rteval: Add measurement and load location to run report
PATCH v2 1/1 rteval: Add measurement and load location to run report
The run report produced at the end of a run does not contain information on load and measurement thread locations.
Adjust MakeReport() functions of LoadModules and MeasurementModules class so that new properties with number of loads and cpu information are added to the XML report and can be read by rteval_text.xsl.
rteval/modules/loads/__init__.py +7 -0
@@ -30,6 +30,7 @@ import libxml2 from rteval.Log import Log from rteval.rtevalConfig import rtevalCfgSection from rteval.modules import RtEvalModules, rtevalModulePrototype +from rteval.systopology import collapse_cpulist, SysTopology class LoadThread(rtevalModulePrototype): def __init__(self, name, config, logger=None):
@@ -131,6 +132,12 @@ class LoadModules(RtEvalModules): def MakeReport(self): rep_n = RtEvalModules.MakeReport(self) rep_n.newProp("load_average", str(self.GetLoadAvg())) + rep_n.newProp("loads", str(self.ModulesLoaded())) + cpulist = self._cfg.GetSection(self._module_config).cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("loadcpus", cpulist) return rep_n
rteval/modules/measurement/__init__.py +7 -1
@@ -24,7 +24,7 @@ import libxml2 from rteval.modules import RtEvalModules, ModuleContainer - +from rteval.systopology import collapse_cpulist, SysTopology class MeasurementProfile(RtEvalModules): """Keeps and controls all the measurement modules with the same measurement profile"""
@@ -189,6 +189,12 @@ measurement profiles, based on their characteristics""" # Get the reports from all meaurement modules in all measurement profiles rep_n = libxml2.newNode("Measurements") + cpulist = self.__cfg.GetSection("measurement").cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("measurecpus", cpulist) + for mp in self.__measureprofiles: mprep_n = mp.MakeReport() if mprep_n:
rteval/rteval_text.xsl +8 -0
@@ -13,6 +13,14 @@ <xsl:value-of select="run_info/date"/><xsl:text> </xsl:text><xsl:value-of select="run_info/time"/> <xsl:text> </xsl:text> + <xsl:text> Loads: </xsl:text> + <xsl:value-of select="loads/@loads"/><xsl:text> loads run on cores </xsl:text><xsl:value-of select="loads/@loadcpus"/> + <xsl:text> </xsl:text> + + <xsl:text> Measurement: </xsl:text> + <xsl:text>measurement threads run on cores </xsl:text><xsl:value-of select="Measurements/@measurecpus"/> + <xsl:text> </xsl:text> + <xsl:text> Run time: </xsl:text> <xsl:value-of select="run_info/@days"/><xsl:text> days </xsl:text> <xsl:value-of select="run_info/@hours"/><xsl:text>h </xsl:text>
Signature · 1 line
2.31.1
Quoted 14 lines · Leah Leshchinsky
The run report produced at the end of a run does not contain information on load and measurement thread locations.
Adjust
MakeReport()functions of LoadModules and MeasurementModules class so that new properties with number of loads and cpu information are added to the XML report and can be read by rteval_text.xsl.Signed-off-by Leah Leshchinsky <[email protected]>rteval/modules/loads/__init__.py +0
-0
rteval/modules/loads/__init__.py import libxml2 · line 33
Quoted 4 lines
rteval/modules/loads/__init__.py +1
-0@@ -30,6 +30,7 @@ import libxml2 from rteval.Log import Log from rteval.rtevalConfig import rtevalCfgSection from rteval.modules import RtEvalModules, rtevalModulePrototype +from rteval.systopology import collapse_cpulist, SysTopology
add CpuList, and you might not need collapse_cpulist, and SysTopology
rteval/modules/loads/__init__.py class LoadModules(RtEvalModules): · line 136
Quoted 9 lines
rteval/modules/loads/__init__.py +2
-0class LoadThread(rtevalModulePrototype): def __init__(self, name, config, logger=None):@@ -131,6 +132,12 @@ class LoadModules(RtEvalModules): def MakeReport(self): rep_n = RtEvalModules.MakeReport(self) rep_n.newProp("load_average", str(self.GetLoadAvg())) + rep_n.newProp("loads", str(self.ModulesLoaded())) + cpulist = self._cfg.GetSection(self._module_config).cpulist
The above value is either going to be all cpus, or the value of --loads-cpulist=LIST or the inverse of --measurement-cpulist=LIST I don't think it can be empty, so the following code after if not cpulist is deadcode
rteval/modules/loads/__init__.py class LoadModules(RtEvalModules): · line 139
Quoted 3 lines
rteval/modules/loads/__init__.py +3
-0+ if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus())
If the above code were to run it would just be a list of all online cpus
Admitedly the systopology api is a bit confusing but since it does a lot, I think all you really need is
cpulist = CpuList(self._cfg.GetSection(self._module_config).cpulist)
rep_n.newProp("loadcpus", str(cpulist))
And that's it. You need to do something similar to the measurement section too.
