|
»
Asynchronous bulk transfer with QuickUSB FX2 (linux) ?
|
View:
New views
14 Messages
—
Rating Filter:
Alert me
|
 |
Asynchronous bulk transfer with QuickUSB FX2 (linux) ?
by ChMeessen
::
Rate this Message:
Hello,
I'm a computer scientist and beginner with USB programming on linux. Our problem is as following. We have a card developped by electronicians in our lab with an FPGA and a QuickUSB controller. It is currently used with a linux computer but has a very low throughput ~3MB/s. We need at least 10MB/s and preferably 30MB/s. As stated in the QuickUSB doc, the driver provided by quickUSB for linux doesn't support asynchronous bulk transfer. Reading the following thread of this mailing list http://www.nabble.com/Write-Linux-USB-Driver-td14913265.html I saw that libusb did not support asynchronous transfer and that quick USB has no interrupt link. Unfortunately I can't change the controller now. Maybe in future card versions. I assume that the linux driver provided with the QuickUSB is just a thin layer on top of libusb and reading the latest libusb specification it is advertised to support isochronous and asynchronous transfer. This would be a great news. On windows we achieved 30MB/s throughput, but the controller didn't stop sending data. It kept sending whatever data was on the controller's data write port. So the question is if there is any code example and tutorial I could look at to learn how to get the best throughput out of this QuickUSB controller. Is this at least feasible ? I assume the linux side is kind of standard way to do things for which I should easily find some example code. One of the questions I'm asking my self is if the interrupt channel is required to get the maximum throughput ? -- Ch. Meessen ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
Re: Asynchronous bulk transfer with QuickUSB FX2 (linux) ?
by Peter Stuge-4
::
Rate this Message:
Hi Christophe,
Christophe Meessen wrote: > We have a card developped by electronicians in our lab with an FPGA and > a QuickUSB controller. That product is really overpriced! Wow. :( > It is currently used with a linux computer but has a very low > throughput ~3MB/s. We need at least 10MB/s and preferably 30MB/s. > > As stated in the QuickUSB doc, the driver provided by quickUSB for > linux doesn't support asynchronous bulk transfer. You have to dump all your technical information on the list. Then we would actually be able to provide help, or ask you for some testing. Right now, you haven't specified the problem very clearly so noone knows exactly what you need to do. > One of the questions I'm asking my self is if the interrupt channel > is required to get the maximum throughput ? No. USB interrupt pipes will not give very high throughput. If you need reliable transfers you should be using bulk. This is specified by the device descriptors, so can not be chosen by the software developer. Feedback to your department is that USB absolutely must be a crossover project with both hardware and software groups working TOGETHER - or the end result will likely be useless hardware and certainly lack of understanding among students. Time wasted. Please read the USB specification to learn more about the USB software stack/model. (The hardware guys have already done this, and you should of course have been doing that together.) USB is completely host centric. The device must follow many rules and specifications to work correctly. This means that software developers actually create some requirements for the hardware developers. This isn't really the norm, so it is important to have both groups working together on some parts. Please send lsusb -v output (the descriptors) for your device. Finally, when writing your driver using libusb, make sure that you have many transfer requests pending at all times, that is the only way you will get the throughput you want. (The request queue travels all the way down to the wire, so if you make sure there are pending requests, the wire utilization will be maximal.) There is also a Cypress FX2 community for people working with those USB chips in Linux. Note that you will quite likely have to throw away whatever software quickusb provided you with, unless they are actually using libusb as outlined above. You should look into what their software does, and maybe you can actually get some support from them - I think that should be included in the high price for their product. //Peter ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
Re: Asynchronous bulk transfer with QuickUSB FX2 (linux) ?
