Re: [Flexradio] FlexSDR software and Multiple CPU's
am running an hp graphics workstation with 2 pent. 3, each at 1gh with 1.5g ram and scusi hard drive with xp-pro with the radio as part of the software. need the xp-pro to pickup the two processors. works for me. tom [EMAIL PROTECTED]
Re: [Flexradio] FlexSDR software and Multiple CPU's
Cecil, I am running an AMD Athlon dual core 4400+ on Win2003. With PowerSDR, logging software, MIXw, VAC, Vcom, Firefox and a Citrix session for e-mail, my total processor utilization is about 10-15% max. About 8% on the average. All of that is on the first processor. Less than 1% is running on the second processor. From what I can tell, PowerSDR is not specifically written to take advantage of dual processor machines. When I look at the processor utilization, ~90% of the total CPU utilization in on the first processor. As you continue to add application load, it will mostly consume the first processor until the load gets to about 50% (assuming no single process exceeds 50%) and then new processes are run on the other processor. This is mostly a function of the operating system and not the application. I would suspect that the multi-processing logic is a little more robust on Win2003 than XP. In my testing, most things being equal, two 1 gig processors will not outperform a 2 gig processor. Multi-processor aware applications specifically choose how to distribute their process between the processors. But, with the current incarnation of PowerSDR not really pushing the processor on very fast machines, for now it is a moot point. Although I'd like to see how a multi-processor aware version of PowerSDR would behave and how specific application processes would be divided among the processors. Having the graphics run on one and the FFT and signal processing operations run on the other would be interesting. -Tim From: [EMAIL PROTECTED] on behalf of KD5NWA Sent: Thu 11/24/2005 11:51 PM To: FlexRadio@flex-radio.biz Subject: [Flexradio] FlexSDR software and Multiple CPU's Will the existing Flex software take advantage of a PC that has multiple CPU's? The dual CPU unit is the fastest PC while running the FlexSDR software, it usually runs between 10% to 15% utilization. I'm debating on which box to install my Delta-44 card, I have two PC's that are candidates one uses a 1700MHz Athalon in a industrial all metal case, a unit that is easy to work with and has lots of open slots, the other box is a Dell with 2X 1000MHz Pentium III's. Both are going to end up on my workbench next to my antenna's and radios. I would use one with the Delta-44 for use with Software Define Radios, the other is to be used as a programming platform and interfacing to ongoing radio projects. Thanks Cecil Bayona KD5NWA www.qrpradio.com I fail to see why doing the same thing over and over and getting the same results every time is insanity: I've almost proved it isn't; only a few more tests now and I'm sure results will differ this time ... ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
Re: [Flexradio] FlexSDR software and Multiple CPU's
At 08:29 AM 11/25/2005, Tim Ellison wrote: Cecil, I am running an AMD Athlon dual core 4400+ on Win2003. With PowerSDR, logging software, MIXw, VAC, Vcom, Firefox and a Citrix session for e-mail, my total processor utilization is about 10-15% max. About 8% on the average. All of that is on the first processor. Less than 1% is running on the second processor. From what I can tell, PowerSDR is not specifically written to take advantage of dual processor machines. When I look at the processor utilization, ~90% of the total CPU utilization in on the first processor. As you continue to add application load, it will mostly consume the first processor until the load gets to about 50% (assuming no single process exceeds 50%) and then new processes are run on the other processor. This is mostly a function of the operating system and not the application. I would suspect that the multi-processing logic is a little more robust on Win2003 than XP. In my testing, most things being equal, two 1 gig processors will not outperform a 2 gig processor. If for no other reason than both processors are sharing one memory access pipe.. Unless the code in one processor is entirely reading from cache, and cache isn't shared, and the other processor isn't writing to memory space covered by the cache. (all these are real hard to control) Multi-processor aware applications specifically choose how to distribute their process between the processors. But, with the current incarnation of PowerSDR not really pushing the processor on very fast machines, for now it is a moot point. Although I'd like to see how a multi-processor aware version of PowerSDR would behave and how specific application processes would be divided among the processors. Having the graphics run on one and the FFT and signal processing operations run on the other would be interesting. Indeed, this would be interesting. Even more interesting would be if the various components were separated enough, with a clean enough interface, that you could run the UI on one computer and the signal processing chunk(s) on other computers. Raw UDP sockets would probably serve, or, one could use something like MPI or PVM to pull it to a higher abstraction level.
