Re: [libusb] Asynchronous transfer using libusb-1.0 on ubuntu

[Ларионов Д. <scu...@ya...>]

> You can't have overlapping transfers. Remember that the model
> anticipates physical devices, which have one read head that can only be
> one place at a time. It's all serialized.

It's almost a decade since the last time I touched Mass Storage Class so I can be entirely wrong, please correct me in this case.
But isn't it just a (very) thin wrapper over the SCSI protocol?
And SCSI *does* support overlapping transfers; in fact, it (theoretically) can even increase the performance because the internal controller of the media will be able to re-arrange requests (e.g. to optimize the movement of HDD heads).

-- 
Даниил Ларионов


16.08.2017, 22:42, "Tim Roberts" <ti...@pr...>:
> Anand Nahar wrote:
>>  Why? Not trying to be snotty, but if you're reading large blocks, the
>>  gain from asynchronous processing is not that large.
>>
>>    * I was wondering, if I have multiple asynchronous transfers lined
>>      up, it would utilize the bandwidth all the time (may be 2 or more
>>      transfers overlap) and not wait for the next transfer to be
>>      initiated. I just wanted to make sure that there is no idle
>>      periods in between 2 consecutive transfers. Right now I have
>>      multiple threads transferring data synchronously to different
>>      devices but this still doesn't guarantee no idle period.
>
> You can't have overlapping transfers. Remember that the model
> anticipates physical devices, which have one read head that can only be
> one place at a time. It's all serialized.
>
>>  "can I submit a second CBW before the first transfer is complete?
>>
>>    * I have actually tried this. I submit multiple CBW’s and read the
>>      data in their callback functions. It doesn’t seem to work and I
>>      understand why that doesn’t work. I was curious on whether there
>>      is some other approach to this.
>
> I'm a little surprised by that. I would have expected you could submit
> CBW, read, CBW, read, in that order, all in advance, and it work just
> fine. However, as I said, I couldn't find support in the spec for that.
>
> --
> Tim Roberts, ti...@pr...
> Providenza & Boekelheide, Inc.
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> libusb-devel mailing list
> lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libusb-devel