Quoted 53 lines
rteval/modules/loads/__init__.py +1
-0+ rep_n.newProp("loadcpus", cpulist) return rep_nrteval/modules/measurement/__init__.py +7
-1@@ -24,7 +24,7 @@ import libxml2 from rteval.modules import RtEvalModules, ModuleContainer - +from rteval.systopology import collapse_cpulist, SysTopology class MeasurementProfile(RtEvalModules): """Keeps and controls all the measurement modules with the same measurement profile"""@@ -189,6 +189,12 @@ measurement profiles, based on their characteristics""" # Get the reports from all meaurement modules in all measurement profiles rep_n = libxml2.newNode("Measurements") + cpulist = self.__cfg.GetSection("measurement").cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("measurecpus", cpulist) + for mp in self.__measureprofiles: mprep_n = mp.MakeReport() if mprep_n:rteval/rteval_text.xsl +8
-0@@ -13,6 +13,14 @@ <xsl:value-of select="run_info/date"/><xsl:text> </xsl:text><xsl:value-of select="run_info/time"/> <xsl:text> </xsl:text> + <xsl:text> Loads: </xsl:text> + <xsl:value-of select="loads/@loads"/><xsl:text> loads run on cores </xsl:text><xsl:value-of select="loads/@loadcpus"/> + <xsl:text> </xsl:text> + + <xsl:text> Measurement: </xsl:text> + <xsl:text>measurement threads run on cores </xsl:text><xsl:value-of select="Measurements/@measurecpus"/> + <xsl:text> </xsl:text> + <xsl:text> Run time: </xsl:text> <xsl:value-of select="run_info/@days"/><xsl:text> days </xsl:text> <xsl:value-of select="run_info/@hours"/><xsl:text>h </xsl:text>Signature · 1 line
2.31.1
Something in the xsl code is not working, I'm not picking up any values in the reports for "Measurements/@measurecpus" for example.
Thanks
Quoted 82 lines · Leah Leshchinsky
The run report produced at the end of a run does not contain information on load and measurement thread locations.
Adjust
MakeReport()functions of LoadModules and MeasurementModules class so that new properties with number of loads and cpu information are added to the XML report and can be read by rteval_text.xsl.Signed-off-by Leah Leshchinsky <[email protected]>rteval/modules/loads/__init__.py +7
-0@@ -30,6 +30,7 @@ import libxml2 from rteval.Log import Log from rteval.rtevalConfig import rtevalCfgSection from rteval.modules import RtEvalModules, rtevalModulePrototype +from rteval.systopology import collapse_cpulist, SysTopology class LoadThread(rtevalModulePrototype): def __init__(self, name, config, logger=None):@@ -131,6 +132,12 @@ class LoadModules(RtEvalModules): def MakeReport(self): rep_n = RtEvalModules.MakeReport(self) rep_n.newProp("load_average", str(self.GetLoadAvg())) + rep_n.newProp("loads", str(self.ModulesLoaded())) + cpulist = self._cfg.GetSection(self._module_config).cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("loadcpus", cpulist) return rep_nrteval/modules/measurement/__init__.py +7
-1@@ -24,7 +24,7 @@ import libxml2 from rteval.modules import RtEvalModules, ModuleContainer - +from rteval.systopology import collapse_cpulist, SysTopology class MeasurementProfile(RtEvalModules): """Keeps and controls all the measurement modules with the same measurement profile"""@@ -189,6 +189,12 @@ measurement profiles, based on their characteristics""" # Get the reports from all meaurement modules in all measurement profiles rep_n = libxml2.newNode("Measurements") + cpulist = self.__cfg.GetSection("measurement").cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("measurecpus", cpulist) + for mp in self.__measureprofiles: mprep_n = mp.MakeReport() if mprep_n:rteval/rteval_text.xsl +8
-0@@ -13,6 +13,14 @@ <xsl:value-of select="run_info/date"/><xsl:text> </xsl:text><xsl:value-of select="run_info/time"/> <xsl:text> </xsl:text> + <xsl:text> Loads: </xsl:text> + <xsl:value-of select="loads/@loads"/><xsl:text> loads run on cores </xsl:text><xsl:value-of select="loads/@loadcpus"/> + <xsl:text> </xsl:text> + + <xsl:text> Measurement: </xsl:text> + <xsl:text>measurement threads run on cores </xsl:text><xsl:value-of select="Measurements/@measurecpus"/> + <xsl:text> </xsl:text> + <xsl:text> Run time: </xsl:text> <xsl:value-of select="run_info/@days"/><xsl:text> days </xsl:text> <xsl:value-of select="run_info/@hours"/><xsl:text>h </xsl:text>Signature · 1 line
2.31.1
Hi John,
I do believe the cpulist variable can be empty. I was able to test this by running rteval without specifying and load and measurement cpus.
