On 8/31/2018 10:20 AM, Himanshu Jha wrote:
> On Fri, Aug 31, 2018 at 10:39:07PM +0530, Himanshu Jha wrote:
>> On Fri, Aug 31, 2018 at 07:55:36AM -0700, David Frey wrote:
>>> On 8/29/2018 10:35 AM, David Frey wrote:
>>>> Hi,
>>>>
>>>> I just want to float an idea and see if it gets any support.
>>>>
>>>> In struct regmap_config, there is a member "bool use_single_rw" which is
>>>> documented as follows:
>>>>
>>>> "If set, converts the bulk read and write operations into a series of
>>>> single read and write operations. This is useful for device that does
>>>> not support bulk read and write."
>>>>
>>>> My idea is to split this into:
>>>> bool use_single_read;
>>>> bool use_single_write;
>>>>
>>>> I think this would be useful because at least one device that I'm using
>>>> (Bosch bmi160 IMU) supports bulk reads, but not bulk writes. It would
>>>> be convenient to be able to use regmap_bulk_read and regmap_bulk_write
>>>> and have only regmap_bulk_write split the writes apart. This change
>>>> would also support the case where a device supports bulk writes, but not
>>>> bulk reads. I don't know of such a device, but I haven't made any
>>>> attempt to find one.
>>>>
>>>> What do you think?
>
> Btw, if device doesn't support a bulk write and instead only supports a bulk
> read as we see in BMI160, wouldn't the following suffice the problem:
>
> .use_single_rw = true,
> .can_multi_write = false, // or probably not setting it up(clear)
>
> 313 * @can_multi_write: If set, the device supports the multi write mode of bulk
> 314 * write operations, if clear multi write requests will be
> 315 * split into individual write operations
I don't think the can_multi_write flag is relevant here. Bulk writes
have the form: {regN, dataN, dataN+1, dataN+2, ...}. Multi-write is
used by regmap_multi_reg_write() and is this form: {regaA, dataA, regF,
dataF, regC, dataC, ...}.
