# Really long spiService call **URL:** \ **Category:** Legato Application Framework\ **Created:** [June 18, 2017, 6:16pm UTC](https://forum.legato.io/t/really-long-spiservice-call/2503 "2017-06-18T18:16:29Z")\ **Posts on this page:** 4\ **Page:** 1
**Author:** ![GuillaumeCarel](https://sea1.discourse-cdn.com/flex019/user_avatar/forum.legato.io/guillaumecarel/32/151_2.png) [@GuillaumeCarel](https://forum.legato.io/u/GuillaumeCarel)\ **Post date:** [June 18, 2017, 6:16pm UTC](https://forum.legato.io/t/really-long-spiservice-call/2503/1 "2017-06-18T18:16:29Z")
Hello everyone, I’m using the iot card extension for can bus develops for Legato, for that I set up the SPI connection with the spiService. But I’m having an issue : the time to recover the interruption from the iot card (when a CAN message is receive) and to get the CAN message through the SPI interface is really really long. I measured with an oscilloscope the different signals on the card and the time needed to get the CAN message is 2000 us on average. With this value I can’t expect to read a CAN bus faster than 500Hz … this is far too low. I measured with a c function the time to execute 2 spi instructions (read , read/write) and it’s 1200us (833 Hz)… this is absurd because an SPI connection is supposed to go to 1 MHz at least. I suppose the delay com from the event Loop that call the differents functions (interruption,spiService). Is there a way to reduce this delay and alloy my connection to be quicker ? ---
**Author:** ![dfrey](https://sea1.discourse-cdn.com/flex019/user_avatar/forum.legato.io/dfrey/32/563_2.png) [@dfrey](https://forum.legato.io/u/dfrey)\ **Post date:** [June 19, 2017, 10:37pm UTC](https://forum.legato.io/t/really-long-spiservice-call/2503/2 "2017-06-19T22:37:06Z")
You could use the `spiLibrary` component directly rather than interacting with `spiService`. Take a look at the [`Component.cdef` for `spiService`](https://github.com/legatoproject/legato-af/blob/master/components/spiService/Component.cdef) to see an example of how to reference `spiLibrary` directly. The interface presented by `spiLibrary` is nearly identical to `spiService` except that you have to obtain the file descriptor to the SPI device yourself and you pass around the fd instead of a `le_spi_DeviceHandleRef_t`. The advantage of using `spiLibrary` directly instead of going through `spiService` is that you won’t pay IPC overhead for each SPI function called. Having said that, I’m not sure if IPC is the limiting factor. If you get a chance to try it and measure the results, please post a reply. ---
**Author:** ![GuillaumeCarel](https://sea1.discourse-cdn.com/flex019/user_avatar/forum.legato.io/guillaumecarel/32/151_2.png) [@GuillaumeCarel](https://forum.legato.io/u/GuillaumeCarel)\ **Post date:** [June 20, 2017, 10:10am UTC](https://forum.legato.io/t/really-long-spiservice-call/2503/3 "2017-06-20T10:10:46Z")
Thank you very much dfrey, It works far better now ! I created my own unsandboxed application (to be able to read the file) then I created a component where I put [le\_spiLibrary.c](https://github.com/legatoproject/legato-af/blob/master/components/spiLibrary/le_spiLibrary.c) and [le\_spiLibrary.h](https://github.com/legatoproject/legato-af/blob/master/components/spiLibrary/le_spiLibrary.h). I added my own files (main.c …). At the beginning the execution time was even worse than before : 3000us (for one le\_spiLib\_WriteReadHD instruction). But then I commented all the LE\_DEBUG and LE\_ERROR in le\_spiLibrary.c. And now my app runs faster: 500us (for one le\_spiLib\_WriteReadHD instruction). (Note that the **SPI communication itself** is about 130 us long) Here is the modified `le_spiLibrary.c`: [le\_spiLibrary.c](https://forum.legato.io/uploads/short-url/3FyVjtbemUwWgc0enM32xpAp2PU.c) (9.2 KB) Regards. Guillaume ---
**Author:** ![GuillaumeCarel](https://sea1.discourse-cdn.com/flex019/user_avatar/forum.legato.io/guillaumecarel/32/151_2.png) [@GuillaumeCarel](https://forum.legato.io/u/GuillaumeCarel)\ **Post date:** [June 20, 2017, 1:41pm UTC](https://forum.legato.io/t/really-long-spiservice-call/2503/4 "2017-06-20T13:41:40Z")
But i still have a delay problem … The CAN controller interruption (when a CAN message is receive) is process by Legato , and it’s too long : 700us. 1. Legato see the interruption 2. Legato put the handler associate to the interruption in the event\_Loop 3. The handler is on top of the loop and begin 4. I use the SPI to look the message … It’s really long and not natural because the interruption go up in the framework and then go down (from hardware to software). I don’t understand why the SPI driver isn’t able to register a handler on an interruption, to be able to stay at the same hardware level and ignore the framework …