (unlikely(!adapter->nic->hdr.valid))
> >  		return 0;
> >
> >  	pkts = bytes = errs = 0;
> > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> >  		struct rte_mbuf *sg;
> >  		void *ptr;
> >
> > -		if (pkt_len > framesz) {
> > +		if (unlikely(pkt_len > framesz)) {
> >  			errs++;
> >  			break;
> >  		}
> > @@ -379,7 +379,7 @@ retry:
> >  			goto retry;
> >  		}
> >
> > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> >  			/*
> >  			 * host freed this and got false positive,
> >  			 * need to recover the status and retry.
> > @@ -388,7 +388,7 @@ retry:
> >  			goto retry;
> >  		}
> >
> > -		if (++idx >= MEMNIC_NR_PACKET)
> > +		if (unlikely(++idx >= MEMNIC_NR_PACKET))
> >  			idx = 0;
> >  		adapter->down_idx = idx;
> >
> > --
> > 1.8.3.1

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Thomas Monjalon]

2014-10-01 00:01, Hiroshi Shimamoto:
> > The patch is ok. For the commit message, is it better
> > "to reduce branch mispredication"?
> 
> yes, that seems more suitable to explain the situation.
> 
> Thomas, what do you think? Can you replace the message when you apply
> this patch?

Yes, I did it ;)

-- 
Thomas

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Neil Horman]

On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> 
> To reduce instruction cache miss, add branch condition hints into
> recv/xmit functions. This improves a bit performance.
> 
> We can see performance improvements with memnic-tester.
> Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
>  size |  before  |  after
>    64 | 5.54Mpps | 5.55Mpps
>   128 | 5.46Mpps | 5.44Mpps
>   256 | 5.21Mpps | 5.22Mpps
>   512 | 4.50Mpps | 4.52Mpps
>  1024 | 3.71Mpps | 3.73Mpps
>  1280 | 3.21Mpps | 3.22Mpps
>  1518 | 2.92Mpps | 2.93Mpps
> 
> Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> ---
>  pmd/pmd_memnic.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
> 
> diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> index 7fc3093..875d3ea 100644
> --- a/pmd/pmd_memnic.c
> +++ b/pmd/pmd_memnic.c
> @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
>  	int idx, next;
>  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
>  
> -	if (!adapter->nic->hdr.valid)
> +	if (unlikely(!adapter->nic->hdr.valid))
>  		return 0;
>  
>  	pkts = bytes = errs = 0;
>  	idx = adapter->up_idx;
>  	for (nr = 0; nr < nb_pkts; nr++) {
>  		p = &data->packets[idx];
> -		if (p->status != MEMNIC_PKT_ST_FILLED)
> +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
>  			break;
>  		/* prefetch the next area */
>  		next = idx;
> -		if (++next >= MEMNIC_NR_PACKET)
> +		if (unlikely(++next >= MEMNIC_NR_PACKET))
>  			next = 0;
>  		rte_prefetch0(&data->packets[next]);
> -		if (p->len > framesz) {
> +		if (unlikely(p->len > framesz)) {
>  			errs++;
>  			goto drop;
>  		}
>  		mb = rte_pktmbuf_alloc(adapter->mp);
> -		if (!mb)
> +		if (unlikely(!mb))
>  			break;
>  
>  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
>  	uint64_t pkts, bytes, errs;
>  	uint32_t framesz = adapter->framesz;
>  
> -	if (!adapter->nic->hdr.valid)
> +	if (unlikely(!adapter->nic->hdr.valid))
>  		return 0;
>  
>  	pkts = bytes = errs = 0;
> @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
>  		struct rte_mbuf *sg;
>  		void *ptr;
>  
> -		if (pkt_len > framesz) {
> +		if (unlikely(pkt_len > framesz)) {
>  			errs++;
>  			break;
>  		}
> @@ -379,7 +379,7 @@ retry:
>  			goto retry;
>  		}
>  
> -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
once and prevent it from getting reloaded from memory at a later time, this is
exactly contrary to that, both in the sense that you're explicitly reloading the
same variable multiple times, and that you're using it as part of a comparison
operation, rather than an asignment operation

Neil

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Hiroshi Shimamoto]

Hi,

> Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> 
> On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> >
> > To reduce instruction cache miss, add branch condition hints into
> > recv/xmit functions. This improves a bit performance.
> >
> > We can see performance improvements with memnic-tester.
> > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> >  size |  before  |  after
> >    64 | 5.54Mpps | 5.55Mpps
> >   128 | 5.46Mpps | 5.44Mpps
> >   256 | 5.21Mpps | 5.22Mpps
> >   512 | 4.50Mpps | 4.52Mpps
> >  1024 | 3.71Mpps | 3.73Mpps
> >  1280 | 3.21Mpps | 3.22Mpps
> >  1518 | 2.92Mpps | 2.93Mpps
> >
> > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > ---
> >  pmd/pmd_memnic.c | 18 +++++++++---------
> >  1 file changed, 9 insertions(+), 9 deletions(-)
> >
> > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > index 7fc3093..875d3ea 100644
> > --- a/pmd/pmd_memnic.c
> > +++ b/pmd/pmd_memnic.c
> > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> >  	int idx, next;
> >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> >
> > -	if (!adapter->nic->hdr.valid)
> > +	if (unlikely(!adapter->nic->hdr.valid))
> >  		return 0;
> >
> >  	pkts = bytes = errs = 0;
> >  	idx = adapter->up_idx;
> >  	for (nr = 0; nr < nb_pkts; nr++) {
> >  		p = &data->packets[idx];
> > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> >  			break;
> >  		/* prefetch the next area */
> >  		next = idx;
> > -		if (++next >= MEMNIC_NR_PACKET)
> > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> >  			next = 0;
> >  		rte_prefetch0(&data->packets[next]);
> > -		if (p->len > framesz) {
> > +		if (unlikely(p->len > framesz)) {
> >  			errs++;
> >  			goto drop;
> >  		}
> >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > -		if (!mb)
> > +		if (unlikely(!mb))
> >  			break;
> >
> >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> >  	uint64_t pkts, bytes, errs;
> >  	uint32_t framesz = adapter->framesz;
> >
> > -	if (!adapter->nic->hdr.valid)
> > +	if (unlikely(!adapter->nic->hdr.valid))
> >  		return 0;
> >
> >  	pkts = bytes = errs = 0;
> > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> >  		struct rte_mbuf *sg;
> >  		void *ptr;
> >
> > -		if (pkt_len > framesz) {
> > +		if (unlikely(pkt_len > framesz)) {
> >  			errs++;
> >  			break;
> >  		}
> > @@ -379,7 +379,7 @@ retry:
> >  			goto retry;
> >  		}
> >
> > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
> PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
> once and prevent it from getting reloaded from memory at a later time, this is
> exactly contrary to that, both in the sense that you're explicitly reloading the
> same variable multiple times, and that you're using it as part of a comparison
> operation, rather than an asignment operation

