[dpdk-dev] ixgbe: fix icc compile error

  Fixes: 06554d381d97 ("ixgbe: speed up non-vector Tx")

icc (icc (ICC) 16.0.1 20151021) is generating following compile error:
"
  CC ixgbe_rxtx.o
  .../drivers/net/ixgbe/ixgbe_rxtx.c(153): error #3656: variable
  "free" may be used before its value is set
                      (nb_free > 0 && m->pool != free[0]->pool)) {
			                         ^
"

Indeed this is a false positive and code is correct.
"nb_free" check prevents the free[] access before its value set.

But the patch is just for compiler to remove compiler error.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
 drivers/net/ixgbe/ixgbe_rxtx.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
  

On Wed, 23 Mar 2016 17:45:58 +0000
Ferruh Yigit <ferruh.yigit@intel.com> wrote:

> Fixes: 06554d381d97 ("ixgbe: speed up non-vector Tx")
> 
> icc (icc (ICC) 16.0.1 20151021) is generating following compile error:
> "
>   CC ixgbe_rxtx.o
>   .../drivers/net/ixgbe/ixgbe_rxtx.c(153): error #3656: variable
>   "free" may be used before its value is set
>                       (nb_free > 0 && m->pool != free[0]->pool)) {
> 			                         ^
> "
> 
> Indeed this is a false positive and code is correct.
> "nb_free" check prevents the free[] access before its value set.
> 
> But the patch is just for compiler to remove compiler error.
> 
> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> ---
>  drivers/net/ixgbe/ixgbe_rxtx.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ixgbe/ixgbe_rxtx.c b/drivers/net/ixgbe/ixgbe_rxtx.c
> index ff6ddb8..ef1a26f 100644
> --- a/drivers/net/ixgbe/ixgbe_rxtx.c
> +++ b/drivers/net/ixgbe/ixgbe_rxtx.c
> @@ -128,7 +128,8 @@ ixgbe_tx_free_bufs(struct ixgbe_tx_queue *txq)
>  	struct ixgbe_tx_entry *txep;
>  	uint32_t status;
>  	int i, nb_free = 0;
> -	struct rte_mbuf *m, *free[RTE_IXGBE_TX_MAX_FREE_BUF_SZ];
> +	struct rte_mbuf *m;
> +	struct rte_mbuf *free[RTE_IXGBE_TX_MAX_FREE_BUF_SZ] = {0};
>  
>  	/* check DD bit on threshold descriptor */
>  	status = txq->tx_ring[txq->tx_next_dd].wb.status;

Although this fixes a compiler bug, it also introduces a performance hit.
The initialization will generate a string instruct (rep; stosz) and this
stalls the multi-execution unit on many Intel CPU's.

Better to add a override to the Makefile (for icc only and with big comment).
  

On 3/23/2016 7:35 PM, Stephen Hemminger wrote:
> On Wed, 23 Mar 2016 17:45:58 +0000
> Ferruh Yigit <ferruh.yigit@intel.com> wrote:
> 
>> Fixes: 06554d381d97 ("ixgbe: speed up non-vector Tx")
>>
>> icc (icc (ICC) 16.0.1 20151021) is generating following compile error:
>> "
>>   CC ixgbe_rxtx.o
>>   .../drivers/net/ixgbe/ixgbe_rxtx.c(153): error #3656: variable
>>   "free" may be used before its value is set
>>                       (nb_free > 0 && m->pool != free[0]->pool)) {
>> 			                         ^
>> "
>>
>> Indeed this is a false positive and code is correct.
>> "nb_free" check prevents the free[] access before its value set.
>>
>> But the patch is just for compiler to remove compiler error.
>>
>> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
>> ---
>>  drivers/net/ixgbe/ixgbe_rxtx.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/ixgbe/ixgbe_rxtx.c b/drivers/net/ixgbe/ixgbe_rxtx.c
>> index ff6ddb8..ef1a26f 100644
>> --- a/drivers/net/ixgbe/ixgbe_rxtx.c
>> +++ b/drivers/net/ixgbe/ixgbe_rxtx.c
>> @@ -128,7 +128,8 @@ ixgbe_tx_free_bufs(struct ixgbe_tx_queue *txq)
>>  	struct ixgbe_tx_entry *txep;
>>  	uint32_t status;
>>  	int i, nb_free = 0;
>> -	struct rte_mbuf *m, *free[RTE_IXGBE_TX_MAX_FREE_BUF_SZ];
>> +	struct rte_mbuf *m;
>> +	struct rte_mbuf *free[RTE_IXGBE_TX_MAX_FREE_BUF_SZ] = {0};
>>  
>>  	/* check DD bit on threshold descriptor */
>>  	status = txq->tx_ring[txq->tx_next_dd].wb.status;
> 
> Although this fixes a compiler bug, it also introduces a performance hit.
> The initialization will generate a string instruct (rep; stosz) and this
> stalls the multi-execution unit on many Intel CPU's.
> 
> Better to add a override to the Makefile (for icc only and with big comment).
> 

Thank you for the comment. Konstantin also expressed similar performance
concern related this patch.

Nack for the patch, we will look for different solution.

Regards,
ferruh