by Marko Cebokli
::
Rate this Message:
Hello Cristophe,
I think your problem is mainly in the system latencies. I play with the same EZ-USB cypress chip: http://lea.hamradio.si/~s57uuu/uuusb/index.htm (see also the software section linked from there) and I think we have the same problem - it is not hard to get a lot of average bandwidth (>20 MB/s), but the problem is when you do not have much device-side buffering, like only the cypress internal buffers, that you experience data loss. With libusb 0.1 I can get cca 2..10 MB/s of quasi reliable transfers (1% lost packets), depending on the machine. (that is with a "relatime" kernel and scheduling) Currently, I am experimenting with libusb 1.0. I found synchronous or asynchronous mode with a single transfer to be quite similar to libusb 0.1. Throwing out more than one transfer at a time does help. I foun the difference between one and two issued asynchronous transfers quite pronounced, but using more than two does not seem to bring very much. I hope to tidy up the code soon and put it at the above website soon... Marko Cebokli On Thursday 08 October 2009 10:15:10 Christophe Meessen wrote: > Hello, > > I'm a computer scientist and beginner with USB programming on linux. > Our problem is as following. > > We have a card developped by electronicians in our lab with an FPGA and > a QuickUSB controller. > It is currently used with a linux computer but has a very low throughput > ~3MB/s. We need at least 10MB/s and preferably 30MB/s. > > As stated in the QuickUSB doc, the driver provided by quickUSB for linux > doesn't support asynchronous bulk transfer. > > Reading the following thread of this mailing list > http://www.nabble.com/Write-Linux-USB-Driver-td14913265.html > > I saw that libusb did not support asynchronous transfer and that quick > USB has no interrupt link. Unfortunately I can't change the controller > now. Maybe in future card versions. > > I assume that the linux driver provided with the QuickUSB is just a thin > layer on top of libusb and reading the latest libusb specification it is > advertised to support isochronous and asynchronous transfer. This would > be a great news. > > On windows we achieved 30MB/s throughput, but the controller didn't stop > sending data. It kept sending whatever data was on the controller's data > write port. > > So the question is if there is any code example and tutorial I could > look at to learn how to get the best throughput out of this QuickUSB > controller. Is this at least feasible ? > > I assume the linux side is kind of standard way to do things for which I > should easily find some example code. One of the questions I'm asking my > self is if the interrupt channel is required to get the maximum > throughput ? > > -- > Ch. Meessen > > --------------------------------------------------------------------------- >--- Come build with us! The BlackBerry(R) Developer Conference in SF, CA is > the only developer event you need to attend this year. Jumpstart your > developing skills, take BlackBerry mobile applications to market and stay > ahead of the curve. Join us from November 9 - 12, 2009. Register now! > http://p.sf.net/sfu/devconference > _______________________________________________ > Libusb-devel mailing list > Libusb-devel@... > https://lists.sourceforge.net/lists/listinfo/libusb-devel ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
Re: Asynchronous bulk transfer with QuickUSB FX2(linux) ?
by Michael Plante
::
Rate this Message:
>> and I think we have the same problem - it is not hard to get a lot of
average >> bandwidth (>20 MB/s), but the problem is when you do not have much >> device-side buffering, like only the cypress internal buffers, that you >> experience data loss. If this is the problem, I imagine it's easy to fix, since he's already using an FPGA. If the data is generated on-demand, you have dedicated logic to do it. If it's a stream of data back to the host from some bursty source, throw a FIFO into the FPGA. But is it the problem in this case? Michael ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
|
|
|
Re: Asynchronous bulk transfer with QuickUSB FX2 (linux) ?