ACCESS_ONCE prevents compiler optimization and ensures load from memory.
There could be multiple threads which read/write that index.
We should compare the value previous and the current value in memory.
In that reason, I use ACCESS_ONCE macro to get value in the memory.

thanks,
Hiroshi

> 
> Neil

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Bruce Richardson]

On Tue, Sep 30, 2014 at 11:52:00PM +0000, Hiroshi Shimamoto wrote:
> Hi,
> 
> > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > 
> > On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> > > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > >
> > > To reduce instruction cache miss, add branch condition hints into
> > > recv/xmit functions. This improves a bit performance.
> > >
> > > We can see performance improvements with memnic-tester.
> > > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> > >  size |  before  |  after
> > >    64 | 5.54Mpps | 5.55Mpps
> > >   128 | 5.46Mpps | 5.44Mpps
> > >   256 | 5.21Mpps | 5.22Mpps
> > >   512 | 4.50Mpps | 4.52Mpps
> > >  1024 | 3.71Mpps | 3.73Mpps
> > >  1280 | 3.21Mpps | 3.22Mpps
> > >  1518 | 2.92Mpps | 2.93Mpps
> > >
> > > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > > ---
> > >  pmd/pmd_memnic.c | 18 +++++++++---------
> > >  1 file changed, 9 insertions(+), 9 deletions(-)
> > >
> > > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > > index 7fc3093..875d3ea 100644
> > > --- a/pmd/pmd_memnic.c
> > > +++ b/pmd/pmd_memnic.c
> > > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> > >  	int idx, next;
> > >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> > >
> > > -	if (!adapter->nic->hdr.valid)
> > > +	if (unlikely(!adapter->nic->hdr.valid))
> > >  		return 0;
> > >
> > >  	pkts = bytes = errs = 0;
> > >  	idx = adapter->up_idx;
> > >  	for (nr = 0; nr < nb_pkts; nr++) {
> > >  		p = &data->packets[idx];
> > > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> > >  			break;
> > >  		/* prefetch the next area */
> > >  		next = idx;
> > > -		if (++next >= MEMNIC_NR_PACKET)
> > > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> > >  			next = 0;
> > >  		rte_prefetch0(&data->packets[next]);
> > > -		if (p->len > framesz) {
> > > +		if (unlikely(p->len > framesz)) {
> > >  			errs++;
> > >  			goto drop;
> > >  		}
> > >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > > -		if (!mb)
> > > +		if (unlikely(!mb))
> > >  			break;
> > >
> > >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > >  	uint64_t pkts, bytes, errs;
> > >  	uint32_t framesz = adapter->framesz;
> > >
> > > -	if (!adapter->nic->hdr.valid)
> > > +	if (unlikely(!adapter->nic->hdr.valid))
> > >  		return 0;
> > >
> > >  	pkts = bytes = errs = 0;
> > > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > >  		struct rte_mbuf *sg;
> > >  		void *ptr;
> > >
> > > -		if (pkt_len > framesz) {
> > > +		if (unlikely(pkt_len > framesz)) {
> > >  			errs++;
> > >  			break;
> > >  		}
> > > @@ -379,7 +379,7 @@ retry:
> > >  			goto retry;
> > >  		}
> > >
> > > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> > Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
> > PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
> > once and prevent it from getting reloaded from memory at a later time, this is
> > exactly contrary to that, both in the sense that you're explicitly reloading the
> > same variable multiple times, and that you're using it as part of a comparison
> > operation, rather than an asignment operation
> 
> ACCESS_ONCE prevents compiler optimization and ensures load from memory.
> There could be multiple threads which read/write that index.
> We should compare the value previous and the current value in memory.
> In that reason, I use ACCESS_ONCE macro to get value in the memory.

Should you not just make the variable volatile? That's the normal way to 
guarantee reads from memory and prevent the compiler caching things in 
registers.

/Bruce

> 
> thanks,
> Hiroshi
> 
> > 
> > Neil
> 

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Hiroshi Shimamoto]

> Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> 
> On Tue, Sep 30, 2014 at 11:52:00PM +0000, Hiroshi Shimamoto wrote:
> > Hi,
> >
> > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > >
> > > On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> > > > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > >
> > > > To reduce instruction cache miss, add branch condition hints into
> > > > recv/xmit functions. This improves a bit performance.
> > > >
> > > > We can see performance improvements with memnic-tester.
> > > > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> > > >  size |  before  |  after
> > > >    64 | 5.54Mpps | 5.55Mpps
> > > >   128 | 5.46Mpps | 5.44Mpps
> > > >   256 | 5.21Mpps | 5.22Mpps
> > > >   512 | 4.50Mpps | 4.52Mpps
> > > >  1024 | 3.71Mpps | 3.73Mpps
> > > >  1280 | 3.21Mpps | 3.22Mpps
> > > >  1518 | 2.92Mpps | 2.93Mpps
> > > >
> > > > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > > > ---
> > > >  pmd/pmd_memnic.c | 18 +++++++++---------
> > > >  1 file changed, 9 insertions(+), 9 deletions(-)
> > > >
> > > > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > > > index 7fc3093..875d3ea 100644
> > > > --- a/pmd/pmd_memnic.c
> > > > +++ b/pmd/pmd_memnic.c
> > > > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> > > >  	int idx, next;
> > > >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> > > >
> > > > -	if (!adapter->nic->hdr.valid)
> > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > >  		return 0;
> > > >
> > > >  	pkts = bytes = errs = 0;
> > > >  	idx = adapter->up_idx;
> > > >  	for (nr = 0; nr < nb_pkts; nr++) {
> > > >  		p = &data->packets[idx];
> > > > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > > > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> > > >  			break;
> > > >  		/* prefetch the next area */
> > > >  		next = idx;
> > > > -		if (++next >= MEMNIC_NR_PACKET)
> > > > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> > > >  			next = 0;
> > > >  		rte_prefetch0(&data->packets[next]);
> > > > -		if (p->len > framesz) {
> > > > +		if (unlikely(p->len > framesz)) {
> > > >  			errs++;
> > > >  			goto drop;
> > > >  		}
> > > >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > > > -		if (!mb)
> > > > +		if (unlikely(!mb))
> > > >  			break;
> > > >
> > > >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > > > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > >  	uint64_t pkts, bytes, errs;
> > > >  	uint32_t framesz = adapter->framesz;
> > > >
> > > > -	if (!adapter->nic->hdr.valid)
> > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > >  		return 0;
> > > >
> > > >  	pkts = bytes = errs = 0;
> > > > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > >  		struct rte_mbuf *sg;
> > > >  		void *ptr;
> > > >
> > > > -		if (pkt_len > framesz) {
> > > > +		if (unlikely(pkt_len > framesz)) {
> > > >  			errs++;
> > > >  			break;
> > > >  		}
> > > > @@ -379,7 +379,7 @@ retry:
> > > >  			goto retry;
> > > >  		}
> > > >
> > > > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > > > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> > > Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
> > > PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
> > > once and prevent it from getting reloaded from memory at a later time, this is
> > > exactly contrary to that, both in the sense that you're explicitly reloading the
> > > same variable multiple times, and that you're using it as part of a comparison
> > > operation, rather than an asignment operation
> >
> > ACCESS_ONCE prevents compiler optimization and ensures load from memory.
> > There could be multiple threads which read/write that index.
> > We should compare the value previous and the current value in memory.
> > In that reason, I use ACCESS_ONCE macro to get value in the memory.
> 
> Should you not just make the variable volatile? That's the normal way to
> guarantee reads from memory and prevent the compiler caching things in
> registers.

