diff mbox

[RFC/RFT,10/10] cpufreq: intel_pstate: enable boost for SKX

Message ID 20180516044911.28797-11-srinivas.pandruvada@linux.intel.com (mailing list archive)
State RFC, archived
Headers show

Commit Message

Srinivas Pandruvada May 16, 2018, 4:49 a.m. UTC
Enable HWP boost on Skylake server platform.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
---
 drivers/cpufreq/intel_pstate.c | 9 +++++++++
 1 file changed, 9 insertions(+)

Comments

Peter Zijlstra May 16, 2018, 7:49 a.m. UTC | #1
On Tue, May 15, 2018 at 09:49:11PM -0700, Srinivas Pandruvada wrote:
> Enable HWP boost on Skylake server platform.

Why not unconditionally enable it on everything HWP ?
Srinivas Pandruvada May 16, 2018, 3:46 p.m. UTC | #2
On Wed, 2018-05-16 at 09:49 +0200, Peter Zijlstra wrote:
> On Tue, May 15, 2018 at 09:49:11PM -0700, Srinivas Pandruvada wrote:
> > Enable HWP boost on Skylake server platform.
> 
> Why not unconditionally enable it on everything HWP ?
Never noticed in any difference performance in clients with HWP nor
anybody complained. Since clients uses single power domain, some other
CPUs always gives boost. 

Thanks,
Srinivas
Peter Zijlstra May 16, 2018, 3:54 p.m. UTC | #3
On Wed, May 16, 2018 at 08:46:24AM -0700, Srinivas Pandruvada wrote:
> On Wed, 2018-05-16 at 09:49 +0200, Peter Zijlstra wrote:
> > On Tue, May 15, 2018 at 09:49:11PM -0700, Srinivas Pandruvada wrote:
> > > Enable HWP boost on Skylake server platform.
> > 
> > Why not unconditionally enable it on everything HWP ?
> Never noticed in any difference performance in clients with HWP nor
> anybody complained. Since clients uses single power domain, some other
> CPUs always gives boost. 

Can't the hardware tell us about per-core stuff? Do we really have to
use tables for that?
Srinivas Pandruvada May 17, 2018, 12:52 a.m. UTC | #4
On Wed, 2018-05-16 at 17:54 +0200, Peter Zijlstra wrote:
> On Wed, May 16, 2018 at 08:46:24AM -0700, Srinivas Pandruvada wrote:
> > On Wed, 2018-05-16 at 09:49 +0200, Peter Zijlstra wrote:
> > > On Tue, May 15, 2018 at 09:49:11PM -0700, Srinivas Pandruvada
> > > wrote:
> > > > Enable HWP boost on Skylake server platform.
> > > 
> > > Why not unconditionally enable it on everything HWP ?
> > 
> > Never noticed in any difference performance in clients with HWP nor
> > anybody complained. Since clients uses single power domain, some
> > other
> > CPUs always gives boost. 
> 
> Can't the hardware tell us about per-core stuff? Do we really have to
> use tables for that?
AFAIK not in a standard way across all platforms. But I will check on
this.
diff mbox

Patch

diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c
index 65d11d2..827c003 100644
--- a/drivers/cpufreq/intel_pstate.c
+++ b/drivers/cpufreq/intel_pstate.c
@@ -1863,6 +1863,11 @@  static const struct x86_cpu_id intel_pstate_cpu_ee_disable_ids[] = {
 	{}
 };
 
+static const struct x86_cpu_id intel_pstate_hwp_boost_ids[] __initconst = {
+	ICPU(INTEL_FAM6_SKYLAKE_X, core_funcs),
+	{}
+};
+
 static int intel_pstate_init_cpu(unsigned int cpunum)
 {
 	struct cpudata *cpu;
@@ -1893,6 +1898,10 @@  static int intel_pstate_init_cpu(unsigned int cpunum)
 			intel_pstate_disable_ee(cpunum);
 
 		intel_pstate_hwp_enable(cpu);
+
+		id = x86_match_cpu(intel_pstate_hwp_boost_ids);
+		if (id)
+			hwp_boost = true;
 	}
 
 	intel_pstate_get_cpu_pstates(cpu);