Iam working on a project wherein an FPGA send streaming data to FX3 via GPIF. We are using syncSlaveFIFO example code as it is to fetch the data and dump onto a Windows PC with Streamer app on it to validate the data. But the socket0 buffer donot get any data. We want to check the GPIF design of this interface but couldn't find the source. WHere can I find GPIF project design file for Sync SlaveFIFO example code? We are using SDK 1.2 version.

Expecting early reply...

Thanks

Sriraam

Hi Sirriram,

download the AN65974.zip file from cypress...

here you get documentation and source code...

http://www.cypress.com/?rID=51581

and here just the zip file with source code

http://www.cypress.com/?docID=39756

regards
lumpi

Thank you Lumpi6. Using the docs you specified we were able to find the throughput rate of FX3 in WIn7.

But we got only 280 MBps. We want to try in ubuntu Linux 12.10 and see if we can get more speed. So, we installed the FX3 SDK for linux.

But, we are unable to find any Streamer app for Linux there. What app should we use to find data throughput in Linux?

Hi Sriraam,

you welcome. The throughput sounds good are you using a PCIx card USB3 controller card or do you have a onboard USB3 with intel's ivy bridge? On an ivy bridge I got almost 400MB/s with the streamer app when configuring burst 14 or 16 and use 64 packets per xfer and 64 packets to queue.

You may open a new topic to get info if there is a streamer app for linux available. I have not found anything for now...

regards,

lumpi

Hi Sriram/Lumpi

You can use the same cyusb_linux tool for measuring the throughput over Bulk/ISO endpoints.

cyusb_linux tool supports all the functionalities supported by control center, bulkloop, streamer applications on windows.

For more details, please refer to the attached pdf.

Thanks,

sai krishna.

Hi Lumpi6,

We are using a Dell Vostro Win7 laptop which has 2 USB 3.0 ports. After checking the device manager, I understand that it uses Renesas PCIe 1.1 to USB 3.0 controller card inside. not a intel ivy bridge. Also, surprisingly, when we checked with the C++ Streamer app ( a standalone .exe) we are getting 320MBps in WIn7 itself. But, what is more surprising is, with theoritical speed of PCIe 1.1 as only 250 MBps, how did we get 320 MBps. And, if my understanding is wrong, and if it is PCIe 2.0  card (with theoritical speed of 500MBps), then we should be able to get atleast 380 MBps. We are confused now! If you have any idea on this, please throw some light on us.  Thanks.

Hi Sai Krishna,

We had checked with the cyusb_linux tool and found that data transfer rate can be checked only for isochronous endpoints and not for Bulk endpoints (which we are using). Please advise us further steps on what should we do.

Thanks,

Sriraam

Hi Sriraam,

both solutions with PCIe 1.1 (using two lanes) or with PCIe 2.0 you be able to transfer 500Mbps so that depend on the internal connection of the renesas chip set. If you have a USB3.0 analyser you may double check if the throughput of the streamer app is correct. In my experience and measures it was all the time almost the same values mesured by streamer app or usb3 analyser. So what you see should be correct.

The reason why it is not more than 320Mbps is, I believe, that the renesas chip has this max bandwidth of the ~300Mbps. Only I got faster transfers with onbaord ivy bridge pc all other USB3 chip were up to now way slower. I think the renesas chip is one of generation 2 or 3. May be you tell me your used driver and firmware version? The newest of renesas generation 2 and 3 driver version is actually 3.0.23.0 and the firmware version is 2018 (you can see it in the windows device manager if you open the properties page of the renesas usb. The generation 1 has a driver like 2.x.xx.x.

regards

lumpi

Hi Sriraam,

both solutions with PCIe 1.1 (using two lanes) or with PCIe 2.0 you be able to transfer 500Mbps so that depend on the internal connection of the renesas chip set. If you have a USB3.0 analyser you may double check if the throughput of the streamer app is correct. In my experience and measures it was all the time almost the same values mesured by streamer app or usb3 analyser. So what you see should be correct.

The reason why it is not more than 320Mbps is, I believe, that the renesas chip has this max bandwidth of the ~300Mbps. Only I got faster transfers with onbaord ivy bridge pc all other USB3 chip were up to now way slower. I think the renesas chip is one of generation 2 or 3. May be you tell me your used driver and firmware version? The newest of renesas generation 2 and 3 driver version is actually 3.0.23.0 and the firmware version is 2018 (you can see it in the windows device manager if you open the properties page of the renesas usb. The generation 1 has a driver like 2.x.xx.x.

regards

lumpi

Hi Lumpi,

I think your reasoning might be correct. My renesas USB3.0 chip driver version is 2.0.32.0 and firmware version is 1343. So, it seems that it is of older Gen 1 chip. Thanks for the gyaan, Lumpi.

By the way, do you use C# Streamer app (or) C++ Streamer app?

Hi Sriraam,

you're welcome..

I am using the c# streamer app of the installer from cypress CyUSBSuite version 1.2.1.0.

The actual drivers of the generation 1 chip (chip name/type 720200 and 720200A) is 2.1.39.0 and the firmware version is 3034 of 720200 and 4020 of 720200A). You may get it from.

http://www.station-drivers.com/page/renesas.htm

regards,

lumpi

Hi Lumpi,

Thanks for the  info. But, when i tried installing the latest driver you specified, my Renesas USB3 controller is not accepting it and is saying that the best driver is already installed. I think the controller may be an older chip of the Gen1 controllers.

Anyway thanks for all the help.

Rgds,

Sriraam

Hi Sai Krishna,

We had checked with the cyusb_linux tool and found that data transfer rate can be checked only for isochronous endpoints and not for Bulk endpoints (which we are using). Kindly clarify.

Also we find that the C# Streamer app is giving us lesser throughput while the C++ Streamer app is far better for the same exact setup and hardware. Why so much difference between them?

Expecting your reply for Both the questions...

Thanks,

Sriraam

Unfortunately, cyusb_linux tool supports throughput measurement only for ISO transfers.

Coming to your second question:

C# is managed code , it does not produce the Native code while build; while running the CLR will produce the native code then execute it. 

C++ Streamer is native code, so it will execute very fast. No CLR involvement.

Please refer to the following page if you need more details regarding CLR:

http://en.wikipedia.org/wiki/Common_Language_Runtime

Regards,

sai krishna.

Thanks Sai krishna for the info about C#/C++ Streamer.

Also, can we expect a cyusb_linux tool update which supports data throughput in Bulk transfers anytime soon?

Thanks,

Sriraam