We don't want always accessing to memory, it could cause performance degradation.
Like linux kernel, I use it in the place only we really load from memory.

thanks,
Hiroshi

> 
> /Bruce
> 
> >
> > thanks,
> > Hiroshi
> >
> > >
> > > Neil
> >

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Neil Horman]

On Wed, Oct 01, 2014 at 09:12:44AM +0000, Hiroshi Shimamoto wrote:
> > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > 
> > On Tue, Sep 30, 2014 at 11:52:00PM +0000, Hiroshi Shimamoto wrote:
> > > Hi,
> > >
> > > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > >
> > > > On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> > > > > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > >
> > > > > To reduce instruction cache miss, add branch condition hints into
> > > > > recv/xmit functions. This improves a bit performance.
> > > > >
> > > > > We can see performance improvements with memnic-tester.
> > > > > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> > > > >  size |  before  |  after
> > > > >    64 | 5.54Mpps | 5.55Mpps
> > > > >   128 | 5.46Mpps | 5.44Mpps
> > > > >   256 | 5.21Mpps | 5.22Mpps
> > > > >   512 | 4.50Mpps | 4.52Mpps
> > > > >  1024 | 3.71Mpps | 3.73Mpps
> > > > >  1280 | 3.21Mpps | 3.22Mpps
> > > > >  1518 | 2.92Mpps | 2.93Mpps
> > > > >
> > > > > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > > > > ---
> > > > >  pmd/pmd_memnic.c | 18 +++++++++---------
> > > > >  1 file changed, 9 insertions(+), 9 deletions(-)
> > > > >
> > > > > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > > > > index 7fc3093..875d3ea 100644
> > > > > --- a/pmd/pmd_memnic.c
> > > > > +++ b/pmd/pmd_memnic.c
> > > > > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> > > > >  	int idx, next;
> > > > >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> > > > >
> > > > > -	if (!adapter->nic->hdr.valid)
> > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > >  		return 0;
> > > > >
> > > > >  	pkts = bytes = errs = 0;
> > > > >  	idx = adapter->up_idx;
> > > > >  	for (nr = 0; nr < nb_pkts; nr++) {
> > > > >  		p = &data->packets[idx];
> > > > > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > > > > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> > > > >  			break;
> > > > >  		/* prefetch the next area */
> > > > >  		next = idx;
> > > > > -		if (++next >= MEMNIC_NR_PACKET)
> > > > > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> > > > >  			next = 0;
> > > > >  		rte_prefetch0(&data->packets[next]);
> > > > > -		if (p->len > framesz) {
> > > > > +		if (unlikely(p->len > framesz)) {
> > > > >  			errs++;
> > > > >  			goto drop;
> > > > >  		}
> > > > >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > > > > -		if (!mb)
> > > > > +		if (unlikely(!mb))
> > > > >  			break;
> > > > >
> > > > >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > > > > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > >  	uint64_t pkts, bytes, errs;
> > > > >  	uint32_t framesz = adapter->framesz;
> > > > >
> > > > > -	if (!adapter->nic->hdr.valid)
> > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > >  		return 0;
> > > > >
> > > > >  	pkts = bytes = errs = 0;
> > > > > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > >  		struct rte_mbuf *sg;
> > > > >  		void *ptr;
> > > > >
> > > > > -		if (pkt_len > framesz) {
> > > > > +		if (unlikely(pkt_len > framesz)) {
> > > > >  			errs++;
> > > > >  			break;
> > > > >  		}
> > > > > @@ -379,7 +379,7 @@ retry:
> > > > >  			goto retry;
> > > > >  		}
> > > > >
> > > > > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > > > > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> > > > Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
> > > > PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
> > > > once and prevent it from getting reloaded from memory at a later time, this is
> > > > exactly contrary to that, both in the sense that you're explicitly reloading the
> > > > same variable multiple times, and that you're using it as part of a comparison
> > > > operation, rather than an asignment operation
> > >
> > > ACCESS_ONCE prevents compiler optimization and ensures load from memory.
> > > There could be multiple threads which read/write that index.
> > > We should compare the value previous and the current value in memory.
> > > In that reason, I use ACCESS_ONCE macro to get value in the memory.
> > 
> > Should you not just make the variable volatile? That's the normal way to
> > guarantee reads from memory and prevent the compiler caching things in
> > registers.
> 
> We don't want always accessing to memory, it could cause performance degradation.
> Like linux kernel, I use it in the place only we really load from memory.
> 
Thats not true at all.  Every single read of adapter->down_idx in
memnic_xmit_pkts() is wrapped in a ACCESS_ONCE call.  Theres no difference in
doing that and just declaring a volitile variable and pointing it to
&adapter->down_idx (save for the increased legibility of the code)

Neil

> thanks,
> Hiroshi
> 
> > 
> > /Bruce
> > 
> > >
> > > thanks,
> > > Hiroshi
> > >
> > > >
> > > > Neil
> > >
> 

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Hiroshi Shimamoto]

> Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> 
> On Wed, Oct 01, 2014 at 09:12:44AM +0000, Hiroshi Shimamoto wrote:
> > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > >
> > > On Tue, Sep 30, 2014 at 11:52:00PM +0000, Hiroshi Shimamoto wrote:
> > > > Hi,
> > > >
> > > > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > > >
> > > > > On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> > > > > > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > > >
> > > > > > To reduce instruction cache miss, add branch condition hints into
> > > > > > recv/xmit functions. This improves a bit performance.
> > > > > >
> > > > > > We can see performance improvements with memnic-tester.
> > > > > > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> > > > > >  size |  before  |  after
> > > > > >    64 | 5.54Mpps | 5.55Mpps
> > > > > >   128 | 5.46Mpps | 5.44Mpps
> > > > > >   256 | 5.21Mpps | 5.22Mpps
> > > > > >   512 | 4.50Mpps | 4.52Mpps
> > > > > >  1024 | 3.71Mpps | 3.73Mpps
> > > > > >  1280 | 3.21Mpps | 3.22Mpps
> > > > > >  1518 | 2.92Mpps | 2.93Mpps
> > > > > >
> > > > > > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > > > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > > > > > ---
> > > > > >  pmd/pmd_memnic.c | 18 +++++++++---------
> > > > > >  1 file changed, 9 insertions(+), 9 deletions(-)
> > > > > >
> > > > > > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > > > > > index 7fc3093..875d3ea 100644
> > > > > > --- a/pmd/pmd_memnic.c
> > > > > > +++ b/pmd/pmd_memnic.c
> > > > > > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> > > > > >  	int idx, next;
> > > > > >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> > > > > >
> > > > > > -	if (!adapter->nic->hdr.valid)
> > > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > > >  		return 0;
> > > > > >
> > > > > >  	pkts = bytes = errs = 0;
> > > > > >  	idx = adapter->up_idx;
> > > > > >  	for (nr = 0; nr < nb_pkts; nr++) {
> > > > > >  		p = &data->packets[idx];
> > > > > > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > > > > > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> > > > > >  			break;
> > > > > >  		/* prefetch the next area */
> > > > > >  		next = idx;
> > > > > > -		if (++next >= MEMNIC_NR_PACKET)
> > > > > > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> > > > > >  			next = 0;
> > > > > >  		rte_prefetch0(&data->packets[next]);
> > > > > > -		if (p->len > framesz) {
> > > > > > +		if (unlikely(p->len > framesz)) {
> > > > > >  			errs++;
> > > > > >  			goto drop;
> > > > > >  		}
> > > > > >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > > > > > -		if (!mb)
> > > > > > +		if (unlikely(!mb))
> > > > > >  			break;
> > > > > >
> > > > > >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > > > > > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > > >  	uint64_t pkts, bytes, errs;
> > > > > >  	uint32_t framesz = adapter->framesz;
> > > > > >
> > > > > > -	if (!adapter->nic->hdr.valid)
> > > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > > >  		return 0;
> > > > > >
> > > > > >  	pkts = bytes = errs = 0;
> > > > > > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > > >  		struct rte_mbuf *sg;
> > > > > >  		void *ptr;
> > > > > >
> > > > > > -		if (pkt_len > framesz) {
> > > > > > +		if (unlikely(pkt_len > framesz)) {
> > > > > >  			errs++;
> > > > > >  			break;
> > > > > >  		}
> > > > > > @@ -379,7 +379,7 @@ retry:
> > > > > >  			goto retry;
> > > > > >  		}
> > > > > >
> > > > > > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > > > > > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> > > > > Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
> > > > > PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
> > > > > once and prevent it from getting reloaded from memory at a later time, this is
> > > > > exactly contrary to that, both in the sense that you're explicitly reloading the
> > > > > same variable multiple times, and that you're using it as part of a comparison
> > > > > operation, rather than an asignment operation
> > > >
> > > > ACCESS_ONCE prevents compiler optimization and ensures load from memory.
> > > > There could be multiple threads which read/write that index.
> > > > We should compare the value previous and the current value in memory.
> > > > In that reason, I use ACCESS_ONCE macro to get value in the memory.
> > >
> > > Should you not just make the variable volatile? That's the normal way to
> > > guarantee reads from memory and prevent the compiler caching things in
> > > registers.
> >
> > We don't want always accessing to memory, it could cause performance degradation.
> > Like linux kernel, I use it in the place only we really load from memory.
> >
> Thats not true at all.  Every single read of adapter->down_idx in
> memnic_xmit_pkts() is wrapped in a ACCESS_ONCE call.  Theres no difference in
> doing that and just declaring a volitile variable and pointing it to
> &adapter->down_idx (save for the increased legibility of the code)

You're right, at this moment there is no reference without ACCESS_ONCE.
I'm not sure adding code to access that variable in the future, but
would like to avoid accidentally a code which causes a performance issue,
I think keeping the declaration in structure without volatile.
As you mentioned, using local variable which points down_idx will be fine.

I will submit a cleanup patch before starting the next development for DPDK v1.8.

thanks,
Hiroshi

> 
> Neil
> 
> > thanks,
> > Hiroshi
> >
> > >
> > > /Bruce
> > >
> > > >
> > > > thanks,
> > > > Hiroshi
> > > >
> > > > >
> > > > > Neil
> > > >
> >

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Neil Horman]