by ChMeessen
::
Rate this Message:
Hello,
thank you very much for all your answers. First a bit more information on the device behind the usb. It is an X ray imaging prototype that we are developing in our lab. It has to download 560x960 16/32bit pixels images fast (~1MB/2MB data). It also has to upload some configurations, but performance is less critical. The current data throughput we get on Linux with the quick usb driver is about 3 MB/s and is too low. The colleague who implemented it is from the hardware side. I'm from the sofware side and I'm trying to figure out how I can get close to 30MB/s data transfer, which seams a reasonable goal according to the USB 2.0 specifications. I have to deal with the existing card and USB device. Looking at the FX2 data-sheet I have the impression it is a good controller. The choice was driven by the rational to have a plug and play USB controller which is true ... to some extend. The current code (3MB/s) uses the bulk synchronous transfer mode. The firmware was chosen for fastest data transfer rate and measured with the oscilloscope to be 480Mb/s. The oscilloscope, also showed huge waiting time between data sending which sounds logic with synchronous data transfer and a driver in user space. Now that libusb 1.0 support asynchronous transfer, I'm planning to give it a try. But I have some probably noob questions to ask. 1. Marko Cebokli reports 1% loss of packets. I was told USB bulk transfer is reliable and robust which means that data is transmitted uncorrupted and without loss to endpoints. If the USB controller detects corrupted packets it takes care to retransmit them. Is that true ? There is enough buffer space in our fpga card, so buffer overflow should not occur. Bus speed is much faster than USB speed. 2. Did anyone achieve to get 30MB/s throughput using libusb 1.0 or is this too ambitious ? Should I better consider writing a kernel driver which, I was told, would have much less latency ? 3. Is libusb 1.0 still depending on usbfs kernel driver ? If yes, is their a constrain on its version ? I'm using Ubuntu with the linux kernel 2.6.28-15 on which libusb 0.1-4 is installed. Here is what I see in the message log file when I plug the controller: Oct 9 11:37:50 xxx kernel: [330902.700012] usb 2-6: new high speed USB device using ehci_hcd and address 11 Oct 9 11:37:50 xxx kernel: [330902.833697] usb 2-6: configuration #1 chosen from 1 choice Here is the output of lsusb -v Bus 002 Device 011: ID 0fbb:0001 Bitwise Systems, Inc. Device Descriptor: bLength 18 bDescriptorType 1 bcdUSB 2.00 bDeviceClass 255 Vendor Specific Class bDeviceSubClass 255 Vendor Specific Subclass bDeviceProtocol 255 Vendor Specific Protocol bMaxPacketSize0 64 idVendor 0x0fbb Bitwise Systems, Inc. idProduct 0x0001 bcdDevice 2.11 iManufacturer 1 Bitwise Systems iProduct 2 QuickUSB QUSB2 Module v2.11.10 (Simple I/O) iSerial 0 bNumConfigurations 1 Configuration Descriptor: bLength 9 bDescriptorType 2 wTotalLength 32 bNumInterfaces 1 bConfigurationValue 1 iConfiguration 0 bmAttributes 0x80 (Bus Powered) MaxPower 500mA Interface Descriptor: bLength 9 bDescriptorType 4 bInterfaceNumber 0 bAlternateSetting 0 bNumEndpoints 2 bInterfaceClass 255 Vendor Specific Class bInterfaceSubClass 255 Vendor Specific Subclass bInterfaceProtocol 255 Vendor Specific Protocol iInterface 0 Endpoint Descriptor: bLength 7 bDescriptorType 5 bEndpointAddress 0x02 EP 2 OUT bmAttributes 2 Transfer Type Bulk Synch Type None Usage Type Data wMaxPacketSize 0x0200 1x 512 bytes bInterval 0 Endpoint Descriptor: bLength 7 bDescriptorType 5 bEndpointAddress 0x86 EP 6 IN bmAttributes 2 Transfer Type Bulk Synch Type None Usage Type Data wMaxPacketSize 0x0200 1x 512 bytes bInterval 0 Device Qualifier (for other device speed): bLength 10 bDescriptorType 6 bcdUSB 2.00 bDeviceClass 255 Vendor Specific Class bDeviceSubClass 255 Vendor Specific Subclass bDeviceProtocol 255 Vendor Specific Protocol bMaxPacketSize0 64 bNumConfigurations 1 Device Status: 0x0000 (Bus Powered) The product Id indicates simpleIO which is the firmware currently installed. This one allows to reach 480Mb/s sending speed. I see two endpoints. Apparently one for input and the other for outpu, each one configured with 512 byte (words?) buffer. In the DX2 datasheet it is stated that there should be four 512 buffers for each end point. What does the 1x512 mean ? Does it mean there is only one 512 buffer ? 4. I downloaded libusb 1.0.3 I would like to give it a try. Is there anything I have to do before being trying to use it ? For instance do something with udev so it somehow attaches usbfs driver to it ? 5. I wrote a piece of code compiled with libusb 0.1(pre-installed on Ubuntu) that locates the QuickUSB device and get its description informations. Apparently it can't get the product identification string unless it is run with root privilege (sudo). Same for lsusb. Is this the normal behavior ? Ch. Meessen ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
Re: Asynchronous bulk transfer with QuickUSB FX2 (linux) ?