This patch is attempting to replicate the behavior found in rteval/__init.py lines 187-201, which also finds the cpulist using
cpulist = self._loadmods._cfg.GetSection("loads").cpulist
and if the cpulist is empty, defaults to reporting the online cpus. I assume this is because the rteval program uses all available cpus to run measurement and load threads if no specific cpus are specified.
Please let me know if I am misunderstanding any of the above.
I will investigate why the xsl code is not working for you. It seems to be working for me at the moment.
Thank you for the review,
Leah
Quoted 86 lines · Leah Leshchinsky
Quoted 82 lines · Leah Leshchinsky
The run report produced at the end of a run does not contain information on load and measurement thread locations.
Adjust
MakeReport()functions of LoadModules and MeasurementModules class so that new properties with number of loads and cpu information are added to the XML report and can be read by rteval_text.xsl.Signed-off-by Leah Leshchinsky <[email protected]>rteval/modules/loads/__init__.py +7
-0@@ -30,6 +30,7 @@ import libxml2 from rteval.Log import Log from rteval.rtevalConfig import rtevalCfgSection from rteval.modules import RtEvalModules, rtevalModulePrototype +from rteval.systopology import collapse_cpulist, SysTopology class LoadThread(rtevalModulePrototype): def __init__(self, name, config, logger=None):@@ -131,6 +132,12 @@ class LoadModules(RtEvalModules): def MakeReport(self): rep_n = RtEvalModules.MakeReport(self) rep_n.newProp("load_average", str(self.GetLoadAvg())) + rep_n.newProp("loads", str(self.ModulesLoaded())) + cpulist = self._cfg.GetSection(self._module_config).cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("loadcpus", cpulist) return rep_nrteval/modules/measurement/__init__.py +7
-1@@ -24,7 +24,7 @@ import libxml2 from rteval.modules import RtEvalModules, ModuleContainer - +from rteval.systopology import collapse_cpulist, SysTopology class MeasurementProfile(RtEvalModules): """Keeps and controls all the measurement modules with the same measurement profile"""@@ -189,6 +189,12 @@ measurement profiles, based on their characteristics""" # Get the reports from all meaurement modules in all measurement profiles rep_n = libxml2.newNode("Measurements") + cpulist = self.__cfg.GetSection("measurement").cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("measurecpus", cpulist) + for mp in self.__measureprofiles: mprep_n = mp.MakeReport() if mprep_n:rteval/rteval_text.xsl +8
-0@@ -13,6 +13,14 @@ <xsl:value-of select="run_info/date"/><xsl:text> </xsl:text><xsl:value-of select="run_info/time"/> <xsl:text> </xsl:text> + <xsl:text> Loads: </xsl:text> + <xsl:value-of select="loads/@loads"/><xsl:text> loads run on cores </xsl:text><xsl:value-of select="loads/@loadcpus"/> + <xsl:text> </xsl:text> + + <xsl:text> Measurement: </xsl:text> + <xsl:text>measurement threads run on cores </xsl:text><xsl:value-of select="Measurements/@measurecpus"/> + <xsl:text> </xsl:text> + <xsl:text> Run time: </xsl:text> <xsl:value-of select="run_info/@days"/><xsl:text> days </xsl:text> <xsl:value-of select="run_info/@hours"/><xsl:text>h </xsl:text>Signature · 1 line
2.31.1Hi John,
Quoted 1 line
I do believe the cpulist variable can be empty. I was able to test this by running rteval without specifying and load and measurement cpus.