On Wed, Oct 01, 2014 at 11:33:23PM +0000, Hiroshi Shimamoto wrote:
> > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > 
> > On Wed, Oct 01, 2014 at 09:12:44AM +0000, Hiroshi Shimamoto wrote:
> > > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > >
> > > > On Tue, Sep 30, 2014 at 11:52:00PM +0000, Hiroshi Shimamoto wrote:
> > > > > Hi,
> > > > >
> > > > > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > > > >
> > > > > > On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> > > > > > > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > > > >
> > > > > > > To reduce instruction cache miss, add branch condition hints into
> > > > > > > recv/xmit functions. This improves a bit performance.
> > > > > > >
> > > > > > > We can see performance improvements with memnic-tester.
> > > > > > > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> > > > > > >  size |  before  |  after
> > > > > > >    64 | 5.54Mpps | 5.55Mpps
> > > > > > >   128 | 5.46Mpps | 5.44Mpps
> > > > > > >   256 | 5.21Mpps | 5.22Mpps
> > > > > > >   512 | 4.50Mpps | 4.52Mpps
> > > > > > >  1024 | 3.71Mpps | 3.73Mpps
> > > > > > >  1280 | 3.21Mpps | 3.22Mpps
> > > > > > >  1518 | 2.92Mpps | 2.93Mpps
> > > > > > >
> > > > > > > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > > > > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > > > > > > ---
> > > > > > >  pmd/pmd_memnic.c | 18 +++++++++---------
> > > > > > >  1 file changed, 9 insertions(+), 9 deletions(-)
> > > > > > >
> > > > > > > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > > > > > > index 7fc3093..875d3ea 100644
> > > > > > > --- a/pmd/pmd_memnic.c
> > > > > > > +++ b/pmd/pmd_memnic.c
> > > > > > > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> > > > > > >  	int idx, next;
> > > > > > >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> > > > > > >
> > > > > > > -	if (!adapter->nic->hdr.valid)
> > > > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > > > >  		return 0;
> > > > > > >
> > > > > > >  	pkts = bytes = errs = 0;
> > > > > > >  	idx = adapter->up_idx;
> > > > > > >  	for (nr = 0; nr < nb_pkts; nr++) {
> > > > > > >  		p = &data->packets[idx];
> > > > > > > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > > > > > > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> > > > > > >  			break;
> > > > > > >  		/* prefetch the next area */
> > > > > > >  		next = idx;
> > > > > > > -		if (++next >= MEMNIC_NR_PACKET)
> > > > > > > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> > > > > > >  			next = 0;
> > > > > > >  		rte_prefetch0(&data->packets[next]);
> > > > > > > -		if (p->len > framesz) {
> > > > > > > +		if (unlikely(p->len > framesz)) {
> > > > > > >  			errs++;
> > > > > > >  			goto drop;
> > > > > > >  		}
> > > > > > >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > > > > > > -		if (!mb)
> > > > > > > +		if (unlikely(!mb))
> > > > > > >  			break;
> > > > > > >
> > > > > > >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > > > > > > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > > > >  	uint64_t pkts, bytes, errs;
> > > > > > >  	uint32_t framesz = adapter->framesz;
> > > > > > >
> > > > > > > -	if (!adapter->nic->hdr.valid)
> > > > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > > > >  		return 0;
> > > > > > >
> > > > > > >  	pkts = bytes = errs = 0;
> > > > > > > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > > > >  		struct rte_mbuf *sg;
> > > > > > >  		void *ptr;
> > > > > > >
> > > > > > > -		if (pkt_len > framesz) {
> > > > > > > +		if (unlikely(pkt_len > framesz)) {
> > > > > > >  			errs++;
> > > > > > >  			break;
> > > > > > >  		}
> > > > > > > @@ -379,7 +379,7 @@ retry:
> > > > > > >  			goto retry;
> > > > > > >  		}
> > > > > > >
> > > > > > > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > > > > > > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> > > > > > Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
> > > > > > PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
> > > > > > once and prevent it from getting reloaded from memory at a later time, this is
> > > > > > exactly contrary to that, both in the sense that you're explicitly reloading the
> > > > > > same variable multiple times, and that you're using it as part of a comparison
> > > > > > operation, rather than an asignment operation
> > > > >
> > > > > ACCESS_ONCE prevents compiler optimization and ensures load from memory.
> > > > > There could be multiple threads which read/write that index.
> > > > > We should compare the value previous and the current value in memory.
> > > > > In that reason, I use ACCESS_ONCE macro to get value in the memory.
> > > >
> > > > Should you not just make the variable volatile? That's the normal way to
> > > > guarantee reads from memory and prevent the compiler caching things in
> > > > registers.
> > >
> > > We don't want always accessing to memory, it could cause performance degradation.
> > > Like linux kernel, I use it in the place only we really load from memory.
> > >
> > Thats not true at all.  Every single read of adapter->down_idx in
> > memnic_xmit_pkts() is wrapped in a ACCESS_ONCE call.  Theres no difference in
> > doing that and just declaring a volitile variable and pointing it to
> > &adapter->down_idx (save for the increased legibility of the code)
> 
> You're right, at this moment there is no reference without ACCESS_ONCE.
> I'm not sure adding code to access that variable in the future, but
> would like to avoid accidentally a code which causes a performance issue,
> I think keeping the declaration in structure without volatile.
> As you mentioned, using local variable which points down_idx will be fine.
So you would like to continue using a macro incorrectly to avoid a possible
performance issue with code that hasn't been written yet?  Thats nonsensical.
What performance issue to see occuring if you created a volatile variable and
then used it in conjunction with ACCESS_ONCE?

Neil

> 
> I will submit a cleanup patch before starting the next development for DPDK v1.8.
> 
> thanks,
> Hiroshi
> 
> > 
> > Neil
> > 
> > > thanks,
> > > Hiroshi
> > >
> > > >
> > > > /Bruce
> > > >
> > > > >
> > > > > thanks,
> > > > > Hiroshi
> > > > >
> > > > > >
> > > > > > Neil
> > > > >
> > >
> 

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Hiroshi Shimamoto]

> Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> 
> On Wed, Oct 01, 2014 at 11:33:23PM +0000, Hiroshi Shimamoto wrote:
> > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > >
> > > On Wed, Oct 01, 2014 at 09:12:44AM +0000, Hiroshi Shimamoto wrote:
> > > > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > > >
> > > > > On Tue, Sep 30, 2014 at 11:52:00PM +0000, Hiroshi Shimamoto wrote:
> > > > > > Hi,
> > > > > >
> > > > > > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > > > > >
> > > > > > > On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> > > > > > > > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > > > > >
> > > > > > > > To reduce instruction cache miss, add branch condition hints into
> > > > > > > > recv/xmit functions. This improves a bit performance.
> > > > > > > >
> > > > > > > > We can see performance improvements with memnic-tester.
> > > > > > > > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> > > > > > > >  size |  before  |  after
> > > > > > > >    64 | 5.54Mpps | 5.55Mpps
> > > > > > > >   128 | 5.46Mpps | 5.44Mpps
> > > > > > > >   256 | 5.21Mpps | 5.22Mpps
> > > > > > > >   512 | 4.50Mpps | 4.52Mpps
> > > > > > > >  1024 | 3.71Mpps | 3.73Mpps
> > > > > > > >  1280 | 3.21Mpps | 3.22Mpps
> > > > > > > >  1518 | 2.92Mpps | 2.93Mpps
> > > > > > > >
> > > > > > > > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > > > > > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > > > > > > > ---
> > > > > > > >  pmd/pmd_memnic.c | 18 +++++++++---------
> > > > > > > >  1 file changed, 9 insertions(+), 9 deletions(-)
> > > > > > > >
> > > > > > > > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > > > > > > > index 7fc3093..875d3ea 100644
> > > > > > > > --- a/pmd/pmd_memnic.c
> > > > > > > > +++ b/pmd/pmd_memnic.c
> > > > > > > > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> > > > > > > >  	int idx, next;
> > > > > > > >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> > > > > > > >
> > > > > > > > -	if (!adapter->nic->hdr.valid)
> > > > > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > > > > >  		return 0;
> > > > > > > >
> > > > > > > >  	pkts = bytes = errs = 0;
> > > > > > > >  	idx = adapter->up_idx;
> > > > > > > >  	for (nr = 0; nr < nb_pkts; nr++) {
> > > > > > > >  		p = &data->packets[idx];
> > > > > > > > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > > > > > > > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> > > > > > > >  			break;
> > > > > > > >  		/* prefetch the next area */
> > > > > > > >  		next = idx;
> > > > > > > > -		if (++next >= MEMNIC_NR_PACKET)
> > > > > > > > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> > > > > > > >  			next = 0;
> > > > > > > >  		rte_prefetch0(&data->packets[next]);
> > > > > > > > -		if (p->len > framesz) {
> > > > > > > > +		if (unlikely(p->len > framesz)) {
> > > > > > > >  			errs++;
> > > > > > > >  			goto drop;
> > > > > > > >  		}
> > > > > > > >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > > > > > > > -		if (!mb)
> > > > > > > > +		if (unlikely(!mb))
> > > > > > > >  			break;
> > > > > > > >
> > > > > > > >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > > > > > > > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > > > > >  	uint64_t pkts, bytes, errs;
> > > > > > > >  	uint32_t framesz = adapter->framesz;
> > > > > > > >
> > > > > > > > -	if (!adapter->nic->hdr.valid)
> > > > > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > > > > >  		return 0;
> > > > > > > >
> > > > > > > >  	pkts = bytes = errs = 0;
> > > > > > > > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > > > > >  		struct rte_mbuf *sg;
> > > > > > > >  		void *ptr;
> > > > > > > >
> > > > > > > > -		if (pkt_len > framesz) {
> > > > > > > > +		if (unlikely(pkt_len > framesz)) {
> > > > > > > >  			errs++;
> > > > > > > >  			break;
> > > > > > > >  		}
> > > > > > > > @@ -379,7 +379,7 @@ retry:
> > > > > > > >  			goto retry;
> > > > > > > >  		}
> > > > > > > >
> > > > > > > > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > > > > > > > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> > > > > > > Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
> > > > > > > PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
> > > > > > > once and prevent it from getting reloaded from memory at a later time, this is
> > > > > > > exactly contrary to that, both in the sense that you're explicitly reloading the
> > > > > > > same variable multiple times, and that you're using it as part of a comparison
> > > > > > > operation, rather than an asignment operation
> > > > > >
> > > > > > ACCESS_ONCE prevents compiler optimization and ensures load from memory.
> > > > > > There could be multiple threads which read/write that index.
> > > > > > We should compare the value previous and the current value in memory.
> > > > > > In that reason, I use ACCESS_ONCE macro to get value in the memory.
> > > > >
> > > > > Should you not just make the variable volatile? That's the normal way to
> > > > > guarantee reads from memory and prevent the compiler caching things in
> > > > > registers.
> > > >
> > > > We don't want always accessing to memory, it could cause performance degradation.
> > > > Like linux kernel, I use it in the place only we really load from memory.
> > > >
> > > Thats not true at all.  Every single read of adapter->down_idx in
> > > memnic_xmit_pkts() is wrapped in a ACCESS_ONCE call.  Theres no difference in
> > > doing that and just declaring a volitile variable and pointing it to
> > > &adapter->down_idx (save for the increased legibility of the code)
> >
> > You're right, at this moment there is no reference without ACCESS_ONCE.
> > I'm not sure adding code to access that variable in the future, but
> > would like to avoid accidentally a code which causes a performance issue,
> > I think keeping the declaration in structure without volatile.
> > As you mentioned, using local variable which points down_idx will be fine.
> So you would like to continue using a macro incorrectly to avoid a possible
> performance issue with code that hasn't been written yet?  Thats nonsensical.