by Alan Stern
::
Rate this Message:
On Fri, 9 Oct 2009, Christophe Meessen wrote:
> 1. Marko Cebokli reports 1% loss of packets. I was told USB bulk > transfer is reliable and robust which means that data is transmitted > uncorrupted and without loss to endpoints. If the USB controller detects > corrupted packets it takes care to retransmit them. Is that true ? Yes. Marko was referring to the possibility that the buffer space on the device might be too small, so that packets overflow the buffer and are lost before they can be sent to the host (IN) or before the device can act on them (OUT). > There is enough buffer space in our fpga card, so buffer overflow should > not occur. Bus speed is much faster than USB speed. > > 2. Did anyone achieve to get 30MB/s throughput using libusb 1.0 or is > this too ambitious ? Should I better consider writing a kernel driver > which, I was told, would have much less latency ? Bus latency for kernel drivers is no smaller than for properly-written asynchronous userspace drivers. I don't know what speeds people have attained using libusb-1.0. In the end, it is probably limited more by the device than by the host. > 3. Is libusb 1.0 still depending on usbfs kernel driver ? Yes. > If yes, is > their a constrain on its version ? I'm using Ubuntu with the linux > kernel 2.6.28-15 on which libusb 0.1-4 is installed. Libusb-1.0 certainly will work with your kernel. While there are constraints, you shouldn't have to worry about them. I imagine libusb would work with any kernel later than 2.6.9, and probably with earlier kernels too. > Here is the output of lsusb -v ... > Endpoint Descriptor: > bLength 7 > bDescriptorType 5 > bEndpointAddress 0x02 EP 2 OUT > bmAttributes 2 > Transfer Type Bulk > Synch Type None > Usage Type Data > wMaxPacketSize 0x0200 1x 512 bytes > bInterval 0 > Endpoint Descriptor: > bLength 7 > bDescriptorType 5 > bEndpointAddress 0x86 EP 6 IN > bmAttributes 2 > Transfer Type Bulk > Synch Type None > Usage Type Data > wMaxPacketSize 0x0200 1x 512 bytes > bInterval 0 > The product Id indicates simpleIO which is the firmware currently > installed. This one allows to reach 480Mb/s sending speed. I see two > endpoints. Apparently one for input and the other for outpu, each one > configured with 512 byte (words?) buffer. Bytes. And the number is not a buffer size; it is the maximum packet size. > In the DX2 datasheet it is > stated that there should be four 512 buffers for each end point. What > does the 1x512 mean ? Does it mean there is only one 512 buffer ? No. The "1x" means that the endpoint is not capable of "high-bandwidth" operation. This is expected, because "high-bandwidth" applies only to periodic endpoints (interrupt and isochronous). In full-speed or low-speed devices, interrupt and isochronous transfers are limited to one transaction per frame. In high-speed devices, non-"high-bandwidth" interrupt and isochronous endpoints are likewise limited to one transaction per microframe (that's what the "1x" refers to). However "high-bandwidth" interrupt and isochronous endpoints are allowed up to three transactions per microframe; that's where the name "high-bandwidth" comes from. By contrast, high-speed bulk endpoints are allowed as many as 13 transactions per microframe. Hence they can enjoy even more bandwidth than the so-called "high-bandwidth" endpoints. > 4. I downloaded libusb 1.0.3 I would like to give it a try. Is there > anything I have to do before being trying to use it ? For instance do > something with udev so it somehow attaches usbfs driver to it ? You don't need to do anything. > 5. I wrote a piece of code compiled with libusb 0.1(pre-installed on > Ubuntu) that locates the QuickUSB device and get its description > informations. Apparently it can't get the product identification string > unless it is run with root privilege (sudo). Same for lsusb. Is this the > normal behavior ? This depends on the permissions for the device file: /dev/bus/usb/BBB/DDD. If write access is restricted to root then you will need root privilege to do anything nontrivial with the device. The device file's permissions are generally determined by a udev rule. If you like, you can add a rule to give your QuickUSB device more relaxed permissions. Alan Stern ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
Re: Asynchronous bulk transfer with QuickUSB FX2 (linux) ?
