Hi Albert,
On 06/25/2012 01:34 PM, Albert ARIBAUD wrote:
Hi Aneesh,
BTW, I agree that enabling un-aligned access is not a bad idea.
Just being "not a bad idea" is not enough for me to accept this. It will have to be the sole sound solution to a problem, and at this point, I do not think it is as far as USB structure mis-alignement issues are concerned.
My point is that enabling un-aligned accesses in itsown merit is not a bad idea, not as a solution to this problem. I have seen it being enabled in HLOS environment. TI's Symbian port for instance had it enabled for OMAP3. I don't know why Linux too shoudln't take advantage of such hw features. Perhaps you don't want to take it at this point of time to force the real solution to the USB problem, which is reasonable.
What is the (non-contrived) problem to which allowing mis-aligned accesses would be a solution?
memcpy() when there is a mismatch in the alignment of source and destination buffers. Let's say the source buffer is 4 byte aligned but the destination buffer is only 2 byte aligned. I believe relaxed alignment requirements will help in writing an accelerated memcpy routine for this case.
Again, my point is that as a platform software provider I would like to enable such features to make maximum things work on my platform, where as the developer of a generic sw module should probably avoid depending on such features to ensure maximum portability.
br, Aneesh