No, I will replace ACCESS_ONCE macro with local volatile variable, then
ACCESS_ONCE macro will disappear.

thanks,
Hiroshi

> What performance issue to see occuring if you created a volatile variable and
> then used it in conjunction with ACCESS_ONCE?
> 
> Neil
> 
> >
> > I will submit a cleanup patch before starting the next development for DPDK v1.8.
> >
> > thanks,
> > Hiroshi
> >
> > >
> > > Neil
> > >
> > > > thanks,
> > > > Hiroshi
> > > >
> > > > >
> > > > > /Bruce
> > > > >
> > > > > >
> > > > > > thanks,
> > > > > > Hiroshi
> > > > > >
> > > > > > >
> > > > > > > Neil
> > > > > >
> > > >
> >

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Neil Horman]

On Thu, Oct 02, 2014 at 02:07:09AM +0000, Hiroshi Shimamoto wrote:
> > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > 
> > On Wed, Oct 01, 2014 at 11:33:23PM +0000, Hiroshi Shimamoto wrote:
> > > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > >
> > > > On Wed, Oct 01, 2014 at 09:12:44AM +0000, Hiroshi Shimamoto wrote:
> > > > > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > > > >
> > > > > > On Tue, Sep 30, 2014 at 11:52:00PM +0000, Hiroshi Shimamoto wrote:
> > > > > > > Hi,
> > > > > > >
> > > > > > > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > > > > > >
> > > > > > > > On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> > > > > > > > > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > > > > > >
> > > > > > > > > To reduce instruction cache miss, add branch condition hints into
> > > > > > > > > recv/xmit functions. This improves a bit performance.
> > > > > > > > >
> > > > > > > > > We can see performance improvements with memnic-tester.
> > > > > > > > > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> > > > > > > > >  size |  before  |  after
> > > > > > > > >    64 | 5.54Mpps | 5.55Mpps
> > > > > > > > >   128 | 5.46Mpps | 5.44Mpps
> > > > > > > > >   256 | 5.21Mpps | 5.22Mpps
> > > > > > > > >   512 | 4.50Mpps | 4.52Mpps
> > > > > > > > >  1024 | 3.71Mpps | 3.73Mpps
> > > > > > > > >  1280 | 3.21Mpps | 3.22Mpps
> > > > > > > > >  1518 | 2.92Mpps | 2.93Mpps
> > > > > > > > >
> > > > > > > > > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > > > > > > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > > > > > > > > ---
> > > > > > > > >  pmd/pmd_memnic.c | 18 +++++++++---------
> > > > > > > > >  1 file changed, 9 insertions(+), 9 deletions(-)
> > > > > > > > >
> > > > > > > > > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > > > > > > > > index 7fc3093..875d3ea 100644
> > > > > > > > > --- a/pmd/pmd_memnic.c
> > > > > > > > > +++ b/pmd/pmd_memnic.c
> > > > > > > > > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> > > > > > > > >  	int idx, next;
> > > > > > > > >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> > > > > > > > >
> > > > > > > > > -	if (!adapter->nic->hdr.valid)
> > > > > > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > > > > > >  		return 0;
> > > > > > > > >
> > > > > > > > >  	pkts = bytes = errs = 0;
> > > > > > > > >  	idx = adapter->up_idx;
> > > > > > > > >  	for (nr = 0; nr < nb_pkts; nr++) {
> > > > > > > > >  		p = &data->packets[idx];
> > > > > > > > > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > > > > > > > > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> > > > > > > > >  			break;
> > > > > > > > >  		/* prefetch the next area */
> > > > > > > > >  		next = idx;
> > > > > > > > > -		if (++next >= MEMNIC_NR_PACKET)
> > > > > > > > > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> > > > > > > > >  			next = 0;
> > > > > > > > >  		rte_prefetch0(&data->packets[next]);
> > > > > > > > > -		if (p->len > framesz) {
> > > > > > > > > +		if (unlikely(p->len > framesz)) {
> > > > > > > > >  			errs++;
> > > > > > > > >  			goto drop;
> > > > > > > > >  		}
> > > > > > > > >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > > > > > > > > -		if (!mb)
> > > > > > > > > +		if (unlikely(!mb))
> > > > > > > > >  			break;
> > > > > > > > >
> > > > > > > > >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > > > > > > > > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > > > > > >  	uint64_t pkts, bytes, errs;
> > > > > > > > >  	uint32_t framesz = adapter->framesz;
> > > > > > > > >
> > > > > > > > > -	if (!adapter->nic->hdr.valid)
> > > > > > > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > > > > > > >  		return 0;
> > > > > > > > >
> > > > > > > > >  	pkts = bytes = errs = 0;
> > > > > > > > > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > > > > > > >  		struct rte_mbuf *sg;
> > > > > > > > >  		void *ptr;
> > > > > > > > >
> > > > > > > > > -		if (pkt_len > framesz) {
> > > > > > > > > +		if (unlikely(pkt_len > framesz)) {
> > > > > > > > >  			errs++;
> > > > > > > > >  			break;
> > > > > > > > >  		}
> > > > > > > > > @@ -379,7 +379,7 @@ retry:
> > > > > > > > >  			goto retry;
> > > > > > > > >  		}
> > > > > > > > >
> > > > > > > > > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > > > > > > > > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> > > > > > > > Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
> > > > > > > > PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
> > > > > > > > once and prevent it from getting reloaded from memory at a later time, this is
> > > > > > > > exactly contrary to that, both in the sense that you're explicitly reloading the
> > > > > > > > same variable multiple times, and that you're using it as part of a comparison
> > > > > > > > operation, rather than an asignment operation
> > > > > > >
> > > > > > > ACCESS_ONCE prevents compiler optimization and ensures load from memory.
> > > > > > > There could be multiple threads which read/write that index.
> > > > > > > We should compare the value previous and the current value in memory.
> > > > > > > In that reason, I use ACCESS_ONCE macro to get value in the memory.
> > > > > >
> > > > > > Should you not just make the variable volatile? That's the normal way to
> > > > > > guarantee reads from memory and prevent the compiler caching things in
> > > > > > registers.
> > > > >
> > > > > We don't want always accessing to memory, it could cause performance degradation.
> > > > > Like linux kernel, I use it in the place only we really load from memory.
> > > > >
> > > > Thats not true at all.  Every single read of adapter->down_idx in
> > > > memnic_xmit_pkts() is wrapped in a ACCESS_ONCE call.  Theres no difference in
> > > > doing that and just declaring a volitile variable and pointing it to
> > > > &adapter->down_idx (save for the increased legibility of the code)
> > >
> > > You're right, at this moment there is no reference without ACCESS_ONCE.
> > > I'm not sure adding code to access that variable in the future, but
> > > would like to avoid accidentally a code which causes a performance issue,
> > > I think keeping the declaration in structure without volatile.
> > > As you mentioned, using local variable which points down_idx will be fine.
> > So you would like to continue using a macro incorrectly to avoid a possible
> > performance issue with code that hasn't been written yet?  Thats nonsensical.
> 
> No, I will replace ACCESS_ONCE macro with local volatile variable, then
> ACCESS_ONCE macro will disappear.
> 
Ah, sorry, misunderstood your intentions.

