Re: [dpdk-dev] [PATCH] net/mlx5: restrict workaround of gcc AVX512F bug

2018-11-12 Thread Jerin Jacob
cal.com, justin.pa...@microsoft.com > Subject: [dpdk-dev] [PATCH] net/mlx5: restrict workaround of gcc AVX512F bug > X-Mailer: git-send-email 2.19.0 > > +/* > + * GCC bug workaround for rte_memcpy broken when optimized for AVX512. > + * Details are in https://bugs.dpdk.org/show_bug.cgi

Re: [dpdk-dev] [PATCH] net/mlx5: restrict workaround of gcc AVX512F bug

2018-11-12 Thread Thomas Monjalon
13/11/2018 01:14, Stephen Hemminger: > On Mon, Nov 12, 2018, 4:01 PM Thomas Monjalon > > A bug was found when the inline function mlx5_tx_complete() > > is optimized with AVX512F instructions. It corrupts an offset > > in the instructions vmovdqu8 of the AVX2 version of rte_mov128(), > > used in

Re: [dpdk-dev] [PATCH] net/mlx5: restrict workaround of gcc AVX512F bug

2018-11-12 Thread Stephen Hemminger
The additional annotations clutter the code. How big a performance hit is it to disable for whole driver? Or just use memcpy instead of rte_memcpy? On Mon, Nov 12, 2018, 4:01 PM Thomas Monjalon A bug was found when the inline function mlx5_tx_complete() > is optimized with AVX512F instructions. I

[dpdk-dev] [PATCH] net/mlx5: restrict workaround of gcc AVX512F bug

2018-11-12 Thread Thomas Monjalon
A bug was found when the inline function mlx5_tx_complete() is optimized with AVX512F instructions. It corrupts an offset in the instructions vmovdqu8 of the AVX2 version of rte_mov128(), used in rte_memcpy(), which is called in rte_mempool_put_bulk(). All the above functions are inline. So the wo