DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] Queries related to Hyperthreading
@ 2014-10-07  3:29 Ravi Kumar Iyer
  2014-10-07  4:17 ` Ramia, Kannan Babu
  0 siblings, 1 reply; 4+ messages in thread
From: Ravi Kumar Iyer @ 2014-10-07  3:29 UTC (permalink / raw)
  To: dev

hello Intel Gurus,

I have some queries related to intel HT technology. I am running an intel dpdk application and  trying to benchmark as well as optimize the same. As of now it is a single threaded application and would like to keep it like that for now. I am running it on an intel sandy bridge arch processor with HT on. Now to increase the performance I have disabled HT, but am still getting the same performance[ just a marginal improvement]. Now as per our understanding so far wrt HT, two Hardware threads share the same physical core. Now if we switch off HT[from bios] shouldn't just one hardware Thread get hold of the complete CPU core.  The application is heavily compute oriented and as such minimal memory access or any IO operations. Appreciate any insight on this.


thanks
ravi


"DISCLAIMER: This message is proprietary to Aricent and is intended solely for the use of the individual to whom it is addressed. It may contain privileged or confidential information and should not be circulated or used for any purpose other than for what it is intended. If you have received this message in error, please notify the originator immediately. If you are not the intended recipient, you are notified that you are strictly prohibited from using, copying, altering, or disclosing the contents of this message. Aricent accepts no responsibility for loss or damage arising from the use of the information transmitted by this email including damage from virus."

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [dpdk-dev] Queries related to Hyperthreading
  2014-10-07  3:29 [dpdk-dev] Queries related to Hyperthreading Ravi Kumar Iyer
@ 2014-10-07  4:17 ` Ramia, Kannan Babu
  2014-10-07  6:05   ` Ravi Kumar Iyer
  0 siblings, 1 reply; 4+ messages in thread
From: Ramia, Kannan Babu @ 2014-10-07  4:17 UTC (permalink / raw)
  To: Ravi Kumar Iyer, dev

Hi Ravi

The Intel HT actually has its own resources except few places like execution units and other resources in the pipeline, hence there are many factors play here for the performance of the application with and without HT feature. In simple words, even when you turned on the HT, and use only one of the HT, then the performance will be more or less be similar to the core with HT turned off. Also assume that if you are running same piece of code in both the HTs, but it doesn't have lots have contention on the common resource (you can figure out this using vTune profiling), then there will be very marginal performance difference between HT turned ON Vs OFF.

Regards
Kannan Babu

-----Original Message-----
From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ravi Kumar Iyer
Sent: Tuesday, October 7, 2014 8:59 AM
To: dev@dpdk.org
Subject: [dpdk-dev] Queries related to Hyperthreading

hello Intel Gurus,

I have some queries related to intel HT technology. I am running an intel dpdk application and  trying to benchmark as well as optimize the same. As of now it is a single threaded application and would like to keep it like that for now. I am running it on an intel sandy bridge arch processor with HT on. Now to increase the performance I have disabled HT, but am still getting the same performance[ just a marginal improvement]. Now as per our understanding so far wrt HT, two Hardware threads share the same physical core. Now if we switch off HT[from bios] shouldn't just one hardware Thread get hold of the complete CPU core.  The application is heavily compute oriented and as such minimal memory access or any IO operations. Appreciate any insight on this.


thanks
ravi


"DISCLAIMER: This message is proprietary to Aricent and is intended solely for the use of the individual to whom it is addressed. It may contain privileged or confidential information and should not be circulated or used for any purpose other than for what it is intended. If you have received this message in error, please notify the originator immediately. If you are not the intended recipient, you are notified that you are strictly prohibited from using, copying, altering, or disclosing the contents of this message. Aricent accepts no responsibility for loss or damage arising from the use of the information transmitted by this email including damage from virus."

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [dpdk-dev] Queries related to Hyperthreading
  2014-10-07  4:17 ` Ramia, Kannan Babu
@ 2014-10-07  6:05   ` Ravi Kumar Iyer
  2014-10-08  6:22     ` Ramia, Kannan Babu
  0 siblings, 1 reply; 4+ messages in thread
From: Ravi Kumar Iyer @ 2014-10-07  6:05 UTC (permalink / raw)
  To: Ramia, Kannan Babu, dev

Thanks Kannan,

Do we have any links/docs where in we could read more about the behavior of Hyperthreads.
We are executing a code which is similar to the following example:


So let us we have very tight loop of code which does basic arithmetic operations[addition, subtraction]. Now the cost to complete these operations is say C clock cycles on a single core[ with HT on]. Now if we run the same code with HT off, shouldn't the instruction pipeline of this code get executed faster [ theoretical at double speed].

Thanks
ravi




-----Original Message-----
From: Ramia, Kannan Babu [mailto:kannan.babu.ramia@intel.com]
Sent: Tuesday, October 07, 2014 9:48 AM
To: Ravi Kumar Iyer; dev@dpdk.org
Subject: RE: Queries related to Hyperthreading

Hi Ravi

The Intel HT actually has its own resources except few places like execution units and other resources in the pipeline, hence there are many factors play here for the performance of the application with and without HT feature. In simple words, even when you turned on the HT, and use only one of the HT, then the performance will be more or less be similar to the core with HT turned off. Also assume that if you are running same piece of code in both the HTs, but it doesn't have lots have contention on the common resource (you can figure out this using vTune profiling), then there will be very marginal performance difference between HT turned ON Vs OFF.