Thanks
Neil

> thanks,
> Hiroshi
> 
> > What performance issue to see occuring if you created a volatile variable and
> > then used it in conjunction with ACCESS_ONCE?
> > 
> > Neil
> > 
> > >
> > > I will submit a cleanup patch before starting the next development for DPDK v1.8.
> > >
> > > thanks,
> > > Hiroshi
> > >
> > > >
> > > > Neil
> > > >
> > > > > thanks,
> > > > > Hiroshi
> > > > >
> > > > > >
> > > > > > /Bruce
> > > > > >
> > > > > > >
> > > > > > > thanks,
> > > > > > > Hiroshi
> > > > > > >
> > > > > > > >
> > > > > > > > Neil
> > > > > > >
> > > > >
> > >
> 

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Neil Horman]

On Wed, Oct 01, 2014 at 09:44:45AM +0100, Bruce Richardson wrote:
> On Tue, Sep 30, 2014 at 11:52:00PM +0000, Hiroshi Shimamoto wrote:
> > Hi,
> > 
> > > Subject: Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> > > 
> > > On Tue, Sep 30, 2014 at 11:14:40AM +0000, Hiroshi Shimamoto wrote:
> > > > From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > >
> > > > To reduce instruction cache miss, add branch condition hints into
> > > > recv/xmit functions. This improves a bit performance.
> > > >
> > > > We can see performance improvements with memnic-tester.
> > > > Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
> > > >  size |  before  |  after
> > > >    64 | 5.54Mpps | 5.55Mpps
> > > >   128 | 5.46Mpps | 5.44Mpps
> > > >   256 | 5.21Mpps | 5.22Mpps
> > > >   512 | 4.50Mpps | 4.52Mpps
> > > >  1024 | 3.71Mpps | 3.73Mpps
> > > >  1280 | 3.21Mpps | 3.22Mpps
> > > >  1518 | 2.92Mpps | 2.93Mpps
> > > >
> > > > Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> > > > Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> > > > ---
> > > >  pmd/pmd_memnic.c | 18 +++++++++---------
> > > >  1 file changed, 9 insertions(+), 9 deletions(-)
> > > >
> > > > diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> > > > index 7fc3093..875d3ea 100644
> > > > --- a/pmd/pmd_memnic.c
> > > > +++ b/pmd/pmd_memnic.c
> > > > @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
> > > >  	int idx, next;
> > > >  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> > > >
> > > > -	if (!adapter->nic->hdr.valid)
> > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > >  		return 0;
> > > >
> > > >  	pkts = bytes = errs = 0;
> > > >  	idx = adapter->up_idx;
> > > >  	for (nr = 0; nr < nb_pkts; nr++) {
> > > >  		p = &data->packets[idx];
> > > > -		if (p->status != MEMNIC_PKT_ST_FILLED)
> > > > +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
> > > >  			break;
> > > >  		/* prefetch the next area */
> > > >  		next = idx;
> > > > -		if (++next >= MEMNIC_NR_PACKET)
> > > > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> > > >  			next = 0;
> > > >  		rte_prefetch0(&data->packets[next]);
> > > > -		if (p->len > framesz) {
> > > > +		if (unlikely(p->len > framesz)) {
> > > >  			errs++;
> > > >  			goto drop;
> > > >  		}
> > > >  		mb = rte_pktmbuf_alloc(adapter->mp);
> > > > -		if (!mb)
> > > > +		if (unlikely(!mb))
> > > >  			break;
> > > >
> > > >  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> > > > @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > >  	uint64_t pkts, bytes, errs;
> > > >  	uint32_t framesz = adapter->framesz;
> > > >
> > > > -	if (!adapter->nic->hdr.valid)
> > > > +	if (unlikely(!adapter->nic->hdr.valid))
> > > >  		return 0;
> > > >
> > > >  	pkts = bytes = errs = 0;
> > > > @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
> > > >  		struct rte_mbuf *sg;
> > > >  		void *ptr;
> > > >
> > > > -		if (pkt_len > framesz) {
> > > > +		if (unlikely(pkt_len > framesz)) {
> > > >  			errs++;
> > > >  			break;
> > > >  		}
> > > > @@ -379,7 +379,7 @@ retry:
> > > >  			goto retry;
> > > >  		}
> > > >
> > > > -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> > > > +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
> > > Why are you using ACCESS_ONCE here?  Or for that matter, anywhere else in this
> > > PMD?  The whole idea of the ACCESS_ONCE macro is to assign a value to a variable
> > > once and prevent it from getting reloaded from memory at a later time, this is
> > > exactly contrary to that, both in the sense that you're explicitly reloading the
> > > same variable multiple times, and that you're using it as part of a comparison
> > > operation, rather than an asignment operation
> > 
> > ACCESS_ONCE prevents compiler optimization and ensures load from memory.
> > There could be multiple threads which read/write that index.
> > We should compare the value previous and the current value in memory.
> > In that reason, I use ACCESS_ONCE macro to get value in the memory.
> 
> Should you not just make the variable volatile? That's the normal way to 
> guarantee reads from memory and prevent the compiler caching things in 
> registers.
> 