by Ray S
::
Rate this Message:
At 01:28 PM 10/9/2009 +0200, you wrote:
The colleague who implemented it is from the hardware side. I'm from the Hello, Look at Marko Cebokli's pages: http://lea.hamradio.si/~s57uuu/uuusb/index.htm#fifo where he implements 30MB/s http://lea.hamradio.si/~s57uuu/uuusb/uuusb_software.htm#bandwidth the main issue he reports finding is the occasional long wait for bulk_read. Since the internal buffer is max of 4k, an external FIFO would be needed for high speed unless you run RTOS. I've implemented his simple_prg_rd.c in Python/libusb, if you're interested in Python. I also have an FX2 (Sermod-100 board) and am implementing a high frame rate camera with his suggestions, and with an external FIFO like the http://www.semiconductorstore.com/cart/pc/viewPrd.asp?idproduct=10565#details Also see this FPGA project with FX2 http://www.myhdl.org/doku.php/users:cfelton:projects:usbp and http://www.acquiredevices.com/usbsoftware.html if you have not looked at their code yet... More bookmarks: http://lea.hamradio.si/~s57uuu/uuusb/uuusb_software.htm http://github.com/mulicheng/fx2lib/tree/master http://allmybrain.com/tag/fx2/ http://volodya-project.sourceforge.net/fx2_programmer.php http://www.fpgaz.com/usbp/ http://www.triplespark.net/elec/periph/USB-FX2/software/ R&D Manager - Cognitive Vision 8580 Production Ave. Ste. B, San Diego, CA 92121 rays@CognitiveVision.com ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
Re: Asynchronous bulk transfer with QuickUSB FX2 (linux) ?