Regards
Kannan Babu

-----Original Message-----
From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ravi Kumar Iyer
Sent: Tuesday, October 7, 2014 8:59 AM
To: dev@dpdk.org
Subject: [dpdk-dev] Queries related to Hyperthreading

hello Intel Gurus,

I have some queries related to intel HT technology. I am running an intel dpdk application and  trying to benchmark as well as optimize the same. As of now it is a single threaded application and would like to keep it like that for now. I am running it on an intel sandy bridge arch processor with HT on. Now to increase the performance I have disabled HT, but am still getting the same performance[ just a marginal improvement]. Now as per our understanding so far wrt HT, two Hardware threads share the same physical core. Now if we switch off HT[from bios] shouldn't just one hardware Thread get hold of the complete CPU core.  The application is heavily compute oriented and as such minimal memory access or any IO operations. Appreciate any insight on this.


thanks
ravi


"DISCLAIMER: This message is proprietary to Aricent and is intended solely for the use of the individual to whom it is addressed. It may contain privileged or confidential information and should not be circulated or used for any purpose other than for what it is intended. If you have received this message in error, please notify the originator immediately. If you are not the intended recipient, you are notified that you are strictly prohibited from using, copying, altering, or disclosing the contents of this message. Aricent accepts no responsibility for loss or damage arising from the use of the information transmitted by this email including damage from virus."


"DISCLAIMER: This message is proprietary to Aricent and is intended solely for the use of the individual to whom it is addressed. It may contain privileged or confidential information and should not be circulated or used for any purpose other than for what it is intended. If you have received this message in error, please notify the originator immediately. If you are not the intended recipient, you are notified that you are strictly prohibited from using, copying, altering, or disclosing the contents of this message. Aricent accepts no responsibility for loss or damage arising from the use of the information transmitted by this email including damage from virus."

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [dpdk-dev] Queries related to Hyperthreading
  2014-10-07  6:05   ` Ravi Kumar Iyer
@ 2014-10-08  6:22     ` Ramia, Kannan Babu
  0 siblings, 0 replies; 4+ messages in thread
From: Ramia, Kannan Babu @ 2014-10-08  6:22 UTC (permalink / raw)
  To: Ravi Kumar Iyer, dev

Please read this document for more detail information.

http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html?wapkw=248966

Also look into the chapter-8 in this document.

Regards
Kannan Babu

-----Original Message-----
From: Ravi Kumar Iyer [mailto:Ravi.Iyer@aricent.com] 
Sent: Tuesday, October 7, 2014 11:35 AM
To: Ramia, Kannan Babu; dev@dpdk.org
Subject: RE: Queries related to Hyperthreading

Thanks Kannan,

Do we have any links/docs where in we could read more about the behavior of Hyperthreads.
We are executing a code which is similar to the following example:


So let us we have very tight loop of code which does basic arithmetic operations[addition, subtraction]. Now the cost to complete these operations is say C clock cycles on a single core[ with HT on]. Now if we run the same code with HT off, shouldn't the instruction pipeline of this code get executed faster [ theoretical at double speed].

Thanks
ravi




-----Original Message-----
From: Ramia, Kannan Babu [mailto:kannan.babu.ramia@intel.com]
Sent: Tuesday, October 07, 2014 9:48 AM
To: Ravi Kumar Iyer; dev@dpdk.org
Subject: RE: Queries related to Hyperthreading

Hi Ravi

The Intel HT actually has its own resources except few places like execution units and other resources in the pipeline, hence there are many factors play here for the performance of the application with and without HT feature. In simple words, even when you turned on the HT, and use only one of the HT, then the performance will be more or less be similar to the core with HT turned off. Also assume that if you are running same piece of code in both the HTs, but it doesn't have lots have contention on the common resource (you can figure out this using vTune profiling), then there will be very marginal performance difference between HT turned ON Vs OFF.

Regards
Kannan Babu

-----Original Message-----
From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ravi Kumar Iyer
Sent: Tuesday, October 7, 2014 8:59 AM
To: dev@dpdk.org
Subject: [dpdk-dev] Queries related to Hyperthreading

hello Intel Gurus,

I have some queries related to intel HT technology. I am running an intel dpdk application and  trying to benchmark as well as optimize the same. As of now it is a single threaded application and would like to keep it like that for now. I am running it on an intel sandy bridge arch processor with HT on. Now to increase the performance I have disabled HT, but am still getting the same performance[ just a marginal improvement]. Now as per our understanding so far wrt HT, two Hardware threads share the same physical core. Now if we switch off HT[from bios] shouldn't just one hardware Thread get hold of the complete CPU core.  The application is heavily compute oriented and as such minimal memory access or any IO operations. Appreciate any insight on this.


thanks
ravi


"DISCLAIMER: This message is proprietary to Aricent and is intended solely for the use of the individual to whom it is addressed. It may contain privileged or confidential information and should not be circulated or used for any purpose other than for what it is intended. If you have received this message in error, please notify the originator immediately. If you are not the intended recipient, you are notified that you are strictly prohibited from using, copying, altering, or disclosing the contents of this message. Aricent accepts no responsibility for loss or damage arising from the use of the information transmitted by this email including damage from virus."


"DISCLAIMER: This message is proprietary to Aricent and is intended solely for the use of the individual to whom it is addressed. It may contain privileged or confidential information and should not be circulated or used for any purpose other than for what it is intended. If you have received this message in error, please notify the originator immediately. If you are not the intended recipient, you are notified that you are strictly prohibited from using, copying, altering, or disclosing the contents of this message. Aricent accepts no responsibility for loss or damage arising from the use of the information transmitted by this email including damage from virus."

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2014-10-08  6:15 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-10-07  3:29 [dpdk-dev] Queries related to Hyperthreading Ravi Kumar Iyer
2014-10-07  4:17 ` Ramia, Kannan Babu
2014-10-07  6:05   ` Ravi Kumar Iyer
2014-10-08  6:22     ` Ramia, Kannan Babu

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).