Re: [gentoo-dev] [PATCH 1/2] rocm.eclass: Fix the xnack feature for gfx90a

2023-12-05 Thread WuYiyang
On Mon, Nov 27, 2023 at 08:08:40AM +0800, Benda Xu wrote: > Hi Yiyang, > > Yiyang Wu writes: > > > Upstream usually ships 2 version: gfx90a:xnack-, gfx90a:xnack+. Although > > a single gfx90a should have maximum compatibility, According to [1,2], > > compile with xnack+/xnack- may have better

Re: [gentoo-dev] [PATCH 1/2] rocm.eclass: Fix the xnack feature for gfx90a

2023-11-26 Thread Benda Xu
Hi Yiyang, Yiyang Wu writes: > Upstream usually ships 2 version: gfx90a:xnack-, gfx90a:xnack+. Although > a single gfx90a should have maximum compatibility, According to [1,2], > compile with xnack+/xnack- may have better performance on xnack > enabled/disabled GPUs. Therefore we ship both the

[gentoo-dev] [PATCH 1/2] rocm.eclass: Fix the xnack feature for gfx90a

2023-11-16 Thread Yiyang Wu
MI210 is xnack disabled by default. Compile with :xnack+ will produce GPU kernels that cannot be run on MI210. Two targets, gfx90a_xnack and gfx90a_noxnack are introduced to replace gfx90a, for the following reason: Upstream usually ships 2 version: gfx90a:xnack-, gfx90a:xnack+. Although a