[PATCH] cpufreq/amd-pstate: Add comment explaining nominal_perf usage for performance policy

Juan Martinez posted 1 patch 3 weeks ago
drivers/cpufreq/amd-pstate.c | 13 +++++++++++++
1 file changed, 13 insertions(+)
[PATCH] cpufreq/amd-pstate: Add comment explaining nominal_perf usage for performance policy
Posted by Juan Martinez 3 weeks ago
Add comment explaining why nominal_perf is used for MinPerf when the
CPU frequency policy is set to CPUFREQ_POLICY_PERFORMANCE, rather than
using highest_perf or lowest_nonlinear_perf.

Signed-off-by: Juan Martinez <juan.martinez@amd.com>
---
 drivers/cpufreq/amd-pstate.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/drivers/cpufreq/amd-pstate.c b/drivers/cpufreq/amd-pstate.c
index c45bc98721d2..88b26f36937b5 100644
--- a/drivers/cpufreq/amd-pstate.c
+++ b/drivers/cpufreq/amd-pstate.c
@@ -636,6 +636,19 @@ static void amd_pstate_update_min_max_limit(struct cpufreq_policy *policy)
 	WRITE_ONCE(cpudata->max_limit_freq, policy->max);
 
 	if (cpudata->policy == CPUFREQ_POLICY_PERFORMANCE) {
+		/*
+		 * For performance policy, set MinPerf to nominal_perf rather than
+		 * highest_perf or lowest_nonlinear_perf.
+		 *
+		 * Per commit 0c411b39e4f4c, using highest_perf was observed
+		 * to cause frequency throttling on power-limited platforms, leading to
+		 * performance regressions. Using lowest_nonlinear_perf would limit
+		 * performance too much for HPC workloads requiring high frequency
+		 * operation and minimal wakeup latency from idle states.
+		 *
+		 * nominal_perf therefore provides a balance by avoiding throttling
+		 * while still maintaining enough performance for HPC workloads.
+		 */
 		perf.min_limit_perf = min(perf.nominal_perf, perf.max_limit_perf);
 		WRITE_ONCE(cpudata->min_limit_freq, min(cpudata->nominal_freq, cpudata->max_limit_freq));
 	} else {
Re: [PATCH] cpufreq/amd-pstate: Add comment explaining nominal_perf usage for performance policy
Posted by Viresh Kumar 2 weeks, 5 days ago
On 16-01-26, 15:45, Juan Martinez wrote:
> Add comment explaining why nominal_perf is used for MinPerf when the
> CPU frequency policy is set to CPUFREQ_POLICY_PERFORMANCE, rather than
> using highest_perf or lowest_nonlinear_perf.
> 
> Signed-off-by: Juan Martinez <juan.martinez@amd.com>
> ---
>  drivers/cpufreq/amd-pstate.c | 13 +++++++++++++
>  1 file changed, 13 insertions(+)
> 
> diff --git a/drivers/cpufreq/amd-pstate.c b/drivers/cpufreq/amd-pstate.c
> index c45bc98721d2..88b26f36937b5 100644
> --- a/drivers/cpufreq/amd-pstate.c
> +++ b/drivers/cpufreq/amd-pstate.c
> @@ -636,6 +636,19 @@ static void amd_pstate_update_min_max_limit(struct cpufreq_policy *policy)
>  	WRITE_ONCE(cpudata->max_limit_freq, policy->max);
>  
>  	if (cpudata->policy == CPUFREQ_POLICY_PERFORMANCE) {
> +		/*
> +		 * For performance policy, set MinPerf to nominal_perf rather than
> +		 * highest_perf or lowest_nonlinear_perf.
> +		 *
> +		 * Per commit 0c411b39e4f4c, using highest_perf was observed
> +		 * to cause frequency throttling on power-limited platforms, leading to
> +		 * performance regressions. Using lowest_nonlinear_perf would limit
> +		 * performance too much for HPC workloads requiring high frequency
> +		 * operation and minimal wakeup latency from idle states.
> +		 *
> +		 * nominal_perf therefore provides a balance by avoiding throttling
> +		 * while still maintaining enough performance for HPC workloads.
> +		 */
>  		perf.min_limit_perf = min(perf.nominal_perf, perf.max_limit_perf);
>  		WRITE_ONCE(cpudata->min_limit_freq, min(cpudata->nominal_freq, cpudata->max_limit_freq));
>  	} else {

Applied. Thanks.

-- 
viresh