by azdes
::
Rate this Message:
Using libusb 1.0, on Linux, with asyncrhonous bulk transfers on a PLX
USB chip, we were able to comfortably sustain between 25MBytes and 28MBytes per second. This was even without any ping ponging of buffers (and writing out to a file on disk on the user side). You just have to make sure your firmware side has the IN data ready to go and can turn around quickly. And for that operation, you may want to keep other USB devices off that particular bus. Our PC side was running java through jni to a C++ dll that interacts with libusb. Good luck. -----Original Message----- From: Christophe Meessen [mailto:meessen@...] Sent: Friday, October 09, 2009 4:28 AM Cc: Libusb List Subject: Re: [Libusb-devel] Asynchronous bulk transfer with QuickUSB FX2 (linux) ? Hello, thank you very much for all your answers. First a bit more information on the device behind the usb. It is an X ray imaging prototype that we are developing in our lab. It has to download 560x960 16/32bit pixels images fast (~1MB/2MB data). It also has to upload some configurations, but performance is less critical. The current data throughput we get on Linux with the quick usb driver is about 3 MB/s and is too low. The colleague who implemented it is from the hardware side. I'm from the sofware side and I'm trying to figure out how I can get close to 30MB/s data transfer, which seams a reasonable goal according to the USB 2.0 specifications. I have to deal with the existing card and USB device. Looking at the FX2 data-sheet I have the impression it is a good controller. The choice was driven by the rational to have a plug and play USB controller which is true ... to some extend. The current code (3MB/s) uses the bulk synchronous transfer mode. The firmware was chosen for fastest data transfer rate and measured with the oscilloscope to be 480Mb/s. The oscilloscope, also showed huge waiting time between data sending which sounds logic with synchronous data transfer and a driver in user space. Now that libusb 1.0 support asynchronous transfer, I'm planning to give it a try. But I have some probably noob questions to ask. 1. Marko Cebokli reports 1% loss of packets. I was told USB bulk transfer is reliable and robust which means that data is transmitted uncorrupted and without loss to endpoints. If the USB controller detects corrupted packets it takes care to retransmit them. Is that true ? There is enough buffer space in our fpga card, so buffer overflow should not occur. Bus speed is much faster than USB speed. 2. Did anyone achieve to get 30MB/s throughput using libusb 1.0 or is this too ambitious ? Should I better consider writing a kernel driver which, I was told, would have much less latency ? 3. Is libusb 1.0 still depending on usbfs kernel driver ? If yes, is their a constrain on its version ? I'm using Ubuntu with the linux kernel 2.6.28-15 on which libusb 0.1-4 is installed. Here is what I see in the message log file when I plug the controller: Oct 9 11:37:50 xxx kernel: [330902.700012] usb 2-6: new high speed USB device using ehci_hcd and address 11 Oct 9 11:37:50 xxx kernel: [330902.833697] usb 2-6: configuration #1 chosen from 1 choice Here is the output of lsusb -v Bus 002 Device 011: ID 0fbb:0001 Bitwise Systems, Inc. Device Descriptor: bLength 18 bDescriptorType 1 bcdUSB 2.00 bDeviceClass 255 Vendor Specific Class bDeviceSubClass 255 Vendor Specific Subclass bDeviceProtocol 255 Vendor Specific Protocol bMaxPacketSize0 64 idVendor 0x0fbb Bitwise Systems, Inc. idProduct 0x0001 bcdDevice 2.11 iManufacturer 1 Bitwise Systems iProduct 2 QuickUSB QUSB2 Module v2.11.10 (Simple I/O) iSerial 0 bNumConfigurations 1 Configuration Descriptor: bLength 9 bDescriptorType 2 wTotalLength 32 bNumInterfaces 1 bConfigurationValue 1 iConfiguration 0 bmAttributes 0x80 (Bus Powered) MaxPower 500mA Interface Descriptor: bLength 9 bDescriptorType 4 bInterfaceNumber 0 bAlternateSetting 0 bNumEndpoints 2 bInterfaceClass 255 Vendor Specific Class bInterfaceSubClass 255 Vendor Specific Subclass bInterfaceProtocol 255 Vendor Specific Protocol iInterface 0 Endpoint Descriptor: bLength 7 bDescriptorType 5 bEndpointAddress 0x02 EP 2 OUT bmAttributes 2 Transfer Type Bulk Synch Type None Usage Type Data wMaxPacketSize 0x0200 1x 512 bytes bInterval 0 Endpoint Descriptor: bLength 7 bDescriptorType 5 bEndpointAddress 0x86 EP 6 IN bmAttributes 2 Transfer Type Bulk Synch Type None Usage Type Data wMaxPacketSize 0x0200 1x 512 bytes bInterval 0 Device Qualifier (for other device speed): bLength 10 bDescriptorType 6 bcdUSB 2.00 bDeviceClass 255 Vendor Specific Class bDeviceSubClass 255 Vendor Specific Subclass bDeviceProtocol 255 Vendor Specific Protocol bMaxPacketSize0 64 bNumConfigurations 1 Device Status: 0x0000 (Bus Powered) The product Id indicates simpleIO which is the firmware currently installed. This one allows to reach 480Mb/s sending speed. I see two endpoints. Apparently one for input and the other for outpu, each one configured with 512 byte (words?) buffer. In the DX2 datasheet it is stated that there should be four 512 buffers for each end point. What does the 1x512 mean ? Does it mean there is only one 512 buffer ? 