Ok, you are correct, cpulist can be empty if the following are not set --loads-cpulist=LIST
--measurement-cpulist=LIST
So you do need some code to initialize the list of it is empty. But there are still problems
For example if you shut off cpu 3
su -c 'echo 0 > /sys/devices/system/cpu/cpu3/online '
And then request measurement on cpus 0-5
su -c './rteval-cmd -d5s --measurement-cpulist=0-5'
The report should say
Measurement: measurement threads run on cores 0-2,4,5 Loads: 3 loads run on cores 6-11 (the inverse) but instead with your patch it says Loads: 3 loads run on cores 1,2,4,6,7,8,9,10,11 Measurement: measurement threads run on cores 0-5
Now there are problems that exist even before your patch, so for example, the initial info to the screen not included in the report doesn't consider offline cpus. (We should fix that too of course, but we can do this stepwise)
And it's also possible that I have introduced some new problems with the changes to systopology that will hopefully make things simpler in the long run but may be interferring with your code. So, there are many things to untangle here.
Quoted 6 lines
This patch is attempting to replicate the behavior found in
rteval/__init.pylines 187-201, which also finds the cpulist using
cpulist = self._loadmods._cfg.GetSection("loads").cpulist
and if the cpulist is empty, defaults to reporting the online cpus. I assume this is because the rteval program uses all available cpus to run measurement and load threads if no specific cpus are specified.
Quoted 10 lines
Please let me know if I am misunderstanding any of the above.
I will investigate why the xsl code is not working for you. It seems to be working for me at the moment.
Thank you for the review,
Leah
Thanks Leah
John
Quoted 134 lines · John Kacur
Quoted 86 lines · Leah Leshchinsky
Quoted 82 lines · Leah Leshchinsky
The run report produced at the end of a run does not contain information on load and measurement thread locations.
Adjust
MakeReport()functions of LoadModules and MeasurementModules class so that new properties with number of loads and cpu information are added to the XML report and can be read by rteval_text.xsl.Signed-off-by Leah Leshchinsky <[email protected]>rteval/modules/loads/__init__.py +7
-0@@ -30,6 +30,7 @@ import libxml2 from rteval.Log import Log from rteval.rtevalConfig import rtevalCfgSection from rteval.modules import RtEvalModules, rtevalModulePrototype +from rteval.systopology import collapse_cpulist, SysTopology class LoadThread(rtevalModulePrototype): def __init__(self, name, config, logger=None):@@ -131,6 +132,12 @@ class LoadModules(RtEvalModules): def MakeReport(self): rep_n = RtEvalModules.MakeReport(self) rep_n.newProp("load_average", str(self.GetLoadAvg())) + rep_n.newProp("loads", str(self.ModulesLoaded())) + cpulist = self._cfg.GetSection(self._module_config).cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("loadcpus", cpulist) return rep_nrteval/modules/measurement/__init__.py +7
-1@@ -24,7 +24,7 @@ import libxml2 from rteval.modules import RtEvalModules, ModuleContainer - +from rteval.systopology import collapse_cpulist, SysTopology class MeasurementProfile(RtEvalModules): """Keeps and controls all the measurement modules with the same measurement profile"""@@ -189,6 +189,12 @@ measurement profiles, based on their characteristics""" # Get the reports from all meaurement modules in all measurement profiles rep_n = libxml2.newNode("Measurements") + cpulist = self.__cfg.GetSection("measurement").cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("measurecpus", cpulist) + for mp in self.__measureprofiles: mprep_n = mp.MakeReport() if mprep_n:rteval/rteval_text.xsl +8
-0@@ -13,6 +13,14 @@ <xsl:value-of select="run_info/date"/><xsl:text> </xsl:text><xsl:value-of select="run_info/time"/> <xsl:text> </xsl:text> + <xsl:text> Loads: </xsl:text> + <xsl:value-of select="loads/@loads"/><xsl:text> loads run on cores </xsl:text><xsl:value-of select="loads/@loadcpus"/> + <xsl:text> </xsl:text> + + <xsl:text> Measurement: </xsl:text> + <xsl:text>measurement threads run on cores </xsl:text><xsl:value-of select="Measurements/@measurecpus"/> + <xsl:text> </xsl:text> + <xsl:text> Run time: </xsl:text> <xsl:value-of select="run_info/@days"/><xsl:text> days </xsl:text> <xsl:value-of select="run_info/@hours"/><xsl:text>h </xsl:text>Signature · 1 line
2.31.1Hi John,
Quoted 1 line
I do believe the cpulist variable can be empty. I was able to test this by running rteval without specifying and load and measurement cpus.