Re: [Flexradio] FlexSDR software and Multiple CPU's
What does it take for an application to be called multi-processor aware? A typical instance of PowerSDR Console has about 30 threads running. Of course, most of them don't do very much. They are more or less system overhead (.NET and various other subsystems running a lot of threads). However, PowerSDR is a true multithreaded application. For example, DSP and graphics do run on separate threads. Exactly how their execution is distributed between different processors is usually managed by the operating system. It would also be possible for the application to have control over this (using a parameter called process or thread affinity). But in most cases it's actually better to leave that to the OS. Yes, there are some difficult optimization issues if you really have to get full 100% from all your CPU's all the time. But in less extreme cases, I think multithreaded applications like PowerSDR should perform pretty well also on multi-CPU systems. 73, Sami OH2BFO On 11/25/05, Jim Lux [EMAIL PROTECTED] wrote: At 08:29 AM 11/25/2005, Tim Ellison wrote: Cecil, I am running an AMD Athlon dual core 4400+ on Win2003. With PowerSDR, logging software, MIXw, VAC, Vcom, Firefox and a Citrix session for e-mail, my total processor utilization is about 10-15% max. About 8% on the average. All of that is on the first processor. Less than 1% is running on the second processor. From what I can tell, PowerSDR is not specifically written to take advantage of dual processor machines. When I look at the processor utilization, ~90% of the total CPU utilization in on the first processor. As you continue to add application load, it will mostly consume the first processor until the load gets to about 50% (assuming no single process exceeds 50%) and then new processes are run on the other processor. This is mostly a function of the operating system and not the application. I would suspect that the multi-processing logic is a little more robust on Win2003 than XP. In my testing, most things being equal, two 1 gig processors will not outperform a 2 gig processor. If for no other reason than both processors are sharing one memory access pipe.. Unless the code in one processor is entirely reading from cache, and cache isn't shared, and the other processor isn't writing to memory space covered by the cache. (all these are real hard to control) Multi-processor aware applications specifically choose how to distribute their process between the processors. But, with the current incarnation of PowerSDR not really pushing the processor on very fast machines, for now it is a moot point. Although I'd like to see how a multi-processor aware version of PowerSDR would behave and how specific application processes would be divided among the processors. Having the graphics run on one and the FFT and signal processing operations run on the other would be interesting. Indeed, this would be interesting. Even more interesting would be if the various components were separated enough, with a clean enough interface, that you could run the UI on one computer and the signal processing chunk(s) on other computers. Raw UDP sockets would probably serve, or, one could use something like MPI or PVM to pull it to a higher abstraction level.
Re: [Flexradio] FlexSDR software and Multiple CPU's
I agree. I have an application that has many threads. Several of these do dsp stuff and several others are reponsible for formatting and recording the data. Under XP Pro, the difference between a single processor and dual is very close to 2:1. In fact, just hyperthreading alone can yield as much as 1.5:1. This is letting the OS decide who to put where. When I initially saw the results I abandoned plans to split my data crunching into odd channels in one thread and even in another. Richard W5SXD - --- Original Message --- - From: [EMAIL PROTECTED] To: FlexRadio@flex-radio.biz Sent: Fri, 25 Nov 2005 20:41:06 What does it take for an application to be called multi-processor aware? A typical instance of PowerSDR Console has about 30 threads running. Of course, most of them don't do very much. They are more or less system overhead (.NET and various other subsystems running a lot of threads). However, PowerSDR is a true multithreaded application. For example, DSP and graphics do run on separate threads. Exactly how their execution is distributed between different processors is usually managed by the operating system. It would also be possible for the application to have control over this (using a parameter called process or thread affinity). But in most cases it's actually better to leave that to the OS. Yes, there are some difficult optimization issues if you really have to get full 100% from all your CPU's all the time. But in less extreme cases, I think multithreaded applications like PowerSDR should perform pretty well also on multi-CPU systems. 73, Sami OH2BFO On 11/25/05, Jim Lux [EMAIL PROTECTED] wrote: At 08:29 AM 11/25/2005, Tim Ellison wrote: Cecil, I am running an AMD Athlon dual core 4400+ on Win2003. With PowerSDR, logging software, MIXw, VAC, Vcom, Firefox and a Citrix session for e-mail, my total processor utilization is about 10-15% max. About 8% on the average. All of that is on the first processor. Less than 1% is running on the second processor. From what I can tell, PowerSDR is not specifically written to take advantage of dual processor machines. When I look at the processor utilization, ~90% of the total CPU utilization in on the first processor. As you continue to add application load, it will mostly consume the first processor until the load gets to about 50% (assuming no single process exceeds 50%) and then new processes are run on the other processor. This is mostly a function of the operating system and not the application. I would suspect that the multi-processing logic is a little more robust on Win2003 than XP. In my testing, most things being equal, two 1 gig processors will not outperform a 2 gig processor. If for no other reason than both processors are sharing one memory access pipe.. Unless the code in one processor is entirely reading from cache, and cache isn't shared, and the other processor isn't writing to memory space covered by the cache. (all these are real hard to control) Multi-processor aware applications specifically choose how to distribute their process between the processors. But, with the current incarnation of PowerSDR not really pushing the processor on very fast machines, for now it is a moot point. Although I'd like to see how a multi-processor aware version of PowerSDR would behave and how specific application processes would be divided among the processors. Having the graphics run on one and the FFT and signal processing operations run on the other would be interesting. Indeed, this would be interesting. Even more interesting would be if the various components were separated enough, with a clean enough interface, that you could run the UI on one computer and the signal processing chunk(s) on other computers. Raw UDP sockets would probably serve, or, one could use something like MPI or PVM to pull it to a higher abstraction level. ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexrad io_flex-radio.biz
[Flexradio] FlexSDR software and Multiple CPU's
Will the existing Flex software take advantage of a PC that has multiple CPU's? The dual CPU unit is the fastest PC while running the FlexSDR software, it usually runs between 10% to 15% utilization. I'm debating on which box to install my Delta-44 card, I have two PC's that are candidates one uses a 1700MHz Athalon in a industrial all metal case, a unit that is easy to work with and has lots of open slots, the other box is a Dell with 2X 1000MHz Pentium III's. Both are going to end up on my workbench next to my antenna's and radios. I would use one with the Delta-44 for use with Software Define Radios, the other is to be used as a programming platform and interfacing to ongoing radio projects. Thanks Cecil Bayona KD5NWA www.qrpradio.com I fail to see why doing the same thing over and over and getting the same results every time is insanity: I've almost proved it isn't; only a few more tests now and I'm sure results will differ this time ...