4. I downloaded libusb 1.0.3 I would like to give it a try. Is there anything I have to do before being trying to use it ? For instance do something with udev so it somehow attaches usbfs driver to it ? 5. I wrote a piece of code compiled with libusb 0.1(pre-installed on Ubuntu) that locates the QuickUSB device and get its description informations. Apparently it can't get the product identification string unless it is run with root privilege (sudo). Same for lsusb. Is this the normal behavior ? Ch. Meessen ------------------------------------------------------------------------ ------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
Asynchronous callbacks for multiple tranfers
by Marko Cebokli
::
Rate this Message:
I am playing now with async transfers, and have one question:
If, since the last call of usb_handle_events, several transfers were completed, does the next call to usb_handle_events call the callback function once for every transfer that was completed in the mean time, or only once, so that I must I call usb_handle_events several times to process multiple completed tranfers? Marko Cebokli ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
Re: Asynchronous callbacks for multiple tranfers
by Marko Cebokli
::
Rate this Message:
ops, one more question...
is it OK to re-submit a transfer from within the calback function? Marko Cebokli On Saturday 10 October 2009 17:08:20 Marko Cebokli wrote: > I am playing now with async transfers, and have one question: > > If, since the last call of usb_handle_events, several transfers were > completed, > > does the next call to usb_handle_events call the callback function once for > every transfer that was completed in the mean time, or only once, so that I > must I call usb_handle_events several times to process multiple completed > tranfers? > > Marko Cebokli > > --------------------------------------------------------------------------- >--- Come build with us! The BlackBerry(R) Developer Conference in SF, CA is > the only developer event you need to attend this year. Jumpstart your > developing skills, take BlackBerry mobile applications to market and stay > ahead of the curve. Join us from November 9 - 12, 2009. Register now! > http://p.sf.net/sfu/devconference > _______________________________________________ > Libusb-devel mailing list > Libusb-devel@... > https://lists.sourceforge.net/lists/listinfo/libusb-devel ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
Re: Asynchronous callbacks for multiple tranfers
by Hans Petter Selasky
::
Rate this Message:
Hi,
On Saturday 10 October 2009 17:21:00 Marko Cebokli wrote: > is it OK to re-submit a transfer from within the calback function? Typically yes, at least on FreeBSD. Except in case the transfer is cancelled by your software, you should not re-submit the transfer. >does the next call to usb_handle_events call the callback function once for >every transfer that was completed in the mean time, or only once, so that I >must I call usb_handle_events several times to process multiple completed >tranfers? That depends on the implementation. You should call the usb_handle_events() function as long as the libusb_wait_for_event() function returns and don't make any assumptions about how many callbacks you get per usb_handle_events() call. --HPS ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
|
|
|
|
|
Re: Asynchronous callbacks for multiple tranfers
by Daniel Drake
::
Rate this Message:
Marko Cebokli wrote:
> I am playing now with async transfers, and have one question: > > If, since the last call of usb_handle_events, several transfers were > completed, > > does the next call to usb_handle_events call the callback function once for > every transfer that was completed in the mean time, or only once, so that I > must I call usb_handle_events several times to process multiple completed > tranfers? Don't read too much into this.. just call it in a loop until the events that you are anticipating have occurred. And yes it is fine to resubmit transfers from the completion handler. Daniel ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Libusb-devel mailing list Libusb-devel@... https://lists.sourceforge.net/lists/listinfo/libusb-devel |
| Free embeddable forum powered by Nabble | Forum Help |