Ok, you are correct, cpulist can be empty if the following are not set --loads-cpulist=LIST
--measurement-cpulist=LISTSo you do need some code to initialize the list of it is empty. But there are still problems
For example if you shut off cpu 3
su -c 'echo 0 >/sys/devices/system/cpu/cpu3/online'And then request measurement on cpus 0-5
su -c './rteval-cmd -d5s --measurement-cpulist=0-5'The report should say
Measurement: measurement threads run on cores 0-2,4,5 Loads: 3 loads run on cores 6-11 (the inverse) but instead with your patch it says Loads: 3 loads run on cores 1,2,4,6,7,8,9,10,11 Measurement: measurement threads run on cores 0-5Now there are problems that exist even before your patch, so for example, the initial info to the screen not included in the report doesn't consider offline cpus. (We should fix that too of course, but we can do this stepwise)
And it's also possible that I have introduced some new problems with the changes to systopology that will hopefully make things simpler in the long run but may be interferring with your code. So, there are many things to untangle here.
Quoted 12 lines
This patch is attempting to replicate the behavior found in
rteval/__init.pylines 187-201, which also finds the cpulist using
cpulist = self._loadmods._cfg.GetSection("loads").cpulist
and if the cpulist is empty, defaults to reporting the online cpus. I assume this is because the rteval program uses all available cpus to run measurement and load threads if no specific cpus are specified.Please let me know if I am misunderstanding any of the above.
I will investigate why the xsl code is not working for you. It seems to be working for me at the moment.
Quoted 8 lines
Quoted 4 lines
Thank you for the review,
LeahThanks Leah
John
Hi John,
Quoted 6 lines
The report should say
Measurement: measurement threads run on cores 0-2,4,5 Loads: 3 loads run on cores 6-11 (the inverse) but instead with your patch it says Loads: 3 loads run on cores 1,2,4,6,7,8,9,10,11 Measurement: measurement threads run on cores 0-5
Now that I'm looking at the rteval program prior to my changes, I see that this issue existed in the initial pre-report posted to stdout. This seems to be a separate bug in the way that 'self._loadmods._cfg.GetSection("loads").cpulist' and self.__cfg.GetSection("measurement").cpulist are calculated.
When I test your example on the rteval main branch, this is the result:
su -c 'echo 0 > /sys/devices/system/cpu/cpu3/online ' su -c './rteval-cmd -d5s --measurement-cpulist=0-5'
started 3 loads on cores 1,2,4,6,7
started measurement threads on cores 0-5
Should we file this as a separate bug, since the scope of it goes beyond my current patch?
Thanks,
Leah
Quoted 161 lines · Leah Leshchinsky
Quoted 134 lines · John Kacur
Quoted 86 lines · Leah Leshchinsky
Quoted 82 lines · Leah Leshchinsky
The run report produced at the end of a run does not contain information on load and measurement thread locations.