Agreed, the ACCESS_ONCE macro does that as part of its defintion, but seeing
ACCESS_ONCE used repeatedly in a loop just isn't easily parsible to human eyes.
Declare a volatile variable local to the function, assign it to the address of
the memory you want to read, and use it as you normally would.

Neil

> /Bruce
> 
> > 
> > thanks,
> > Hiroshi
> > 
> > > 
> > > Neil
> > 
> 

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Xie, Huawei]

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Hiroshi Shimamoto
> Sent: Tuesday, September 30, 2014 7:15 PM
> To: dev@dpdk.org
> Cc: Hayato Momma
> Subject: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit
> 
> From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> 
> To reduce instruction cache miss, add branch condition hints into
> recv/xmit functions. This improves a bit performance.
> 
> We can see performance improvements with memnic-tester.
> Using Xeon E5-2697 v2 @ 2.70GHz, 4 vCPU.
>  size |  before  |  after
>    64 | 5.54Mpps | 5.55Mpps
>   128 | 5.46Mpps | 5.44Mpps
>   256 | 5.21Mpps | 5.22Mpps
>   512 | 4.50Mpps | 4.52Mpps
>  1024 | 3.71Mpps | 3.73Mpps
>  1280 | 3.21Mpps | 3.22Mpps
>  1518 | 2.92Mpps | 2.93Mpps
> 
> Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> Reviewed-by: Hayato Momma <h-momma@ce.jp.nec.com>
> ---
>  pmd/pmd_memnic.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
> 
> diff --git a/pmd/pmd_memnic.c b/pmd/pmd_memnic.c
> index 7fc3093..875d3ea 100644
> --- a/pmd/pmd_memnic.c
> +++ b/pmd/pmd_memnic.c
> @@ -289,26 +289,26 @@ static uint16_t memnic_recv_pkts(void *rx_queue,
>  	int idx, next;
>  	struct rte_eth_stats *st = &adapter->stats[rte_lcore_id()];
> 
> -	if (!adapter->nic->hdr.valid)
> +	if (unlikely(!adapter->nic->hdr.valid))
>  		return 0;
> 
>  	pkts = bytes = errs = 0;
>  	idx = adapter->up_idx;
>  	for (nr = 0; nr < nb_pkts; nr++) {
>  		p = &data->packets[idx];
> -		if (p->status != MEMNIC_PKT_ST_FILLED)
> +		if (unlikely(p->status != MEMNIC_PKT_ST_FILLED))
>  			break;
>  		/* prefetch the next area */
>  		next = idx;
> -		if (++next >= MEMNIC_NR_PACKET)
> +		if (unlikely(++next >= MEMNIC_NR_PACKET))
On IA,  compiler can use add, cmp and cmov to avoid branch.
But If MEMNIC_NR_PACKET is always power of 2, 
	it is better just next = (next + 1) & (MEMNIC_NR_PACKET - 1)

>  			next = 0;
>  		rte_prefetch0(&data->packets[next]);
> -		if (p->len > framesz) {
> +		if (unlikely(p->len > framesz)) {
>  			errs++;
>  			goto drop;
>  		}
>  		mb = rte_pktmbuf_alloc(adapter->mp);
> -		if (!mb)
> +		if (unlikely(!mb))
>  			break;
> 
>  		rte_memcpy(rte_pktmbuf_mtod(mb, void *), p->data, p->len);
> @@ -350,7 +350,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
>  	uint64_t pkts, bytes, errs;
>  	uint32_t framesz = adapter->framesz;
> 
> -	if (!adapter->nic->hdr.valid)
> +	if (unlikely(!adapter->nic->hdr.valid))
>  		return 0;
> 
>  	pkts = bytes = errs = 0;
> @@ -360,7 +360,7 @@ static uint16_t memnic_xmit_pkts(void *tx_queue,
>  		struct rte_mbuf *sg;
>  		void *ptr;
> 
> -		if (pkt_len > framesz) {
> +		if (unlikely(pkt_len > framesz)) {
>  			errs++;
>  			break;
>  		}
> @@ -379,7 +379,7 @@ retry:
>  			goto retry;
>  		}
> 
> -		if (idx != ACCESS_ONCE(adapter->down_idx)) {
> +		if (unlikely(idx != ACCESS_ONCE(adapter->down_idx))) {
>  			/*
>  			 * host freed this and got false positive,
>  			 * need to recover the status and retry.
> @@ -388,7 +388,7 @@ retry:
>  			goto retry;
>  		}
> 
> -		if (++idx >= MEMNIC_NR_PACKET)
> +		if (unlikely(++idx >= MEMNIC_NR_PACKET))
>  			idx = 0;
>  		adapter->down_idx = idx;
> 
> --
> 1.8.3.1

Re: [dpdk-dev] [memnic PATCH v2 6/7] pmd: add branch hint in recv/xmit

[Thomas Monjalon]

2014-09-30 14:38, Xie, Huawei:
> > -		if (++next >= MEMNIC_NR_PACKET)
> > +		if (unlikely(++next >= MEMNIC_NR_PACKET))
> 
> On IA,  compiler can use add, cmp and cmov to avoid branch.
> But If MEMNIC_NR_PACKET is always power of 2, 
> 	it is better just next = (next + 1) & (MEMNIC_NR_PACKET - 1)

Power of 2 is not enforced for MEMNIC_NR_PACKET.

-- 
Thomas