On 11/13/2019 2:31 AM, Ye Xiaolong wrote:
> On 11/12, Bruce Richardson wrote:
>> When compiling with 16B descriptor support enabled, clang compiles gave an
>> error, complaining that the final parameter of _mm256_blend_epi32() had to
>> be an immediate value (i.e. compile-time constant). While it appears that
>> GCC was able to convert the constant variable value "fdir_blend_mask" into
>> the blend call, clang was not doing so. To guarantee the use of an
>> immediate we convert the variable value to a #define.
>>
>> Fixes: 7d087a0a8b8e ("net/i40e: support flow director on AVX Rx")
>> Cc: harry.van.haa...@intel.com
>>
>> Signed-off-by: Bruce Richardson <bruce.richard...@intel.com>
>> ---
>> drivers/net/i40e/i40e_rxtx_vec_avx2.c | 10 +++++-----
>> 1 file changed, 5 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/net/i40e/i40e_rxtx_vec_avx2.c
>> b/drivers/net/i40e/i40e_rxtx_vec_avx2.c
>> index b9f1a240c..3bcef1363 100644
>> --- a/drivers/net/i40e/i40e_rxtx_vec_avx2.c
>> +++ b/drivers/net/i40e/i40e_rxtx_vec_avx2.c
>> @@ -529,6 +529,7 @@ _recv_raw_pkts_vec_avx2(struct i40e_rx_queue *rxq,
>> struct rte_mbuf **rx_pkts,
>> * identifies an FDIR ID match, and zeros the RSS value
>> * in the mbuf on FDIR match to keep mbuf data clean.
>> */
>> +#define FDIR_BLEND_MASK ((1 << 3) | (1 << 7))
>>
>> /* Flags:
>> * - Take flags, shift bits to null out
>> @@ -557,9 +558,8 @@ _recv_raw_pkts_vec_avx2(struct i40e_rx_queue *rxq,
>> struct rte_mbuf **rx_pkts,
>> * otherwise the mb0_1 register RSS field is zeroed.
>> */
>> const __m256i fdir_zero_mask = _mm256_setzero_si256();
>> - const uint32_t fdir_blend_mask = (1 << 3) | (1 << 7);
>> __m256i tmp0_1 = _mm256_blend_epi32(fdir_zero_mask,
>> - fdir_mask, fdir_blend_mask);
>> + fdir_mask, FDIR_BLEND_MASK);
>> __m256i fdir_mb0_1 = _mm256_and_si256(mb0_1, fdir_mask);
>> mb0_1 = _mm256_andnot_si256(tmp0_1, mb0_1);
>>
>> @@ -575,7 +575,7 @@ _recv_raw_pkts_vec_avx2(struct i40e_rx_queue *rxq,
>> struct rte_mbuf **rx_pkts,
>> __m256i tmp2_3 = _mm256_alignr_epi8(fdir_mask,
>> fdir_mask, 12);
>> __m256i fdir_mb2_3 = _mm256_and_si256(mb2_3, tmp2_3);
>> tmp2_3 = _mm256_blend_epi32(fdir_zero_mask, tmp2_3,
>> - fdir_blend_mask);
>> + FDIR_BLEND_MASK);
>> mb2_3 = _mm256_andnot_si256(tmp2_3, mb2_3);
>> rx_pkts[i + 2]->hash.fdir.hi =
>> _mm256_extract_epi32(fdir_mb2_3, 3);
>> rx_pkts[i + 3]->hash.fdir.hi =
>> _mm256_extract_epi32(fdir_mb2_3, 7);
>> @@ -583,7 +583,7 @@ _recv_raw_pkts_vec_avx2(struct i40e_rx_queue *rxq,
>> struct rte_mbuf **rx_pkts,
>> __m256i tmp4_5 = _mm256_alignr_epi8(fdir_mask,
>> fdir_mask, 8);
>> __m256i fdir_mb4_5 = _mm256_and_si256(mb4_5, tmp4_5);
>> tmp4_5 = _mm256_blend_epi32(fdir_zero_mask, tmp4_5,
>> - fdir_blend_mask);
>> + FDIR_BLEND_MASK);
>> mb4_5 = _mm256_andnot_si256(tmp4_5, mb4_5);
>> rx_pkts[i + 4]->hash.fdir.hi =
>> _mm256_extract_epi32(fdir_mb4_5, 3);
>> rx_pkts[i + 5]->hash.fdir.hi =
>> _mm256_extract_epi32(fdir_mb4_5, 7);
>> @@ -591,7 +591,7 @@ _recv_raw_pkts_vec_avx2(struct i40e_rx_queue *rxq,
>> struct rte_mbuf **rx_pkts,
>> __m256i tmp6_7 = _mm256_alignr_epi8(fdir_mask,
>> fdir_mask, 4);
>> __m256i fdir_mb6_7 = _mm256_and_si256(mb6_7, tmp6_7);
>> tmp6_7 = _mm256_blend_epi32(fdir_zero_mask, tmp6_7,
>> - fdir_blend_mask);
>> + FDIR_BLEND_MASK);
>> mb6_7 = _mm256_andnot_si256(tmp6_7, mb6_7);
>> rx_pkts[i + 6]->hash.fdir.hi =
>> _mm256_extract_epi32(fdir_mb6_7, 3);
>> rx_pkts[i + 7]->hash.fdir.hi =
>> _mm256_extract_epi32(fdir_mb6_7, 7);
>> --
>> 2.21.0
>>
>
> Acked-by: Xiaolong Ye <xiaolong...@intel.com>
>
> Applied to dpdk-next-net-intel, Thanks.
>
Adding following build error into commit log while pulling to next-net, thanks
to Bruce:
.../i40e_rxtx_vec_avx2.c:561:21: error: argument to '__builtin_ia32_pblendd256'
must be a constant integer
__m256i tmp0_1 = _mm256_blend_epi32(fdir_zero_mask,
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
