Spi write functions return too late

[adi1133]

I want to use hardware spi, I only need to write, I want to use to dump the data into the write buffer and then go on with the main program without waiting to do the full write before resuming. Is there any way to do this besides inserting ASM ?

This is the test application I made, under an oscilloscope I can see that PIN_A1 toggles only after the transfer is done, I'd like this to happen while transferring.

[Ttelmah]

[adi1133]

Thank you for the fast response, your response is working as intended ,indeed a selectable option for this would be nice.
I guess this works for i2c too, since it uses the same SSP hardware module.
Does getenv("SFR:xxxxx") work with all the registers found in the datasheet of the device ?

[Ttelmah]

It should do (both for SSP, and the use of getenv).

Only problem sometimes is that the names may not be 100% obvious.....
Years ago, I moaned (both here, and to CCS), about the behaviour of the SSP code, and over time, the change was made for interrupt use.
Might be worth another person pointing this out to them.

The amazing thing (to me), is how few people seem to realise the limitations of the standard approach....

Best Wishes

[asmboy]

[FvM]

Most SPI applications are controlling a CS pin in software and thus need to wait for completion of the frame,
as the present spi_write() implementation does. I fear that creating another configuration parameter for
this feature would promote user confusion.

I think. for those applications with different requirements writing your own code isn't asking too much.
Just my five cents.

[Ttelmah]

I don't see that it needs to be confusing _provided it is properly documented_.... (unfortunately, something that CCS is not 'great' at....).

However look (for example), at the read_adc function. Default behaviour to do everything. trigger the read,wait for it to complete, and return the value.
However one is given the option of adding a parameter like 'ADC_START_ONLY', and it then just starts the ADC.
So something like:

spi_write(val);

Behaves exactly as the current implementation, but:

spi_write(val,DO_NOT_WAIT);

Would be simple to implement (I've actually done it, 'overloading' the existing function, since the compiler allows this now), and would mean that one didn't have to do your own functions.

I synchronise SPI, with a slave select for the whole block, rather than for the individual bytes - this is common for a lot of third party chips, so want to get on with other things while the writing takes place.

Best Wishes

[FvM]