Adjust
MakeReport()functions of LoadModules and MeasurementModules class so that new properties with number of loads and cpu information are added to the XML report and can be read by rteval_text.xsl.Signed-off-by Leah Leshchinsky <[email protected]>rteval/modules/loads/__init__.py +7
-0@@ -30,6 +30,7 @@ import libxml2 from rteval.Log import Log from rteval.rtevalConfig import rtevalCfgSection from rteval.modules import RtEvalModules, rtevalModulePrototype +from rteval.systopology import collapse_cpulist, SysTopology class LoadThread(rtevalModulePrototype): def __init__(self, name, config, logger=None):@@ -131,6 +132,12 @@ class LoadModules(RtEvalModules): def MakeReport(self): rep_n = RtEvalModules.MakeReport(self) rep_n.newProp("load_average", str(self.GetLoadAvg())) + rep_n.newProp("loads", str(self.ModulesLoaded())) + cpulist = self._cfg.GetSection(self._module_config).cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("loadcpus", cpulist) return rep_nrteval/modules/measurement/__init__.py +7
-1@@ -24,7 +24,7 @@ import libxml2 from rteval.modules import RtEvalModules, ModuleContainer - +from rteval.systopology import collapse_cpulist, SysTopology class MeasurementProfile(RtEvalModules): """Keeps and controls all the measurement modules with the same measurement profile"""@@ -189,6 +189,12 @@ measurement profiles, based on their characteristics""" # Get the reports from all meaurement modules in all measurement profiles rep_n = libxml2.newNode("Measurements") + cpulist = self.__cfg.GetSection("measurement").cpulist + if not cpulist: + st = SysTopology() + cpulist = collapse_cpulist(st.online_cpus()) + rep_n.newProp("measurecpus", cpulist) + for mp in self.__measureprofiles: mprep_n = mp.MakeReport() if mprep_n:rteval/rteval_text.xsl +8
-0@@ -13,6 +13,14 @@ <xsl:value-of select="run_info/date"/><xsl:text> </xsl:text><xsl:value-of select="run_info/time"/> <xsl:text> </xsl:text> + <xsl:text> Loads: </xsl:text> + <xsl:value-of select="loads/@loads"/><xsl:text> loads run on cores </xsl:text><xsl:value-of select="loads/@loadcpus"/> + <xsl:text> </xsl:text> + + <xsl:text> Measurement: </xsl:text> + <xsl:text>measurement threads run on cores </xsl:text><xsl:value-of select="Measurements/@measurecpus"/> + <xsl:text> </xsl:text> + <xsl:text> Run time: </xsl:text> <xsl:value-of select="run_info/@days"/><xsl:text> days </xsl:text> <xsl:value-of select="run_info/@hours"/><xsl:text>h </xsl:text>Signature · 1 line
2.31.1Hi John,
Quoted 1 line
I do believe the cpulist variable can be empty. I was able to test this by running rteval without specifying and load and measurement cpus.
Ok, you are correct, cpulist can be empty if the following are not set --loads-cpulist=LIST
--measurement-cpulist=LISTSo you do need some code to initialize the list of it is empty. But there are still problems
For example if you shut off cpu 3
su -c 'echo 0 >/sys/devices/system/cpu/cpu3/online'And then request measurement on cpus 0-5
su -c './rteval-cmd -d5s --measurement-cpulist=0-5'The report should say
Measurement: measurement threads run on cores 0-2,4,5 Loads: 3 loads run on cores 6-11 (the inverse) but instead with your patch it says Loads: 3 loads run on cores 1,2,4,6,7,8,9,10,11 Measurement: measurement threads run on cores 0-5Now there are problems that exist even before your patch, so for example, the initial info to the screen not included in the report doesn't consider offline cpus. (We should fix that too of course, but we can do this stepwise)
And it's also possible that I have introduced some new problems with the changes to systopology that will hopefully make things simpler in the long run but may be interferring with your code. So, there are many things to untangle here.
Quoted 12 lines
This patch is attempting to replicate the behavior found in
rteval/__init.pylines 187-201, which also finds the cpulist using
cpulist = self._loadmods._cfg.GetSection("loads").cpulist
and if the cpulist is empty, defaults to reporting the online cpus. I assume this is because the rteval program uses all available cpus to run measurement and load threads if no specific cpus are specified.Please let me know if I am misunderstanding any of the above.
I will investigate why the xsl code is not working for you. It seems to be working for me at the moment.
Quoted 8 lines
Quoted 4 lines
Thank you for the review,
LeahThanks Leah
John
Hi John,
Quoted 6 lines
The report should say
Measurement: measurement threads run on cores 0-2,4,5 Loads: 3 loads run on cores 6-11 (the inverse) but instead with your patch it says Loads: 3 loads run on cores 1,2,4,6,7,8,9,10,11 Measurement: measurement threads run on cores 0-5Now that I'm looking at the rteval program prior to my changes, I see that this issue existed in the initial pre-report posted to stdout. This seems to be a separate bug in the way that 'self._loadmods._cfg.GetSection("loads").cpulist' and self.__cfg.GetSection("measurement").cpulist are calculated.
When I test your example on the rteval main branch, this is the result:
su -c 'echo 0 >
/sys/devices/system/cpu/cpu3/online' su -c './rteval-cmd -d5s --measurement-cpulist=0-5'
Quoted 8 lines
started 3 loads on cores 1,2,4,6,7
started measurement threads on cores 0-5Should we file this as a separate bug, since the scope of it goes beyond my current patch?
Thanks,
Leah
Please send a separate patch to fix this.
John