On 12/10/18 4:24 PM, Aubrey Li wrote:
> AVX-512 components usage could cause core turbo frequency drop. So
> it's useful to expose AVX-512 components usage as a heuristic hint
> for the user space job scheduler to cluster the AVX-512 using tasks
> together.
>
> Example:
> $ cat /proc/pid/status | grep AVX512_hint
> AVX512_hint: 1
>
> The hint number '0' indicates the task recently didn't use AVX-512
> components thus unlikely has frequency drop issue. And the number '1'
> indicates the task recently used AVX-512 components thus could cause
> core frequency drop. User space tools may want to further check by:
>
> $ perf stat --pid <pid> -e core_power.lvl2_turbo_license -- sleep 1
>
> Performance counter stats for process id '3558':
>
> 3,251,565,961 core_power.lvl2_turbo_license
>
> 1.004031387 seconds time elapsed
>
> Non-zero counter value confirms that the task causes frequency drop.
>
> Signed-off-by: Aubrey Li <aubrey...@linux.intel.com>
> Cc: Peter Zijlstra <pet...@infradead.org>
> Cc: Andi Kleen <a...@linux.intel.com>
> Cc: Tim Chen <tim.c.c...@linux.intel.com>
> Cc: Dave Hansen <dave.han...@intel.com>
> Cc: Arjan van de Ven <ar...@linux.intel.com>
> ---
> arch/x86/kernel/fpu/xstate.c | 19 +++++++++++++++++++
> fs/proc/array.c | 5 +++++
> 2 files changed, 24 insertions(+)
>
> diff --git a/arch/x86/kernel/fpu/xstate.c b/arch/x86/kernel/fpu/xstate.c
> index 87a57b7642d3..98baa47c97b0 100644
> --- a/arch/x86/kernel/fpu/xstate.c
> +++ b/arch/x86/kernel/fpu/xstate.c
> @@ -7,6 +7,7 @@
> #include <linux/cpu.h>
> #include <linux/mman.h>
> #include <linux/pkeys.h>
> +#include <linux/seq_file.h>
>
> #include <asm/fpu/api.h>
> #include <asm/fpu/internal.h>
> @@ -1245,3 +1246,21 @@ int copy_user_to_xstate(struct xregs_state *xsave,
> const void __user *ubuf)
>
> return 0;
> }
> +
> +/*
> + * Report CPU specific thread state
> + */
> +void arch_task_state(struct seq_file *m, struct task_struct *task)
> +{
> + /*
> + * Check the processor and build option if AVX512 is supported.
> + */
> + if (!cpu_feature_enabled(X86_FEATURE_AVX512F))
> + return;
> + /*
> + * Report AVX-512 components usage:
> + */
> + seq_put_decimal_ull(m, "AVX512_hint:\t",
> + task->thread.fpu.avx512_usage ? 1 : 0);
I believe you need some kind of documentation of this new interface
in kernel's Documentation. So an admin doesn't need to look into the
kernel code to figure out what this /proc/pid/status is supposed to do.
Tim
