Re: [PATCH v4 2/2] sched/topology: Improve load balancing on AMD EPYC

2019-10-09 Thread Guenter Roeck
On 10/9/19 5:04 AM, Matt Fleming wrote: On Mon, 07 Oct, at 08:28:16AM, Guenter Roeck wrote: This patch causes build errors on systems where NUMA does not depend on SMP, for example MIPS and PPC. For example, building mips:ip27_defconfig with SMP disabled results in mips-linux-ld:

Re: [PATCH v4 2/2] sched/topology: Improve load balancing on AMD EPYC

2019-10-09 Thread Matt Fleming
On Mon, 07 Oct, at 08:28:16AM, Guenter Roeck wrote: > > This patch causes build errors on systems where NUMA does not depend on SMP, > for example MIPS and PPC. For example, building mips:ip27_defconfig with SMP > disabled results in > > mips-linux-ld: mm/page_alloc.o: in function

Re: [PATCH v4 2/2] sched/topology: Improve load balancing on AMD EPYC

2019-10-07 Thread Guenter Roeck
Hi, On Thu, Aug 08, 2019 at 08:53:01PM +0100, Matt Fleming wrote: > SD_BALANCE_{FORK,EXEC} and SD_WAKE_AFFINE are stripped in sd_init() > for any sched domains with a NUMA distance greater than 2 hops > (RECLAIM_DISTANCE). The idea being that it's expensive to balance > across domains that far

[PATCH v4 2/2] sched/topology: Improve load balancing on AMD EPYC

2019-08-08 Thread Matt Fleming
SD_BALANCE_{FORK,EXEC} and SD_WAKE_AFFINE are stripped in sd_init() for any sched domains with a NUMA distance greater than 2 hops (RECLAIM_DISTANCE). The idea being that it's expensive to balance across domains that far apart. However, as is rather unfortunately explained in commit