On 2023-05-22 22:09, Qi.Chen@windriver.com wrote:
From: Chen Qi <Qi.Chen@windriver.com>

The current logic is not correct because if the time interval
between the current check and the last check is very small, the PSI
checker is not likely to block things even if the system is heavy
loaded.

It's not good to calculate the value too often. So we change to a 1s
check. As a build will usually take at least minutes, using the 1s value
seems reasonable.

I don't have time to check today but I'm not convinced that this is true.

I suspect that in the typical case, it would be okay but there will be
workloads (perhaps rare workloads) where you want to check more often.

The 1 second intervals where we no longer launch new work may
result in a significantly longer build. IIRC, our benchmark was building
core-image-minimal on a 24 core system without PSI regulation and then
with it set to 300. I'd look at the elapsed time and the buildchart/bootchart: 
   scripts/pybootchartgui/README.pybootchart

Ola, have you been able to test this and if not, can you?

All for today.

../Randy



Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
---
 bitbake/lib/bb/runqueue.py | 29 ++++++++++++++---------------
 1 file changed, 14 insertions(+), 15 deletions(-)

diff --git a/bitbake/lib/bb/runqueue.py b/bitbake/lib/bb/runqueue.py
index 02f1474540..4d49d25153 100644
--- a/bitbake/lib/bb/runqueue.py
+++ b/bitbake/lib/bb/runqueue.py
@@ -179,6 +179,7 @@ class RunQueueScheduler(object):
                     self.prev_memory_pressure = memory_pressure_fds.readline().split()[4].split("=")[1]
                     self.prev_pressure_time = time.time()
                 self.check_pressure = True
+                self.psi_exceeded = False
             except:
                 bb.note("The /proc/pressure files can't be read. Continuing build without monitoring pressure")
                 self.check_pressure = False
@@ -191,6 +192,10 @@ class RunQueueScheduler(object):
         BB_PRESSURE_MAX_{CPU|IO|MEMORY} are set, return True if above threshold.
         """
         if self.check_pressure:
+            now = time.time()
+            tdiff = now - self.prev_pressure_time
+            if tdiff < 1.0:
+                return self.psi_exceeded
             with open("/proc/pressure/cpu") as cpu_pressure_fds, \
                 open("/proc/pressure/io") as io_pressure_fds, \
                 open("/proc/pressure/memory") as memory_pressure_fds:
@@ -198,21 +203,15 @@ class RunQueueScheduler(object):
                 curr_cpu_pressure = cpu_pressure_fds.readline().split()[4].split("=")[1]
                 curr_io_pressure = io_pressure_fds.readline().split()[4].split("=")[1]
                 curr_memory_pressure = memory_pressure_fds.readline().split()[4].split("=")[1]
-                now = time.time()
-                tdiff = now - self.prev_pressure_time
-                if tdiff > 1.0:
-                    exceeds_cpu_pressure =  self.rq.max_cpu_pressure and (float(curr_cpu_pressure) - float(self.prev_cpu_pressure)) / tdiff > self.rq.max_cpu_pressure
-                    exceeds_io_pressure =  self.rq.max_io_pressure and (float(curr_io_pressure) - float(self.prev_io_pressure)) / tdiff > self.rq.max_io_pressure
-                    exceeds_memory_pressure = self.rq.max_memory_pressure and (float(curr_memory_pressure) - float(self.prev_memory_pressure)) / tdiff > self.rq.max_memory_pressure
-                    self.prev_cpu_pressure = curr_cpu_pressure
-                    self.prev_io_pressure = curr_io_pressure
-                    self.prev_memory_pressure = curr_memory_pressure
-                    self.prev_pressure_time = now
-                else:
-                    exceeds_cpu_pressure =  self.rq.max_cpu_pressure and (float(curr_cpu_pressure) - float(self.prev_cpu_pressure)) > self.rq.max_cpu_pressure
-                    exceeds_io_pressure =  self.rq.max_io_pressure and (float(curr_io_pressure) - float(self.prev_io_pressure)) > self.rq.max_io_pressure
-                    exceeds_memory_pressure = self.rq.max_memory_pressure and (float(curr_memory_pressure) - float(self.prev_memory_pressure)) > self.rq.max_memory_pressure
-            return (exceeds_cpu_pressure or exceeds_io_pressure or exceeds_memory_pressure)
+                exceeds_cpu_pressure =  self.rq.max_cpu_pressure and (float(curr_cpu_pressure) - float(self.prev_cpu_pressure)) / tdiff > self.rq.max_cpu_pressure
+                exceeds_io_pressure =  self.rq.max_io_pressure and (float(curr_io_pressure) - float(self.prev_io_pressure)) / tdiff > self.rq.max_io_pressure
+                exceeds_memory_pressure = self.rq.max_memory_pressure and (float(curr_memory_pressure) - float(self.prev_memory_pressure)) / tdiff > self.rq.max_memory_pressure
+                self.prev_cpu_pressure = curr_cpu_pressure
+                self.prev_io_pressure = curr_io_pressure
+                self.prev_memory_pressure = curr_memory_pressure
+                self.prev_pressure_time = now
+                self.psi_exceeded = exceeds_cpu_pressure or exceeds_io_pressure or exceeds_memory_pressure
+            return self.psi_exceeded
         return False
 
     def next_buildable_task(self):


-- 
# Randy MacLeod
# Wind River Linux