From 2c15fb2a5bdf74b3039ea6cb463e4001274157fa Mon Sep 17 00:00:00 2001
From: Richard Hsu <hsuy@google.com>
Date: Tue, 17 May 2022 18:45:22 -0700
Subject: [PATCH] [Tuning] Disable CPU Idle in NNAPI workload with PMQoS CPU
 DMA Latency

To improve the return path latency, we want to keep the CPU at at least WFI state (Idle_1). The PMQos cpu_dma_latency knob prevents the CPU from falling below WFI state. This makes the return path CPU wakeup latency very good. Check with wvw@, the power impact shouldn't be too significant.

The average energy cost per inference dropped from 3.85 to 3.47 mJ. The reason why the power number is lower WITH disable-idle is that, due to better latency, we get to run more inferences given the same amount of time. This makes the average power consumption lower.

Measurement:

MLPerf IC model Latency (ms) Power (mW) Energy/inference (mJ) MLPerf scores
Default 1.35 2837 3.85 560
Disable CPU Idle 0.98 3539 3.47 826

https://docs.google.com/presentation/d/1zx7sLkhOClmuRTCrq8-l3N1mZrrv7f-CtcdMuzV0eaI/edit?pli=1#slide=id.g12dd9e50b4b_0_0

Bug: 232183574
Test: MLPerf on Android T. Performance improved. Verified on Perfetto.
Change-Id: Ia807bf0849e4d9b0b0e8c9510335129ca89e791f
---
 powerhint-cheetah-a0.json | 15 +++++++++++++++
 powerhint-cheetah.json    | 15 +++++++++++++++
 powerhint-panther-a0.json | 15 +++++++++++++++
 powerhint-panther.json    | 15 +++++++++++++++
 4 files changed, 60 insertions(+)

diff --git a/powerhint-cheetah-a0.json b/powerhint-cheetah-a0.json
index 79ef6ff..603f259 100644
--- a/powerhint-cheetah-a0.json
+++ b/powerhint-cheetah-a0.json
@@ -172,6 +172,15 @@
       "DefaultIndex": 0,
       "ResetOnInit": true
     },
+    {
+      "Name": "PMQoSCpuDmaLatency",
+      "Path": "/dev/cpu_dma_latency",
+      "Values": [
+        "44",
+        "1000"
+      ],
+      "HoldFd": true
+    },
     {
       "Name": "CDPreferIdle",
       "Path": "/proc/vendor_sched/cam_prefer_idle",
@@ -1655,6 +1664,12 @@
       "Duration": 2000,
       "Value": "512"
     },
+    {
+      "PowerHint": "ML_ACC",
+      "Node": "PMQoSCpuDmaLatency",
+      "Duration": 2000,
+      "Value": "44"
+    },
     {
       "PowerHint": "DEVICE_IDLE",
       "Node": "RestrictedCpuset",
diff --git a/powerhint-cheetah.json b/powerhint-cheetah.json
index 8e963dd..7a1f75b 100644
--- a/powerhint-cheetah.json
+++ b/powerhint-cheetah.json
@@ -190,6 +190,15 @@
       "DefaultIndex": 0,
       "ResetOnInit": true
     },
+    {
+      "Name": "PMQoSCpuDmaLatency",
+      "Path": "/dev/cpu_dma_latency",
+      "Values": [
+        "44",
+        "1000"
+      ],
+      "HoldFd": true
+    },
     {
       "Name": "CDPreferIdle",
       "Path": "/proc/vendor_sched/cam_prefer_idle",
@@ -1677,6 +1686,12 @@
       "Duration": 2000,
       "Value": "512"
     },
+    {
+      "PowerHint": "ML_ACC",
+      "Node": "PMQoSCpuDmaLatency",
+      "Duration": 2000,
+      "Value": "44"
+    },
     {
       "PowerHint": "DEVICE_IDLE",
       "Node": "RestrictedCpuset",
diff --git a/powerhint-panther-a0.json b/powerhint-panther-a0.json
index 6e6d422..3954149 100644
--- a/powerhint-panther-a0.json
+++ b/powerhint-panther-a0.json
@@ -172,6 +172,15 @@
       "DefaultIndex": 0,
       "ResetOnInit": true
     },
+    {
+      "Name": "PMQoSCpuDmaLatency",
+      "Path": "/dev/cpu_dma_latency",
+      "Values": [
+        "44",
+        "1000"
+      ],
+      "HoldFd": true
+    },
     {
       "Name": "CDPreferIdle",
       "Path": "/proc/vendor_sched/cam_prefer_idle",
@@ -1639,6 +1648,12 @@
       "Duration": 2000,
       "Value": "512"
     },
+    {
+      "PowerHint": "ML_ACC",
+      "Node": "PMQoSCpuDmaLatency",
+      "Duration": 2000,
+      "Value": "44"
+    },
     {
       "PowerHint": "DEVICE_IDLE",
       "Node": "RestrictedCpuset",
diff --git a/powerhint-panther.json b/powerhint-panther.json
index 964eb40..a6eb0da 100644
--- a/powerhint-panther.json
+++ b/powerhint-panther.json
@@ -190,6 +190,15 @@
       "DefaultIndex": 0,
       "ResetOnInit": true
     },
+    {
+      "Name": "PMQoSCpuDmaLatency",
+      "Path": "/dev/cpu_dma_latency",
+      "Values": [
+        "44",
+        "1000"
+      ],
+      "HoldFd": true
+    },
     {
       "Name": "CDPreferIdle",
       "Path": "/proc/vendor_sched/cam_prefer_idle",
@@ -1661,6 +1670,12 @@
       "Duration": 2000,
       "Value": "512"
     },
+    {
+      "PowerHint": "ML_ACC",
+      "Node": "PMQoSCpuDmaLatency",
+      "Duration": 2000,
+      "Value": "44"
+    },
     {
       "PowerHint": "DEVICE_IDLE",
       "Node": "RestrictedCpuset",